<%BANNER%>

Spatial and Demographic Modeling Techniques Applied to the Longleaf Pine (Pinus palustris) Ecosystem of North Central Florida

xml version 1.0 encoding UTF-8
REPORT xmlns http:www.fcla.edudlsmddaitss xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.fcla.edudlsmddaitssdaitssReport.xsd
INGEST IEID E20101122_AAAABP INGEST_TIME 2010-11-23T02:40:37Z PACKAGE UFE0012041_00001
AGREEMENT_INFO ACCOUNT UF PROJECT UFDC
FILES
FILE SIZE 494 DFID F20101122_AAAUVG ORIGIN DEPOSITOR PATH loudermilk_e_Page_01.txt GLOBAL false PRESERVATION BIT MESSAGE_DIGEST ALGORITHM MD5
e72efe5bd3ab439184654fc4ed9cb91b
SHA-1
0adcda35170c4b9e2dab6acb3c93c904ba158099
5813 F20101122_AAAUUS loudermilk_e_Page_59.pro
3a74e8d91cee5ae409db90b599960907
d3676b1a68d8e1eb3c32c62854ac460c7d393c13
10305 F20101122_AAAVBB loudermilk_e_Page_83.QC.jpg
ae313729d4067a9f6b03126689ef38cb
225603ed547465ab0d5644d290b785eb72688e00
23068 F20101122_AAAVAN loudermilk_e_Page_71.QC.jpg
874501f9009de408027d50e76afff5e8
ad080d7542d79b41f46a472ee9cbbbf888baf6c0
84 F20101122_AAAUVH loudermilk_e_Page_02.txt
603045c45abdff2ae68a8b292a188e83
fefab5590aa8c3daf5c940dd11c3dda48244d04c
51795 F20101122_AAAUUT loudermilk_e_Page_61.pro
a284e50c024a21093b4110026fd92ee9
a9ed4c7f770dd59db099d9345d4efdea24f99c3e
3133 F20101122_AAAVBC loudermilk_e_Page_83thm.jpg
762c857e380f606619e8418edb139b95
a8961c254f073b7235ffd6c53a039a5b0af3e9dc
6459 F20101122_AAAVAO loudermilk_e_Page_71thm.jpg
afcf5f8e4a0e794448f44d2655d24ed8
95a7924a19f334a228e74916d15684376fd8eb6e
1397 F20101122_AAAUVI loudermilk_e_Page_04.txt
79f90dc8b29b1667e0390ae84426d5fb
a7d23e592887e22febd432a5a615952d9ef0c3f0
50459 F20101122_AAAUUU loudermilk_e_Page_62.pro
d086453a83710a4bbc36b8b54ebaac22
8443d9624c259215f8a66633f0a7f54c13cd6d90
10881 F20101122_AAAVBD loudermilk_e_Page_84.QC.jpg
35b4eee383406ed894b3773866724784
c3c61d9c0d0533567db910da8e560e870f211ced
1122 F20101122_AAAUVJ loudermilk_e_Page_07.txt
5259965dfabeae363f6da4bff58125f8
1f565436571fbc4a28ea7d88f23dfa35cc6081fd
51705 F20101122_AAAUUV loudermilk_e_Page_65.pro
ebf9f5155545c2e7edaf16c2bf157252
bb47c6390d501751d22e05993f1bdded00b7cb24
22206 F20101122_AAAVBE loudermilk_e_Page_85.QC.jpg
5727f69b973a60c41510db36f0d3c068
8af241acc93b1b7ad4fc335c53736408b23bcf5b
24403 F20101122_AAAVAP loudermilk_e_Page_72.QC.jpg
323d87d10b249f38ade800853f9b2337
578850c4f3e77c31f90b15fd060fc6d065ee09a6
2573 F20101122_AAAUVK loudermilk_e_Page_08.txt
869e227307db90c142cdf04711b7d71b
4256f6cebe8f7d7532afbc2f26f04bfcc9e15a20
52578 F20101122_AAAUUW loudermilk_e_Page_66.pro
697647b8664b88a13a85aa484d0e4852
7d162b906fe96f14aa322fddb74f60ee9656f19b
6125 F20101122_AAAVBF loudermilk_e_Page_85thm.jpg
6084fe089d68d508ad7a8f6bc9122ed1
5615db1adcf1bdee1ae3928989887a02e596bf04
6571 F20101122_AAAVAQ loudermilk_e_Page_72thm.jpg
2eb7b0bfbf6590290fda455a0f603b74
fdca29f82c9ba20e9490e67da613d85797a71b45
30877 F20101122_AAAUUX loudermilk_e_Page_69.pro
1bb39424c99e385dbe1840b96d4ba7c7
0691057350a46fbc183b45a8f0472f64d57ebbed
26007 F20101122_AAAVBG loudermilk_e_Page_86.QC.jpg
e3512cd87307dc4dfd6872ad32d070da
0061fb3576e0d0cbe267947494dcd5a4c5684e86
1033 F20101122_AAAUWA loudermilk_e_Page_37.txt
c6d27f72bdc6df3838665637889c7189
6116c4e4f59df016033c708e47458d71a3f7c153
11177 F20101122_AAAVAR loudermilk_e_Page_74.QC.jpg
ba55f2cc6b012775ad633f1d2bc3b92f
76027987ac5ec105ea9581561700a2a75ace615d
1989 F20101122_AAAUVL loudermilk_e_Page_11.txt
f940cff08a6f1b3b16c0b97ab97533b6
dfecc32d0f5493971eb514f4fc26f584a6b31957
47305 F20101122_AAAUUY loudermilk_e_Page_71.pro
1af1bef23b4b120efdb55f0fcc03ead2
7cb332d43ec7f7af2164e2055078220b03955428
24913 F20101122_AAAVBH loudermilk_e_Page_87.QC.jpg
55f77413f11e628b28080ba7b7acf89f
2d1813c3c9ae03c84153f56814051fb15eb0e572
1523 F20101122_AAAUWB loudermilk_e_Page_38.txt
644dfb64e06ed6f3214f87b1b6da0668
f18f61b40722eab8320493e5e9d0566454941846
3420 F20101122_AAAVAS loudermilk_e_Page_74thm.jpg
35400eb4a2da626d20516e3be16517d8
0144f794255bfe2c9d344a294f89bdc0f77a72ab
1760 F20101122_AAAUVM loudermilk_e_Page_14.txt
86d6b2a5d22049f74a8a3f7ab11badcc
9daadac7bc86dcfc1dd49b1e5a1781ea2232cabf
51709 F20101122_AAAUUZ loudermilk_e_Page_72.pro
2048903643b9e33d644389b81e90554a
9fa3ca22d31bab88d321b01854989d72bf6f679c
7032 F20101122_AAAVBI loudermilk_e_Page_87thm.jpg
fc2d68a9391b8c198fedeb05f8141441
52e753dd134e01eb7821d19e584965cc0e33f1b9
1437 F20101122_AAAUWC loudermilk_e_Page_39.txt
da215c04eea327945622381084fd4b56
f7f86394eefff0dc63d025d988b0c1123720f92e
4913 F20101122_AAAVAT loudermilk_e_Page_75thm.jpg
678947c4c92c6401c2521ec25077a7d9
07b4c4c51d24c0fefdd2f8208e2c0e558a55309a
1894 F20101122_AAAUVN loudermilk_e_Page_15.txt
f6f34b3d26b6aff38971a46f899b6c1b
c7a4f3430bb6a858cd5aab84e041139f7c81b7be
26916 F20101122_AAAVBJ loudermilk_e_Page_88.QC.jpg
90bdec2c4d6c4572657ed97049b0552c
aa800b856909c3693abb2730263c23a0116eed7f
2046 F20101122_AAAUWD loudermilk_e_Page_41.txt
3521226b75a41b6685403e9b71bca296
0de16b768466f5f75fd9f31557eed65a25661ec8
18414 F20101122_AAAVAU loudermilk_e_Page_76.QC.jpg
d338dc09e545b78051c61017a3c0eb99
d9c9d5296e26d6be85e29e032777663b1afc8214
2031 F20101122_AAAUVO loudermilk_e_Page_17.txt
3f92f8ea3b0f5194a42fde4bd6f8914d
574105480d97f1bf84ae1a6439901c1f2a8299ce
6834 F20101122_AAAVBK loudermilk_e_Page_88thm.jpg
a9a3cea18e540d2ca648c105dabff3f3
b489ecbb86b8f4c263a7b8519bb6cb2386ba4717
1756 F20101122_AAAUWE loudermilk_e_Page_43.txt
c6e64bc3285f6228d79e41fb897117a6
1236289a87bc0bdf3cf8e616f3e87ccfb03a6e9a
5572 F20101122_AAAVAV loudermilk_e_Page_76thm.jpg
c5ffd8ab4feaa47223cc18b3ccc33db5
9d5d7119fbf7f2446ff51b11ff3983a134c6022e
1925 F20101122_AAAUVP loudermilk_e_Page_18.txt
76d0ab8e51767c758c3578857291d4f1
3a2559e2aff709a267dcdadacd06c463be2b191c
6901 F20101122_AAAVBL loudermilk_e_Page_89thm.jpg
8e28867e5576f09cc2585f91ab16bdf8
80ebbd66251b626a5a6f272f8d92f15e25b9acdb
1982 F20101122_AAAUWF loudermilk_e_Page_48.txt
dfb7b33cda527edeeed93ea219125a64
8a06246d45abe0c71bcfe390d89083db9f07b96a
10665 F20101122_AAAVAW loudermilk_e_Page_79.QC.jpg
1cf64b7e5cc254be5a888fc988c05977
27bc69e06df3e7a022cdf097ebfbe99dda7fd047
1932 F20101122_AAAUVQ loudermilk_e_Page_22.txt
ac679178971544169874f66ade0a1f78
de067715e635e4df8791aecbb6fee169111388f7
4042 F20101122_AAAVBM loudermilk_e_Page_90thm.jpg
1abd2abdd11ed8beb3d76b3924df203b
c7c4d50964b93e592e47827cdcd2d7e245592b1b
16758 F20101122_AAAVAX loudermilk_e_Page_80.QC.jpg
b653d13de6ddd3d6a6d02ab5fae8b247
aa1e21fb1ec5100e97a6e073e11590fcf54db9b4
1225 F20101122_AAAUVR loudermilk_e_Page_24.txt
7f73706e014530ab40d81e5d144ef5ef
266cf4acd6221e9d364ba0093b81232dcb8cac5b
107744 F20101122_AAAVBN UFE0012041_00001.mets FULL
ce70018372804fdde3a8736c78fa944f
81972eb19f43e9da70f5f60d97a420bb6f00b5bb
1771 F20101122_AAAUWG loudermilk_e_Page_51.txt
78b140bfb82aff64a3fbfd40b2ddf21f
8d2772454df414ef7cf77fa1fdb7c3c11f5319eb
4617 F20101122_AAAVAY loudermilk_e_Page_80thm.jpg
61f022bb63cd84993cfc359d3b2bbb07
edaf761b0bc734d9a630f29a9afc6ddcdc8516bd
2034 F20101122_AAAUVS loudermilk_e_Page_25.txt
94b78c490ec1b847e39733658150d3a3
fe5f0a3a66e857681b31a40878f16da5b7926c3f
1843 F20101122_AAAUWH loudermilk_e_Page_53.txt
e5937821e6ff73c8b9b472e850535817
b586ccc9b27e396b2879357add80c37bb42b8c20
8200 F20101122_AAAVAZ loudermilk_e_Page_82.QC.jpg
089e57bbe9c180368d0220c5249c19e6
0c896459bd69d543ca84c9941f3c0df0610e793d
1765 F20101122_AAAUVT loudermilk_e_Page_27.txt
9a66fc5e75c42a6536cd0a4ed2385f7f
e28edbfa5a2ad597f6ecffbffa0a7b61abbb5f98
659 F20101122_AAAUWI loudermilk_e_Page_54.txt
c19802fcb1e2f9ad264ac1a4e1deaf58
72720d33a3499373698af73303acba2005c9b36e
1913 F20101122_AAAUVU loudermilk_e_Page_29.txt
6a758b7e27f7467f1459fb076bf913fc
f8ddf59e6c1b4873e6601506be7a13287895a5dc
1622 F20101122_AAAUWJ loudermilk_e_Page_56.txt
de3ca79f922db24dfe18b09837e06aa3
34ada8296c818a5783cf05c9d11d323522a36208
2218 F20101122_AAAUVV loudermilk_e_Page_30.txt
a36b4e012a70a5a0f1957886cb14388e
ca7b06b3903a66f6ac3055e9a9e027329cc72602
1940 F20101122_AAAUWK loudermilk_e_Page_58.txt
f1f87d73192b9d7ffdaa22a1e640d226
d14d8e9fe9fa7a619a6eda98f753309e54e3cf80
1689 F20101122_AAAUVW loudermilk_e_Page_32.txt
32e111e786cd51d0e8ba411a7a8dcf85
e7160e6c39c465de29b94f79df359762569f0d5c
1915 F20101122_AAAUXA loudermilk_e_Page_83.txt
dd99d26929cc2081db8dae6d5a122374
9f613e6cbd72063aa523d20360fbcd85dfe11ccd
1991 F20101122_AAAUWL loudermilk_e_Page_62.txt
e30d6ea703eea984de30b701c0589168
ea8deef519aa5e8c77b7da9198d6caf116d499fd
1275 F20101122_AAAUVX loudermilk_e_Page_34.txt
c847c9835dc9b0d77bbc108b5a68c209
3c6b369c73643007f733927cf67caf3dc0573a73
1135 F20101122_AAAUXB loudermilk_e_Page_84.txt
e114ec168da791f7ac0ea8267355bc9f
71094c125125ec8966a149ed187cb7cf5c7cf692
1938 F20101122_AAAUWM loudermilk_e_Page_63.txt
96cbfe47374682c45b59f53701efb04f
dfff7dfdc19e02174a73f791d8a12fc0275508cd
1743 F20101122_AAAUVY loudermilk_e_Page_35.txt
39d0570fcdddfc02eae1248fda793182
3ea8c8ae7b588fb412d5522efcccf091a78dd3d2
2602 F20101122_AAAUXC loudermilk_e_Page_86.txt
97719560c4b2d056d962cf3910910117
43e663f45e0caacea0ebe742b3bb6dd4a6c2c047
1272 F20101122_AAAUWN loudermilk_e_Page_64.txt
47aa9922dd1f4a7a39aabb388babb32b
6c4480096da8af4f9035453df9488d70a27af943
2000 F20101122_AAAUVZ loudermilk_e_Page_36.txt
3676fb1d41bc6f33c8c64221b5f70dab
e253131c425d99a18937cc00ae0c9357226c3de8
2592 F20101122_AAAUXD loudermilk_e_Page_87.txt
cf13a76c84aee3392c7bb2ac76c135c3
ad6faff9195931dc87aea70bcf0a12599c8dd567
2030 F20101122_AAAUWO loudermilk_e_Page_65.txt
195f86f5396fdde56da76b9362e4e715
1496ac15fcfc2a6a5ab0d5f3bf8fcaf9f0ea028d
2698 F20101122_AAAUXE loudermilk_e_Page_88.txt
6cb08c5605292168dadf8c3502b0f6c9
60ba882180cf9254a437e0468364a84f6e3d553d
2061 F20101122_AAAUWP loudermilk_e_Page_66.txt
275466e5b57bf8bbbb8d4d28ff57d518
e5f5ad724a77826d0ab16e84bab1bf1284dca25e
952383 F20101122_AAAUXF loudermilk_e.pdf
fc7af6d3c79602d3aa6cc54736776b7e
9dd1f43f9b2b3b1928cbac2c2250725832d1a381
1515 F20101122_AAAUWQ loudermilk_e_Page_68.txt
ef02d9b14abf915282e89bbc01672b7a
79a9be76cac68934412455ac3e92f04fc90e78d1
3337 F20101122_AAAUXG loudermilk_e_Page_02.QC.jpg
6359adc376d3d02d31326cb596ba9975
62a9d96512566b74f97deaff74213b44e26ebacb
1537 F20101122_AAAUWR loudermilk_e_Page_69.txt
f35a26ac3f4de497365cba83e321e9c1
e9e1c7bafa28e4fa2abfdceff56dc99e4859f3e4
1869 F20101122_AAAUWS loudermilk_e_Page_71.txt
a0986fc86cecba23380c6957da481187
e509f9f02664ce7e83ef3e8f227975e9ccde76c1
1406 F20101122_AAAUXH loudermilk_e_Page_02thm.jpg
25fd6fd1e6076ffb19c0972bbada9f18
23ca812f5a64598eb376d85e25ffac23f73fd07b
1510 F20101122_AAAUWT loudermilk_e_Page_75.txt
5f6f397fb06f586021fac053951a94e7
4eecd282072dc47693460738eb4e0a962a1fd348
3275 F20101122_AAAUXI loudermilk_e_Page_03.QC.jpg
96c91c4cdef34a0f0adc0e9358b03db3
597874b39578d9af2b3a0723c1a28e8ea7177be7
1639 F20101122_AAAUWU loudermilk_e_Page_76.txt
9c0d364ed678dbfeac4ba3540b356dd4
726ed6f5f158bfa2e196ac83607c4acf60d1a482
1382 F20101122_AAAUXJ loudermilk_e_Page_03thm.jpg
7abe0e31ba9104e707d9b4b257bd7e1c
7168240a7ac3aaa487b9db34beaac63297122257
2013 F20101122_AAAUWV loudermilk_e_Page_77.txt
fc811bc66849247a3f3112127927a5f6
744e82326739187faf9cacf4bddc3f4ec44cc830
17533 F20101122_AAAUXK loudermilk_e_Page_05.QC.jpg
7439d4ab2cc4f5b2ba771bb146db9064
4f3f63e6deb09956af26ed8d3193a4eebe1d9c2c
788 F20101122_AAAUWW loudermilk_e_Page_79.txt
940413ec54064e259a373da49dd691db
c6bbcc2ce05247a837041214a63240b49b6dc871
4585 F20101122_AAAUXL loudermilk_e_Page_05thm.jpg
24dfef4121505d23dce2e6f7dda2e07b
87f754f0522b20ecdbcd5087adf23fb99d94ad60
2085 F20101122_AAAUWX loudermilk_e_Page_80.txt
8ca40254a09710ca083a0276223c38d8
8bc65efc1b04e65009cc1a6ba2d32e54690956c0
22624 F20101122_AAAUYA loudermilk_e_Page_18.QC.jpg
b02ec6604edad9f955db9441c829963f
f1fa55e996b2cd65f848e981c8468a1bd96e7c95
22760 F20101122_AAAUXM loudermilk_e_Page_06.QC.jpg
28d31b04df649442f489d75861f8c597
a1a023c8e8984448248c1263db4747882925ce66
728 F20101122_AAAUWY loudermilk_e_Page_81.txt
70ddebb5746d35010b9cb1c98b03028f
7c95f8e9dfcfef49cd60ce403eac615bffd24014
6178 F20101122_AAAUYB loudermilk_e_Page_18thm.jpg
64e4dd8ddd56b8ac36556fbf989db4cc
512b9b66209a32c31888f8bb48ecb26d77b111e0
3304 F20101122_AAAUXN loudermilk_e_Page_07thm.jpg
8c17239de62115ca7f83aa6ad6b8d6e2
e4f103b61b92691b64a88afcc306819791f60adb
719 F20101122_AAAUWZ loudermilk_e_Page_82.txt
6043c36127e731f34f1da07ee8bc2895
5d95d83a2912669a4dc560812c961f923e7f66c0
21627 F20101122_AAAUYC loudermilk_e_Page_19.QC.jpg
71c206adac4cf681624d9379191dd7ec
de2643ccf0d8c28cdb4dac439a4c9bc6a8b07ca4
6084 F20101122_AAAUXO loudermilk_e_Page_08thm.jpg
bbeafb69122cd3b9cf71daf95adb2303
2a9f3916ca42f110f452e3e965bd4d8709b460e2
22164 F20101122_AAAUYD loudermilk_e_Page_20.QC.jpg
8e9cd7414517d292cba3d6b8786158dc
dacf01252e8fbd6d28cedf53eb1b2e406597f956
18151 F20101122_AAAUXP loudermilk_e_Page_10.QC.jpg
cd8eadc5abf98dbcc1bba0560f424d92
3cca88c836264f4e05b4821e78a8d2b383d9866e
6565 F20101122_AAAUYE loudermilk_e_Page_21thm.jpg
4333f4c656115ce3c77553b870355834
1c73489219aad95282b6c2ef2a62eb97af894044
5346 F20101122_AAAUXQ loudermilk_e_Page_10thm.jpg
f5fd52755e8ba1df7bdf558a1db5b8e1
fb36addfc0f8486f22d4157a8eef7a6f726a4377
22398 F20101122_AAAUYF loudermilk_e_Page_22.QC.jpg
ccf5fd02fe6b9bd3e5b8ba93cd354566
9cbb39c37d68096a5cd91b9802f1c81cdebc7ab4
21242 F20101122_AAAUXR loudermilk_e_Page_12.QC.jpg
2231ffdc5263b55b4be50144efcd2942
1a886785f2df109beb48a25a4a59d0b64e06cbf9
22802 F20101122_AAAUYG loudermilk_e_Page_23.QC.jpg
afee1c2a6241644e14abd51b148711bd
be7ecfea45f75d4535d332a1a9da0d26b496afb6
5948 F20101122_AAAUXS loudermilk_e_Page_12thm.jpg
5f5b5395870225495111b05a1f33872f
e4c15999c182392379ddfccc95b9d6452a94f7e3
6354 F20101122_AAAUYH loudermilk_e_Page_23thm.jpg
62d78d584399576275e5b78714e6badf
67c53045a67e8d9b1a6640c627150c90d666532e
14914 F20101122_AAAUXT loudermilk_e_Page_13.QC.jpg
0553c976347f60e1ab013a9cce5f9f9c
026bef08c0ed83226ac415350b189f77e3535bbe
4460 F20101122_AAAUXU loudermilk_e_Page_13thm.jpg
f03f5df7991f5cf157020aeac2c342ba
9a828085fe74f64f246725af901eb8f562fdfba1
16080 F20101122_AAAUYI loudermilk_e_Page_24.QC.jpg
5818723f558b86bbde1bcce04805f414
78d4ba513c2ec145b9fe2953d68c11d734bca243
22724 F20101122_AAAUXV loudermilk_e_Page_15.QC.jpg
74eff7358eb71b14aff7f32b16e97fd3
934f8b119dbb90c42f56d0a1a827af73d136611a
6643 F20101122_AAAUYJ loudermilk_e_Page_25thm.jpg
89d5f0e69ba6aa6de2fa772bb11f512d
ff4948307583bbbce31570c9ad204a868c4208ef
6351 F20101122_AAAUXW loudermilk_e_Page_15thm.jpg
a49b023dd7d073d9641103e28741e507
9a2237eaef2ac39bbdb1b798161ced4013f99d11
20738 F20101122_AAAUYK loudermilk_e_Page_26.QC.jpg
3ba910fdfce47d3c97ebd7192fd019f8
045589a12c1893443ab21094fd6419eea6809c34
6810 F20101122_AAAUXX loudermilk_e_Page_16thm.jpg
f84767801c90ee555ee0a0d3292684f0
aedc02becfe81eecfed4b44dd669cb56ffdaeef5
7079 F20101122_AAAUZA loudermilk_e_Page_38thm.jpg
9c02c25d1838d74bd35888876409c29b
7d2942be58427029c455754059881d1ff00f894b
6020 F20101122_AAAUYL loudermilk_e_Page_26thm.jpg
ef22b4e08f502f0a60a5b0439a837c99
7ce096711242853d2f6c59aa84c94748f4817db3
23978 F20101122_AAAUXY loudermilk_e_Page_17.QC.jpg
a459615da486fb874d35ed3c6652a0c5
92bf00bbb9bc84cfaa77d8e398950fa8a12057d5
20534 F20101122_AAAUZB loudermilk_e_Page_39.QC.jpg
758ec1cab6ad78bbbedcd18975e57d19
102fb7bf3515fa337b6dd8564c4cc32aa30c783e
21287 F20101122_AAAUYM loudermilk_e_Page_27.QC.jpg
5d553de4e0143a27a46f103acda04a6f
c02110653d2619fd7cf47efce0460d86a9907c89
6664 F20101122_AAAUXZ loudermilk_e_Page_17thm.jpg
bd96775dd14cdaeaceaa2201e503eb76
5580d17c583ac7de2b6dff22ab9a35aad346a179
6292 F20101122_AAAUZC loudermilk_e_Page_39thm.jpg
0bf77d585d46a645183e25a0dbe7775c
63f93224236cb56aac6c6d97ecc71384e3301979
5909 F20101122_AAAUYN loudermilk_e_Page_27thm.jpg
30cf43a1619b4670c4866e68f9019383
011c8bfbc29b00bc23aeb73a4f45caf2338f8add
23641 F20101122_AAAUZD loudermilk_e_Page_41.QC.jpg
56577754af672f858d18dd0454d9556d
32d82bbf28975f4883159bbca0ef9f74814b88d4
6221 F20101122_AAAUYO loudermilk_e_Page_28thm.jpg
9021b3fb8fcd511367164b913125ac50
bcecc82ef5fbc0ade29ad0a16c3ea06fbd65b7ea
7405 F20101122_AAAUZE loudermilk_e_Page_42.QC.jpg
f351fcec9fdeb5f9882cd237916edb2a
e1453c5e55026649148396e895ef0c012eb830ed
22028 F20101122_AAAUYP loudermilk_e_Page_29.QC.jpg
89041801a1d01637528a691008cd1469
3898432b602018f7270db90a198d388aec8187f8
2377 F20101122_AAAUZF loudermilk_e_Page_42thm.jpg
8d46b953a5282420604d5ce73921ef1f
fef8c54dfe814ef288cbea1fa9106c23ab4f2e54
6360 F20101122_AAAUYQ loudermilk_e_Page_30thm.jpg
30db9d8aba406631aee6df433388718e
e7432831215f8888ca7b9a8eee7a039c5de8f382
5771 F20101122_AAAUZG loudermilk_e_Page_43thm.jpg
d46020af2f568ae9ee905910fa3025e5
1599a48a8aa3f8f7d9faa6c942286e6f26004cfd
6739 F20101122_AAAUYR loudermilk_e_Page_31thm.jpg
ac8ade6498fdd200539a42a0ac3c4f33
d87d4c1d038e169d20068e449d9e6319f6144643
24834 F20101122_AAAUZH loudermilk_e_Page_44.QC.jpg
6e64e1f80643bbbb97548616b3498562
4249d605271bd9ac414e84945bb19ef272b1cfe0
21147 F20101122_AAAUYS loudermilk_e_Page_32.QC.jpg
73e9683d13e724673d1e065796f33ce5
2aad248b03554a84df2893853fb235902d56c626
6133 F20101122_AAAUZI loudermilk_e_Page_45thm.jpg
1082178b2f012df51c74531f2c2a0edc
ad4b89860221a90d14e70eb16229feeaa231a513
2907 F20101122_AAAUYT loudermilk_e_Page_33thm.jpg
8c56633649d1870e8a57e0007ff9083c
febe07ecea5d19cbe33fb6091fd93084ba673118
4973 F20101122_AAAUYU loudermilk_e_Page_34thm.jpg
c3bea9c58897669649d52945dd0dff35
ce7b457669505000c8599120fff0f7f9cdc2ddd1
6706 F20101122_AAAUZJ loudermilk_e_Page_46thm.jpg
ce24d4e911b9de0a75668cd7de45ebc9
29bccf40f99bde334cabf52bb31a8a762920c5a1
19060 F20101122_AAAUYV loudermilk_e_Page_35.QC.jpg
b235cc2c3243eabbffee66fbc9287648
0757cd89866a4f02b2bd813784b616da9e7df0f1
22985 F20101122_AAAUZK loudermilk_e_Page_47.QC.jpg
ebe99496dd836260f50eb25ae41246b9
3b957051b845ef90241134bceb57f4d4cbfa7a43
5472 F20101122_AAAUYW loudermilk_e_Page_35thm.jpg
42373c9a0c5fe7a81731156ff7f77d20
bf9cbd51a362f45c246a77370d173f8627736a83
6516 F20101122_AAAUZL loudermilk_e_Page_47thm.jpg
53211c7cbc89d886350f073ec03b0804
f4faf7dc4b5229957b6f3d5abb3f9c65071e45b4
23935 F20101122_AAAUYX loudermilk_e_Page_36.QC.jpg
eb9613f7c837c28099a39bb698e6e695
af86ecc85904703ba20c75cf0dbc716f77b0e9d9
22731 F20101122_AAAUZM loudermilk_e_Page_48.QC.jpg
7a72c9f25a5cd7a0ceef5c4f93dc51ea
9f26de6f4a0987864267e16555ec4a4f015386c0
6425 F20101122_AAAUYY loudermilk_e_Page_36thm.jpg
ac9a8129d175d33783d189e1c498568a
0c5c4264199671c34e88d909901b179a93a4ee7f
6467 F20101122_AAAUZN loudermilk_e_Page_48thm.jpg
6a0d24ee574336ae3fd6a5f7a2d07e7b
4f2719c2a438217776c9806e9d0ade4631b1d56d
26221 F20101122_AAAUYZ loudermilk_e_Page_38.QC.jpg
72420563cabf9dcc12d516227e381208
2a603cbfdd63129fe059e32314b5a86b0adeae2c
9131 F20101122_AAAUZO loudermilk_e_Page_49.QC.jpg
0e1ed656bee62b71d706e312e8091fea
a70f1c4420578e0b2294371c95c8db8d376850d1
2968 F20101122_AAAUZP loudermilk_e_Page_49thm.jpg
14e57dacef33dbb798c44ed4df0d8d2a
b6504aa90a5199d7c394b344c32d6897c08f1748
8109 F20101122_AAAUZQ loudermilk_e_Page_50.QC.jpg
52b03c9ec434ba008873d623f4e0910a
bf7713639d3c0eca5107cae7a60c6f175f696384
21867 F20101122_AAAUZR loudermilk_e_Page_53.QC.jpg
fe6a1709802ab0371eae76595a040f16
d221d6139b9cc80a9ad7a78e1969d397b4bb131d
6291 F20101122_AAAUZS loudermilk_e_Page_53thm.jpg
dd059c6991e163b3cbd0e9a47127e269
3885b02a6f6ec8984ce5802cc7a0f8eab3467683
2840 F20101122_AAAUZT loudermilk_e_Page_54thm.jpg
a92fa197f656d16408b883baf40da991
51025a248ff4d79ba32b3da01baa76b4182326b1
25202 F20101122_AAAUDA loudermilk_e_Page_01.jpg
2176285f66074a06c85e2959a0379b83
b3425a374626cd896f46aaac3507a6b93d104be1
20413 F20101122_AAAUZU loudermilk_e_Page_55.QC.jpg
2cc08ecff8d2e55cb44c0896ebf49fe0
ea27d04d3feecdf66623997918f263f2e15cd344
866 F20101122_AAAUDB loudermilk_e_Page_74.txt
641909829450dcfdfabf48b9454914ba
d9c5977ba4840815643a6830a250682b15330fd6
18755 F20101122_AAAUZV loudermilk_e_Page_56.QC.jpg
52df21ca1a437f32983dd3c7e30ab9e4
bb01ea05f756a5f129399fb3c23cf3af8b4f2205
13719 F20101122_AAAUDC loudermilk_e_Page_90.QC.jpg
8d90111daa0d6d17b0d38819679e288c
32fe5c03fcfed40b486d687918768c85ea411941
22381 F20101122_AAAUZW loudermilk_e_Page_57.QC.jpg
68784f9e00b4db19b60edbdbb0a2cf92
741139d58e69a46c3fd6348894e82464b5b59ab1
639920 F20101122_AAAUDD loudermilk_e_Page_24.jp2
0a714c816724f6e451ec15741323c8fd
bd24df4690bb089b1e4266859ec73d6d63a0aa51
22504 F20101122_AAAUZX loudermilk_e_Page_58.QC.jpg
2d8b3e176bcc016ef840336b4299e27a
d246757a2a94c8b28e189fc5fca2f5d576a81fa6
1053954 F20101122_AAAUDE loudermilk_e_Page_04.tif
87103cc28265956cb8eda1072e1dc60b
b81622b088566be3025ce08762cc7cbde1e4488e
6391 F20101122_AAAUZY loudermilk_e_Page_58thm.jpg
6d0c5d9e79df6871ebf82c57dbcf16de
b5844c0a3695aa5fc194bf1805b81e3573eadd41
871750 F20101122_AAAUDF loudermilk_e_Page_56.jp2
37805f90b9d5bb5cd3579d434099fb82
cf2e40f6aa91d3af6cbdd1f9b175ddbfa68933c7
5410 F20101122_AAAUZZ loudermilk_e_Page_59.QC.jpg
b23e69ff4da860f8707c25594adf04ba
ad5070ed2bff4de2f01c6d2c53a8434d73653a95
74159 F20101122_AAAUDG loudermilk_e_Page_25.jpg
f794d79249a0996bb18d353804c6c351
c8d4325275cee7c6c7b60381aca442b5b9125a83
116337 F20101122_AAAUDH loudermilk_e_Page_44.jp2
4a7e93c5477d937854681f5034cdb23d
83996efcf0a7185fdc4f454e9819b3f619187499
1824 F20101122_AAAUDI loudermilk_e_Page_12.txt
01d260b80cb901560985967b7602ecea
2aa07f641bd07c00e9757f1a5073206dca959563
F20101122_AAAUDJ loudermilk_e_Page_23.tif
dcac37c4b688c88622b38280954739e5
fe9d6d683913ffb1e2a34603c2c90d3a8cf17cec
65625 F20101122_AAAUDK loudermilk_e_Page_12.jpg
a5b90ec9a66cbfdf497300e78cd57472
20cf37b9173c1a9d6c1aed412bff3f3fb8ff822a
17334 F20101122_AAAUDL loudermilk_e_Page_04.QC.jpg
53ef1a50249f0a351c8cba001e076f2b
bc4c32fc42074bb8eaed5ab5a15ffe0ce622d9cf
112696 F20101122_AAAUCW loudermilk_e_Page_17.jp2
15ad9f4f4b8a91ad7e1f9ced6fafab6a
b30cd5a598ab2abf229e983c917ad471c0b2710e
91822 F20101122_AAAUDM loudermilk_e_Page_43.jp2
e1a623e29d352bb66fda25effcf486e5
623a2e53c69e871df5561350f35e89998f4ad321
5832 F20101122_AAAUCX loudermilk_e_Page_60thm.jpg
ee4bc6e0dbfdac7897973d55470381de
aaf600fd3f959d5798c7b878b2d706d8ead07f72
53231 F20101122_AAAUEA loudermilk_e_Page_04.jpg
90fe1191fa5d7be8ee8558aa6c3f79fd
96f5c8ac7724213e88328f40079dba710f790b09
4596 F20101122_AAAUDN loudermilk_e_Page_24thm.jpg
5b8b8441f6a9a65c963d2ef1e3b841bc
1aee11a0d0819e79ee22553f0d1de8a6d7dc0263
2040 F20101122_AAAUCY loudermilk_e_Page_16.txt
fab3656ba2196e0e870b1b142222080c
50f589869de80e235b11eafa5c8547672e718897
17272 F20101122_AAAUEB loudermilk_e_Page_74.pro
0eecb3d800c81ba4d28f6b046b44a315
a3e4d12358aff185db09541828946d85b47ded68
25271604 F20101122_AAAUDO loudermilk_e_Page_40.tif
042793afc0f5f37e408ec606376bf6cd
00d54f13ce1fcf1eb7c13a5a04d3759e3c106af6
8423998 F20101122_AAAUCZ loudermilk_e_Page_67.tif
871809de5cdf04c8debfdf4a24f577c5
38c206d5a42841cd81bd0c001106e375af99a697
33809 F20101122_AAAUEC loudermilk_e_Page_04.pro
f72eba856ec7126c82f093277abb5e0f
437051845be0b5fe6defbb65631dcd4502d53873
68740 F20101122_AAAUDP loudermilk_e_Page_20.jpg
8c43a57ce2c891810442895015b7e355
a58ad5950ce3c79d1cfddea653a032bb013f216c
11449 F20101122_AAAUED loudermilk_e_Page_91.QC.jpg
718f74c4cd605b32e5c1971c09faa0f3
8e5691ad6af3ab225b667190a880dfae695ee06a
1961 F20101122_AAAUDQ loudermilk_e_Page_47.txt
c96b96c62fd911c5c0d0cbfaeb8ddb98
ed5a1ca6e28bbb996009b69ccf9b257fe7c87b24
F20101122_AAAUEE loudermilk_e_Page_46.tif
8d049ef963a7c3f4354082a40695e77b
72df99581616e30ef756f11059e5580bb1ef58a8
1051954 F20101122_AAAUDR loudermilk_e_Page_89.jp2
4d60578b486582a8c0f718bb98f0caed
f15f373700e41e840cfaec4e963644236a1e879e
F20101122_AAAUEF loudermilk_e_Page_14.tif
a78f714a4cb030c06093e02014fbe410
f6c263f501221a326423664a0973a56151a8cacc
8798 F20101122_AAAUDS loudermilk_e_Page_81.QC.jpg
50c1d8334321895d1f607e3b5dd285e5
fc7a203056eda0a018532bddf0e65de2f3e19484
6573 F20101122_AAAUEG loudermilk_e_Page_11thm.jpg
e6978a194d08b35464e3ce5c3ee0f36f
f109a01156999f814435a7a6fac7f6a806cc1dff
563 F20101122_AAAUEH loudermilk_e_Page_33.txt
bd083351a3f354dd31c7c184fc877dd3
f9097ad346190f2567632ef2743eb28dc1faf5f2
1901 F20101122_AAAUDT loudermilk_e_Page_59thm.jpg
f62964c0de66209169f735de1f87b0a5
484e25b349c588696bb056430c253881773d2d6f
48382 F20101122_AAAUEI loudermilk_e_Page_58.pro
6e2030c830ee33910fc9e09ed5e0a291
9da8efa1d0c84ff16eeb106f76298ed8db1e22d6
763313 F20101122_AAAUDU loudermilk_e_Page_80.jp2
43b0fa012e726901c2948caf2dccf22a
5975dd45d4ea9e02ec92d4d3a6f8bf36ee030551
1051944 F20101122_AAAUEJ loudermilk_e_Page_08.jp2
85055b53cb6317c8fc0955a0148e7c7a
9e81afdbdedba4e31307fcd476313f16967bdae6
67842 F20101122_AAAUDV loudermilk_e_Page_71.jpg
ac37565efd06bd7f67e54f10f1c007d9
56a451e89d1c7a6988242fa42b68f07932021bdd
6250 F20101122_AAAUEK loudermilk_e_Page_73thm.jpg
4fc6651f8f7069af58e2671ff97ea401
333218d1dcae0ede670c3a88cbacf48553c6a29e
31356 F20101122_AAAUDW loudermilk_e_Page_75.pro
f8199fb91a115bc44d8767acd20db552
045087e0579dda492cc9420b7adc48df2f875a5e
51922 F20101122_AAAUEL loudermilk_e_Page_85.pro
90a54eedcdb6f30666b895c445cdc459
e430e64627ed2f22297c763bba633b90207266db
37961 F20101122_AAAUDX loudermilk_e_Page_35.pro
cae501b167428b53ecdab396aacdc476
61bd46dd3456a8468d47f0ddd0cdcf02c837d8fc
1051980 F20101122_AAAUFA loudermilk_e_Page_38.jp2
67e8a462d189ff2b19381f96e7f9a425
2acfff1682bcecb9a3ede493237ccfab75661086
105948 F20101122_AAAUEM loudermilk_e_Page_18.jp2
8a00e45c7e30c641ac12a5b649cbce85
2367e0b2ab198e84aba646775b17f49dab2da0e1
F20101122_AAAUDY loudermilk_e_Page_61.tif
05fe0d5fb087a5fa9356d0e008d93be1
d218e39f8d8a400cf3b44e6917d3498a28d8aea3
F20101122_AAAUFB loudermilk_e_Page_28.tif
4ff8503b5ad6d34b2134a12747577cc5
c3b827efd3b41621183e3bb832c8973f7c68c7b9
31210 F20101122_AAAUEN loudermilk_e_Page_67.pro
b59d38739d2195dc2e6247fc3c33202b
f28efe48441984a70bf4929c250851a115f842df
448805 F20101122_AAAUDZ loudermilk_e_Page_74.jp2
18f019b90954ef2770324ce6ab176631
22ad998bba0b273d5a2117a1b90ef0642563e3da
F20101122_AAAUFC loudermilk_e_Page_02.tif
d8a07186fec2a80c8d8ce637d78b1b88
e43a4def9bac5e4674e3b9bd27eace217166f2f4
F20101122_AAAUEO loudermilk_e_Page_72.txt
585d7e425d49b170f722301409f251dc
135a577b9989379d6e1aa190d5c2b59b0ff06065
100056 F20101122_AAAUFD loudermilk_e_Page_06.pro
594393e0b3e9d06a2be07d14bd84fb5e
3b250ba33eff5051082420338e16ea8eca868734
21696 F20101122_AAAUEP loudermilk_e_Page_78.QC.jpg
9a9b0930ff98a855728244d8e1c02bc6
380c8cfa93bb36831331fb1dcd5fe6dbf13840f6
2090 F20101122_AAAUFE loudermilk_e_Page_46.txt
dcd0d44c744828aff90ff98daf1172e9
53177fcd58d90ba757233167a956c9ae50bbcdf8
6701 F20101122_AAAUEQ loudermilk_e_Page_77thm.jpg
6bd9ca4b62bee220887c577beb68ab5c
ea6a322ad18233c8b8cd618abb418cd75e42005d
2985 F20101122_AAAUFF loudermilk_e_Page_05.txt
a9443b3c0623730c4236ce6a55f7b7ca
650b14a626cb7cd56e79915497dac71609b776f8
10214 F20101122_AAAUER loudermilk_e_Page_09.QC.jpg
fa14d39e92f70f5d778f2085eb9008f8
f07b6d5e60faa2515d2ca24e40ee579436d619fa
14697 F20101122_AAAUFG loudermilk_e_Page_37.QC.jpg
32071947b2e76072ad1434f9a64f5b0b
abd102c93e88c78948b0ec13247d2141b639f0a6
74538 F20101122_AAAUES loudermilk_e_Page_66.jpg
1e48435479b6fae6cfc9460dec2c84f5
0e08b9d5c00d0cb3908eda060040c8fe728ca2fd
112063 F20101122_AAAUFH loudermilk_e_Page_25.jp2
b7703e611e29e3f53a96a927d774781d
b73dadd5df10e2a9f3ef2678b2ddeee87b19ac05
22939 F20101122_AAAUET loudermilk_e_Page_08.QC.jpg
115e4a9e463ba7b82a9fb9fc3a0b0802
4129247830245cc7e54536f40138dc83c7bf2ebf
111399 F20101122_AAAUFI loudermilk_e_Page_62.jp2
e565bed58f595abc2b55f6095b752e57
43ba327ff6a2ce3d360ab561db8ccebe7e13d450
58027 F20101122_AAAUFJ loudermilk_e_Page_50.jp2
c1e92a8469366991ca1b252f22288907
c2b7d9fa57a99fca8301086b4c86b3d8ddce3e3d
1051958 F20101122_AAAUEU loudermilk_e_Page_06.jp2
bb4fc32e74165f7b2dec01f22eda6ff9
61bb6e8d67b298016622362fc6355e436d9fdce5
5244 F20101122_AAAUFK loudermilk_e_Page_51thm.jpg
84e703e9207f7676a9fe641732dcebb0
686d62047e5c828a38b842a6362c548981c7adb6
F20101122_AAAUEV loudermilk_e_Page_65.tif
8b60fbf7353a0a113dadf5e24c8b3f4c
e6d6a3d0c18e8d70460a999f1b577ae5e71883ce
21877 F20101122_AAAUFL loudermilk_e_Page_28.QC.jpg
3280bb9e233cd1d8d99426ca1cc77e6a
7f6e668874a1bedc0a44d41d2cacc14958055caa
49675 F20101122_AAAUEW loudermilk_e_Page_47.pro
0d1ac3a14087fa17d458a40aefe316ed
124728f501be9905bf3a9fc163e5c80e08116ec8
45561 F20101122_AAAUGA loudermilk_e_Page_73.pro
16725fc168d8ef46468471f14164021a
db50cde689fb69a17c80e203171ff16ece67335e
F20101122_AAAUFM loudermilk_e_Page_68.tif
277dd7ceb7be4d289a5ac893b3949772
fc92fe57e1fa1e51459cc913c6af10d023e2d830
7950 F20101122_AAAUEX loudermilk_e_Page_01.QC.jpg
261d7aa125525e8ac53e08ad79903b21
e32014a892c78000cac5a696ea6ebb94ecf49d9c
F20101122_AAAUGB loudermilk_e_Page_49.tif
511e6137b6bd8992571936fb8f8ecf68
4e505f8bbd3af0341be0500b07e5cf82d199a23c
59550 F20101122_AAAUFN loudermilk_e_Page_52.jpg
4ee7c434693a9d6382852040cce8d074
e7680f9bf6c9fe21e85bc12188b556d205ec43e6
6208 F20101122_AAAUEY loudermilk_e_Page_78thm.jpg
0ae2d9d524b3f3633ebc721be8f68309
734c3932d5da341ba515d728946a70e589a7e501
102734 F20101122_AAAUGC loudermilk_e_Page_71.jp2
f3d4bdf7442271e82450f171e8bd15f2
10dac64113a2c8fb5dad813cdc9c620b099111f0
66156 F20101122_AAAUFO loudermilk_e_Page_78.jpg
03d5e2345925bf1e94e67808d56ec069
f072028a550f546f74def80c1bff3808c93c7b4a
35339 F20101122_AAAUEZ loudermilk_e_Page_84.jpg
16373a73b93ca6de48cff38f8a5fc905
7555c3e07c8dcf811e7e1542de20575218d0519a
6413 F20101122_AAAUGD loudermilk_e_Page_57thm.jpg
2ce6b080d2007a53313d85cf7e7e703f
722eed3f42aaa3933d7d67a88fdee6d2eeb05ff6
1156 F20101122_AAAUFP loudermilk_e_Page_13.txt
192cc0331e2e9766c0ee04ae4790f5bc
2970493c1bdd60d2f47429d637e3de8151d043ef
273 F20101122_AAAUGE loudermilk_e_Page_59.txt
9d90d1b89672d880cc8612f380d9d4ee
1fd580fe9049e96780410dbe6e70390ab01e55e1
6899 F20101122_AAAUFQ loudermilk_e_Page_44thm.jpg
c82144f29536645453e5f018821f0026
c76db9514ddf49350693ebfb0a5d18e6d1fddbf8
1653 F20101122_AAAUGF loudermilk_e_Page_10.txt
3fd2f70a23ba4db9d662e1a0e367e937
21e04fb3f448df90817220fd59278307ee63c907
20499 F20101122_AAAUFR loudermilk_e_Page_14.QC.jpg
7d3c214bc8134ea86971dadcc7c4bac4
847a327c64d527b53fef95048105ecfa0f8cf02b
478 F20101122_AAAUGG loudermilk_e_Page_42.txt
076b41b1e4aeda57558c3698a77746cc
88ce12808cf990bb28db47ee1ae3ed8db9c1656f
5476 F20101122_AAAUFS loudermilk_e_Page_55thm.jpg
820b7d456849690af46e4a277ad89b44
4f63871f7fc26c3345fa712a10007d8e4b92d6c0
41779 F20101122_AAAUGH loudermilk_e_Page_14.pro
ba84938bfc25e268a6f66561dbcab7fb
525c5d2621e80599a820412bee7f644235669c6b
24099 F20101122_AAAUFT loudermilk_e_Page_25.QC.jpg
9adfb8faaa045bd7c4d938c601bd34b8
a23baef2f0946ec5764630a04935a31b2347d6c9
1912 F20101122_AAAUGI loudermilk_e_Page_57.txt
bb6070447d8c482d769a282bece8ddb5
6a55ed3f9aacee1d0eb6ef5d7223f84abeecd1ef
21676 F20101122_AAAUFU loudermilk_e_Page_45.QC.jpg
8b2ad7b1daa85a0ee5283ddb29e4a595
14ba250e25527e7408079e5054e19179ee62665c
50174 F20101122_AAAUGJ loudermilk_e_Page_30.pro
1a7ae3036fb65e91e5c99c8b897ba6a6
62a8fc27238b58953daa4e67f0afd1ef97db8253
F20101122_AAAUGK loudermilk_e_Page_91.tif
bf2ef22c970f7d4ee83665e498c0d160
ed2ca77c0a277ccb3478f16c93874440d0251f46
29041 F20101122_AAAUFV loudermilk_e_Page_70.jpg
0b2e6b9f13bc76b85264a707e2ccd3ea
aab413bc897640da20df378166f22336d0b38674
9379 F20101122_AAAUGL loudermilk_e_Page_54.QC.jpg
8522f89a8cde21b885160785829b7710
e703a00f7b2b7bba14bac03345f9ab1634dd01eb
F20101122_AAAUFW loudermilk_e_Page_06.tif
00764f0de8ed55aacdcff10dbc44d928
a5ea2790884b9c793d28ce1b072d29c6ab4c4c0d
46681 F20101122_AAAUGM loudermilk_e_Page_53.pro
79ecfa96e03ddf47784d36851a5aa2e4
a9dfcc2ac87e27bd4ef4bd10179ea34c36453239
4058 F20101122_AAAUFX loudermilk_e_Page_06.txt
4843a7316fe4d3d8e2f61c0f8fd2840e
04c2099e0c59bca11b255fed4e9548a34f47c325
47539 F20101122_AAAUHA loudermilk_e_Page_37.jpg
aaef8f8cff561afffc04138189820a68
900af4bd11a3da8173e1f37eb417f48f61a34fa5
1802 F20101122_AAAUGN loudermilk_e_Page_50.txt
344f36ed1b957e980266afe440551609
eecbda53ecb83f7164184abadb32473c407c97c4
2418 F20101122_AAAUFY loudermilk_e_Page_31.txt
e0c09c27198c1d00129c51801d2f7081
ff91e178f1ebaf801bddb44d61264e8c2c42829f
14802 F20101122_AAAUHB loudermilk_e_Page_70.pro
f28e3d7daed356c48a0dc5a8a3d60169
4e48ee8d3889373e54331486700c67ed4d02e0b6
23514 F20101122_AAAUGO loudermilk_e_Page_21.QC.jpg
e8707d8582265e5f6cd40794f775c8dd
c3f8bec35d1f4d7b4f034e29ac8e82e328922740
1865 F20101122_AAAUFZ loudermilk_e_Page_19.txt
a22514b7c4aefa720d042c745b910cd5
36c533ebc5bc3ff6edbe82e2972510e47e4dd958
52438 F20101122_AAAUHC loudermilk_e_Page_21.pro
540efcc36479f318d7b9e2359e2fedf5
45246e4b0dce37d32555f734810d3925ed484cee
115137 F20101122_AAAUGP loudermilk_e_Page_46.jp2
3b1ed2f6343ceb99e9b369c551bf355f
9cb13e5f72f73bc43dd9a2300633785d3ef89e4f
117 F20101122_AAAUHD loudermilk_e_Page_03.txt
d8d87b2588aa996fdf405e978f36b911
9e7db295147b46bd54bb2f74b6517d8eafcc3b4c
92762 F20101122_AAAUGQ loudermilk_e_Page_60.jp2
c5c1312042075ef52078a34bdfefd432
c98f466c4d337873cab0bb5ac19e4884be2028d9
F20101122_AAAUHE loudermilk_e_Page_82.tif
f4cdfc6d73e828fc4b26d53b7565d25e
c23adaae2047fc83fb83e707ddfa05aed035b2c7
46415 F20101122_AAAUGR loudermilk_e_Page_28.pro
d6144f795335c823e69185a92668c4ae
327ffd3efc28758cb7a415f3baa68311c69f7c34
94089 F20101122_AAAUHF loudermilk_e_Page_32.jp2
e611bbf1ce4b6dd935c41054ebf99ef6
519bc5d77c06917184b771e9166918231e0b3da4
1400 F20101122_AAAUGS loudermilk_e_Page_67.txt
4eb02ea667e0e46fba545197997e67ce
188c218a155e9a25c1f43f4808b273f82a01e2f6
5809 F20101122_AAAUHG loudermilk_e_Page_32thm.jpg
76f99f42f3532b9c3ecf038056a855ee
c02c37efe2daa9f75d082ce8c36fef38e1a04372
62496 F20101122_AAAUGT loudermilk_e_Page_80.jpg
db4fcd148303e20289f05a82b658af52
00b8439fadc2f4b6a4a5769a59c0821fdd1ddc5c
252449 F20101122_AAAUHH loudermilk_e_Page_82.jp2
9ba576c043a946b1506bee1e5d81b757
3e15633b9ac6ed1b67fc7d48b1867a351d53a714
3502 F20101122_AAAUHI loudermilk_e_Page_91thm.jpg
3f049c2e09798d539c3d95dc0b40e753
48cff0ad9429b4f5386fa736202a5361ec0ec763
102715 F20101122_AAAUGU loudermilk_e_Page_28.jp2
012898fb21d96e5e0cf9b768fa51eba7
6546c124f6b8e6d42ec302114f7b44fd47ed2ea3
5227 F20101122_AAAUHJ loudermilk_e_Page_04thm.jpg
20c3d81c46db8d190852d06d4d25efd2
4dd0af5caf12bc3047744d1426a9a60ee516a3c8
2196 F20101122_AAAUGV loudermilk_e_Page_52.txt
808ce1345dc050a8e4321b75663a6257
d65497770cfaa72b4f74a01a93468e0eba9624f2
604 F20101122_AAAUHK loudermilk_e_Page_49.txt
4d20d5893fb55baee37ebd71db582bc4
490c6e73509c6d416e3e88415f69126f15f60bcd
1333 F20101122_AAAUHL loudermilk_e_Page_02.pro
d524317f242d969939e4a6f598c459c9
dddebf631f4c71eb5b9d912d79a88e7bd1364234
71436 F20101122_AAAUGW loudermilk_e_Page_05.pro
318f201aa8a1c47338e33313b4256bb0
9c9f7b67e86bc0f8f6fe2e5458b4e2ac80e5aa64
25810 F20101122_AAAUIA loudermilk_e_Page_89.QC.jpg
facb1f2332f7a1fb5f8da162be5454c8
8f0cbdf52463920571d57fd9310bc2fda406e458
17203 F20101122_AAAUHM loudermilk_e_Page_68.QC.jpg
250214b92b24b2f85434e9622dc575b8
1aba99089cab07df9dd05bea101810e94f372409
29825 F20101122_AAAUGX loudermilk_e_Page_68.pro
4a77bf8098eda5557a941dc537db3edc
2a747bee4648a23f2e09c9abfea81dcca4b5bc84
2777 F20101122_AAAUIB loudermilk_e_Page_50thm.jpg
7b857fa620b375ef6b8047074d6ac27e
51924940c3028f2aa64adff759fe2c0b9080237a
2689 F20101122_AAAUHN loudermilk_e_Page_01thm.jpg
cfe75db051070aaff0411f178c374940
64f6b2db137a8254205fbb996fd5755b0dfb7280
6315 F20101122_AAAUGY loudermilk_e_Page_19thm.jpg
ab991bcf046671830a0f85eda3c5edcd
06ce8ea64f0ffb8e3684ab926ff8185e2de76e63
104656 F20101122_AAAUIC loudermilk_e_Page_29.jp2
a77805bcd47fa917156b5cdcea03d82a
d3ce5d1a964a7e5b65262ff6a7ded46e3296b0c1
6442 F20101122_AAAUHO loudermilk_e_Page_22thm.jpg
5d3263f360346737af40c5b7b03a60e5
8d307c64ccf970f1c397affdec7c364bc14ba2d2
69193 F20101122_AAAUGZ loudermilk_e_Page_18.jpg
f35f511a4b4b8a31c797bd9d47293de7
9e7d305f91bf90f24f7fc73e5c91e5bd4da45a70
68036 F20101122_AAAUID loudermilk_e_Page_90.jp2
df96e2c1403b3bf5b67ebe4f15602ecb
1e2055f2b9c7a788180e70cdcc14c7a15c302ff8
61550 F20101122_AAAUHP loudermilk_e_Page_51.jpg
3175fee11079d9e33bf5595cc377c472
846d46ef2b236d7c308b302cd965f47a5ffa9dd4
F20101122_AAAUIE loudermilk_e_Page_31.tif
1c47e95a2bceaa2b4dce3b861499002a
2120b23432e2480e197ee881689ee5a9de8edcee
63304 F20101122_AAAUHQ loudermilk_e_Page_60.jpg
9e5b54364ad4dfb5cc086b50bc36015d
984f98037291f6bb9e904eaa029dc21db022b7d6
F20101122_AAAUIF loudermilk_e_Page_15.tif
a3e2aebcb38f168c5580313a800090c5
c8691c5ee310bc1c54ff8efd481e74b46e73e452
F20101122_AAAUHR loudermilk_e_Page_48.tif
9d82a7def5fee9c519c46ad27b02206d
80dab3ebf7a39564a078e91f35d8a7ad09a4d376
23627 F20101122_AAAUIG loudermilk_e_Page_11.QC.jpg
550e99de48d2fe38d556d3a7ae027526
cb9ea1c415f916bc6b90943ddff9d9da49c28ea6
F20101122_AAAUHS loudermilk_e_Page_25.tif
3aaf935836c836ef0453c0d69f8ddc4e
b7f34f2b9383d1072770bc58145c5bef718df9a2
F20101122_AAAUIH loudermilk_e_Page_76.tif
a812baa657972ddc34e397d73eebc3d2
2edcaddef485b69f7e81ab7042baa4f9d71788d1
24163 F20101122_AAAUHT loudermilk_e_Page_66.QC.jpg
a480cf71afe083dcb255d4182fedde0f
b3829b537b72ff07a18a70bf88a51c84857bcfd5
76042 F20101122_AAAUII loudermilk_e_Page_46.jpg
c26cd0b0c726682acba5d96f90178be7
d9456d8bde6b3e89f3f09e21f3dcad7b555f9572
46799 F20101122_AAAUHU loudermilk_e_Page_52.pro
152613f316faadd8aa3bfcf3f6003fa8
7273fe42100fddb79b4811fdb6d57f68af19047b
68185 F20101122_AAAUIJ loudermilk_e_Page_22.jpg
20f8421621c316911986fce9081e002c
090e92886dff7ffae94bab22f2d73b8cdfff50db
F20101122_AAAUHV loudermilk_e_Page_64.tif
a8390e4ffb41c718d9a4ec2993ea31fb
2a3fa056d41862a48f4c25f09a435495567b5a38
6791 F20101122_AAAUIK loudermilk_e_Page_65thm.jpg
2ed1cfc526034d133f4024a604428c9e
f11b97b973f8563dcf623d0916f46a1c5e8056e2
66644 F20101122_AAAUHW loudermilk_e_Page_05.jpg
b72ebe64ec641460ce57c8e2bffb6fee
e60bc6cd535076aa6311bf6b0d3910dadba6cbe2
47333 F20101122_AAAUIL loudermilk_e_Page_83.pro
8f3bd2cd9cbc4b6c077db12d92e7585f
4fca854058382350d23dd5fa429342fa8a6a4168
F20101122_AAAUIM loudermilk_e_Page_10.tif
761049ab748356ef7b80ed9dedf65ca5
08252cbaa77a65baf6731808d82c22a4a9a6767d
48275 F20101122_AAAUHX loudermilk_e_Page_63.pro
839c32e1a2cbbeb94c950353f4413269
45040150bbc71349b594262cc2a6a81a333a369a
50685 F20101122_AAAUJA loudermilk_e_Page_64.jpg
57c49eea4b37c16289cb6490336e52c2
d59580c4fac73a8373fd019f048eb533c0edeb8a
23665 F20101122_AAAUIN loudermilk_e_Page_31.QC.jpg
ba5948d8c834d218bd6c95e2610a514b
5419feb819cd4a686a935e781d2e8553543d2225
F20101122_AAAUHY loudermilk_e_Page_08.tif
2cb5dabc179d10d00e35856f8f8dbb8b
dacd9efe8d7b17f673ea6509d02bd20647736326
108331 F20101122_AAAUJB loudermilk_e_Page_21.jp2
ec384bdb85bd2c83a7d7dd3f1a663c94
29e3bc456aaf7bc9f736039be177273866f592e6
64339 F20101122_AAAUIO loudermilk_e_Page_86.pro
009dc6208b6dfbdf0113cfd29cd78c07
fb72cc074b2fdeafefa82f2f6db5e37892635845
18692 F20101122_AAAUHZ loudermilk_e_Page_52.QC.jpg
1548a55da8c5ee4e4acdc97235012b1e
39325ec62058c11d1714ae74c5eaec98e799abfb
F20101122_AAAUJC loudermilk_e_Page_13.tif
f4b05b87653b895edf88cac592ba7aff
c13747f7c24053aa12984e043ce339f39dd0d0e6
6332 F20101122_AAAUIP loudermilk_e_Page_29thm.jpg
946d9c7e081cb41dafc0cf5e5bdd6c48
0264e5cffb3290da6b3b76294cb7f3b7d7ba14d1
16262 F20101122_AAAUJD loudermilk_e_Page_64.QC.jpg
4e553aa103a9e0f70851c73bc4621f97
8402aa1ebe24e70eaa3d1905d0cf34f99263fe05
5750 F20101122_AAAUIQ loudermilk_e_Page_52thm.jpg
4f0d4a716bc3d8767ed3c6ae125f9893
a5038c3db14fe52e8d19703cccde5711bfb12476
87695 F20101122_AAAUJE loudermilk_e_Page_06.jpg
e08fa44359f9865cee62d7e62453c4c4
9222cc8707699514149ddb41f78fe6db94f4b4e0
103758 F20101122_AAAUIR loudermilk_e_Page_20.jp2
760b4f84982d90a3a920ba7061623af7
7006eb60ed56c75e5e993d74ddb8479cd8a354a6
73447 F20101122_AAAUJF loudermilk_e_Page_16.jpg
f5e70421e911003a7448ff21cc6c3265
3d5ae18f80c4d93bbb4f319dcc026b335e9d1bce
1807 F20101122_AAAUIS loudermilk_e_Page_78.txt
c7c14b951e97e7471201bc54ae998d81
a7489d9394b090e994402a3d89809e01fe8c8d4c
2106 F20101122_AAAUJG loudermilk_e_Page_21.txt
2cf6b4f9edc0ef43304870b6679a4357
860a27e5dbe7d1544e4b1f40b92b40920b9349e4
2035 F20101122_AAAUIT loudermilk_e_Page_61.txt
6901951086cc286858185391779721a6
f4adc1d85f1b5530e402aecc74be78a8ddd04b00
2119 F20101122_AAAUJH loudermilk_e_Page_85.txt
6799adc379b39cc030799b5d9405b902
e225b938703e042f32f3a5a7ada86621e9f443e7
3483 F20101122_AAAUIU loudermilk_e_Page_84thm.jpg
058f89da1d82a195227749d912f27cbf
37dcb18f43dec9cea564ce01830b5d11c9947a9a
66399 F20101122_AAAUJI loudermilk_e_Page_39.jpg
6a5af0e87504ad112f9cf4b02a5a776a
ec3310a492d8d125c0de37fd62d266fb4928207a
F20101122_AAAUIV loudermilk_e_Page_70.tif
bc74c641fa619c3ea9ea52a5489bbfe9
a03d416b7095d956d7fe685325792b42392a449a
101835 F20101122_AAAUJJ loudermilk_e_Page_30.jp2
a031578701dc50780300d74abcbc6128
be652803c370eaae08e49b2353b2f1288c867c45
47949 F20101122_AAAUIW loudermilk_e_Page_15.pro
81e9222a88bb0145d67488e419012361
22d70ac4065cedfd0abaad1fec17999ebb8ebd23
50581 F20101122_AAAUJK loudermilk_e_Page_11.pro
be895954674c5fb4711ca3af1b9736df
902d7d7dd2f5679448f279961405f5d7663b1ea9
6850 F20101122_AAAUIX loudermilk_e_Page_86thm.jpg
887a7981d4f8a9ad8875e80e745c0520
0ac3754e20ee97f8bd959cc669c6c6df08f9d54e
835 F20101122_AAAUJL loudermilk_e_Page_91.txt
980d91e707233667e263a007f461889e
f881116e52b79271811e45b578feace9a683190e
10388 F20101122_AAAUKA loudermilk_e_Page_02.jpg
1bbb31c6aabf4d64219f00f2d369c887
7259e00c9554ddc24ae894302601579d00bb0377
4347 F20101122_AAAUJM loudermilk_e_Page_37thm.jpg
3a7b75d236ec394f9c137de89c54a78a
6daf4ec2588f893b31fbc4a9a49d516a6b29f038
18831 F20101122_AAAUIY loudermilk_e_Page_51.QC.jpg
d8bcf73ea549d8139af8f710e008c4a5
3eb5ba48d85a9d9f9f77f2f901a08860638f8708
16035 F20101122_AAAUKB loudermilk_e_Page_75.QC.jpg
ad09695039409a5e5070628bc5d4fb53
0dd506580754a1c374e1fbd3ae5246f0019d04a3
685047 F20101122_AAAUJN loudermilk_e_Page_68.jp2
c786392cae7f9fae3052ad1562f0f654
68f8f956cdc57c6345dd01074648fba1152966c9
54667 F20101122_AAAUIZ loudermilk_e_Page_67.jpg
26f03e2c90a165657d10d66f755815bf
2fb3304c266921670d4675115a4605965de1ee27
20308 F20101122_AAAUKC loudermilk_e_Page_60.QC.jpg
08891cd098f52db6a9cd1861c6f1d4a3
8742e8b1c0936b2971c764291880978282c86d62
68807 F20101122_AAAUJO loudermilk_e_Page_29.jpg
68dd4977d853a4bcfc38b1dbaaf2b653
6179a009e43d42b6b5ad9435a883f867448d8b94
991208 F20101122_AAAUKD loudermilk_e_Page_40.jp2
5b2582614d325aa7b55be1f98a73e442
2080e4da444a31a75f63f07acf240030267e47c5
47026 F20101122_AAAUJP loudermilk_e_Page_90.jpg
ea3ec91a811fcc14fb5c237c850c2748
17a8026c4f65da3373494909583e07740fae6a06
64129 F20101122_AAAUKE loudermilk_e_Page_87.pro
b10f5e9576b57f3032993ca5fc964c03
280cd494898ff0f82f6271e0b9e8f32ad101cf5a
53337 F20101122_AAAUJQ loudermilk_e_Page_68.jpg
fa6d242cdd949d8c64ed5f4ad40f2e84
06f539fa7dd5c4003d18185b5a82c013603a939c
F20101122_AAAUKF loudermilk_e_Page_78.tif
6601f46877c3944b44e07ede78bb5f7d
aab750b97fdaced9d2348a6d13e06b82446c351f
1784 F20101122_AAAUJR loudermilk_e_Page_45.txt
92c282114322c08c73bdd8a2b4663f48
c9b4f7aaa2787fd7c864cb28a8fe93032431dd03
1957 F20101122_AAAUKG loudermilk_e_Page_23.txt
80b5357e6f75171fbafaf1da4594dc4e
6b204a1975756ffadf04bdd32cbda806bc5d0c36
72425 F20101122_AAAUJS loudermilk_e_Page_41.jpg
59aac6d573118daead3b35e8905c9a26
2fc80a71035b1354a5edca07d701599442c36172
F20101122_AAAUKH loudermilk_e_Page_87.tif
b1e3520952da5d2fb6e5b9a5e7eb908e
8cbe6b86b1fcf54da7b8c7ef72c17423134b83ae
1879 F20101122_AAAUJT loudermilk_e_Page_20.txt
47b7a0788058fd7758742815213a1cfa
739cf7c3a8fbe7ca4b2342d7d2346f726d71a379
6455 F20101122_AAAUKI loudermilk_e_Page_62thm.jpg
bd66a3fd681da8d0e818d38f86395744
852105c600eda341ce7c5a9416a0216e89deccf3
93278 F20101122_AAAUJU loudermilk_e_Page_89.jpg
fba5b20550009e81eba976186442225b
c0ff53deb73d3b9ceb5f6a021937b296d8415318
260705 F20101122_AAAUKJ loudermilk_e_Page_81.jp2
a26ed53473abc74d7f8955aeba4645c5
7979b34b069fb83d6f35a1f184cee74b09d40e7b
52005 F20101122_AAAUJV loudermilk_e_Page_16.pro
14b515e84ddf9fbc72b12897f511a06e
06024bc349176674317157f03a08670a07069040
58874 F20101122_AAAUKK loudermilk_e_Page_10.jpg
6bb215d374e2880a7aa0fb5493b7cdb7
940d2d6583f8432b815db863929bd6e5d50250e9
3010 F20101122_AAAUJW loudermilk_e_Page_09thm.jpg
e706c77515e590d383a96589a73e760c
5010bbfd6bf96b592b41cce61c14cf17d5c64368
F20101122_AAAUKL loudermilk_e_Page_34.tif
8b331b97e1122752598ca98068439d3a
deea3b0b915c014ce4dee6662bfa8a18fb22d844
9325 F20101122_AAAUJX loudermilk_e_Page_33.QC.jpg
04bac2d32994e029eb526104158df242
bc29a1426a45f9092fec2864938468c614630d06
110792 F20101122_AAAULA loudermilk_e_Page_41.jp2
3fe19de261267f609f381c743b7017e4
dd8b4e2dee125ffd54268521ec0dbe59a76db6a7
78976 F20101122_AAAUKM loudermilk_e_Page_08.jpg
8b1b23eb706eeca524c23a42af4eafc1
6cb7d287243b7ebe0dae48a263033da946bbeed9
111524 F20101122_AAAUJY loudermilk_e_Page_77.jp2
97e9b998ded369c57629913f1a62f051
b70fe61205d71acd5b5347b230e8e9aa7913c064
F20101122_AAAULB loudermilk_e_Page_33.tif
d675b22c5b66f8f90beb431c3a4805ea
26000ef93a48dcc4cdcce16fd7aaf3eed52c537f
11043 F20101122_AAAUKN loudermilk_e_Page_42.pro
d62b294c11d196c8bc546d952134f73a
5ddf4ca0a3becff8d22e7c82d92604f4aeb6bdd4
44880 F20101122_AAAULC loudermilk_e_Page_55.pro
06ab1f7cda114a7c692dd83572fa9083
28e6e4a0fa4e91edbb9068792749bf6e71653655
F20101122_AAAUKO loudermilk_e_Page_18.tif
74089a6ec689200bf5ebca52c388904c
df735c96774b26f8e4e61a310250b32f29e1f89e
41954 F20101122_AAAUJZ loudermilk_e_Page_60.pro
cf9c68e10a314842ee33b407959e0150
3f9c3871167085e7ea8763b93a761298b2af5b97
2787 F20101122_AAAULD loudermilk_e_Page_81thm.jpg
0a6353a402156df33ceea424ea94a359
681e46cd4f1a46220418a9a0ceb6b8b0820013ee
24608 F20101122_AAAUKP loudermilk_e_Page_16.QC.jpg
8540b50830fca0c98defe298ad3e7169
c186b48c7a732eb7624ea6f8f04e369fbfa82672
27038 F20101122_AAAULE loudermilk_e_Page_64.pro
ff23a763da206435e109012e8e0772b4
0e8c0a05ec358c0d531b4858e451375b745beb61
1929 F20101122_AAAUKQ loudermilk_e_Page_26.txt
4fe8f60ef5f2c702e4382116f005bf8a
3a81a7f36bb0773da4c0ff34d99bf3718d0f1218
3376 F20101122_AAAULF loudermilk_e_Page_79thm.jpg
af7dcc39ec89c3cffd72a5a576efa30a
23e51f92b8f36dfa3d5ab36a3f8b515be011b5ac
19505 F20101122_AAAUKR loudermilk_e_Page_40.QC.jpg
0458a9f1ae1ec8f87f7a6238bbc39425
9fb130f666676be06674f7d22acfae0945124e10
6445 F20101122_AAAULG loudermilk_e_Page_41thm.jpg
0fedde72683f0f41a6b8cc5375164d66
af62a3194440125d09d0cfa664e0ebc5ffe7aa9d
863685 F20101122_AAAUKS loudermilk_e_Page_35.jp2
76ea84e82a322e09734764a89a063b6f
0effea6fd96f14104893955210b31bff9d5be4ca
68027 F20101122_AAAULH loudermilk_e_Page_55.jpg
018099a599603e3705515aa6e55b0209
10ab4a288920caeb8c2a74acd0b85a7bd8efc3d2
F20101122_AAAUKT loudermilk_e_Page_32.tif
848a39ae1d2720495306d4919ddfd385
782cdbb91c97869ba2b682456eacdf2d8eb479b0
1348 F20101122_AAAULI loudermilk_e_Page_40.txt
064945f4aab8077f52d846103bf5f8b8
c245b0436c7030d9ed28627e11771788fcfb18e2
27610 F20101122_AAAUKU loudermilk_e_Page_42.jp2
0fa6b0928e08c94a23cbf16ea3130538
651d6d2b21a92e0006c76708d8979fa10126245c
16814 F20101122_AAAULJ loudermilk_e_Page_34.QC.jpg
9b7d27325dd90162c819f163b3937c38
016920896b8030dfdbce748bf7fe1df171b01325
F20101122_AAAUKV loudermilk_e_Page_38.tif
50db02d6b79e0cdb91e01fa6ea1e3e3e
3723e83a19064de869639fff7a15b9bd2c94d14a
17035 F20101122_AAAULK loudermilk_e_Page_81.pro
922abb3d31d63baf5b1bafc29cda9ac0
90a508df5ead3dd520cc711a5306f7bedbeb7ca3
1840 F20101122_AAAUKW loudermilk_e_Page_73.txt
76794711b8e9a3341dcefdba00537ae4
3c9bfd914a31f21efde6be65af812b14660bbe48
2082 F20101122_AAAULL loudermilk_e_Page_44.txt
9ec0e23ab66caaa45340d183c6f96729
9011dd88bd116ed848e5c8470d1fe20da6f4ed91
100586 F20101122_AAAUKX loudermilk_e_Page_27.jp2
c4e736ca570d5da3397dfedca0c7f101
2a0ad1805a1d6d873a166048a4a0598850fc15ce
F20101122_AAAULM loudermilk_e_Page_39.tif
62a45d71ad0b80f971ac671ebe977b67
9215554521d6392e31e5cf50b34a5e1a55e33545
114028 F20101122_AAAUKY loudermilk_e_Page_16.jp2
6de418a8bc961cb48172fa8931cf1a84
af5349acc5726359db5f99f2d1f0232bc3d4fcc1
38411 F20101122_AAAUMA loudermilk_e_Page_76.pro
0f1412ed4ba636a0b1e154bbd7a9875e
5dd526e93b9405344871eec3a8ba77c714a12e60
105900 F20101122_AAAULN loudermilk_e_Page_63.jp2
711f5befeefd04bada958848a82b5b24
3c364ef89d9fc96b371f3908c5fc9168555e7377
32551 F20101122_AAAUKZ loudermilk_e_Page_38.pro
28c97a3523e80bccb40e8c28e2d26938
a421b1d6f26797a82f9538beb0f3ad60ba4ed2ad
4943 F20101122_AAAUMB loudermilk_e_Page_64thm.jpg
2fe28c0737fccb5baeec286def084923
398b969781c44038a2081ddc5777f7661b3fccbe
F20101122_AAAULO loudermilk_e_Page_26.tif
070f4a2fc6c510cdfe579736192d5dfa
1065135d89860b5920415b956b909084c2b53e80
20146 F20101122_AAAUMC loudermilk_e_Page_43.QC.jpg
e14a712c8cac9e2cb12ae65a12cc5b56
3461e007846160dd43aa347652b561602609eafa
62739 F20101122_AAAULP loudermilk_e_Page_56.jpg
0f0647fe211330587e2841af932bd77a
eaaf93ba4dca5bf93729ce7298d626d548876ffe
22161 F20101122_AAAUMD loudermilk_e_Page_30.QC.jpg
0deef831640bd42003901eaf65dfc9b7
b566d4115baf0dff7acd49512fc73d1a5294bae0
34649 F20101122_AAAULQ loudermilk_e_Page_91.jpg
de77c398ca344f6eb41120538f4cc893
eb8a4cc6e5c66569e19dec1bb4b43b0b7e9b9651
65859 F20101122_AAAUME loudermilk_e_Page_13.jp2
c8d5c5241ccb4f30c3ec0d5b8801e118
02deab16ed1401cf2e2c826c2aee3a870e223176
68603 F20101122_AAAUMF loudermilk_e_Page_28.jpg
23e60cde40543b8c6de8614e4830c835
c5f6565d6238b025d01eafdb702520280817d0d0
74862 F20101122_AAAULR loudermilk_e_Page_44.jpg
7b6a2657c9fc21694c35c8dc84601738
99ddacab936e6e7b3b39ff6c56a5a56ec097ed83
F20101122_AAAUMG loudermilk_e_Page_84.tif
38d6afc3ee21d5e9d9376a5386e4e8f7
8f7f1f25631efc947ab10a92918a3007a45f3cb5
69008 F20101122_AAAULS loudermilk_e_Page_58.jpg
dc4f4b9cfe031ef7f284749175303395
b3c4fbbcb2901b83833c3c2f29af44c3144fb2f3
26556 F20101122_AAAUMH loudermilk_e_Page_01.jp2
741535c53b781461a3b7e40532724c3c
b18b954d09a089e546760776a5240d6dd8f425ae
F20101122_AAAULT loudermilk_e_Page_41.tif
6b51eb3477d94942b0c89002523b5515
885d6d88170b35e9f0227cb4bc6c53884540b8cf
24637 F20101122_AAAUMI loudermilk_e_Page_61.QC.jpg
4cd6e29591278554d281f1220e00f9f7
5a7e21b8863a473e06f2c583e0a7537a80ca5dd8
19649 F20101122_AAAULU loudermilk_e_Page_79.pro
ee3a468c1474cdd306f4a21d7534f135
b775d6c8372a7f81a59b2c6edb7d6e9d4e41fadc
16363 F20101122_AAAUMJ loudermilk_e_Page_82.pro
104c17ad794218e8eedf31e846924a83
aec8b0bfa65249ca6b478fc77b4abe9635fd68f4
309481 F20101122_AAAULV loudermilk_e_Page_54.jp2
df89f181f664e480e2dce5e25c43b9a2
c1e4a6c62149b652dcd0f9cb0968301f04d3d93e
11608 F20101122_AAAUMK loudermilk_e_Page_07.QC.jpg
afb7c1bf71ceb889940c8342aad480ae
54129304ee826dcf6b3bd72909700bb936fe415e
5386 F20101122_AAAULW loudermilk_e_Page_56thm.jpg
e6c6cb556d20d8fa80e9ca9b106e4f49
68117cac74f585bb4f9673b71d32508c592e627e
F20101122_AAAUML loudermilk_e_Page_85.tif
0fb41df818407b65786de88208049ba1
d4d9ea5915e6752235eb137e6184fc89a84ad4ac
31277 F20101122_AAAULX loudermilk_e_Page_90.pro
983f6859b9c73be3d3f1160fde1ece4a
5ddccecb521a482afbca09879747b3cc557cf426
33187 F20101122_AAAUNA loudermilk_e_Page_79.jpg
4fda64a0be81be4aa4f268bf8f0472a1
6168a34c4caf3f638eb9966b141da1d2e250ec6b
F20101122_AAAUMM loudermilk_e_Page_89.txt
ee9d4865f16a9e3e5145a1d3b6835670
61df5c90bfb6fcf3da93c3449663474a12646efe
794 F20101122_AAAULY loudermilk_e_Page_70.txt
aeec0fff01b6cb7596b31a7e2f203d17
40ce6515403323c5de363de9a5d8d04950a7fd49
F20101122_AAAUNB loudermilk_e_Page_69.tif
6a206d89af90bf4b59ba02153a40939f
c1c67a79b1f8ee4ba3cc08bcc6fd29928e6771da
5846 F20101122_AAAUMN loudermilk_e_Page_14thm.jpg
706ed6048eb6098cf1bb11b928d27fc9
2ce22312ce0f23d851a54c26b17f527e8fc6a475
85474 F20101122_AAAULZ loudermilk_e_Page_10.jp2
f53d2e393d09bd82396f81c649e55e77
a7b03f6ea37b5b0b40bc63cd4ee06869f9265a18
100600 F20101122_AAAUNC loudermilk_e_Page_78.jp2
c5bbb996f66ecae9c0e9db53aacd1af8
bd3bb9e9319d74e9910081f12a5935d55e808f8a
1833 F20101122_AAAUMO loudermilk_e_Page_28.txt
0ae9e957548e25ec883881772fee468a
c16090b9c7a0bb87a18066768da54311e4d555a8
129401 F20101122_AAAUND loudermilk_e_Page_87.jp2
7acaba1df054503446bcce21a36e5919
9205fe75272b6972ec535ac97c877938be2446e5
F20101122_AAAUMP loudermilk_e_Page_16.tif
8b80db789ecdb102e6e900446cf56f1c
955fc63519e157be54853239d2becf51040c6a61
965 F20101122_AAAUNE loudermilk_e_Page_09.txt
204af8e539924b7bae0ffaeeafb202f5
2931d1cf62569c0638c206faed296c5b922205f2
69368 F20101122_AAAUMQ loudermilk_e_Page_63.jpg
1b7fab9e0096bb45eaab2942dfb50235
66885a365e1a8904c287086b83f8b3a7db690dfe
2177 F20101122_AAAUNF loudermilk_e_Page_55.txt
d54941f16d4ca8e47afee578e6e3661a
5b170a7c9a3732a6119b577f3212665aabf9974e
5663 F20101122_AAAUMR loudermilk_e_Page_06thm.jpg
e0476c63171f10851c546b0c5ca81023
aa2cc0b5da514f1e9713a23c7d7c903f4d5b1265
9192 F20101122_AAAUNG loudermilk_e_Page_01.pro
018141badee7830542f9e24509584349
9ac96319cbacd990a10187dcd20780e8541fb2d1
1286 F20101122_AAAUMS loudermilk_e_Page_90.txt
609c39e8114d3a37534e1c46b6beab4a
4ce620770208fb380d68d12ad75e1488f4cdab33
50809 F20101122_AAAUNH loudermilk_e_Page_36.pro
f221b113465b8541b1178b46b3f41784
183ec2fd29cea517468e8e11f0571969c8712464
74261 F20101122_AAAUMT loudermilk_e_Page_72.jpg
d3cb0eac1455ca54a9af9b6a307c9be0
77344068525dd4a6314b19cc9a04db4630b09b6d
24311 F20101122_AAAUNI loudermilk_e_Page_77.QC.jpg
ec3373dfcdcbdc65541831ed1ac24517
92941f2e90f52bb4a843a2c63af705c7ab6252d6
21491 F20101122_AAAUMU loudermilk_e_Page_73.QC.jpg
df2cd9cee74516b8fae83a78cc976f5b
062f862aeb603ff6c311015dd101004ee8b0d30c
77895 F20101122_AAAUNJ loudermilk_e_Page_04.jp2
9465ac7df44edc15ffe06f6e20557109
43651604707d3948eafae81f9b51949112ad869b
48240 F20101122_AAAUMV loudermilk_e_Page_29.pro
127a9e013f6727e1dcdfd1089ba0f52d
884f280ac9cbd01d3660dfcef8ca43be851f0eaf
6164 F20101122_AAAUNK loudermilk_e_Page_20thm.jpg
bb7a249e050a1eec3886a4cd0d521fca
544d173b00ca54c59091f58a094326034bb298e5
F20101122_AAAUMW loudermilk_e_Page_22.tif
2817a95258af0bef17f61c68e0db7335
0192df3cdad4ee2f9f9044d53ac23e55bcde344a
F20101122_AAAUNL loudermilk_e_Page_19.tif
2b85c1f6e6a5d4c03b6f40d10558e38e
5d73c51b6ae90e814c2633b9f4743c9768b56903
24321 F20101122_AAAUMX loudermilk_e_Page_46.QC.jpg
dee6939b5a7cd687ef95d13957ebfc83
d4344536341ad047955e53eff035ff38ea7c9569
133112 F20101122_AAAUNM loudermilk_e_Page_86.jp2
26e342b096bcb9ce726fd98c4505e331
3346ddbc7a2440b54f0459de61ac588d584d84f2
6073 F20101122_AAAUMY loudermilk_e_Page_40thm.jpg
67220aefd1224875aadace24810d99c2
06625bd076c80fa46091943724c7268c1d6fac3f
72510 F20101122_AAAUOA loudermilk_e_Page_17.jpg
2eca09dd6f9366ab5ea479d3fe7ad949
6869eb7b614b0e107e38f1c2c41e46875e8a612b
66849 F20101122_AAAUNN loudermilk_e_Page_73.jpg
0342aecf38f9a9d33ea9810681b035e1
54fef0cdba879325a90681034f3d0615e4ffec93
1762 F20101122_AAAUMZ loudermilk_e_Page_60.txt
a02dbcef754593058f9d09c95efafbdc
cc63f54bfe6744ed1e69f158a8a78a128553ddb0
68063 F20101122_AAAUOB loudermilk_e_Page_19.jpg
4c5bc63f73e81d46c512024ac2cae647
fd66188270652d0948b8c21bb0f379426695fe65
51014 F20101122_AAAUNO loudermilk_e_Page_77.pro
0f2cdaf062e0067e141f8c6d5b366025
d781f0c590202fa32aa48d95e831690a1dfc55d4
73445 F20101122_AAAUOC loudermilk_e_Page_21.jpg
9721ee99cb8576285d8d55addfb7d35e
29cb4f912e1207ce86dc5f0a574de4ac7fbb5c2c
733538 F20101122_AAAUNP loudermilk_e_Page_67.jp2
9d9971b1a9a14c440bf425e1863e4cdd
ffd1a0047271be913bfda818f2ed4a0c39d99c63
70308 F20101122_AAAUOD loudermilk_e_Page_23.jpg
c21df4742c055131b117fe18ac7844fb
edb0c166acd47ea6c5ddd0628d41934cc13cb172
139646 F20101122_AAAUNQ UFE0012041_00001.xml
d8e4f7186e2bc834c6cd5192acc1be00
81288787c2f6fd64b33e26fbd6b91a45e315e379
47611 F20101122_AAAUOE loudermilk_e_Page_24.jpg
db2cf4e8ee25b8b2bafa7e7a55cc2d9d
7ed6e533a5823bad433169e80600bd32adb7351d
63411 F20101122_AAAUOF loudermilk_e_Page_26.jpg
aa777132dbae338a100f992075b00f52
2cf9da982d678bba6b9072e7fc190de65770d079
65700 F20101122_AAAUOG loudermilk_e_Page_27.jpg
75f38c43e8a648e3ffc93e9df5c1d306
f7023fae7ab0e9dd2d22b99c8fb09ec4fff6bd0e
10893 F20101122_AAAUNT loudermilk_e_Page_03.jpg
44665fe80ab0fc7afd24efd08fb63393
685b9c65ee093db5b7fb9dbe8beee14b5e2f7ee1
68084 F20101122_AAAUOH loudermilk_e_Page_30.jpg
46fd48124a049c64a208a5b00ff3f614
e5a27cf19b97998e1344fbf434056824ef4d81a9
38914 F20101122_AAAUNU loudermilk_e_Page_07.jpg
32bc6c0c9ff236e3b6414cc2def81f9d
7aec7b7bbe520141b3fde04a72acdc8ae1cd6718
75272 F20101122_AAAUOI loudermilk_e_Page_31.jpg
85670867d600a16d9ed7a66a66f0b306
ca9b2f2b6458efe4a874004c2a76e932dcecd2cc
35540 F20101122_AAAUNV loudermilk_e_Page_09.jpg
b9c40c65a5d8b3e76518b36f2d1f9d49
f03b90f0a90faa2d89b9ec620f19e6d0efe85b31
63409 F20101122_AAAUOJ loudermilk_e_Page_32.jpg
a10f72ecbb74baa6442dc0a88b44f445
cedb9eeef20041f3709acca202e4ad3f3f229247
73091 F20101122_AAAUNW loudermilk_e_Page_11.jpg
9db2f997c349866890c6b1a84cf957ee
a83b9cb61301feb8539fe14996c7bcd44852bd76
26860 F20101122_AAAUOK loudermilk_e_Page_33.jpg
8dc3f911aed50b8a806c66b20f8afbca
43b15fc1e970f6c2f8f4364188afc867929cce14
45570 F20101122_AAAUNX loudermilk_e_Page_13.jpg
04b067792125ba0613d378da0eda3d2d
c4a2247d1a05979d59b3160ee0eeb18c9aabe43b
51681 F20101122_AAAUOL loudermilk_e_Page_34.jpg
ec7308a01a82db6709694ee5f1c92c9f
d8e1329b17356693fddc17cbe7234b7c2c2aadfe
64530 F20101122_AAAUNY loudermilk_e_Page_14.jpg
7785364a20a35ad8f91bfb9a5618d38c
ae62b33dd196a8f56fa1b10a3a4857b6632d66a8
15812 F20101122_AAAUPA loudermilk_e_Page_59.jpg
24b3e46316765a49b12a9d65d3059a0d
c08e0ba2996f4dbb897f4a54459bffabe0dc4e5f
60889 F20101122_AAAUOM loudermilk_e_Page_35.jpg
845ba898ad2a551073fc279491981084
4d99e392e03aef1a1f312838f7d17de8ff47e251
70650 F20101122_AAAUNZ loudermilk_e_Page_15.jpg
4e0f9053e2ce5c1c656b4590fb9037bd
c649a2f1ca1f42bfb0907f4fe6df0cfb7d8d5524
73793 F20101122_AAAUPB loudermilk_e_Page_61.jpg
831fd5df8fe3367ab5409922332bb329
6d44add59f5aabbaf50cd614bafe3fef0dc5f92e
72307 F20101122_AAAUON loudermilk_e_Page_36.jpg
11d778fddd27c63165273d2f58d1b438
78f23be685759f8a9d0826bee03b42ef621a0650
72474 F20101122_AAAUPC loudermilk_e_Page_62.jpg
c72d820570965a19d88664d1e9ba2512
4ac8a9f2ddb725057acf0c7d6fd82db48bc82bd6
87421 F20101122_AAAUOO loudermilk_e_Page_38.jpg
af92601d5d80c7ea618e117a7f26c229
c4590452a2c333c037adbd1f0df0b9a24ecf6e61
73629 F20101122_AAAUPD loudermilk_e_Page_65.jpg
137183ef7bf7e4cc3fed7c768882caef
60aad5193990bbd44ff9bf9f5195562bf5d5ce63
60510 F20101122_AAAUOP loudermilk_e_Page_40.jpg
40c0470c7d6366ac10f209bac76bd040
186c268559e12bcacee6d5b3eb9d7d0b0d38971b
54165 F20101122_AAAUPE loudermilk_e_Page_69.jpg
188962327536b04b991f60b2598e218d
f453fc965e1abb1a49a094e3324a7f44529d2afc
22210 F20101122_AAAUOQ loudermilk_e_Page_42.jpg
4e97864b6937d4d43b63afb0f7caf76a
f67f86dddc0c6c15484323b49e572b64f892cc7d
36986 F20101122_AAAUPF loudermilk_e_Page_74.jpg
1be54c2b9c8c3e2370baa35037a68e1a
d2d4ec6db97c18fc985745f1a1f9f7dae6d484e3
62527 F20101122_AAAUOR loudermilk_e_Page_43.jpg
49494118c0bb47ec8a4e4ffdaeaae01c
cff633127c327783748ef61ba4056a6925279fa5
47493 F20101122_AAAUPG loudermilk_e_Page_75.jpg
0a5b401edc4fdb7ad765281d31499cf5
d26a396d40ee68a7b0a5e0cf1424c7c0d7d09a6d
64740 F20101122_AAAUOS loudermilk_e_Page_45.jpg
e90e86e0d94c3c45d4a423aea0b06148
0e719e1cc07f45567a252e38381440783b8bb8cd
56400 F20101122_AAAUPH loudermilk_e_Page_76.jpg
a2e437075bec67bf90c926c2fa9c90a2
4fd2b24b08c2cef9477abb38f1108d7a668e266c
71566 F20101122_AAAUOT loudermilk_e_Page_47.jpg
0b9fba8ed30d0015da860928319dda60
f867ea12052d45f12c8fe01cb138a79eda74b814
73602 F20101122_AAAUPI loudermilk_e_Page_77.jpg
16c02ca2bf06a9c21a60a3619afb26bc
75bdef3d5a44d650278538edb35e59770d3b96a1
70722 F20101122_AAAUOU loudermilk_e_Page_48.jpg
0fa1b7d8819aaaea99f3085893ca976b
5eb8ccdb471ab9be8e13eac1e93f12b877fb6ce4
29233 F20101122_AAAUPJ loudermilk_e_Page_81.jpg
9fc827578b59ed59e5c390ea2d0b30a3
3020eb43ffba065b29b3a761965a9f9cb48001c3
28950 F20101122_AAAUOV loudermilk_e_Page_49.jpg
046928d133c35fe3399fa99dae264e40
24efef5168f3351021a160fddc9fdf8b611fe59e
27411 F20101122_AAAUPK loudermilk_e_Page_82.jpg
bb8bcdb5539be59872c14f542f0933bb
d1795258dd846f4b00cc9ee83b7babc24073ba14
28067 F20101122_AAAUOW loudermilk_e_Page_50.jpg
11f55dc99edaf60baf043aabcd8c8bea
ec6fda7e7f209292d355f6a97463c05a7917dfff
39222 F20101122_AAAUPL loudermilk_e_Page_83.jpg
bfd459b0a9097271fafd1d5c595fbf87
1094ada2e0772dd2ab5c7c4af876b7e696af9145
66427 F20101122_AAAUOX loudermilk_e_Page_53.jpg
f8a90e39975f90728d851d4462bd5576
94970f9c02cc86f038ef5821a67988e0131d0f77
103765 F20101122_AAAUQA loudermilk_e_Page_22.jp2
705f94d756b555bee904747db739dc5c
1d72e181eea94bfff26878f140477a452e585e8c
75690 F20101122_AAAUPM loudermilk_e_Page_85.jpg
938cb83a6d68c22a7f9f88ccc136b778
ac16d82333c7e820ecd1ff505b11c8887af3f8a6
29279 F20101122_AAAUOY loudermilk_e_Page_54.jpg
22d7281ee42c9eeda8eda10777380e3c
12d592ce9428116b79badf4f5dafccfe0a67653d
109558 F20101122_AAAUQB loudermilk_e_Page_23.jp2
048a34be02d23ea4d5df373113d1b7b8
6c807bcd52a41f07dfc6999adc6b5365811bac9f
91093 F20101122_AAAUPN loudermilk_e_Page_86.jpg
db12f01260f7d51c8f7cf56675d3dda0
713f950fd9e8a997c65fc6d4fe52a98e1f9672d1
69329 F20101122_AAAUOZ loudermilk_e_Page_57.jpg
a5ce32d706680e48dea6cdadd0922839
e5b013b5de553ced7dc6dae1b8a6b72e66e46da5
93225 F20101122_AAAUQC loudermilk_e_Page_26.jp2
221596c6afdbcccbd06875f86eb31d46
54f0877dc3f4e86ffc8c203659a7e311929355a8
86430 F20101122_AAAUPO loudermilk_e_Page_87.jpg
1fb16a7a0e27967fdedc7fb3736f0f9a
bd662f03562daf3648eb186f08236b1731e87315
109724 F20101122_AAAUQD loudermilk_e_Page_31.jp2
cc2a91703d06e52fdfb0e2c1b6cb4262
540afa72bea34446c0b86cce08f2f452373752c9
97054 F20101122_AAAUPP loudermilk_e_Page_88.jpg
df295a4ef475ee5db1d70aadb69f44f3
d98021216861e5e5712f7fd35f40c3042d91693c
281289 F20101122_AAAUQE loudermilk_e_Page_33.jp2
440cf8494739dddec9344e572eefb5f3
acf1ef7a415760d105906dbe1c1beff7a7d0cfc7
5955 F20101122_AAAUPQ loudermilk_e_Page_02.jp2
dc0947b9888a891242deade3eb3b9b53
ac71ab9961c34878151c6f6e14e0a7d06d4ec14b
687963 F20101122_AAAUQF loudermilk_e_Page_34.jp2
55fcaeb6019356338eacb82677c7677c
f834b63d1198499cc2510a8bfeec0367b2837497
6587 F20101122_AAAUPR loudermilk_e_Page_03.jp2
648c2baa4667708e38daffb5daca6e09
9efd037691f9f9aab1f9c9725f5ac10024481134
108259 F20101122_AAAUQG loudermilk_e_Page_36.jp2
d5d373f29af024fc16e7bc4245632584
ce1625c6b8e0059cea314f439bb590fb14ad0036
1051967 F20101122_AAAUPS loudermilk_e_Page_05.jp2
a37e76c8ab800c85b441781ef24b1163
a91b88acba1de5975ade6c3472e12ee246148a02
637237 F20101122_AAAUQH loudermilk_e_Page_37.jp2
19511d8528fd150454abba0aca1e53d6
fc5506cdb2976b7eca3a14583c7ef4d48fea359d
985376 F20101122_AAAUPT loudermilk_e_Page_07.jp2
1ac308fc4b4c7cb223f934f2852538f0
6c08b8af2ed041eef6ef68d22c0624cca76fe6ef
1051981 F20101122_AAAUQI loudermilk_e_Page_39.jp2
fb1de66ed0461dfd3124b02ac7fbfc30
fc49687009c8062f72e75e8472d031e5af9f166d
946178 F20101122_AAAUPU loudermilk_e_Page_09.jp2
0bbb7408e4b72af42dd8a79385f80539
65c18da88d1cf538bc95819c6a7ddfadaa059ae0
98259 F20101122_AAAUQJ loudermilk_e_Page_45.jp2
1156232ebdeb2052fbfe4440a3e5b9ed
9dc58602aa2e9803a3990cb3837a575f6a49fa44
110211 F20101122_AAAUPV loudermilk_e_Page_11.jp2
3ef5883ead8b9ac70f769015efd6e27b
a3d991478000c3d56b7c77df85911742ef0b83be
108350 F20101122_AAAUQK loudermilk_e_Page_47.jp2
3618fec47c16c27e0d3d0aaf589ef458
081688ae5721009772f4ef0c9d0d9c5ba777f365
98531 F20101122_AAAUPW loudermilk_e_Page_12.jp2
b1de8726365201888cbb1777590c5f88
59b5d699b28122dc8b689ba4930d995a6eb39c77
105072 F20101122_AAAUQL loudermilk_e_Page_48.jp2
b32ad0c9609b8530c2ed1e80a408b066
5a77208148da165c45d9ad8064472e742d6b1d6b
94706 F20101122_AAAUPX loudermilk_e_Page_14.jp2
615ea6cd8c6a465b4704293356958865
1d64c5ce0930f8cc844034fb54c00150c4b1cacd
112853 F20101122_AAAURA loudermilk_e_Page_72.jp2
30cf15277112b00c9a83b9a8821cffd6
3a0184b73b5db670ae8499ee55e3fafa4df8b7ca
312688 F20101122_AAAUQM loudermilk_e_Page_49.jp2
66d37c59cdd6068e226cd9b8236a3498
ae651ee7b70095fb097224164862622292c2026b
106469 F20101122_AAAUPY loudermilk_e_Page_15.jp2
b7468712ada2203c8059ed16bb54bda6
2056d9c99bfc1d2d771cadeb8f5b5798c3823d9c
862952 F20101122_AAAUQN loudermilk_e_Page_51.jp2
01dcf60849671424195609f11853f3e0
5e1b8892bd471a8feb854ab00c209ff67f927d61
101802 F20101122_AAAUPZ loudermilk_e_Page_19.jp2
3486b637f49f77b3ff4667a572bc2ced
c40340bf51034b7d1a9cf691b95ba6017a5c6826
100637 F20101122_AAAURB loudermilk_e_Page_73.jp2
656b77ea7783159e8da52e5bdd9be75b
b2825991de1d446435209f13cf52d068f9637b99
70634 F20101122_AAAURC loudermilk_e_Page_75.jp2
473444be99f3af115aee4155dfe35227
352ce85071b54d408232df2d56f2772155b88fa5
90040 F20101122_AAAUQO loudermilk_e_Page_52.jp2
c11601825577cf96e1e34bb61c3fc618
c1f4ac1d2d052d61f892d6e5f7d43e0990ddb73d
84313 F20101122_AAAURD loudermilk_e_Page_76.jp2
b8b6e299d0b237e07b0caa44cd05e480
9561b7547ad5b600ff0884d4109883bacaa30146
100531 F20101122_AAAUQP loudermilk_e_Page_53.jp2
6d3e1b3132d778cc9b124bfc08d56906
e30b92d508a9e4e19229c49de406df550c2a1e0b
45672 F20101122_AAAURE loudermilk_e_Page_79.jp2
88bc30dd7735a0499d56336d5eafc430
b56c250ec6913afd44694461fd28d1b493d8bac4
899919 F20101122_AAAUQQ loudermilk_e_Page_55.jp2
009a11fdf3e7bfe49ceee92d429ef0fd
e8bc39cb9fbf9cb3debb65441826452a8ba0802b
534019 F20101122_AAAURF loudermilk_e_Page_83.jp2
60bba10d90f1e2f4dcff3127ea1fc123
aa714a36b0ed0b6dbfb30a1b255555618ca3d712
105693 F20101122_AAAUQR loudermilk_e_Page_57.jp2
bd2715b06ba87cfa9534d102ad1b3ef8
49f4e4165c68118876b36ef979fbf5f46713df34
384715 F20101122_AAAURG loudermilk_e_Page_84.jp2
7498fcad7dc0de50d870c2298170bdbe
ad683fa1e5dedb5060da8daf7a5445fdb3ee06f8
105178 F20101122_AAAUQS loudermilk_e_Page_58.jp2
98fb954baef3963dbafe2214704b81b1
baef8dac0405b47889319c0ddce6e4ebe6a22df1
111011 F20101122_AAAURH loudermilk_e_Page_85.jp2
0ca1564285df4a674543f11c5ebb050c
e6f1298bc56abf7d78a9e652c402f91e47129d0d
16104 F20101122_AAAUQT loudermilk_e_Page_59.jp2
ed843e7e2b76f75222ad3d451bbe64ae
63813d8a24f74dc499753e0570de496a039c5c02
136363 F20101122_AAAURI loudermilk_e_Page_88.jp2
8581da786c1a0489529015b679fd6472
5ba624384bf1632bb86dbbfe4a3dcd7c2604ab16
114986 F20101122_AAAUQU loudermilk_e_Page_61.jp2
f918eb6c20f91726bbff150bd4d377dd
635495b56553da13dff9b1f3fa9653f4479d2bb6
46845 F20101122_AAAURJ loudermilk_e_Page_91.jp2
d38f6c00c580078c54d1350a2c9ab133
bd83dde861b818f5e40801b60ef2b638fed263b6
691469 F20101122_AAAUQV loudermilk_e_Page_64.jp2
5b913fe0ecbe17c6f405623e19c053e7
768bcf424f3c530c8aa7b20695f1fb5b3c2e78d7
F20101122_AAAURK loudermilk_e_Page_01.tif
7402ce55fea6b4e467b848744e8fb452
9f922fc0d975425b80f4ddab7adc6f375ab0300c
112527 F20101122_AAAUQW loudermilk_e_Page_65.jp2
c0c9cd3b133753d4faf1527ea65e8d94
d79843721cb865089d136899ea55351a31cc8831
F20101122_AAAUSA loudermilk_e_Page_37.tif
615019891ca7774704ab709898cc9919
d53c0d316d8c8167e0e849e8f3ee3bf18d183e5b
F20101122_AAAURL loudermilk_e_Page_03.tif
1cc95c9df96f963d4ca528e1baaaca1e
a490cb6d398749069af220dec8354f0b74eb8507
112057 F20101122_AAAUQX loudermilk_e_Page_66.jp2
8ace9c2bb00e55b67cc36ef3d3d7aaf0
827a0e1cbd7ccaedf0b5eb3d73bb8b2b809d294a
F20101122_AAAUSB loudermilk_e_Page_42.tif
db31edaf87e6bf921cb6f1347e1876da
d4e088dada90dabc0444b0dc6cc120bec4d3a945
F20101122_AAAURM loudermilk_e_Page_05.tif
daedaf5b70adeab6882d0379719ffc99
c1d43f9e45041f77bea80cfe144b7d24a2abfcce
743034 F20101122_AAAUQY loudermilk_e_Page_69.jp2
115384d25aff23e95afb0d471da65441
5e57be1182aef54dfadb350d606df699b3e4560c
F20101122_AAAURN loudermilk_e_Page_07.tif
690c6d6ec1b9189f9153cf1b331ef9ec
8bba81f60a4ff2bc67110831dc38c2ab5d783bdb
374224 F20101122_AAAUQZ loudermilk_e_Page_70.jp2
f8241ad54eb87a6b615c065f84c0eb25
40a5fb6ff98ceea60d7e42413c75444928585c7a
F20101122_AAAUSC loudermilk_e_Page_43.tif
c73f607fa1022e70160e40b085ca22c8
175f06dae392b34023bf4fe274f5119e52aa0176
F20101122_AAAURO loudermilk_e_Page_09.tif
11a5c823f6247b6bfe841cd0557773ad
47360988ab5303798b11da208650db33fe4afdf3
F20101122_AAAUSD loudermilk_e_Page_44.tif
828960abb844f299acea2f19a82f6c98
6fa7e2c59426fe28145fba7b6cd5c34906cae25d
F20101122_AAAURP loudermilk_e_Page_11.tif
712b4c8204ad95d33f3fe3a82a228135
553369f50d122ed7c123694c9cb34cbd6157daba
F20101122_AAAUSE loudermilk_e_Page_45.tif
ade3e36ac09e68c7d588f2b0dd7024ce
e9403bf628e4f9f899b39bbe2c5994261869dd5e
F20101122_AAAURQ loudermilk_e_Page_12.tif
21f8ab7b9fc20e0e0c8a5661d524be3f
828d9ee81c7d485ec4798885d4b673d7386bfbc6
F20101122_AAAUSF loudermilk_e_Page_47.tif
5f181f639f3a766812c64561fccbb995
2cbfaa91b0e1b28d332e692a000dea697476b2a5
F20101122_AAAURR loudermilk_e_Page_17.tif
53e84b3e2e4bad9ede80841a954ebb28
dd300ff2d4a2d72f3761028e360f4e461d9a6b62
1054428 F20101122_AAAUSG loudermilk_e_Page_50.tif
10a15d8e0c7fbd2d3694e3f00e5546aa
3c85bd1ac28290230146ba856c29c3f64a725318
F20101122_AAAURS loudermilk_e_Page_20.tif
407b8b2747c28134521dc283f3e54b3a
e2ce92a7b57d0608605311b987dd23bde477a9f4
F20101122_AAAUSH loudermilk_e_Page_51.tif
7c124d327c3021962b18659ad776ecb2
d124699c25e4652718bc8e4013adf7cfea45b1ca
F20101122_AAAURT loudermilk_e_Page_21.tif
c1eed429f53c28bad1494b7b9a3270ac
9ca1e90ace6ff976f70803b2b11ee2576e4d028e
F20101122_AAAUSI loudermilk_e_Page_52.tif
cf5d19c4cbfab1d218e0a53c8d7ec536
5eb1fefb0c73a7e0098eea07619fe94606d952a0
F20101122_AAAURU loudermilk_e_Page_24.tif
f3ab0e472ab6dae6a2944aa38135f9ba
2d29586b760939ebcda01847d4b34cacf4518812
F20101122_AAAUSJ loudermilk_e_Page_53.tif
a809d0a0710e1919ddd7efde8a540fb1
da491a61b204aa0d7f3bd1abc27c51fee4af199d
F20101122_AAAURV loudermilk_e_Page_27.tif
a7331056ce75d2579a1abd54cf922d71
dd16ab71b84a7f8e440aec744c1ba6f1e2d70787
F20101122_AAAUSK loudermilk_e_Page_54.tif
d21f2784cd7b2b2bc967a53807d01933
c5e497af6d9017205f84cfd06e78f22db6d5c755
F20101122_AAAURW loudermilk_e_Page_29.tif
1647169b5c3ddbcd6c5d2d4a52179017
4e58a71fbe8b9971a3c795d1d04668310571d6fa
F20101122_AAAUSL loudermilk_e_Page_55.tif
ebf18aabae458a039db52a0a69dd256a
949be641f1b759d079985ac6c387d371d0af93b0
F20101122_AAAURX loudermilk_e_Page_30.tif
585485c6b0957a26cfb667887b73cfc3
825a689ae8672d092c82af06e65c6bac87be7026
F20101122_AAAUTA loudermilk_e_Page_79.tif
3075cc89bbff66ac58770601bffbc413
680a82a69e7efbe08cb9609c44d0da0f387e2ce3
F20101122_AAAUSM loudermilk_e_Page_56.tif
2e2178041d69e4a7e535805cdaf93251
afcbe763ad72b33b018de652ea857b574dd91464
F20101122_AAAURY loudermilk_e_Page_35.tif
e762759ea9a1799d5dfa726f35bb819a
677a795a9438e9604cfc3f3285a106de395085bc
F20101122_AAAUTB loudermilk_e_Page_80.tif
e9cca24a5e6b2c99691ff7391a273a93
f95d7cf9bf15cb442d53736968c9ebd698c50285
F20101122_AAAUSN loudermilk_e_Page_57.tif
a0e1754b9b01256bcb3c4b8417f65249
7e4ff34cd5bad17834af41292d6363651635bd17
F20101122_AAAURZ loudermilk_e_Page_36.tif
41a6482e36afb72e07cb5eacbabf3000
a5e7f67bfa4e51106ce805869665bd1684431f26
F20101122_AAAUTC loudermilk_e_Page_81.tif
a846b35c432c50d6531fd195f93822d1
bdbf36dea51fc94966f2cc4e4be66389b02e0221
F20101122_AAAUSO loudermilk_e_Page_58.tif
bc3ed3de2117bab31e24fdb0e7c1c53d
bd3c4bb39aca720d736dcbaf59ba7b69be5843ba
F20101122_AAAUSP loudermilk_e_Page_59.tif
8103e5874c40be32d6299411bdf9e929
686cc2613197d110cd5d204874473a02a80b2b43
F20101122_AAAUTD loudermilk_e_Page_83.tif
e536b90431b15f728cdff3f5beb2c84d
8f6edb0613f347100ce46090a60c99bbb866d16b
F20101122_AAAUSQ loudermilk_e_Page_60.tif
3e18ee48471a6b10cdce0c4bbd106b81
e466f64e5902c61fbcedbfaee5b0a65850e532cb
F20101122_AAAUTE loudermilk_e_Page_86.tif
4413775a8dc4feff67d5dd1e72ffb0f5
d9fa7e8e738014c25a2467f8fef57da5745cbd59
F20101122_AAAUSR loudermilk_e_Page_62.tif
3d2b2abf704a8e1a05b70f73b8b37929
eee48de2a874deb681f9533b31e4905154eb45b1
F20101122_AAAUTF loudermilk_e_Page_88.tif
b6f975fb4930679370adc059cf0d1bb6
1780fcd34ead4085a2398c3a73cc16cb0e6edd08
F20101122_AAAUSS loudermilk_e_Page_63.tif
2a35a287550d5ad8f82aadb6aa9ecb43
b7167867608bbb02fe3b7d20d4f3daed39fd9098
F20101122_AAAUTG loudermilk_e_Page_89.tif
f08e0ddb1ac04ddb99ec35cdc083b7a2
49a56e2d33d3365f9c28db56fd0c63f1dfd4ed59
F20101122_AAAUST loudermilk_e_Page_66.tif
67099b9c9e53a1fd021a1fa0eac93929
a2c0722aba536b00b3e35f900fcf541a908f6ddb
F20101122_AAAUTH loudermilk_e_Page_90.tif
9dfd2cd4f7e68d7ec221e61b98770795
b22acba4cab23aa7ee443491d03b2e03a343c23c
F20101122_AAAUSU loudermilk_e_Page_71.tif
9d931324c2eaf9e176d46f4e5cd79def
396b274719ae314a81358907a0761743563e73b3
1641 F20101122_AAAUTI loudermilk_e_Page_03.pro
6732c4cdb0a53165febbf1c19d63726a
e10b6aff2815b720fb4159c047dc58c47d8aceea
F20101122_AAAUSV loudermilk_e_Page_72.tif
972bc0316b1f6f79d40de33b7c1037a3
b751fd18049fbfb18d061dc3fe0fe690439014e6
26752 F20101122_AAAUTJ loudermilk_e_Page_07.pro
1e092a26109758a7b58442fdcdd2d6f0
b516a14e6bc6980da80e805f6fb83df8266ef34f
F20101122_AAAUSW loudermilk_e_Page_73.tif
74663f80e5a111d86c6d1863943ae79f
67c305b78f91ce131804561da51e8cc4fce444ca
62820 F20101122_AAAUTK loudermilk_e_Page_08.pro
7533826b12bc09a5a4cd997831e8ea1d
702c5cc26142821cb3b77e504fb75464d5e0633c
F20101122_AAAUSX loudermilk_e_Page_74.tif
f823a6c2fc98a520d63274a15a3e2a15
84c8b5c22d094bda2eb96b720d5cfdf6231d69d9
42472 F20101122_AAAUUA loudermilk_e_Page_32.pro
941acb8d15624af001f2184fed4fd3b3
2273f28cd5ee00fdd70ea30fe98ac6e303a968f6
23857 F20101122_AAAUTL loudermilk_e_Page_09.pro
043381d2241b0d8568819b58efd0607d
297eb1e83931f5fb4c2a5a0cd10786242a44228a
F20101122_AAAUSY loudermilk_e_Page_75.tif
3da1d60757a26c9c5b270b35f08775cd
2b694c75888470a84601d5aacc06bb4767fa1278
11831 F20101122_AAAUUB loudermilk_e_Page_33.pro
4c0801a846f190e08719283cb36356b0
fa2afda7252daf6b5b3a306694d03cae076f812b
37891 F20101122_AAAUTM loudermilk_e_Page_10.pro
27ebb81d9901569e90767101e12ecf24
e61368afd6a91dd142c671eed7900edbbdf699c2
F20101122_AAAUSZ loudermilk_e_Page_77.tif
e82ddbb2815f249b7ece9999773052e3
0c0ef4f9754bb18dfd1c9d74ae1ffa7ea3e7b36c
29134 F20101122_AAAUUC loudermilk_e_Page_34.pro
644005e44ddbc8736954a9c6934e496a
2f9318392344775464fe40d92bbde237af91284a
44470 F20101122_AAAUTN loudermilk_e_Page_12.pro
434415255fe56f83b6d809691ad70124
225850b6bcc978e43f6245f6922bdd31b2aa8930
22876 F20101122_AAAUUD loudermilk_e_Page_37.pro
78e14a75b262126d9234b5dad145c6f2
3a5fdf7586940eb8e22cde827c926b22d34011e9
28773 F20101122_AAAUTO loudermilk_e_Page_13.pro
b6ade95f8c71969495549abce6edbf87
03b6f23ee2c85dbd232fe90d4418701dc84fc1e2
51741 F20101122_AAAUTP loudermilk_e_Page_17.pro
0fa7c78300be2718dbaf9f242c4122ab
91d19474cfaaccc637f68e79e5a63c63b62a6930
31528 F20101122_AAAUUE loudermilk_e_Page_39.pro
358414662ef10c858080dc8b6f436d94
a054f3e4caec5d4bbce221982606ac79eec9b4ca
48565 F20101122_AAAUTQ loudermilk_e_Page_18.pro
70918033cc03251e24611884f8cbdb5d
cacb3dbb66e7518ae68494fa628a5c9b5af4d7ff
28324 F20101122_AAAUUF loudermilk_e_Page_40.pro
816ac266bd0ff9ee7d30ee18e377e87a
0fd3cd3dc6016b4ad4441301cbe161194eb4ad28
46622 F20101122_AAAUTR loudermilk_e_Page_19.pro
2a124e283546b67c19a566b147d50177
effb68293a59716b057f5a9e9d61811abddc55fd
6781 F20101122_AAAVAA loudermilk_e_Page_61thm.jpg
d18db553b7935c623702e62554e1117d
bfbda25388d4b2f424eae38fd19f89262eb5edbe
51316 F20101122_AAAUUG loudermilk_e_Page_41.pro
fdab90b48993b2e252c49910abdc6d49
6fc2e7e4359a47efeefc541672e5f4f741793619
47666 F20101122_AAAUTS loudermilk_e_Page_20.pro
d05f4749c99c88563e50895bdbed8a3b
30e2a52b6f39a5fa48935c0589e796ce2bbf56a7
23735 F20101122_AAAVAB loudermilk_e_Page_62.QC.jpg
5487e72373ab0a2ee6aaf4f7658a161b
da53ebb1ccddd8f4a8864f94fe015d64c703bd94
41610 F20101122_AAAUUH loudermilk_e_Page_43.pro
3467ec3f905a40f0da1360fdd924f02f
73dc02522a315aa213b3f2e854f0d41f6d0f2f83
47813 F20101122_AAAUTT loudermilk_e_Page_22.pro
55a0c0f0000560869f62b055e8d820b8
3b30aab79429548715dd449a71d1bd3ce80041a8
22201 F20101122_AAAVAC loudermilk_e_Page_63.QC.jpg
b0299f969abdbe8984e77d9cd08b2709
e4a3a53ee788c5e1df162a7f6275266d46c71ebc
53065 F20101122_AAAUUI loudermilk_e_Page_44.pro
c9e29ce7a8fc33907b4e92a9a0d8ba1d
fe0214a133ef9a49685940eaccb3f6a1ce239ec1
49679 F20101122_AAAUTU loudermilk_e_Page_23.pro
83b7a1a850322d71039d3bc6c62a8729
f4ccfae69cd6aae026c3162a32da75f5f6b59fdc
6482 F20101122_AAAVAD loudermilk_e_Page_63thm.jpg
93603441a8db02f0138bcb56e9edfebf
6704723eb6bd9284bd0b323c47304f7392edbb31
44123 F20101122_AAAUUJ loudermilk_e_Page_45.pro
a5b3c7e80c62dc7f26491e3c2a2d799e
eeab889cbac1e2227db315b73e9ad2e8bbcf4e39
24824 F20101122_AAAUTV loudermilk_e_Page_24.pro
e2cfd1c1e855d5c199fe6a3b75106b40
b79a69cfef4f01272d95e99863b3b882b49be83b
24239 F20101122_AAAVAE loudermilk_e_Page_65.QC.jpg
71ba3e89a6169044cc8456867390d9f5
886bc9d27feb8a1756e1c19739f4b6b2b0ec9e51
53267 F20101122_AAAUUK loudermilk_e_Page_46.pro
0ac1fedce94647bae49d14cc34999f70
223a9e4c23b5d699179d9a858f032a890d8f530c
51885 F20101122_AAAUTW loudermilk_e_Page_25.pro
f3cec8dcbf486fc8116af61781546ffc
d3f1f8710900f8764f77efcddbc4901fb4120907
6605 F20101122_AAAVAF loudermilk_e_Page_66thm.jpg
ac9a2c2b0ecb13ca7231de4e80826e7f
bfd5245637859cad14c32c8915e64fd8e324db41
44902 F20101122_AAAUVA loudermilk_e_Page_78.pro
4979b93065fde15be78dbd703392fa59
aaccddfb8c5b67dd3cd6523a1faf4c34ed21782d
49638 F20101122_AAAUUL loudermilk_e_Page_48.pro
7152b6935d395ed0bc422385a1c6428d
89bb737487a4f47401a3d42d553839ccd70f1328
42964 F20101122_AAAUTX loudermilk_e_Page_26.pro
d05048a801cb7b23ae408ff304d70c92
59f9d3200ae027ae84fea47a8b065ab8d9ca6fae
16822 F20101122_AAAVAG loudermilk_e_Page_67.QC.jpg
3b5695d9d0ce27a16b7a5915f248ef7e
f45d4821a830927304d5a7d2d82ca8de055e6a6f
50553 F20101122_AAAUVB loudermilk_e_Page_80.pro
6f1fa22c6c2c0f1d0a2aa6eec2544226
2047c8aeb613e0db1cd2a112d79f031db56fc131
12692 F20101122_AAAUUM loudermilk_e_Page_49.pro
5a6c3ab228908ac3f693541ff3e011ec
03a0eaca2adadc447f37e7aef5cdc63a886cdede
44593 F20101122_AAAUTY loudermilk_e_Page_27.pro
40a49bdcae927547438287e4a196d8fd
98a607d5e1d692db0cdfa6c309dddd86dbd2f026
5042 F20101122_AAAVAH loudermilk_e_Page_67thm.jpg
fb5479077626d043f0a855cd0ab447ce
13991149d4eed11288f9e60b128646c4a88f33b1
24052 F20101122_AAAUVC loudermilk_e_Page_84.pro
9db99af91bae71812794eb28e046b9e6
8cd4c2df55d43f697fe18966642902636be15383
35119 F20101122_AAAUUN loudermilk_e_Page_50.pro
e44c29f1ea837e25f939698d74b76df3
8d330e9757a0c66cea6a34b50dc7f8315c3aea54
54006 F20101122_AAAUTZ loudermilk_e_Page_31.pro
a755bf77556eadfa7f07a348df133b1c
b4817d0c88a499f2fa9adb797350d3091a70dea2
4855 F20101122_AAAVAI loudermilk_e_Page_68thm.jpg
eadbefc058a8d781c1c5c9340de9e6ed
b0763a0996b3a5a2c5df2ba14a77dc3d23bb008f
66757 F20101122_AAAUVD loudermilk_e_Page_88.pro
82d555d89d9d8289af9e8227d2ac23d1
35df5e6294c6520499caf86fc57b59b3bf7830b9
39701 F20101122_AAAUUO loudermilk_e_Page_51.pro
f715ba404ba5ae60c6581c00423806d2
9eead1e8d4a71780106228d9ef4021cddaa89294
16998 F20101122_AAAVAJ loudermilk_e_Page_69.QC.jpg
77acf1ba0b4264ad9be7d3d42efd574c
a354fb52a8796a2800f5c16f9538ba520a2b9fd6
58267 F20101122_AAAUVE loudermilk_e_Page_89.pro
6c2f8c5a33f03098edbb119cf4257e94
bbfe496d8d0e88cb7d1648e503bca63132f4f70e
11248 F20101122_AAAUUP loudermilk_e_Page_54.pro
fc4ad99b606af6364bd518661a147a48
46faea5040d7ebe5705b87510ad5d927f6a02b10
4729 F20101122_AAAVAK loudermilk_e_Page_69thm.jpg
66c9f235b7da86ef90db70dd971d67a1
b1701463a30f68c15b6468e73eabe091b6f54225
35598 F20101122_AAAUUQ loudermilk_e_Page_56.pro
fa26791f64151dfa9651dcb1cc5d0ee9
b9a9b1a30ac68bea7d92da958e29d75977860b8b
9336 F20101122_AAAVAL loudermilk_e_Page_70.QC.jpg
93e9a10b21ac01f4a0fecdf32a17056e
89c41cab7ce1927a96578f3b548b57940d13328b
19847 F20101122_AAAUVF loudermilk_e_Page_91.pro
a29fc9f719f85134f7e5f9266f5b1b19
c8541abadd3bd81eb50b2f916c4b6a0d49ee0c47
48588 F20101122_AAAUUR loudermilk_e_Page_57.pro
24b7347392fbe9ab239fb3e02d2b91f1
7722c6f979967bfeb96326a576fcd85e29996f2c
2774 F20101122_AAAVBA loudermilk_e_Page_82thm.jpg
91e498092bdc903d6da3d48f40fc442d
070a4f0361e25983143700327f292bde26ea9279
2917 F20101122_AAAVAM loudermilk_e_Page_70thm.jpg
48218c1da52fb26b30dfc165e4868d9f
ce165a0dd583b7af1340a334a6e6f11bcd64ea18



PAGE 1

SPATIAL AND DEMOGRAPHIC MODELI NG TECHNIQUES APPLIED TO THE LONGLEAF PINE ( Pinus Palustris ) ECOSYSTEM OF NORTH CENTRAL FLORIDA By EVA LOUISE LOUDERMILK A THESIS PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLOR IDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF SCIENCE UNIVERSITY OF FLORIDA 2005

PAGE 2

Copyright 2005 by Eva Louise Loudermilk

PAGE 3

This document is dedicated to my family and friends.

PAGE 4

ACKNOWLEDGMENTS I foremost thank my two advisors, Wendell Cropper and Loukas Arvanitis, for there support, dedication, and belief in my graduate career. My other committee members, Michael Binford and Jane Southworth, provided invaluable help in all aspects of my research. Jack Putz, Eric Jokela, and Alan Long have given their expertise in various details of my project. I thank the School of Forest Resources and Conservation for financial support as well as George Blakeslee and Tim White for their dependable guidance. The opportunities and amenities at the Forest Information Systems laboratory have been immeasurable. Mike Rasser, Balaji Ramachandran, and Valentina Boycheva gave me guidance early in my graduate research, while I have been able to confide in my trustworthy peer and lab partner, Doug Shoemaker, throughout the years. Special thanks go to Libby Zimmerman, Mike Penn, Amy Copeland, and Bobby Cahal of the Goethe State Forest staff and Robert Mitchell and Stephen Pecot from the Joseph W. Jones Ecological Center who have provided valuable data for my research. This experience would not have been complete without the encouragement and support from my family and friends, especially my husband, Erick, and my devoted parents, Eva and Staffan Lundberg. iv

PAGE 5

TABLE OF CONTENTS page ACKNOWLEDGMENTS .................................................................................................iv LIST OF TABLES ............................................................................................................vii LIST OF FIGURES .........................................................................................................viii ABSTRACT .........................................................................................................................x CHAPTER 1 INTRODUCTION............................................................................................................1 2 SPATIAL MODELING OF FOREST SUCCESSION AND FIRE DISTURBANCE IN A LONGLEAF PINE (Pinus Palustris) MANAGED ECOSYSTEM IN NORTH-CENTRAL FLORIDA..................................................................................3 Introduction...................................................................................................................3 Modeling Forest Succession..................................................................................3 Longleaf Pine Habitat: Sandhill............................................................................4 Objectives..............................................................................................................8 LANDIS Model Description.........................................................................................8 General Characteristics..........................................................................................8 Fire Disturbance Characteristics..........................................................................10 Methods......................................................................................................................11 Study Area...........................................................................................................11 Parameterization of LANDIS..............................................................................13 Fire Simulation Scenarios and Parameters..........................................................17 Calibration and Sensitivity Analyses...................................................................18 Results and Discussion...............................................................................................19 Calibration and Sensitivity Analyses...................................................................19 Fire Scenario Effects on Species Abundance......................................................23 Management Implications...................................................................................25 Future Model Development.................................................................................29 Conclusions.................................................................................................................30 v

PAGE 6

3 DEVELOPMENT OF A DEMOGRAPHIC MATRIX MODEL OF LONGLEAF PINES (Pinus Palustris).............................................................................................32 Introduction.................................................................................................................32 Longleaf Pine.......................................................................................................32 Population Matrix Modeling...............................................................................33 Objectives............................................................................................................34 Methods......................................................................................................................34 Data Sources........................................................................................................34 Transition Matrix Development..........................................................................35 Results and Discussion...............................................................................................40 Error Analysis......................................................................................................40 Elasticity Analysis...............................................................................................41 Fire Frequency Analysis......................................................................................42 Future Model Development.................................................................................45 Implications of Using Multiple Data Sources.....................................................47 Conclusions.................................................................................................................47 4 HYBRID APPROACH TO SPATIAL AND DEMOGRAPHIC MODELING OF LONGLEAF PINES (Pinus Palustris).......................................................................49 Introduction.................................................................................................................49 Longleaf Pine.......................................................................................................49 Hybrid Modeling.................................................................................................50 Objectives............................................................................................................52 Methods......................................................................................................................52 Study Area...........................................................................................................52 Outside Data Sources..........................................................................................53 Hybrid Modeling Techniques: The Spatial and Demographic Components.......54 Results and Discussion...............................................................................................57 Fire Effects on Individual Populations................................................................57 Implications of Hybrid Approach........................................................................60 Future Model Development.................................................................................61 Conclusions.................................................................................................................62 5 CONCLUSIONS.............................................................................................................64 General Conclusions...................................................................................................64 Limitations, Implications, and Future Work...............................................................65 Broad Implications of Research..................................................................................67 APPENDIX: LANDIS PARAMETER FILES..................................................................69 LIST OF REFERENCES...................................................................................................74 BIOGRAPHICAL SKETCH.............................................................................................80 vi

PAGE 7

LIST OF TABLES Table page 2-1. Attributes of the four species studied using LANDIS................................................19 2-2. Calibration analysis of mean fire size (MS) in each fire regime scenario..................20 2-3. Calibration analysis of mean fire return intervals (MI) in each fire regime scenario.....................................................................................................................20 2-4. Statistics on species abundance within each fire scenario.........................................29 3-1. Transition matrices developed from multiple data sources including the base transition matrix created by Platt et al. (1988).........................................................39 3-2. Error analysis comparing one simulated time step of the fire matrix model to real five year growth data from the Ichwauy Preserve, FL.............................................41 3-3. Descriptive statistics of the various fire return intervals...........................................44 vii

PAGE 8

LIST OF FIGURES page 2-1. Map of the study site: Watermelon Pond Unit of the Goethe State Forest, Florida, USA..........................................................................................................................13 2-2. Fire fuel accumulation curve for all three active habitats at the WPU......................15 2-3. Proportion of area damaged by fire and burn frequency in each habitat using both fire scenarios in LANDIS.........................................................................................22 2-4. Sensitivity analysis: Mean area damaged by fire each model year as the mean fire size (MS) changed by 20% in both fire scenarios....................................................23 2-5. Sensitivity analysis: Mean area damaged by fire each model year with change in overall mean fire return interval (MI) in the IFM....................................................24 2-6. Fire scenario effects on species abundance in a single LANDIS model run.............26 2-7. P. palustris (PIPA) behavior following fire disturbance in a southeastern portion of the WPU...............................................................................................................27 2-8. Comparison of dynamics of tree species dominance between the two fire scenarios in the east portion of the WPU.................................................................28 2-9. Map of all fires, their distribution, and severity at the WPU throughout the 500 year simulation.........................................................................................................29 3-1. Example of the function used to determine the values of the matrix elements (trees >50cm dbh) assuming a maximum growth rate (=1.05)..............................38 3-2. Negative exponential function of seedling establishment after the onset of fire suppression...............................................................................................................40 3-3. Elasticity of various lambda values ()....................................................................43 3-4. Total number of individuals of P. palustris during various fire intervals.................44 3-5. Total number of P. palustris seedlings during various fire intervals........................45 4-1. Map of the study site: Watermelon Pond Unit of the Goethe State Forest, Florida, USA..........................................................................................................................53 viii

PAGE 9

4-2. The three areas within the study site with P. palustris populations used in the hybrid model analysis...............................................................................................56 4-3. Diagram of the time step discrepancy between the matrix population model (4-year) and LANDIS (10-year)...................................................................................57 4-4. Total abundance changes within the small population in relation to the seedling abundance changes...................................................................................................58 4-5. Total population changes of the matrix population model assuming fires occur either once or every iteration (2 to 3 times) within a LANDIS time step................59 4-6. Total population changes of the matrix population model assuming fires occur every iteration (2 to 3 times) within a LANDIS time step.......................................63 ix

PAGE 10

Abstract of Thesis Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Master of Science SPATIAL AND DEMOGRAPHIC MODELING TECHNIQUES APPLIED TO THE LONGLEAF PINE (Pinus Palustris) ECOSYSTEM OF NORTH CENTRAL FLORIDA By Eva Louise Loudermilk December 2005 Chair: Loukas G. Arvanitis Cochair: Wendell P. Cropper Major Department: Forest Resources and Conservation Modeling longleaf pine (Pinus palustris) response to fire requires addressing processes at both the landscape and population scale. The LANDIS (LANdscape DIsturbance and Succession) model simulates tree behavioral traits and stochastic fire disturbances on a large but coarse temporal and spatial scale. A matrix population model simulates growth and survival of a species through transition probabilities. Using specific elements from each model provides an opportunity to examine the population dynamics of P. palustris in a large managed forest. The purpose of this research was to a) spatially model forest successional patterns of a P. palustris ecosystem in a north-central Florida forest using LANDIS; b) develop a density dependent matrix population model of P. palustris; and c) utilize the linked outputs of LANDIS and the matrix population model to examine individual P. palustris populations at the Watermelon Pond Unit (WPU) of the Goethe State Forest, FL. Remote x

PAGE 11

and ground reference data aggregated with spatial and image analysis techniques were explicit elements used to develop the LANDIS model. An existing four year P. palustris transition matrix model was extended to include a seedling size class, fecundity estimates, and a density dependent function to develop the demographic model. Two fire scenarios were applied to both LANDIS and the demographic model, simulating a fire-maintained P. palustris ecosystem and a fire suppressed one. The fire regimes simulated in LANDIS were used to determine which one of the two transition matrices was employed for several individual P. palustris populations at the WPU through time. LANDIS simulated certain traits of P. palustris dynamics well, but the results were not entirely realistic: Fire frequency was still much lower than what is required in a fire-maintained ecosystem, especially within individual populations. Land managers of the WPU may use LANDIS to visualize the adverse effects that hardwoods may have on P. palustris. Further fine-tuning of parameters and smaller time steps may improve this model for visualizing P. palustris dynamics. At the scale of individual populations, the matrix model may be used for management of P. palustris. Elasticity and model behavior analyses have demonstrated that the matrix population model plausibly represented P. palustris dynamics in this ecosystem. A dual transition matrix technique (with and without fire) was used to simulate the response of this conifer to fire scenarios. The hybrid technique provided for the missing elements of one model to be supplied by the other, but comparing outputs from each model was not possible due to their incompatible output formats. Research on P. palustris population dynamics, coupled with advanced modeling techniques, provides for opportunities to continue understanding this complex conifer and its habitat. xi

PAGE 12

CHAPTER 1 INTRODUCTION Seemingly endless longleaf pine (P. palustris) savannahs once dominated the southeastern United States. Logging, development, and fire suppression have depleted this ecosystem, and now only a few remaining natural old growth longleaf pine areas are left (Platt et al., 1988; Hartnett and Krofta, 1989; Myers, 1990). One of these areas is found in the Goethe State Forest (Remnants, 2002) located in Levy and Alachua counties of Florida, USA. One portion, known as the Watermelon Pond Unit (WPU), was once predominantly longleaf pine habitat. It has since been severely degraded or has succeeded into other biological communities, such as scrub and hardwood forests (Hardin, 2004). Information on patterns and ecological functions of this system on various scales are increasingly needed for designing sound forest management plans (He et al., 2002), including prescribed burning, which is essential for this ecosystems maintenance and restoration requirements (Andrews, 1917; Platt et al. 1988; Boyer, 1990; Hardin, 2004). In this research, we intended to model this ecosystem on two different levels; a landscape and population scale, to determine the advantages that each has for long-term land management plans. Modeling successional patterns and the fire regime on a spatially explicit landscape scale and over long time periods is an effective way to gain knowledge of complex processes utilizing existing ecosystem-level inputs (He et al., 2002). The LANDIS (LANdscape DIsturbance and Succession; Mladenoff et al., 1996) model was used in this study to develop a management tool for visualizing longleaf pine dynamics at the WPU 1

PAGE 13

2 and the adverse affects that hardwoods have on its community. Two model fire regimes were developed to simulate a fire maintained longleaf pine ecosystem and a fire suppressed one. Modeling longleaf pine dynamics on a population scale, such as with a density-dependent transition matrix approach, may provide managers with more detailed information about the system, specifically size-class rates of survival, mortality, and fecundity (Caswell, 2001). The population model was parameterized based on literature estimates of this conifers response to fire occurrence and fire suppression. The model framework included two separate transition matrices representing each fire condition. These two modeling techniques provide managers with two methods for examining the long-term dynamics of this system on two very different scales. These two models may now be brought together: Information from LANDIS, such as simulated fire frequency, may be used to employ the transition matrix model to analyze the individual longleaf pine populations at the WPU during various fire regimes over time.

PAGE 14

CHAPTER 2 SPATIAL MODELING OF FOREST SUCCESSION AND FIRE DISTURBANCE IN A LONGLEAF PINE (Pinus Palustris) MANAGED ECOSYSTEM IN NORTH-CENTRAL FLORIDA Introduction Modeling Forest Succession Modeling forest succession is essential in understanding the relatively slow changes of large heterogeneous forests for long-term land management plans. Attempts have been made to physically measure forest succession quantitatively. However, this is limited in practice due to for labor-intensive sampling over long time periods. Individual-based models of tree competition and succession have been a common approach for this issue (e.g., Botkin, 1993). Forest simulation models are effective over various spatial and temporal scales and useful when comparing the affects of different land management practices, predicting land-use, climate change, and forest production over time (Peng, 2000). Managing fire-dependent ecosystems in Florida, such as longleaf pines (Pinus palustris) savannahs, has become increasingly important to reduce wildfire outbreaks and ensure future establishment of this sensitive tree species. Mechanistic and stochastic models of fire and fuel dynamics have aided in this type of management. Mechanistic models include FARSITE (Finney, 1998) and BEHAVE (Andrews, 1986) that require specific physical and chemical properties of the landscape to determine the rate and shape of fire spread. The stochastic model DISPATCH (Baker et al., 1991) simulates the effects of changing climate or shifts in fire disturbance regimes on relatively large 3

PAGE 15

4 landscapes. LANDIS (LANdscape DIsturbance and Succession, Mladenoff et al., 1996) models landscapes over large spatial and temporal scales requiring relatively coarse input data of dominant tree species and fire characteristics of the land (He and Mladenoff, 1999). This model is spatially interactive, implementing variables such as tree species, seeding characteristics, and cohort-based growth capabilities, as well as fire ignition, spread, and fuel traits. These features enable users to gain insight into the overall dynamics of the forest system visually and quantitatively over decades to several centuries (Mladenoff and He, 1999). The focus of this study was to spatially model long-term forest succession in a longleaf pine (Pinus palustris) ecosystem through two fire disturbance regimes utilizing the LANDIS model and assessing its practicality as a management tool in the Goethe State Forest, Florida. Longleaf Pine Habitat: Sandhill Description The sandhill community is found on rolling hills of sandy infertile soils of north-central Florida. Sparse longleaf pines (Pinus palustris) dominate the canopy with a low understory of wiregrass (Aristida stricta) as well as other grasses and forbs. The mid-canopy consists mainly of turkey oak (Quercus laevis) and blue-jack oak (Quercus incana) in varying densities (Myers, 1990; Greenberg and Simons, 1999). Some oaks produce oak domes, dense thickets of sand-live oaks (Quercus geminata) with little or no vegetative ground cover (Guerin, 1993). History and Current State Sandhill was once one of the largest forest types in the southeastern coastal plain of the United States, but has diminished by over 98% due to harvesting and fire suppression (Myers, 1990). The highly dispersed pine and low understory provided for an easily

PAGE 16

5 traveled and nicely shaded route for early European settlers (Hall, 1829). Overtime, this ecosystem became fragmented by clear-cutting for timber, development, introduction of livestock, and agricultural practices. Many of these fragments have suffered from fire suppression, resulting in encroachment by faster growing, shade tolerant tree species (Frost, 1993; Rasser, 2003). Restoration and preservation of P. palustris dominated ecosystems has become a well known practice in the southeast which contributes to recreation, aesthetics, reducing the onset of pest infestation, and conservation for endangered and often endemic species that thrive in this habitat (Boyer, 1993; Neel, 1993). Land managers plan for frequent low-intensity fires (~3-5 years) that keep encroaching hardwoods in check (Hartnett and Krofta, 1989; Platt and Rathbun, 1993). These fires also create an environment that supports P. palustris recruitment and open savannah like parks for recreation including hunting and easily traveled trails (Myers, 1990). Several endemic species thrive in these habitats, including the Shermans fox squirrel (Sciurus niger), red-cockaded woodpecker (Picoides borealis), brown-headed nuthatch (Sitta pusilla), Bachmans sparrow (Aimophila aestivalis), and the gopher tortoise (Gopherus polyphemus) (Engstrom, 1993). Unique Traits P. palustris is the characteristic fire-dependent conifer of sandhill communities. Frequent low-intensity fires remove accumulated leaf litter, overgrown grasses, and encroaching oaks exposing bare-mineral soil required for seed germination and establishment (Croker, 1975; Boyer, 1990; Myers, 1990). Another distinctive feature about this pine is its seedlings slow growing grass stage. Here it develops a long thick tap root (Myers, 1990), increases in diameter (Platt et al., 1988), and is protected from fire through dense tufts of long, moisture rich needles. It may stay in this stage for 5-15

PAGE 17

6 years. Within one growing season, P. palustris can grow over 1m in height avoiding lethal temperatures of fire near the ground surface while protecting the terminal bud. A maturing tree develops fire-resilient scaly plates, which define it as the most fire-resistant pine. Furthermore, the ecosystem has vegetation with pyrogenic traits that facilitate the ignition and spread of fire (Fonda, 2001; Myers, 1990). Without frequent fire, the stable P. palustris dominated forest disappears due to failure of seedling establishment and incursion of hardwood upland species (Myers, 1990). Competitive Oaks If fire is suppressed on sandhill sites for less than a decade, oaks (Quercus spp.) or other hardwood species (Prunus serotina, P. caroliniana) from adjacent communities may encroach and often out-compete the shade intolerant P. palustris (Myers, 1990). Southern hardwood species often resprout, even if they appear to be killed by fire or other disturbances. Most flourish under dense canopy cover and have chemical and physical traits that reduce the onset of fire (Platt and Schwartz, 1990). Three southeastern United States oak species (Q. geminata, Q. hemisphaerica, Q. laevis) are focused on in this study as competitors with P. palustris. Turkey oak (Q. laevis) is a deciduous species uniquely adapted to fire-maintained communities of the southeastern U.S. They are found in dry sandy areas and well drained ridges of the coastal plain (Berg and Hamrick, 1994). Sandhill areas are their favored habitat, sometimes out-competing longleaf pines eventually forming turkey oak barrens. Q. laevis life history traits, such as vegetative propagation, fast growth, and vigorous resprouting subsequent to fire, are important factors in the interaction with P. palustris. It does not contend well with other more shade tolerant oaks (e.g., Q. geminata, Q. hemisphaerica), which are also examined in this study, increasing the risk

PAGE 18

7 of upland forest encroachment during fire inhibition (McGinty and Christy, 1977; Putz, personal communication, 2005). The commonly dominant laurel oak (Q. hemisphaerica) is found in a diversity of north and central Florida habitats which exhibit a broad range of moisture and fertility gradients (Platt and Schwartz, 1990). The moisture rich environments created by hardwood habitats with Q. hemisphaerica decrease the likelihood and intensity of fire, although fires during dry periods may be detrimental to this tree and habitat (Wade et al., 1980). It is frequently considered a facultative seral species on sandhill sites, with a high tolerance of shade, clonal traits and resprouting capabilities subsequent to moderately intense burns (Putz, personal communication, 2005). Boundaries between hardwood forests and sandhills are often kept in check by low-intensity prescribed burns ignited in sandhill areas and allowed to creep into the edges of hardwood forests (Platt and Schwartz, 1990; Hardin, 2004). Sand-live oak (Q. geminata) tolerates highly infertile sandy soils typical of Floridas scrub ecosystem as well as xeric hardwood forests of north-central Florida. Its traits are similar to Q. virginiana (live oak), but are smaller in size and produces acorns in pairs. Q. geminata is not quite as tolerant of shade as Q. hemisphaerica and is slightly more fire tolerant, but thrives as a clonal propagator and strong resprouter after fire. Q. geminata form the aforementioned clonal oak domes within sandhill communities that can spread and dominate the landscape as scrub or xeric hammock if not regulated by frequent burning (Guerin, 1993). Sandhill, scrub, and hardwood forests are mutually exclusive ecosystems that respond very differently to fire. Sandhill habitats have vegetation with pyrogenic traits

PAGE 19

8 that facilitate the ignition and spread of fire. Scrub and sandhill and are closely linked ecologically but the fire-frequency is much lower in scrub and its principal components are invasive (Myers, 1990). Hardwood forests create moisture rich environments containing less flammable leaf litter deterring the onset of fire and reducing its intensity (Platt and Schwartz, 1990). Objectives The purpose of this research was to use the LANDIS model to study the spatial relationships of these four species and their habitats in response to various fuel treatments. Two scenarios were developed for this study: (I) the Frequent Fire Model (FFM), simulating a fire-maintained P. palustris ecosystem; and (II) the Infrequent Fire Model (IFM), simulating a historically fire suppressed landscape. Following are questions proposed for this research: 1) Can the unique fire regime of the P. palustris habitat be modeled using LANDIS? 2) What are the implications of using LANDIS as a long-term management tool for the P. palustris habitat? 3) Are the fire and fuel characteristics of LANDIS sufficient to simulate the short and long mean fire return intervals of various habitats in the study site? LANDIS Model Description General Characteristics LANDIS (LANdscape DIsturbance and Succession) was introduced in 1996 and has been used extensively for spatial analysis and modeling of large heterogeneous landscapes and long time frames (Mladenoff et al., 1996, Mladenoff and He, 1999). Originally, LANDIS was designed to simulate the forests of the northern lake states (Mladenoff et al., 1996, He and Mladenoff, 1999). Since then, it has been used for forest succession modeling in many parts of the world. Successional patterns of an undisturbed

PAGE 20

9 boreal forest reserve in northeast China were compared to surrounding disturbed areas using dominant tree species in LANDIS (He et al., 2002). In Finland, Pennanen and Kuuluvainen (2002) manipulated the simulated fire behavior and successional processes of LANDIS to form FIN-LANDIS. Instead of using specific tree species of a forest, Franklin et al. (2001) used plant functional groups. These groups were ascertained from Noble and Slayters (1980) vital attribute fundamentals describing the successional role of plant species. He et al. (2004) created various fuel treatments to assess the risk of wildfire occurrence in the Missouri Ozark Highlands. Overall, LANDIS is used to simulate large forested areas that have limited inputs. The presence/absence nature and coarse scale of LANDIS avoids the need for precise measurements and prediction of species abundance (Mladenoff and He, 1999). The spatial scale is user defined but is often in 10s to 100s of square meter pixels. The temporal scale is restricted to 10-year time intervals. LANDIS combines the ability to simulate landscape-sized forest successional patterns while implementing specific tree species attributes and probabilistic disturbance regimes, such as fire, wind, and harvesting. Forest succession is simulated in LANDIS based on seed dispersal, seedling establishment, sprouting traits, competition (fire and shade tolerance), growth, and mortality characteristics of each tree species modeled (Gustafson et al., 2004). For instance, a shade-intolerant species cannot establish easily on a site where shade tolerant species are well established. These species will remain dominant unless other attributes, disturbances or environmental factors alter the conditions throughout the simulation (Mladenoff and He, 1999).

PAGE 21

10 Fire Disturbance Characteristics Fire disturbance in LANDIS affects tree establishment, mortality, and inter-specific competition. The probability of establishment for each tree species is based on fire tolerance, age, and site characteristics. The stochastic and spatially explicit nature of fire is represented through mathematically defined distributions and various algorithms. Fire disturbance parameters influence the size, probability of fire occurrence on the land, return interval, spread, and severity of the fire regime. The fire tolerance levels (1-5) of each tree species are categorical representations of their relational tolerances to fire. Fire probability (P) in LANDIS follows a negative exponential distribution and is calculated in LANDIS with the following equation: P = b x If x MI (e+2) (2-1) MI is the mean fire return interval (years) for each landtype. The parameter If represents the year since the last fire on each landtype. A larger If represents a higher probability that a fire will occur on that landtype. The fire probability coefficient (b) represents the chance a cell randomly selected for ignition check will ignite and is used during the calibration process (He et al., 2003; Mladenoff and He, 1999). The fire probability function works directly with ignition and spread of fire in LANDIS. Fire is ignited based on the percentage of cells checked for ignition (chosen randomly) and on the probability that a fire will occur. It is also a function of map size (Ignition = ignition coefficient x total number of cells). LANDIS is able to ignite and therefore simulate more than one fire disturbance within an iteration, but the chance is exponentially decreased through time (He and Mladenoff, 1999). Fires spread in LANDIS as a function of fire size, fire probability, species attributes (e.g., fire tolerance and age), and spatial arrangement of adjacent cells. A fire will continue to spread until

PAGE 22

11 the pre-determined maximum fire size is reached, the adjacent cells cannot burn based on probability estimates, or it reaches the edge of forested land (He and Mladenoff, 1999; Mladenoff and He, 1999). Fuel accumulation for each landtype is established from a conceptual fire curve representing a relationship between fuel quantities, based on detritus accumulation and decomposition characteristics (Gustafson et al., 2004). It is dependent on the time since a burn last occurred (He et al., 2003) and the landtype severity and species susceptibility to fire act together to determine which species and age-cohorts are killed during a fire disturbance (Mladenoff et al.,1996). As such, the fire disturbance in LANDIS is not entirely stochastic as some areas within a landscape may be more susceptible to fire than others (He and Mladenoff, 1999). The fire disturbance file is exclusively designed for simulating fire size on the entire landscape and is a determining factor of fire spread, fire probability, fire and forest susceptibility, and spatial configuration of forest conditions (Gustafson et al., 2004). Fire size also follows the notion that large disturbances are less frequent than small disturbances through a log-normal distribution (Mladenoff and He, 1999). These fire disturbance parameters and their calculations are described in detail elsewhere (He and Mladenoff, 1999; Mladenoff and He, 1999; He et al., 2003). Methods Study Area The Watermelon Pond Unit (WPU) of the Goethe State Forest is a 1938-ha tract located in Levy and Alachua counties of Florida, USA (latititude 29 o 32 42 , longitude 82 o 36 20; Figure 2-1). Its checker-board structure makes management difficult, especially for prescribed burning. Agricultural land owners, state and county roads, and

PAGE 23

12 two small towns divide the forest, further complicating management. Before European settlement, the land was a homogeneous landscape of the sandhill community type dominated by P. palustris. During the last century, the area was fragmented by anthropogenic effects, specifically phosphate mining and non-sustainable timber harvesting. Many of these areas were unmanaged causing dramatic successional changes in this ecosystem with few P. palustris trees left today (Rasser, 2003). A main goal of land managers is the immediate implementation of prescribed fire plans for restoration and preservation of sandhill areas (Hardin, 2004). Sandhills of the WPU are classified into four general groups, ranked in order of quality: 1) areas with scattered mature P. palustris (>=30 ft 2 basal area) with subdominant Q. laevis and abundant natural P. palustris regeneration, 2) stands where Q. laevis predominate with P. palustris being a minor component (10-20 ft 2 ba) and sparse P. palustris regeneration, 3) pure Q. laevis and scrub oak stands, and 4) areas newly planted with P. palustris seedlings. Wiregrass (Aristata stricta) and other natural herbaceous groundcover are found in these areas, which are considered strong indicators of community health (Hardin, 2004). Note that Q. laevis is an abundant tree species found at the WPU, and is considered a dominant component of some historic sandhill sites (Greenberg and Simons, 1999). As such, it provides a rare opportunity for the WPU staff to manage Q. laevis dominated sandhills. To maintain a level of biodiversity, it is important to keep intact the recently developed scrub and hardwood communities while restraining further encroachment into sandhill sites with prescribed burning plans. Scrub is Floridas most endangered ecosystem with many rare and threatened endemic species. Prescribed fire initiated in

PAGE 24

13 sandhill areas will be allowed to burn into the edges of these scrub and hardwood forests as part of the land management plans (Hardin, 2004). Archer, FL Bronson, FLSR 24CR 337Legend Goethe State Forest Boundary State of Florida Alachua County Levy CountyK 0241Kilometers Figure 2-1. Map of the study site: Watermelon Pond Unit of the Goethe State Forest, Florida, USA. Parameterization of LANDIS Spatial inputs Spatial data specific to the WPU used to develop both the landtype and the forest composition map and their attributes are the: a) biological communities delineated from high-resolution digital aerial photography taken in February 2002, b) geo-referenced ground data on tree species from 2002 (Rasser, 2003), c) geo-referenced ground data on biological communities from 2003 and d) temporal aerial photography. The high-resolution digital aerial photography (0.3m ground sampling distance) was donated from

PAGE 25

14 Mpower3/Emerge (Andover, MA). These datasets provided baseline information representing the areas vegetative structure and composition. ArcGIS 8.x, 9.0 and IMAGINE 8.x were used to develop the final Erdas GIS file format datasets with a spatial resolution of 20m. All input parameter files for LANDIS, except the spatial files, are included in the Appendix. Landtypes. The WPU was stratified into four distinct habitats unique in their fire and fuel traits: a) current P. palustris sandhill habitat, b) hardwoods, c) scrub, and d) wet areas. These landtypes were created from digitally delineated biological communities on habitat type and their characteristic tree species. The ground reference data was used to perform an accuracy assessment on the landtype map. The fire accumulation and severity curves for each landtype were qualitatively assessed based on literature and current site conditions of each habitat and used for both fire scenarios (Figure 2-2). Qualities, such as highly flammable longleaf pine needles (Fonda, 2001) and wiregrass, dry sandy soils, and low decomposition rates create the characteristic sandhill environment with highly frequent (<10 years) low-intensity fires. Scrub is considered a pyrogenic community that accumulates fuels slowly and often has a high heat of ignition, but produces catastrophic fires at a rather low MI (50 years) (Myers, 1990). Furthermore, the abundance of the shrub Ceratiola ericoides, a highly flammable and characteristic scrub species, contributes to the pyrogenic nature of this habitat. Hardwood forests have rather long MIs (80 years) with a moisture rich environment and fast decomposition rates that greatly reduce the intensity of these periodic fires (Platt and Schwartz, 1990). The wet areas are defined as portions of the WPU with hydric soils, currently abundant with water, and/or prairie type habitat with fluctuating water levels. Since the wet areas have little or no

PAGE 26

15 trees, they were not actively modeled in LANDIS. The approximate areas for each landtype are as follows: current P. palustris habitat (1050 ha), scrub (375 ha), hardwoods (400 ha), and wet areas (165 ha). Even though the WPU has distinct habitats, it is small and relatively flat, which does not restrict the establishment of any of these tree species. Therefore, the establishment coefficients were set to one for each tree species on all landtypes in the FFM and instead their successional behavior was purely designed based on their attribute parameters and response to fire. The establishment coefficients in the IFM were also set to one, except P. palustris was reduced to 0.25 because of its weak competitive and establishment abilities during fire suppression. 01234501030507090100Years of Fuel AccumulationFire Severity Category Sandhill Hardwood Scrub Figure 2-2. Fire fuel accumulation curve for all three active habitats at the WPU. Forest composition. The forest composition map and associated attributes were derived from ground reference data, temporal aerial photography, and managerial records and correspondence. Forest classes required for LANDIS were distributed using the biological community spatial dataset as a foundation. The vegetative data designated the tree species found in each forest class. The ground reference data was used to perform an accuracy assessment on the forest composition map. For simplification, only the four most abundant tree species at the WPU were used in the LANDIS model, namely

PAGE 27

16 longleaf pine (P. palustris), turkey oak (Q. laevis), sand-live oak (Q. geminata), and laurel oak (Q. hemisphaerica). Historic aerial photography and forest management records were used to determine tree age-class distributions. Approximately one set of aerial photographs of the study site per decade back to 1949 were used. The final map consisted of 17 classes with various combinations of species and their age distributions within the study site (see Appendix). Species attributes Demographic parameters for the four tree species modeled were obtained from literature (Duncan and Duncan, 1988; Boyer, 1990; Young and Young, 1992; Preston and Braham, 2002) and experts opinion (Long, personal communication, 2004; Jokela, personal communication, 2004; Putz, personal communication, 2005; Table 2-1.). Unlike other studies using LANDIS (Pennanen and Kuuluvainen, 2002; Wimberly, 2004), longevity was input as the typical lifespan of a given species in this geographic region rather than its maximum known lifespan. This creates a more realistic mortality inference of each species. A seeding distance was input to represent the range that a tree species can effectively seed out, with the assumption that other natural effects, e.g., wild animals, influence where a seed may fall and germinate. The Quercus spp. modeled are readily able to resprout after a disturbance at a very young age. With a long seeding range and early propagation, some Quercus spp. establish and spread effortlessly in LANDIS. The propagation probability coefficient was developed by understanding its sensitivity (see below) within LANDIS and qualitatively shaping this variable for each Quercus spp.

PAGE 28

17 Fire Simulation Scenarios and Parameters Two fire scenarios were created to simulate a fire-maintained P. palustris ecosystem (Frequent Fire Model: FFM) and a fire suppressed one (Infrequent Fire Model: IFM), each with unique fire return intervals and fire size parameters. All fire parameters for both models (Table 2-2., Table 2-3.) were obtained from the Goethe State Forest management plans (Hardin, 2004) and communication with management personnel (Cahal, personal communication, 2005; Copeland, personal communication, 2005; Penn, personal communication, 2005; Zimmerman, personal communication, 2005). Because the P. palustris ecosystem requires frequent low intensity fires, the mean fire return interval (MI) of this habitat in the FFM was set to <10 years to ensure fires every LANDIS time step. The P. palustris ecosystem in the IFM followed historic records noting the fire suppressed period dating back to 1890 with a MI of 45 years (Zimmerman, personal communication, 2005). The scrub and hardwood habitats followed the same landtype parameterization methods mentioned in the Spatial Input: Landtypes section. Fire size variables for both models were estimated from delineated prescribed burns for the WPU, their management plans, and known wildfire outbreaks. The burn size parameters of the FFM followed that of known sizes for prescribed burns and future management plans. The burn size parameters of the IFM were directly related to the only known wildfire that occurred in 1980. Each simulation was run for 50 LANDIS time steps (500 years) to ensure reaching model stabilization. Species abundance was recorded and used to compare the effectiveness of each fire scenario focusing on P. palustris and its establishing abilities.

PAGE 29

18 Calibration and Sensitivity Analyses The fire parameters of LANDIS were calibrated by adjusting the associated fire coefficients interactively to achieve a desired mean fire return interval for all landtypes and mean fire sizes for each of the two fire scenarios within a 500 year simulation. Default values were initially used for the fire size (50) and fire probability [b = MI of each landtype (see Equation 2-1)] coefficients and manipulated in intervals of 10 depending on the scale and direction of error (He et al., 2003). For each treatment, these values were systematically varied until the parameters met the simulated values or until no further improvements could be made. Many of the tree species attributes were interactively manipulated to develop relational species behavioral traits. For instance, Q. hemisphaerica and Q. geminata were set to have rather similar dominance traits if no fire was introduced and the former with stronger seeding capabilities, while P. palustris thrived and established itself subsequent to a burn. Q. laevis was targeted as a transitional species that dominates areas with P. palustris if fire is somewhat infrequent or recently ceased, but may be replaced by invading Q. hemisphaerica and Q. geminata. P. palustris was also set to not have the ability to vegetatively propagate or establish well during fire suppression. The vegetation propagation probability coefficients of the Q. spp. were set equal to one another because of their similar re-sprouting abilities after a burn. Only a small (1% to <6%) change in the abundance of any of the four species was recorded from the simultaneously adjusted vegetation propagation coefficients over a range from 0.1 to 0.9 of the three Q. spp. To avoid bias, the median coefficient (0.5) was chosen for all three Q. spp.

PAGE 30

19 Sensitivity analyses were performed on the mean fire size and mean fire return interval to measure their contribution to the model response. Each parameter was adjusted 20% from the base run and P. palustris abundance was recorded during 500 simulation years as well as the simulated mean fire size on the landscape in each interval (Mladenoff and He, 1999). The MI analysis was only performed on the IFM to illustrate the effects of a 20% change in the MI of all landtypes simultaneously, which could not be done with the already minimal fire return interval (<10 yrs.) of the FFM. Table 2-1. Attributes of the four species studied using LANDIS. PIPA QUGE QULA QUHE Longevity (yrs.) 200 130 100 100 Age of Maturity (yrs.) 20 20 20 20 Shade Tolerance (1-5) 1 4 3 4 Fire Tolerance (1-5) 5 1 4 1 Effective Seeding Distance (m) 30 50 50 70 Maximum Seeding Distance (m) 50 70 70 90 Propagation Probability After Fire (0-1) 0 0.5 0.5 0.5 Minimum Resprout Age (yrs.) 0 5 5 5 PIPA: Pinus palustris; QUGE: Quercus geminata; QULA: Quercus laevis; QUHE: Quercus hemisphaerica Results and Discussion Calibration and Sensitivity Analyses Mean fire size was calibrated for both the FFM and IFM, with a -12% and -16% error, respectively (Table 2-2.). Error is calculated by the dividing the difference between observed fire size and the expected fire size by the expected fire size. The simulated MS was accurate to within <0.7% between both models. The IFM was more accurately calibrated than the FFM in terms of expected mean fire return intervals (Table 2-3.). The low MIs of the hardwood and scrub habitats could not be calibrated accurately in the FFM due to the high MI (<10 yrs.) of the P. palustris sandhill habitat. This result

PAGE 31

20 appeared realistic as fires may ignite within sandhill areas and spread into adjacent habitats. The total amount of area burned in hardwood and scrub sites were relatively low in the FFM while sandhill areas burned readily (Figure 2-3.). The hardwood sites were the most difficult to correctly calibrate for either scenario because the low intensity fire after long durations of fire suppression increased the probability that fire could occur again in LANDIS. As such, the increase MI of the sandhill areas in the IFM contributed to the decrease in fire frequency of both the hardwood and scrub habitats. The hardwood habitat MI became more accurately simulated while scrub habitat resulted in an exceedingly larger MI. Ultimately, the fire and fuel characteristics in LANDIS were sufficient to simulate the diverse fire regimes of theses three habitats during longer fire return intervals. Table 2-2. Calibration analysis of mean fire size (MS) in each fire regime scenario. Frequent Fire Model Infrequent Fire Model Predetermined MS, ha (% area) 120 (5.9%) 106 (5.3%) Simulated MS, ha (% area) 105 (5.2%) 101 (5.0%) Standard Deviation (% area) 59 (2.9%) 28 (1.4%) Error -12% -16% Table 2-3. Calibration analysis of mean fire return intervals (MI) in each fire regime scenario. Frequent Fire Model Infrequent Fire Model Sandhill Hardwoods Scrub Sandhill Hardwoods Scrub Predetermined MI 10 80 50 45 80 50 Simulated MI 13 15 25 38 46 77 Standard Deviation 6 6 16 19 27 71 Error 31% -81% -50% -15% -43% 53% The mean fire size (MS) parameters were generally more sensitive in the IFM than the FFM (Figure 2-4.). In the IFM, there was a 17% increase and 7% decrease in MS with the direct 20% change in MS. The strong increase in fire size was directly related to the average increase of 36% in P. palustris abundance throughout the simulation. This

PAGE 32

21 simulation reflects a management goal of fire re-introduction at the WPU. The FFM resulted in a 2% and 4% increase in MS with a 20% increase and decrease in the MS, respectively. These small changes correlated with the small average changes in P. palustris abundance, ranging from 0.5% to 3%. The lack of change in the simulated MS of the FFM is the direct result of the influence of the time since the last fire in LANDIS. Once a fire has occurred, fire probability and severity diminish on that site. Since the MS showed little change in the FFM, the P. palustris population was only slightly influenced. The MIs, MS, and P. palustris abundance changes in the IFM were more sensitive to a 20% increase in MI within all landtypes rather than a decrease (Figure 2-5). There was a 13% increase in overall simulated MI with an increase in overall MI. The increase in MI resulted in two fires dropped in years 90 and 320. MS did not show much change (-2%) with increased MI, but the total amount of area burned increased by 5%. P. palustris abundance reacted with a 16% increase in average abundance with increased MI. A large fire on the sandhill habitat in year 400 brought on new P. palustris growth and influenced the calculated increase in average abundance greatly. There was no change in simulated MIs, MS, or P. palustris abundance throughout the simulation when the MIs were reduced by 20%. These changes are complex (Mladenoff and He, 1999) and largely due to fire probability estimates calculated for each landtype as a function of MI, time since last fire, and a calibrated fire probability coefficient.

PAGE 33

22 Frequent Fire Model0.0000.0500.1000.1500.2000.2500.3000100200300400500 Infrequent Fire Model0.0000.0500.1000.1500.2000.2500.3000100200300400500Model Year Sandhill Scrub Hardwoods Proportion of Study Area Figure 2-3. Proportion of area damaged by fire and burn frequency in each habitat using both fire scenarios in LANDIS.

PAGE 34

23 Frequent Fire Model01002003004005006000100200300400500 Infrequent Fire Model01002003004005006000100200300400500Model Year +20% MS Base MS -20% MS Mean Area Damaged (ha) Figure 2-4. Sensitivity analysis: Mean area damaged by fire each model year as the mean fire size (MS) changed by 20% in both fire scenarios. Fire Scenario Effects on Species Abundance Species abundance changes throughout the LANDIS simulations illustrate the effects of the two fire regimes on each species. The results were analyzed after model stabilization (~yr. 200), unless otherwise stated. The most dramatic effects between the two fire regimes was the change in abundance of P. palustris (Figure 2-6.). In the FFM, the population was able to recover and gradually increase in number through regular recruitment. Gaps were formed from trees that died of old age or killed by fires (Figure 2-7.). As fires occurred, mature P. palustris were able to seed into these gaps and within

PAGE 35

24 a few decades become a dominant species. The moderately fire resistant Q. laevis becomes more dominant on the land as well during a frequent fire regime, while Q. hemisphaerica and Q. geminata are suppressed. At year 500, the higher mean population abundance (34%) of P. palustris in the FFM demonstrates the ability of this species to thrive in a fire-maintained community (Figure 2-8.). Over 500 years, the mean abundance of P. palustris and Q. laevis, the two more fire tolerant tree species increased by 28% and 22%, respectively, when using the FFM compared to the IFM. The other two oaks, Q. hemisphaerica and Q. geminata, only revealed a slight increase in mean abundance (4% and 2%, respectively) in the FFM. Fires do not entirely extirpate any of these three oak populations because of their ability to vegetatively propagate subsequent to fire at a relatively young age (Boyer, 1990; Long, personal communication, 2004, Table 2-1.). Their quick re-establishment capabilities make P. palustris ecosystems difficult to restore after long periods of successional changes caused by fire suppression (Hartnett and Krofta, 1989). 01002003004005006000100200300400500Model YearMean Area Damaged (ha) +20% MI Base MI -20% MI Figure 2-5. Sensitivity analysis: Mean area damaged by fire each model year with change in overall mean fire return interval (MI) in the IFM.

PAGE 36

25 The FFM demonstrated more stability in the tree populations over time than the IFM, illustrated by the overall lower variation of species abundance throughout the 500 year simulations (Table 2-4.) This is caused by the highly intermittent fires in the IFM (Figure 2-3.). Even though the FFM maintained a relatively stable population of P. palustris, the outcomes were not entirely realistic. The northern portions of the WPU with P. palustris were burned only 1 to 10 times over the 50 LANDIS time steps. A typical sandhill community requires prescribed fires every 3-5 years (Hardin, 2004), or at least a 10 year return time (Platt et al., 1991). As such, the entire community at the WPU should burn every LANDIS time step. The model may have been affected by the fragmented structure of the WPU. Fires burned more readily (20 fires) within the non-restrictive boundaries of the southeastern portion of the WPU (Figure 2-9.) suggesting fire spread constraints in the smaller confined areas of the northern WPU. Management Implications Using this LANDIS model as a management tool for the WPU is plausible for understanding the adverse effects that some of Floridas oak species have on P. palustris ecosystems within just a few decades. The strong seeding and re-sprouting abilities, shade tolerance, fast growth, and diverse site characteristics of Q. hemisphaerica make it a difficult tree to control on fire suppressed sandhill sites and may require other methods of control (Myers, 1990; Putz, personal communication, 2005). Q. geminata may also be difficult to control, is tolerant of the typical xeric conditions of the sandhill and scrub communities, and therefore is a strong competitor with P. palustris (Myers, 1990). Both of these Q. spp. are regularly kept in check by frequent fires that are allowed to creep into their borders (Hardin, 2004), but are opportunistic during fire suppression.

PAGE 37

26 Frequent Fire Model05001000150020000100200300400500 Infrequent Fire Model05001000150020000100200300400500Model Year PIPA QUGE QULA QUHE Species Abundance (ha) Figure 2-6. Fire scenario effects on species abundance in a single LANDIS model run. PIPA: Pinus palustris; QUGE: Quercus geminata; QULA: Quercus laevis; QUHE: Quercus hemisphaerica Q. laevis tolerates fires well and re-sprouts vigorously following fire (McGinty and Christy, 1977), but was the least abundant species on the landscape during the model runs. In LANDIS, this species was replaced by the other hardwoods due to its lack of shade tolerance and short longevity. P. palustris was able to out-compete Q. laevis with frequent fire in spite of the oaks re-sprouting abilities. As such, the species is complex

PAGE 38

27 and difficult to model. Otherwise, LANDIS is known to be capable of adequately simulating forest dynamics when only a few species are included (Franklin et al., 2001; Pennanen and Kuulavainen, 2002). K 05001,000250Meters 01,0002,000500MetersDominant SpeciesAge Cohort Distribution of PIPAy380 PIPA QUGE QUHE Open Area QULAy390y400 Area without PIPA Younger Age Cohorts Older Age CohortsFire between years 380 & 390Fires between years 390 & 400 Figure 2-7. P. palustris (PIPA) behavior following fire disturbance in a southeastern portion of the WPU. The top row represents the dominant species of each cell. The bottom row illustrates the recruitment of young PIPA in openings created by fire induced tree mortality. PIPA: Pinus palustris; QUGE: Quercus geminata; QULA: Quercus laevis; QUHE: Quercus hemisphaerica Even though the P. palustris population did stabilize and continued to increase in abundance in the FFM, this version (3.7) of LANDIS is applicable for only a few of the management plans for the WPU. LANDIS maintained an aesthetically pleasing mosaic type landscape pattern and illustrated the invading capabilities of particular hardwood species. It also provided the spatial and temporal aspects of fire and tree species

PAGE 39

28 behavior with minimal inputs (Mladenoff et al., 1996; Mladenoff and He, 1999). The seeding characteristics of all four species were effectively modeled: They all took advantage of gaps created by fire or age induced mortality and the oaks readily re-sprouted subsequent to fire disturbance. P. palustris typically became dominant shortly after a fire through tolerance and recruitment (Figure 2-7.). The short fire return interval of 3-5 years (Hardin, 2004) required to maintain and/or restore sandhill communities could not be modeled with LANDIS 3.7 because of the 10 year time step. Furthermore, the fragmented structure of the WPU restricted fire spread across the landscape. y400y450y500Frequent Fire Model Infrequent Fire ModelK 0120.5KilometersDominant Tree Species PIPA QUGE QUHE Open Area QULA Figure 2-8. Comparison of dynamics of tree species dominance between the two fire scenarios in the east portion of the WPU. PIPA: Pinus palustris; QUGE: Quercus geminata; QULA: Quercus laevis; QUHE: Quercus hemisphaerica

PAGE 40

29 Table 2-4. Statistics on species abundance within each fire scenario. Frequent Fire Model Infrequent Fire Model Mean Std Mean Std PIPA 924 125 664 279 QUGE 1,319 108 1,285 128 QULA 745 357 583 413 QUHE 876 81 837 103 Frequent Fire ModelInfrequent Fire Model 0120.5Kilometers 0120.5KilometersUpper ScenesLower ScenesKFire Severity No Fires 1 2 3 4 5 Figure 2-9. Map of all fires, their distribution, and severity at the WPU throughout the 500 year simulation. Darker reds are more severe fires. The most recent fire is displayed on top of older ones. Future Model Development Modeling techniques that may further advance this studys use of LANDIS includes implementing smaller time steps, further calibration, and fine-tuning of the spatial inputs. The next version of LANDIS (LANDIS II) will have one to several year time steps, which may enhance the realism of P. palustris behavior to fire and its

PAGE 41

30 competitive tree species. Since the fragmented structure of the WPU restricted fire spread in this LANDIS model, simulating each section separately may induce more frequent fires on each individual population of P. palustris. The fire and fuel characteristics of LANDIS were sufficient to simulate the various fire regimes desired in this study, especially during higher fire return intervals and severe fires. Although, additional calibration may increase accuracy of the fire disturbance parameters. Other tree species present at the WPU may be added to the model to study more complex interactions within the landscape. Wiregrass (Aristida Stricta), a grass species and an indicator of P. palustris sandhill community health (Hardin, 2004) may be included as a relational species (Gustafson et al., 2004). Additional historical data or tree cores from the WPU may provide for more accurate age distributions of species. Although long-term validation procedures are difficult to establish, they would support the quality and viability of this model. Conclusions Several of this studys objectives were met. Two fire regimes were created to assess the landscape over time in relation to P. palustris and its competitive hardwoods in an ideal fire-maintained habitat and a fire suppressed one. Even though this model did demonstrate certain characteristics of P. palustris dynamics well, the results were not entirely realistic: Fire frequency was still considerably lower in the FFM, especially within individual populations, than what is required to maintain a viable P. palustris ecosystem. Q. hemisphaerica and Q. geminata were quite accurately depicted on the landscape, while some of the simulated dynamics of Q. laevis lacked realism. LANDIS II may provide the details to more precisely simulate the fire regime desired for this study, although the fragmented structure of the WPU may still adversely affect the

PAGE 42

31 models fire spread processes. The WPU may be able to use LANDIS as a management tool to visualize the adverse effects hardwoods may have on P. palustris sandhill communities within just a few decades. With some fine-tuning of parameters and implementation of smaller time steps, this model may dramatically improve and be applicable for management to visualize P. palustris population dynamics through time.

PAGE 43

CHAPTER 3 DEVELOPMENT OF A DEMOGRAPHIC MATRIX MODEL OF LONGLEAF PINES (Pinus Palustris) Introduction Longleaf Pine P. palustris is a unique fire-dependent tree species of the southeastern United States. Frequent low-intensity fires remove accumulated leaf litter, overgrown grasses, and encroaching oaks exposing bare-mineral soil required for seed germination and establishment (Croker, 1975; Boyer, 1990; Myers, 1990). Another distinctive feature about this pine is its seedlings slow growing grass stage. Here it develops a long thick tap root (Myers, 1990), increases in diameter (Platt et al., 1988), and is protected from fire through dense tuft of long, moisture rich needles. It may stay in this stage for 5-15 years. Within one growing season, P. palustris can grow over 1m in height avoiding lethal temperatures of fire near the ground surface while protecting the terminal bud. A maturing tree develops fire-resilient scaly plates, which define it as the most fire-resistant pine. Furthermore, the ecosystem has vegetation with pyrogenic traits that facilitate the ignition and spread of fire (Fonda, 2001; Myers, 1990). Without frequent fire, the stable P. palustris dominated forest disappears due to failure of seedling establishment and incursion of hardwood upland species (Myers, 1990). Sandhill, a P. palustris dominated ecosystem, was once one of the largest forest types in the southeastern coastal plain of the United States, but has diminished by over 90% due to harvesting and fire suppression. The highly dispersed pine and low 32

PAGE 44

33 understory provided for an easily traveled and nicely shaded route for early European settlers (Hall, 1829). The ecosystem became fragmented by clear-cutting for timber, development, introduction of livestock, and agricultural practices. Many of these fragments have suffered from fire suppression resulting in encroachment by more fast growing, shade tolerant tree species (Frost, 1993; Rasser, 2003). Restoration and preservation of P. palustris dominated ecosystems has become a well known practice in the southeast which contributes to recreation, aesthetics, reducing the onset of pest infestation, and conservation for endangered and often endemic species that thrive in this habitat (Boyer, 1993; Neel, 1993). Land managers plan for frequent low-intensity fires (~3-5 years) that keep encroaching hardwoods in check (Hartnett and Krofta, 1989; Platt and Rathbun, 1993). These fires also create an environment that supports P. palustris recruitment and open savannah like parks for recreation including hunting and easily traveled trails (Myers, 1990). Several endemic species thrive in these habitats, including the Shermans fox squirrel (Sciurus niger), red-cockaded woodpecker (Picoides borealis), brown-headed nuthatch (Sitta pusilla), Bachmans sparrow (Aimophila aestivalis), and the gopher tortoise (Gopherus polyphemus) (Engstrom, 1993). Population Matrix Modeling Population matrix models are an ideal starting point for analyzing tree population dynamics (Escalante et al., 2004; Zuidema and Zagt, 2000). Population matrix models often use size-based classes and transition probabilities developed from individual tree measurements over time. They are mechanistically designed to simulate survival, growth, death, and fecundity in user defined time steps assuming that the environment in which their data was collected is constant over time (Caswell, 2001). The labor required to census and map the data needed to estimate these probabilities is intense (Silvertown et

PAGE 45

34 al., 1996). Estimation of size-specific fecundity is especially difficult because the process of seedling establishment depends on seed production, seed viability, dispersal, and site quality on multiple scales. Objectives The purpose of this research was to study P. palustris demographics in order to develop comprehensive density-dependent transition matrix models using multiple data sources. Two transition matrices were developed: (I) A P. palustris population transition matrix for a fire-maintained ecosystem and; and (II) A P. palustris population transition matrix for an ecosystem during fire suppression. These matrices were combined into one model to assess its response to various fire frequencies. Methods Data Sources Data were compiled on longleaf pine demography from several sources for the development of the fire-maintained transition matrix model. A population transition matrix built by Platt et al. (1988) for longleaf pines (P. palustris) at the Wade Tract, Thomas County, Georgia was used as a base for our model. Longleaf pine seedling and mature tree growth data recorded by the Joseph W. Jones Ecological Center at the Ichwauy Preserve, Baker County, Georgia (Jones et al., 2003; Palik et al., 2003) were used to extend the Platt model. Longleaf pine seedling survivorship data were extracted from the Jones Center data and a seedbed preparation study performed at the Escambia Experimental Forest in southwest Alabama (Croker, 1975). Simulated fire suppression effects on longleaf pine recruitment were estimated from a study in Chinsegut Hill McCarty-Woods in northern Glades County, Florida (Hartnett and Krofta, 1989).

PAGE 46

35 Transition Matrix Development We extended an existing four year transition matrix model (Platt et al., 1988) with a new 0-2cm diameter size class (Table 3-1a.). The model was implemented in the python programming language. The transition matrix created by Platt et al. was specifically designed for a frequently burned P. palustris population (1-4 years), and therefore an ideal starting point. The four year transition probabilities were in eight size classes, starting with 2-10cm and continuing in 10cm increments (e.g., 10-20cm) until reaching the 70cm+ size class. The Platt model lacked seedling (0-2cm) size class transition probabilities and fecundity estimates for a fire maintained P. palustris ecosystem. An intensive gap analysis on P. palustris seedling survivorship at the Ichwauy Preserve from 1998 through 2003 provided a complete population state vector (Jones et al., 2003; Palik et al., 2003). In this study, ground measurements were collected on adult longleaf pines [>4cm dbh (diameter at breast height)] and randomly planted plots of seedlings. Seedling survivorship (34%) was estimated in two frequently burned P. palustris sites (Croker, 1975; Jones et al., 2003; Palik et al., 2003) which was similar to Boyers (1963) seedling survival rate (38%). Although the seedling mortality rate (including significant fire scorch) seems high, the effects of fire suppression on seedling establishment are profound (Boyer, 1963; Boyer, 1990; Myers, 1990). The current population of P. palustris at the Ichwauy Preserve is essentially at an ideal maximum density or K (basal area of 11.58 m 2 /ha) according to Boyers (1999) estimates for maintaining a healthy and naturally regenerating P. palustris population and therefore applied as such to the model. We assumed that fecundity and seedling survival were functions of tree density. Size-specific relative cone production from Platt et al. (1988) was used to scale fecundity estimates. Fecundity values in the matrix represent

PAGE 47

36 the size-specific per capita number of newly established seedlings that enter the population. Because seedling establishment depends on a complex process that includes cone production, seed viability, seed dispersal, and germination safe sites, these parameters are difficult to estimate. In order to estimate fecundity parameters, we assumed a maximum population growth rate () of 1.05. The maximum growth rate would only occur in a population with no density limitation. Size-specific fecundity values were estimated with an iterative procedure (Cropper and DiResta, 1999) using the Bisection method. We assumed that fecundity scaled with size according to the ratio of 1:1:3:3:5:5:5 for the seven largest size classes. Model behavior is not sensitive to the assumption of how fecundity scales with size (Cropper and DiResta, 1999). This is also consistent with the relatively low elasticity (asymptotic growth rate sensitivity) for fecundity values in the matrix model (see below). Longleaf pine seedling growth (McGuire et al., 2001; Gagnon et al., 2004) and distribution pattern in gaps (Platt et al., 1988; Brockway and Outcalt, 1998) indicate that seedlings are competitively inhibited by larger trees. Although the functional relationship between adult density and seedling establishment and survival cannot be quantified, long-term simulations require density-dependence for realism. A population characterized by our matrix model growing at = 1.05 could increase from 200 trees per ha to 1,662 trees per ha in 80 years. For comparison, the density of a virgin longleaf pine stand in Louisiana in 1935 was reported to be between 191 and 233 trees per ha (Penfound and Watkins, 1937). In order to simulate seedling density-dependence, we assumed that the fecundity and seedling survival matrix elements were functions of the number of trees greater than

PAGE 48

37 10cm diameter. We also assumed a local population carrying capacity (K) of 500 trees greater than 10cm diameter (100m radius plot). Calibrating the density function starts with two sets of matrix parameters; one, the matrix for the maximum growth rate (= 1.05, or other assumed values); the other being a matrix that represents stable growth at K ( = 1.00). Fecundity and seedling survival values are calculated as: f(N) = Max e (-kd N) (3-1) where f(N) is the value of the matrix element at population size N, Max is the value of the matrix element in the maximum growth matrix, and kd is a constant (e.g., a fecundity value of 2.63 for trees greater than 50 cm at = 1.05; Figure 3-1.). The value of kd is determined by solving the equation for a value of f(N) equal to the = 1 matrix element and a value of N set at K, which is at a stable equilibrium in this model. This density-dependent model has been used to model fisheries (Caswell, 2001) and a Poa population in the UK (Case, 2000) based on measured fecundity and seedling survival at different densities. A second transition matrix model was created to represent the effects of fire suppression on P. palustris establishment (Table 3-1b.). It has been determined that after just 15 years without fire disturbance, P. palustris recruitment ceases (Hartnett and Krofta, 1989) due to lack of bare mineral soil, low shade tolerance, and poor competitive abilities of seedlings (Myers, 1990). Even just a few years of fire suppression can dramatically reduce the amount of recruitment and vigorous herbaceous encroachment may adversely affect growth and survival of young longleaf pines (Hartnett and Krofta, 1989; Platt and Rathbun, 1993). As such, we assumed that seedling number would decline exponentially as fires are suppressed (Figure 3-2.). We were able to achieve this

PAGE 49

38 by simply using the same transition probabilities as the first model and setting the size-specific fecundity parameters to zero. 00.511.522.531000Value of Matrix Element 0200400600800Population Density Figure 3-1. Example of the function used to determine the values of the matrix elements (trees >50cm dbh) assuming a maximum growth rate (=1.05). The dotted line represents the population at K, which is at a stable equilibrium.

PAGE 50

39 Table 3-1. Transition matrices developed from multiple data sources including the base transition matrix created by Platt et a l. (1988). The measurements refer to the diameter of the tr ees (at breast height for those applicable). a.) Fire Transition Matrix Year 0 (cm) Year 4 (cm) 0 to 2 2 to 10 10 to 20 20 to 30 30 to 40 40 to 50 50 to 60 60 to 70 70 + 0 to 2 0.114 0 0.123 0.123 0.369 0.369 0. 615 0.615 0.615 2 to10 0.228 0.756 10 to20 0.082 0.834 20 to30 0.138 0.865 30 to40 0.127 0.834 40 to50 0.152 0.899 50 to60 0.077 0.875 60 to70 0.084 0.891 70+ 0.054 0.882 Deaths 0.658 0.162 0.028 0. 008 0.014 0.024 0.041 0.055 0.118 b.) No Fire Transition Matrix Year 0 (cm) Year 4 (cm) 0 to 2 2 to 10 10 to 20 20 to 30 30 to 40 40 to 50 50 to 60 60 to 70 70 + 0 to2 0.114 0 0 0 0 0 0 0 0 2 to10 0.228 0.756 10 to20 0.082 0.834 20 to30 0.138 0.865 30 to40 0.127 0.834 40 to50 0.152 0.899 50 to60 0.077 0.875 60 to70 0.084 0.891 70+ 0.054 0.882 Deaths 0.658 0.162 0.028 0. 008 0.014 0.024 0.041 0.055 0.118

PAGE 51

40 0500100015002000250004812Model YearSeedling Abundance 16 Figure 3-2. Negative exponential function of seedling establishment after the onset of fire suppression. Seedling numbers are below zero by year 16. Results and Discussion Error Analysis A simulated population with one fire in each four year time step was compared to real data from the Ichwauy Preserve collected on a five year time interval (1998-2003) for an error analysis (Table 3-2.). Only the size-classes from 2cm on were used in this analysis because the same individuals were measured in 1998 and 2003. This does not account for additional individuals entering the population, which is otherwise the case in the fire matrix model using fecundity and seedling transition probabilities. As a result, there was a high percentage of error (118%) in the 2-10cm size class. The 60-70cm size classes resulted in the most percentage error (303%) due to the small population and associated changes in that size class. Otherwise, the % error ranged from -11 to 6%. Determining P. palustris growth from diameter measurements within five years is difficult because of the slow and variable growth characteristics of P. palustris. As the habitat is burned, the pines scaly plates protect the tree, but may stunt its growth and allow the bark to slough off causing growth patterns to seem absent or even reversed (e.g., lower dbh) during short periods (Myers,1990; Putz, personal communication,

PAGE 52

41 2005). As such, it is also relatively safe to compare P. palustris growth measurements with a one year discrepancy. Table 3-2. Error analysis comparing one simulated time step of the fire matrix model to real five year growth data from the Ichwauy Preserve, FL. Size-Class (cm) % Error Difference 2 10 118 474 10 20 -11 -105 20 30 6 55 30 40 -7 -88 40 50 4 32 50 60 -5 -7 60 70 303 6 70 + 0 0 Elasticity Analysis An elasticity analysis of the transition matrix was performed to evaluate the relative sensitivity of to changes in the matrix elements (Caswell, 2001). Elasticity (e ij ) is calculated using the following equation: VWwvaejiijij, (3-2) Where a ij are elements of the stage population matrix, W is the right eigenvector of the dominant eigenvalue (), V is the left eigenvector, and is the scalar product of the two eigenvectors. The population matrix elements were classified as P, the probability of surviving and remaining in the same size class; G, the probability of growing into the next largest size class; and F, the per capita production of established seedlings. Overall, the elasticity values (Figure 3-3.) demonstrated that P. palustris population growth was far more sensitive to P (range: 0.003 0.17), while G and F were much less sensitive (range: 0.001 0.04). This suggests that accurate estimates for P are necessary for representing the population growth rate (Doak et al., 1994). G values for smaller size class trees were higher, signifying the relative importance of P. palustris growth of younger trees. As s

PAGE 53

42 increase, G and F begin to play a stronger role in the model as P weakens. In conclusion, this elasticity analysis follows similar analyses of other natural populations by summarizing the comparative effects of small changes in demographic rates (Doak et al., 1994; Escalante et al., 2004). Fire Frequency Analysis Simulated fire frequency was varied to assess model behavior (Figure 3-4.; Figure 3-5.). With constant fires (1-4 year fire return interval) the population remained steady and at K. Lower standard deviation and standard error of the total population with frequent fires suggested a more stable population through time (Ott and Longnecker, 2001; Table 3-3.). As the fire return intervals became larger, spikes in species abundance became more dramatic with each onset of fire but the population remained well below K and continued to deplete with proliferation of fire suppression (Figure 3-4.). These spikes in population increase are the direct result of seedling recruitment when fires are brought back into the system (Figure 3-5.). It is not clear how well this simulates reality. A healthy P. palustris population that has experienced fire suppression lasting 40 years may not be able to re-establish itself from a newly introduced fire regime due to the large fuel load that has most likely accumulated and other vegetative encroachment (Hartnett and Krofta, 1989; Gilliam and Platt, 1999). However, the model does assume that fire occurs 1-4 times in each time step. As such, some recruitment is likely especially if cone production is in a mast year and conditions are favorable (Platt et al., 1988; Boyer, 1999). The shorter fire return intervals then depict the more realistic behavior of P. palustris in a fire maintained community.

PAGE 54

43 Lambda = 1.0500.020.040.060.080.10.120.140.160.18123456789 F P G Lambda = 1.0300.020.040.060.080.10.120.140.160.18123456789Elasticity of Lambda F P G Lambda = 1.0000.020.040.060.080.10.120.140.160.18123456789Size Class F P G Figure 3-3. Elasticity of various lambda values (). F: fecundity; P: probability of remaining in the same size class; G: probability of growing to the next size class. Size classes are in order from smallest to largest.

PAGE 55

44 Table 3-3. Descriptive statistics of the various fire return intervals. Every four year fire return interval represents 1-4 years of low-intensity burns in one matrix time step. Fire Return Intervals (yrs.) 4 8 12 20 40 No Fire Mean 5,940 4,994 4,620 4,226 3,957 3,537 Standard Error 35 125 153 181 192 199 Standard Deviation 177 639 779 920 980 1,015 Sample Variance 31,447 407,997 606,372 847,299 960,623 1,029,834 Minimum 5,776 4,027 3,444 2,909 2,715 2,247 Maximum 6,756 6,756 6,756 6,756 6,756 6,756 01000200030004000500060007000800002040608010Model YearAbundance 0 4 yr. 12 yr. 40 yr. No Fire Figure 3-4. Total number of individuals of P. palustris during various fire intervals. The peaks in population growth are from seedlings establishing during years of fire occurrence. When the population was indefinitely subjected to fire suppression, the seedling size class effectively disappears within four time steps (16 years) in a negatively exponential fashion (Figure 3-5.), while the total population continued to decline until it reached below one at year 536. With pines reaching maximum and mean ages of 400-500 (Boyer, 1999) and 200-250 (Putz, personal communication, 2005), respectively, and assuming no anthropogenic (e.g., logging) or catastrophic disturbances (e.g., hurricanes), this estimate of the youngest tree reaching 536 years of age is unlikely. Although, the

PAGE 56

45 population is under 100 by year 316, suggesting a struggling population in route to its demise. 050010001500200025003000020406080100Model YearSeedling Abundance 4 yr. 12 yr. 40 yr. No Fire Figure 3-5. Total number of P. palustris seedlings during various fire intervals. The peaks coincide with population growth rates in Figure 3-4. An advantage to the frequent fire regime is the consistent uneven age structure of the population. Infrequent fires that stimulate new P. palustris growth tend to create gaps of missing age-classes in the population. This may influence cone production and seed crops with the lack of necessary mature trees at particular periods in time. Gap sizes or densities of adult trees may therefore fluctuate dramatically through time as a result. Unnecessarily large gaps allow some P. palustris recruitment, but also allow for faster growing competing vegetation to take over the landscape. If gaps are not large enough, seedling establishment and success may be inadequate for a sustainable population (Grace and Platt, 1995; Palik et al., 2003). Future Model Development Several additions to the transition matrix and more in-depth validation procedures may be used to further develop this model. Since most P. palustris forests are managed

PAGE 57

46 to some degree and are often in a restoration stage, there are many variations in population structure and composition, understory and site conditions, and fuel traits, which may influence population development. The model should include validation with site specific management strategies, such as burning initiatives, seedbed preparation, and thinning. These managed forests spanning from east Texas to the east coast of Virginia (Frost, 1993) provide for an opportunity to use and compare this model with their P. palustris population dynamics. More specifically, K may vary between sites and between size and even age classes within a population. In our model, fecundity estimates were assumed based on a hypothetical growth rate. Applying empirical fecundity values in a stage based distribution may improve these estimates. Seedling parameters are difficult to estimate due to their complex growth and survival patterns related to environmental influences, such as fire and nutrient competition from adult P. palustris or encroaching vegetation (Grace and Platt, 1995; Gagnon et al., 2004). Their growth rate in diameter was especially difficult. For instance, Grace and Platt (1995) showed that many P. palustris seedlings do not even reach 2cm in diameter in four years. As such, it may be required to separate this size class into smaller size classes for particular P. palustris populations. This modeling technique may be further improved by acquiring more long-term data on this slow growing and long-lived species. Unfortunately, funding and research opportunities are often rather short-term and therefore may influence the models utility. For instance, the data sources used in this study did not have measurements recorded beyond a six year period. Other studies (e.g., Veno, 1976) used past research and data

PAGE 58

47 collected to continue understanding forest dynamics of a site on a larger temporal scale (>20 years). Implications of Using Multiple Data Sources It would have been ideal to collect all of the data for model development and testing within the Goethe State Forest. This was not practical because of the incredible time, money, and effort required to collect and analyze the raw data needed to create such a comprehensive demographic model on P. palustris. On the other hand, using multiple data sources may produce large errors associated with environmental and anthropogenic factors. P. palustris populations may vary by site condition, such as vegetative structure and composition, landscape history, climatic conditions, soil and available nutrients, and high genetic variation (Boyer, 1990). Management and research objectives impact these populations as well, through variation in the fire regime, physical manipulation of the landscape (e.g., planting, thinning), and having different data collection methods and goals. Conclusions This model has the potential to become practical and useful for further understanding of the population dynamics of this long-lived conifer. The essential elements of a complete transition matrix have been estimated from reliable resources. A K was implemented into the model as a spatial restriction or density-dependent function. The unique survivorship characteristics of P. palustris seedlings and fecundity estimates were applied to the transition matrix to develop a more comprehensive modeling scheme. Elasticity and model behavior analyses provided for model responsiveness to its inherent elements and validation procedures resulted in relatively low error. Using the dual transition matrix technique applied the behavioral response of this conifer to fire. The

PAGE 59

48 abundant research done on P. palustris population dynamics (e.g., Boyer, 1990) coupled with population modeling techniques provide opportunities to continue understanding this complex conifer and its habitat.

PAGE 60

CHAPTER 4 HYBRID APPROACH TO SPATIAL AND DEMOGRAPHIC MODELING OF LONGLEAF PINES (Pinus Palustris) Introduction Longleaf Pine P. palustris is a unique fire-dependent tree species of the southeastern United States. Frequent low-intensity fires remove accumulated leaf litter, overgrown grasses, and encroaching oaks exposing bare-mineral soil required for seed germination and establishment (Croker, 1975; Boyer, 1990; Myers, 1990). Another distinctive feature about this pine is its seedlings slow growing grass stage. Here it develops a long thick tap root (Myers, 1990), increases in diameter (Platt et al., 1988), and is protected from fire through dense tuft of long, moisture rich needles. It may stay in this stage for 5-15 years. Within one growing season, P. palustris can grow over 1m in height avoiding lethal temperatures of fire near the ground surface while protecting the terminal bud. A maturing tree develops fire-resilient scaly plates, which define it as the most fire-resistant pine. Furthermore, the ecosystem has vegetation with pyrogenic traits that facilitate the ignition and spread of fire (Fonda, 2001; Myers, 1990). Without frequent fire, the stable P. palustris dominated forest disappears due to failure of seedling establishment and incursion of hardwood upland species (Myers, 1990). Sandhill, a P. palustris dominated ecosystem, was once one of the largest forest types in the southeastern coastal plain of the US, but has diminished by over 98% due to harvesting and fire suppression (Myers, 1990). The highly dispersed pine and low 49

PAGE 61

50 understory provided for an easily traveled and nicely shaded route for early European settlers (Hall, 1829). The ecosystem became fragmented by clear-cutting for timber, development, introduction of livestock, and agricultural practices. Many of these fragments have suffered from fire suppression resulting in encroachment by more fast growing, shade tolerant tree species (Frost, 1993; Rasser, 2003). Restoration and preservation of P. palustris dominated ecosystems has become a well known practice in the southeast which contributes to recreation, aesthetics, reducing the onset of pest infestation, and conservation for endangered and often endemic species that thrive in this habitat (Boyer, 1993; Neel, 1993). Land managers plan for frequent low-intensity fires (~3-5 years) that keep encroaching hardwoods in check (Hartnett and Krofta, 1989; Platt and Rathbun, 1993). These fires also create an environment that supports P. palustris recruitment and open savannah like parks for recreation including hunting and easily traveled trails (Myers, 1990). Several endemic species thrive in these habitats, including the Shermans fox squirrel (Sciurus niger), red-cockaded woodpecker (Picoides borealis), brown-headed nuthatch (Sitta pusilla), Bachmans sparrow (Aimophila aestivalis), and the gopher tortoise (Gopherus polyphemus) (Engstrom, 1993). Hybrid Modeling Creating hybrid ecosystem process models provide an opportunity to employ important aspects of two or more models to create a new more robust model. Hybrid models can be constructed with components that include simulation models, statistical models, and empirical spatial models [Geographic Information System (GIS) models]. Robinson and Ek (2003) utilized at least six various empirical and physiological based modules to develop an individual-tree growth and dynamics model. Logofet and Korotkov (2002) incorporated spatially explicit forest inventory data to apply a

PAGE 62

51 geographical component within a Markov chain model measuring natural succession of forest types. Lamar and McGraw (2005) compared two matrix population models of eastern hemlock (Tsuga canadensis L.), one created from field data and one created from remote sensing techniques using low elevation aerial photography. Each of these studies used the inherent strengths from each model to establish a more viable model of forest dynamics. In this study, density-dependent matrix population models use particular outputs from the spatially explicit LANDIS (LANdscape DIsturbance and Succession; Mladenoff et al., 1996) model to assess P. palustris population dynamics. Population matrix models are an ideal starting point for analyzing tree population dynamics (Escalante et al., 2004; Zuidema and Zagt, 2000). They have become powerful tools used in the natural sciences and have gone on to be used in hybrid model research (Logofet and Korotkov, 2002; Lamar and McGraw, 2005). Population matrix models often use size-based classes and transition probabilities developed from individual tree measurements over time. They are mechanistically designed to simulate survival, growth, death, and fecundity in user defined time steps assuming that the environment in which their data were collected is constant over time (Caswell, 2001). The labor required to census and map the data needed to estimate these probabilities is intense (Silvertown et al., 1996). Estimation of size-specific fecundity is especially difficult because the process of seedling establishment depends on seed production, seed viability, dispersal, and site quality on multiple scales. LANDIS on the other hand, models landscapes over large spatial and temporal scales requiring relatively coarse input data of dominant tree species and fire characteristics of the land (He and Mladenoff, 1999). The model is spatially interactive,

PAGE 63

52 implementing variables such as tree species seeding characteristics and cohort-based growth capabilities, as well as fire ignition, spread, and fuel traits. These features enable the user to gain insight into the overall dynamics of the forest system visually and quantitatively over decades to several centuries (Mladenoff and He, 1999). Objectives The purpose of this research was to utilize the linked outputs of a spatially explicit forest succession model (LANDIS), and a density dependent transition matrix model to examine P. palustris population dynamics over time. Temporal fire data from LANDIS was used to select between two transition matrices: (I) A P. palustris population transition matrix for a fire-maintained ecosystem and; (II) A P. palustris population transition matrix during fire suppression. The ultimate goal was to tie P. palustris population numbers to specific management units or independent populations. Methods Study Area The Watermelon Pond Unit (WPU) of the Goethe State Forest is a 1938-ha tract located in Levy and Alachua counties of Florida, USA (latititude 29 o 32 42 , longitude 82 o 36 20; Figure 4-1.). It has a checker-board structure that makes land management difficult, especially for prescribed burning. Agricultural land owners, state and county roads, and two small towns divide the forest, further complicating management. Before European settlement, the land was a homogeneous landscape of the sandhill community type dominated by P. palustris. During the last century, the area was fragmented by anthropogenic effects, specifically phosphate mining and non-sustainable timber harvesting. Many of these areas were unmanaged causing dramatic successional changes in this ecosystem with few P. palustris trees left today (Rasser, 2003). Now, one of the

PAGE 64

53 main management goals is the immediate implementation of prescribed burning initiatives for restoration and preservation of the remaining sandhill areas (Hardin, 2004). Archer, FL Bronson, FLSR 24CR 337Legend Goethe State Forest Boundary State of Florida Alachua County Levy CountyK 0241Kilometers Figure 4-1. Map of the study site: Watermelon Pond Unit of the Goethe State Forest, Florida, USA. Outside Data Sources Data were collected on longleaf pine demographics from several outside sources for the development of the fire-maintained transition matrix model used in this study. A population transition matrix built by Platt et al. (1988) for longleaf pines (Pinus palustris) at the Wade Tract, Thomas County, Georgia was used as a base for our model. Longleaf pine seedling and mature tree growth data recorded by the Joseph W. Jones Ecological Center at the Ichwauy Preserve, Baker County, Georgia were used for assessing density parameters (Jones et al., 2003; Palik et al., 2003). Longleaf pine seedling survivorship

PAGE 65

54 data were extracted from the Jones Center data and a seedbed preparation study performed at the Escambia Experimental Forest in southwest Alabama (Croker, 1975). Simulated fire suppression effects on longleaf pine recruitment were estimated from a study in Chinsegut Hill McCarty-Woods in northern Glades County, Florida (Hartnett and Krofta, 1989). Hybrid Modeling Techniques: The Spatial and Demographic Components Modeling P. palustris response to fire requires addressing processes at both the landscape scale and at the population scale. It is certainly possible to construct a single model that simulates the landscape fire regime as well as seedling establishment and tree growth and mortality. A hybrid approach, however, has the advantage of using models that have already been developed (LANDIS, P. palustris matrix population model), which are operating at different time steps, to meet the model objective with significantly less model development effort. The principal problem of the hybrid model approach is passing information between the individual models. LANDIS outputs, and other data sources, were used to estimate the: a) initial P. palustris size-classes and fire distributions; b) P. palustris carrying capacity; and c) fire disturbance frequencies, all within individual P. palustris populations of the WPU. These populations were chosen because of their size and relative fire disturbance frequency and distribution simulated within LANDIS (Figure 4-2.). A small isolated population in the northern territory (270ha) of the WPU was used as well as the much larger southeastern section (WPU East; 792ha). For comparison, the entire population (1874ha) at the WPU was also simulated assuming each fire disturbance in LANDIS affects the entire sites P. palustris population. From a graphical analysis of age and size relationships in Platt et al. (1998), the initial age-class distributions created for LANDIS

PAGE 66

55 (see Chapter 1) were used to determine the size-class distributions found in each of the three P. palustris populations. The number of individuals within each population was calculated from: a) the known area of P. palustris sandhill habitat within each section; b) the size-class distribution within each section; and c) the P. palustris population density of the Ichwauy Preserve (basal area of 11.58 m 2 /ha). This population was essentially at an ideal maximum density or carrying capacity (K) according to Boyers (1999) estimates for maintaining a healthy and naturally regenerating P. palustris population. We assumed that fecundity and seedling survival were functions of tree density. Size-specific relative cone production from Platt et al. (1988) was used to scale fecundity estimates (see Chapter 2). Fire frequency distributions were purely based on the occurrence of fire within each of the three populations for each model created in LANDIS. The Frequent Fire Model (FFM) simulated fires every one to two LANDIS time steps (10-20 years) and the Infrequent Fire Model (IFM) simulated a more fire suppressed ecosystem with a fire return interval of 1-7 LANDIS time steps (10-70 years). The size, intensity, and frequency of burns varied across the landscape which were influenced by land characteristics and tree species interand intraspecific behavior. Fire was simulated with the matrix population model through application of two size class transition matrices representing population dynamics with fire or with no fire during each ten year time step in LANDIS. The development and analysis of the two matrices are described in detail elsewhere (see Chapter 2). Essentially, one four year transition matrix represents a P. palustris population subjected to a regular low-intensity burning regime (1-4 years). The other four year transition matrix represents a P. palustris population subjected to fire suppression with fecundity set to zero. When the fire

PAGE 67

56 transition matrix is chosen, fecundity and seedling survival rates are density-dependent, with equilibrium at the carrying capacity (K). Density-dependence is not applied in the fire suppression transition matrix because of the rapid decline in recruitment in the absence of frequent burning (Hartnett and Kofta, 1989). "Small" Isolated Population "Large" Population:WPU East 0120.5Kilometers 0120.5Kilometers Figure 4-2. The three areas within the study site with P. palustris populations used in the hybrid model analysis: 1) The entire WPU (1874ha); 2) The southeastern portion of the WPU, namely WPU East (792ha); 3) A small isolated population in the northern section of the WPU (270ha). Refer to Figure 4-1. for the layout of the WPU. Two application methods were proposed to handle the time step discrepancy between LANDIS and the matrix model. The hybrid matrix model runs an alternate of 2 to 3 times per LANDIS time step, so that five 4-year time steps equal two LANDIS time steps (Figure 4-3.). When a fire occurs in LANDIS, it only disturbs an area once and therefore, the fire matrix model should only run once per LANDIS time step. The

PAGE 68

57 positioning of whether the fire matrix runs first, second, or third, is randomly selected. For comparison, a second matrix model was designed to employ the fire transition matrix 2 or 3 times when a fire occurs in LANDIS. Figure 4-3. Diagram of the time step discrepancy between the matrix population model (4-year) and LANDIS (10-year). Results and Discussion Fire Effects on Individual Populations The overall simulated longleaf pine population abundance changes were a direct result of seedling recruitment fluctuations caused by fire disturbance and fire suppression (Figure 4-4.) As fires occur on the landscape, the accumulated fuel (e.g., leaf litter, herbaceous growth) burns to expose bare mineral soil for the establishment and germination of P. palustris seedlings, assuming good seed production and site conditions (Gilliam and Platt, 1999). Without fires, seedling recruitment decreases exponentially until they cease to establish 15 years (~ 4 times steps) from the onset of fire suppression (Hartnett and Krofta, 1989). Matrix Model LANDIS 0 10 20 Time Steps Time (yrs) Overlap

PAGE 69

58 010000200003000040000500000100200300400500Model YearAbundance Total Seedlings Figure 4-4. Total abundance changes within the small population in relation to the seedling abundance changes. The spikes relate directly between total population numbers and the recruitment of seedlings. The two application methods for using the fire transition matrix (e.g., 1 or 2/3 times per LANDIS time step) were compared (Figure 4-5.). Having the fire transition matrix run 2 to 3 times more results in a clear percent increase in mean (41-59%) and abundance at year 500 (74-84%) in all three populations. The entire WPU and WPU East had higher variation (31% and 28%, respectively) in population numbers using the transition matrix 2/3 times per LANDIS time step, while the isolated portion of the WPU did not show much difference (0.03%). In the isolated population, there was a time lag that allowed for fire suppression to take effect for both application methods. In the two larger populations, the fires were frequent enough to allow the 2/3 fire method to keep a steady increase or maintained P. palustris population (e.g., smooth curve).

PAGE 70

59 Entire WPU0200000400000600000800000100000004080120160200 1 Fire 2/3 Fires WPU East02000040000600008000010000004080120160200Abundance Isolated Population0100002000030000400005000004080120160200Model Year Figure 4-5. Total population changes of the matrix population model assuming fires occur either once or every iteration (2 to 3 times) within a LANDIS time step. Scale may be different between graphs.

PAGE 71

60 P. palustris population dynamics were compared over 500 years assuming fires occurred in all transition matrices (all 2/3 times steps) during fire disturbance in LANDIS (Figure 4-6.). The fire frequencies of the FFM resulted in relatively stable numbers in the two larger populations, while the isolated population continued to decline in both the FFM and the IFM. The 62-80% increase in fire frequency in all populations resulted in 90-96% larger numbers at year 500 using the FFM. The high fluctuations in the WPU East numbers was due to the large number of fires from the FFM that were spaced far enough apart for multiple reactions to fire suppression and seedling recruitment following fires. Comparing P. palustris simulation outputs of LANDIS and the matrix model was not possible due to their incompatible output formats. LANDIS outputs the presence of a species in particular locations (cells) across a landscape and through time, lacking population numbers. The matrix model outputs a species population in size-classes through time, lacking a spatial component. However, utilizing the output fire frequencies of LANDIS in the matrix model to provide P. palustris numbers in individual populations at the WPU offered an opportunity for the missing elements of one model to be supplied by the other. Implications of Hybrid Approach This modeling approach demonstrates the ability to combine strengths of ecosystem models and compensate for their weaknesses (Kimmins et al., 1999). Here we utilized the P. palustris transition probabilities derived from field work done by Platt et al. (1988) and others (Croker, 1975; Hartnett and Krofta, 1989; Jones et al., 2003, Palik et al., 2003) and added a density dependent function (see Chapter 2). A spatial component

PAGE 72

61 was built-in to specifically look at individual populations within the WPU as well as the fire frequencies and distributions created in LANDIS. Although this was an opportunity to use the elements of two models at two different scales, each model simulates species dynamics quite differently, especially related to fecundity and population changes. LANDIS simulates fecundity from tree age-cohorts reaching maturity and dispersing seeds into adjacent cells (He et al., 2003), while the fire matrix model utilizes fecundity transition probabilities (Caswell, 2001). LANDIS simulates tree growth dynamics through a presence/absence approach with interacting tree age-cohorts and influential disturbances and site conditions (He et al., 2003), while matrix models use size-based transition probabilities of one population of species assuming no competition, disturbance, or site variability (Caswell, 2001). Future Model Development The matrix population model approach is not spatially explicit, which may be a major limitation for longleaf pine simulation. Seedling establishment is clearly sensitive to overstory density (Boyer, 1993; Grace and Platt, 1995; McGuire et al,. 2001). The hybrid modeling approach, using LANDIS to provide some of the missing processes, is not a complete solution. Although long-term data is often difficult to obtain, utilizing GIS data may provide the long-term information needed for calibration of such demographic models (Logogfet and Korotkov, 2002). Acquiring and comparing demographic data collected from field-work and remotely sensed imagery (Lamar and McGraw, 2005) may be another calibration and validation option. Obtaining site-specific demographic data is an ideal approach to enhancing this modeling scheme. Since the data sources used to create these models are from a variety of projects and sites, demographic data from the WPU would be critical for model improvement. Other details

PAGE 73

62 that are essential to the integration of these two types of models include understanding size and age relationships of trees such as P. palustris (Platt et al., 1988) and measuring fecundity and seedling growth in sites with different fire histories. The time step discrepancy may be solved by using the next version of LANDIS (LANDIS II), which incorporates one to several year time steps. This may enhance the realism of P. palustris behavior to fire and its competitive tree species in LANDIS while providing analogous time steps between the matrix models and LANDIS. As a healthy P. palustris population requires frequent (3-5 years) low-intensity fires (Hardin, 2004), the fire return interval may decrease dramatically using LANDIS II, especially in the many isolated populations of P. palustris within the WPU. Conclusions The hybrid techniques used in this study were successful. The temporal fire frequency and spatial data from LANDIS provided for inputs to simulate the P. palustris matrix models using various fire regimes. Individual populations, or management units within the WPU, were simulated and analyzed. The outputs of these two models were not compared because of their incompatible output formats [e.g., presence of species (LANDIS) vs. population numbers (matrix model)]. Using LANDIS II, spatial technologies, such as GIS and remote sensing, and more site-specific long-term ground data on growth and fecundity of P. palustris may improve justification procedures. The abundant research done on P. palustris population dynamics (e.g., Boyer, 1990) together with demographic and hybrid modeling techniques provide opportunities to continue understanding this complex conifer and its habitat.

PAGE 74

63 Entire WPU020000040000060000080000010000000100200300400500 FFM IFM WPU East0200004000060000800001000000100200300400500Abundance Isolated Population 010000200003000040000500000100200300400500Model Year Figure 4-6. Total population changes of the matrix population model assuming fires occur every iteration (2 to 3 times) within a LANDIS time step. Scale is different between graphs. FFM: Frequent Fire Model; IFM: Infrequent Fire Model.

PAGE 75

CHAPTER 5 CONCLUSIONS General Conclusions Results, conclusions, implications, and overall applicability for each portion of this research are discussed here. Conclusions from the spatial modeling portion of the research include: Two fire regimes were created in relation to P. palustris and its competitive hardwoods in an ideal fire-maintained habitat and a fire suppressed one. LANDIS demonstrated certain characteristics of P. palustris dynamics well: recruitment of young after fire, seeding distance, and fire tolerance. Results were not entirely realistic: Fire frequency was lower in the FFM than what is required to maintain a viable P. palustris ecosystem. Q. hemisphaerica and Q. geminata were accurately depicted, while dynamics of Q. laevis lacked realism. The demographic modeling led to the following conclusions: The essential elements of a complete P. palustris transition matrix were estimated from data and literature. Additions to the P. palustris matrix population model developed by Platt et al. (1988) included: o A density-dependent function o Survivorship and growth rates of P. palustris seedlings (<2cm dbh) o Fecundity estimates 64

PAGE 76

65 Model sensitivity and dynamics were analyzed through elasticity and scenario analyses. Validation procedures resulted in relatively low error. A dual transition matrix technique (with and without fire) applied the behavioral response of this conifer to fire. The hybrid modeling approach, consisting of the combined application of LANDIS and the demographic model, led to the following conclusions: The temporal fire frequency and spatial data from LANDIS provided inputs to simulate the P. palustris matrix models using various fire regimes. General responses to fire suppression were realistic. Individual populations within the WPU were simulated and analyzed. The incompatible output formats of the two models [e.g., presence of species (LANDIS) vs. population numbers (matrix model)] did not allow for model comparison. Limitations, Implications, and Future Work Spatial Modeling There are several limitations with the spatial model developed using LANDIS. Q. laevis tolerates fires well and re-sprouts vigorously following fire (McGinty and Christy, 1977), but was the least abundant species on the landscape during the model runs. The short fire return interval of 3-5 years (Hardin, 2004) required to maintain and/or restore sandhill communities could not be modeled with LANDIS 3.7 because of the 10 year time step. Furthermore, the fragmented structure of the WPU restricted fire spread across the landscape.

PAGE 77

66 With some fine-tuning of parameters and implementation of smaller time steps, this model may dramatically improve and be applicable for management to visualize P. palustris population dynamics through time. Species behavior for Q. laevis could be improved and additional calibration may fine-tune the model. LANDIS II may provide a superior framework to simulate the fire regime desired for this study, although the fragmented structure of the WPU may still adversely affect the models fire spread processes. Simulating each section separately may induce more frequent fires on each individual population of P. palustris. Future work may also include adding more trees species found at the WPU to LANDIS, obtaining more historical data, collecting tree cores for ground referenced species age distributions, and applying long-term validation procedures to support the quality and viability of the model. Demographic Modeling There are several limitations to the development of the P. palustris demographic model. The model should include site specific validation where management strategies vary. More specifically, density-dependent dynamics may vary between sites and between size and even age classes within a population. In our model, fecundity estimates were assumed based on a hypothetical growth rate. Applying empirical fecundity values in a stage based distribution may improve these estimates. Obtaining data on seedling growth rates in diameter was difficult and seemed to vary highly within and between sites (Grace and Platt, 1995). It may be required to separate this size class into smaller size classes for particular P. palustris populations. This modeling technique may be further improved by acquiring more long-term data on this slow growing and long-lived species and collecting all of the data for model development and testing within the Goethe State Forest would have been ideal. This

PAGE 78

67 deemed impractical because of the incredible time, money, and effort required to collect and analyze the raw data needed to create such a comprehensive demographic model on P. palustris. Hybrid Modeling Approach There are several limitations to the hybrid modeling approach. Although this method demonstrates the ability to combine strengths of ecosystem models, each model simulates species dynamics quite differently, especially related to fecundity and population changes. Furthermore, using LANDIS to provide some of the missing processes is not a complete solution. The time step discrepancy may be solved by using LANDIS II which incorporates one to several year time steps. Using LANDIS II, spatial technologies, such as GIS and remote sensing, and more site-specific long-term ground data on growth and fecundity of P. palustris may improve justification procedures. Broad Implications of Research Forest managers should understand the long-term implications of their management plans on forests as a whole. Using a broad-scaled model, such as LANDIS, provides managers with a glimpse of how their forests are changing over time in respect to tree behaviors and interactions, site conditions, and disturbances without exhaustive input parameters. They may easily change parameters, such as fire disturbances or species attributes in LANDIS using its graphical user interface. To study the finer details of a forest, a comprehensive demographic model aids to assess one species over time within specific populations of a forest. Although these inputs may be difficult to collect, using outside data sources on a commonly researched species may provide for the information needed.

PAGE 79

68 Much of this research and data collected may be used for studies in other natural areas. Although the spatial inputs and site conditions used in LANDIS were specifically designed for the WPU, the attributes for each tree species may be used elsewhere. As the four tree species and their habitats are very common in north-central Florida forests, the methodology may be utilized to study landscape level processes in other managed landscapes. As mentioned earlier, because the P. palustris demographic model used multiple data sources, the model may be tested in various areas in the southeast to assess its viability. The site conditions and management regimes within these areas may refine the demographic model for use in a specific area.

PAGE 80

APPENDIX LANDIS PARAMETER FILES Main Parameter File Species Attribute File 69

PAGE 81

70 Landtype Attribute File: FFM

PAGE 82

71 Landtype Attribute File: IFM

PAGE 83

72 Map Attribute File

PAGE 84

73 Map Index File Age Index File Disturbance File: FFM Disturbance File: IFM

PAGE 85

LIST OF REFERENCES Andrews, E.F., 1917. Agency of fire in propagation of longleaf pines. Botanical Gazette. 64, 497-508. Andrews, P.L., 1986. BEHAVE: Fire behavior prediction and fuel modeling system BURN subsystem, part 1. In: Mladenoff, D.J., Baker, W.L. (Eds.), Spatial Modeling of Forest Landscape Change: Approaches and Applications. Cambridge University Press, Cambridge, UK, pp. 172-174. Baker, W.L., Egbert, S.L., Frazier, G.F., 1991. A spatial model for studying the effects of climatic change on the structure of landscapes subject to large disturbances. Ecological Modelling. 56, 109-125. Berg, E.E., Hamrick, J.L., 1994. Spatial and genetic structure of two sandhills oaks: Quercus laevis and Quercus margaretta (Fagaceae). American Journal of Botany. 81, 7-14. Botkin, D.B., 1993. Forest Dynamics: An Ecological Model. Oxford University Press, New York, NY, USA, 309 pp. Boyer, W.D., 1963. Development of Longleaf Pine Seedlings under Parent Trees. USDA Forest Service Research Paper SO-4. Southern Forest Experiment Station, New Orleans, LA, USA, 7 pp. Boyer, W.D., 1990. Pinus palustris Mill. longleaf pine. In: Burns, R, M., Honkala, B.H. (technical coordinators), Silvics of North America, Volume 1, Conifers. USDA Forest Service Agricultural Handbook 654. Washington D.C., USA, pp. 405-412. Boyer, W.D., 1993. Regenerating longleaf pine with natural seeding. In: Herman, S.M. (Ed.), The Longleaf Pine Ecosystem: Ecology, Restoration, and Management. Proceedings, 18 th of the Tall Timbers Fire Ecology Conference. Tall Timbers, Inc., Tallahassee, FL, USA, pp. 299-310. Boyer, W.D., 1999. Longleaf pine: natural regeneration and management. Alabama Treasured Forests. 18, 7-9. Brockway, D.G., Outcalt, K.W., 1998. Gap-phase regeneration in longleaf pine wiregrass ecosystems. Forest Ecology and Management. 160, 125-139. Case, T.D., 2000. An Illustrated Guide to Theoretical Ecology. Oxford press, Oxford, NY, USA, 449 pp. 74

PAGE 86

75 Caswell, H., 2001. Matrix Population Models: Construction, Analysis, and Interpretation. 2 nd Ed. Sinauer Associates, Inc., Sunderland, MA, USA, 722 pp. Croker, T.C., 1975. Seedbed Preparation Aids Natural Regeneration of Longleaf Pine. USDA Forest Service Research Paper SO-112. Southern Forest Experiment Station, Ashville, NC, USA, 6 pp. Cropper, Jr. W.P., DiResta, D., 1999. Simulation of a commercial sponge population in Biscayne Bay, Florida: recovery following Hurricane Andrew and management implications. Ecological Modelling. 118, 1-15. Doak, D., Kareiva, P., Klepetka, B., 1994. Modeling population viability for the desert tortoise in the Western Mojave Desert. Ecological Applications. 4, 446-460. Duncan, W.H., Duncan, M.B., 1988. Trees of Southeastern United States. University of Georgia Press, Athens, GA, USA, 322 pp. Engstrom, R.T., 1993. Characteristic mammals and birds of longleaf pine forests. In: Herman, S.M. (Ed.), The Longleaf Pine Ecosystem: Ecology, Restoration, and Management. Proceedings, 18 th of the Tall Timbers Fire Ecology Conference. Tall Timbers, Inc., Tallahassee, FL, USA, pp. 127-138. Escalante, S., Montana, C., Orellana, R., 2004. Demography and potential extractive use of the liana palm, Desmoncus orthacanthos Martius (Arecaceae), in southern Quintana Roo, Mexico. Forest Ecology and Management. 187, 3-18. Finney, M.A., 1998. FARSITE: Fire Area Simulator Model Development and Evaluation. USDA Forest Service Research Paper RMRS-RP-4. Rocky Mountain Research Station, Ogden, UT, USA, 47 pp. Fonda, R. W., 2001. Burning characteristics of needles from eight pine species. Forest Science. 47, 390-396. Franklin, J., Syphard, A.D., Mladenoff, D.J., He, H.S., Simons, D.K., Martin, R.P., Deutschman, D., OLeary, J.F., 2001. Simulating the effects of different fire regimes on plant functional groups in Southern California. Ecological Modelling. 142, 261-283. Frost, C.C., 1993. Four centuries of changing landscape patterns in the longleaf pine ecosystem. In: Herman, S.M. (Ed.), The Longleaf Pine Ecosystem: Ecology, Restoration, and Management. Proceedings, 18 th of the Tall Timbers Fire Ecology Conference. Tall Timbers, Inc., Tallahassee, FL, USA, pp. 17-44. Gagnon, J. L., Jokela, E.J., Moser, W.K., Huber, D. A., 2004. Characteristics of gaps and natural regeneration in mature longleaf pine flatwoods ecosystems. Forest Ecology and Management. 187, 373-380.

PAGE 87

76 Gilliam, F.S., Platt, W.J., 1999. Effects of long-term fire exclusion on tree species composition and stand structure in an old-growth Pinus palustris (longleaf pine) forest. Plant Ecology. 140, 15-26. Grace, S.L., Platt, W.J., 1995. Effects of adult tree density and fire on the demography of pregrass stage juvenile longleaf pine (Pinus palustris Mill.). Journal of Ecology. 83, 75-86. Greenberg, C.H., Simons, R.W., 1999. Age, composition, and stand structure of old-growth oak sites in the Florida High Pine Landscape: implications for ecosystem management and restoration. Natural Areas Journal. 19, 30. Guerin, D.N., 1993. Oak dome clonal structure and fire ecology in a Florida longleaf pine dominated community. Bulletin of the Torrey Botanical Club. 120, 107-144. Gustafson, E.J., Zollner, P.A., Sturtevant, B.R., He, H.S., Mladenoff, D.J., 2004. Influence of forest management alternatives and land type on susceptibility to fire in northern Wisconsin, USA. Landscape Ecology. 19, 327-341. Hall, B., 1829. Travels in North America, in the Years 1827 and 1828. In: Myers, R.L., Ewel, J.J. (Eds.), 1990. Ecosystems of Florida. University of Central Florida Press, Orlando, FL, USA, pp. 153. Hardin, D., 2004. Watermelon Pond Restoration Plan: the First Five Years. Division of Forestry, Forest Management Division, Tallahassee, FL, USA, 27 pp. Hartnett, D.C., Krofta, D.M., 1989. Fifty-five years of post-fire succession in a southern mixed hardwood forest. Bulletin of the Torrey Botanical Club. 116, 107-113. He, H.S., Hao, Z., Larsen, D.R., Dai, L., Hu, Y., Chang, Y., 2002. A simulation study of landscape scale forest succession in northeastern China. Ecological Modelling. 156, 153-166. He, H.S., Mladenoff, D.J., 1999. Spatially explicit and stochastic simulation of forest-landscape fire disturbance and succession. Ecology. 80, 81-99. He, H.S., Mladenoff, D.J., Nimerfro, K.K., Gustafson, E.J., 2003. LANDIS 3.7 Users Guide. School of Natural Resources, University of Missouri-Columbia, Columbia, MO, USA, 63 pp. He, H.S., Shang, B.Z., Crow, T.R., Gustafson, E.J., Shifley, S.R., 2004. Simulating forest fuel and fire risk dynamics across landscapes LANDIS fuel module design. Ecological Modelling. 180, 135-151. Jones, R.H., Mitchell, R.J., Stevens, G.N., Pecot, S.D., 2003. Controls of fine root dynamics across a gradient of gap sizes in a pine woodland. Oecologia. 134, 132-143.

PAGE 88

77 Kimmins, J.P., Mailly, D., Seel, B., 1999. Modelling forest ecosystem net primary production: the hybrid simulation approach used in FORECAST. Ecological Modelling. 122, 195-224. Lamar, W.R., McGraw, J.B., 2005. Evaluating the use of remotely sensed data in matrix population modeling for eastern hemlock (Tsuga canadensis L.). Forest Ecology and Management. 212, 50-64. Logofet D.O., Korotkov V.N., 2002. Hybrid optimization: a heuristic solution to the Markov-chain calibration problem. Ecological Modelling. 151, 51-61. McGinty, D.T., Christy, E.J., 1977. Turkey oak ecology on a Georgia sandhill. American Midland Naturalist. 98, 487-491. McGuire, J.P., Mithcell, R.J., Moser, E.B., Pecot, S.D., Gjerstad, D.H., Hedman, C.W., 2001. Gaps in a gappy forest: plant resources, longleaf pine regeneration, and understory response to tree removal in longleaf pine savannas. Canadian Journal of Forest Research. 31, 765-778. Mladenoff, D.J., He, H.S., 1999. Design and behavior of LANDIS, an object-oriented model of forest landscape disturbance and succession. In: Mladenoff, D.J., Baker, W.L. (Eds.), Spatial Modeling of Forest Landscape Change: Approaches and Applications. Cambridge University Press, Cambridge, UK, pp. 125-162. Mladenoff, D.J., Host, G.E., Boeder, J., Crow, T.R., 1996. LANDIS: A spatial model of forest landscape disturbance, succession, and management. In: Goodchild, M.F., Steyaert, L.T., Parks, B.O. (Eds.), GIS and Environmental Modeling: Progress and Research Issues. GIS World Books, Fort Collins, CO, USA, pp. 175-180. Myers, R.L., 1990. Scrub and high pine. In: Myers, R.L., Ewel, J.J. (Eds.), Ecosystems of Florida. University of Central Florida Press, Orlando, FL, USA, pp. 150-193. Neel, W.L., 1993. An ecological approach to longleaf pine forestry. In: Herman, S.M. (Ed.), The Longleaf Pine Ecosystem: Ecology, Restoration, and Management. Proceedings, 18 th of the Tall Timbers Fire Ecology Conference. Tall Timbers, Inc., Tallahassee, FL, USA, pp. 337-338. Noble, I.R., Slayter, R.O., 1980. The use of vital attributes to predict succession changes in plant communities subject to recurrent disturbances. Vegetatio. 43, 5-21. Ott, R.L., Longnecker, M., 2001. An Introduction to Statistical Methods and Data Analysis. 5 th Ed. Duxbury, Pacific Grove, CA, USA, 1152 pp. Palik, B., Mitchell, R.J., Pecot, S., Battaglia, M., Pu, M., 2003. Spatial distribution of overstory retention influences resources and growth of longleaf pine seedlings. Ecological Applications. 13, 674-686.

PAGE 89

78 Penfound, W.T., Watkins, A.G., 1937. Phytosociological studies in the pinelands of southeastern Louisiana. American Midland Naturalist. 18, 661-682. Peng C., 2000. Quantitative methods: understanding the role of forest simulation models in sustainable forest management. Environmental Impact Assessment Review. 20, 481-501. Pennanen, J., Kuuluvainen, T., 2002. A spatial simulation approach to natural forest landscape dynamics in boreal Fennoscandia. Forest Ecology and Management. 164, 157-175. Platt, W.J., Glitzenstein, J.S., Streng, D. R., 1991. Evaluating pyrogenicity and its effects on vegetation in longleaf pine savannas. In: High Intensity Fire in Wildlands: Management, Challenges, and Options. Proceedings, 17 th of the Tall Timbers Fire Ecology Conference. Tall Timbers, Inc., Tallahassee, FL, USA, pp. 143-161. Platt, W.J., Gregory, W.E., Rathbun, S.L., 1988. The population dynamics of a long-lived conifer (Pinus palustris). The American Naturalist. 131, 491-525. Platt, W.J., Rathbun, S.L., 1993. Dynamics of an old-growth longleaf pine population. In: Herman, S.M. (Ed.), The Longleaf Pine Ecosystem: Ecology, Restoration, and Management. Proceedings, 18 th of the Tall Timbers Fire Ecology Conference. Tall Timbers, Inc., Tallahassee, FL, USA, pp. 275-298. Platt, W.J., Schwartz, M.W., 1990. Temperate hardwood forests. In: Myers, R.L., Ewel, J.J. (Eds.), 1990. Ecosystems of Florida. University of Central Florida Press, Orlando, FL, USA, pp. 150-193. Preston, R.J., Braham, R.R., 2002. North American Trees. Iowa State Press, Ames, IA, USA, 520 pp. Rasser, M.K., 2003. Monitoring and Assessing Longleaf Pine Ecosystem Restoration: A Case Study in North-Central Florida [Theses]. University of Florida. University Microfilms, Gainesville, FL. Available from: University library electronic archives or [http://purl.fcla.edu/fcla/etd/UFE0001256]. 60 pp. Remnants of Floridas Longleaf Forest, 2002. The Longleaf Alliance. < http://www.longleafalliance.org/ecosystem/map/flmap.htm > Data accessed: July, 2005. Robinson, A.P., Ek, A.R., 2003. Description and validation of a hybrid model of forest growth and stand dynamics for the Great Lakes region. Ecological Modelling. 170, 73-104.

PAGE 90

79 Silvertown, J., Franco, M., Menges, E., 1996. Interpretation of elasticity matrices as an aid to the management of plant populations for conservation. In: Lamar, W.R., McGraw, J.B., Evaluating the use of remotely sensed data in matrix population modeling for eastern hemlock (Tsuga canadensis L.). Forest Ecology and Management. pp. 51. Veno, P.A., 1976. Successional relationships of five Florida plant communities. Ecology. 57, 498-508. Wade, D., Ewel, J., Hofstetter, R., 1980. Fire in South Florida Ecosystems. USDA Forest Service General Technical Report SE-17. Southeastern Forest Experiment Station, Asheville, NC, USA, 125 pp. Wimberly, M.C., 2004. Fire and forest landscapes in the Georgia Piedmont: an assessment of spatial modeling assumptions. Ecological Modelling. 180, 41-56. Young, C.G., Young, J.A., 1992. Seeds of Woody Plants in North America. Dioscorides Press, Portland, OR, USA, 407 pp. Zuidema, P. A., Zagt, R. J., 2000. Using population matrices for long-lived species: a review of published models for 35 woody plants. In: Zuidema, P.A. Demography of Exploited Tree Species in the Bolivian Amazon [Dissertation]. Utrecht University, Utrecht, NL, pp. 159-182.

PAGE 91

BIOGRAPHICAL SKETCH Eva Louise Loudermilk was born a first generation American in 1976 with a strong Swedish heritage. With a B.S. degree from the University of Florida (UF) in animal science in 2001, she worked as an assistant curator of ornithology at the Florida Museum of Natural History as well as a field assistant monitoring several animal and plant species throughout the state. She became affiliated with the Forest Information Systems Laboratory at the School of Forest Resources and Conservation at UF and began her graduate career in 2003 with an emphasis in Geographical Information Systems. She plans to continue her education within the interdisciplinary ecology PhD program at the School of Natural Resources and Environment at UF. 80


Permanent Link: http://ufdc.ufl.edu/UFE0012041/00001

Material Information

Title: Spatial and Demographic Modeling Techniques Applied to the Longleaf Pine (Pinus palustris) Ecosystem of North Central Florida
Physical Description: Mixed Material
Copyright Date: 2008

Record Information

Source Institution: University of Florida
Holding Location: University of Florida
Rights Management: All rights reserved by the source institution and holding location.
System ID: UFE0012041:00001

Permanent Link: http://ufdc.ufl.edu/UFE0012041/00001

Material Information

Title: Spatial and Demographic Modeling Techniques Applied to the Longleaf Pine (Pinus palustris) Ecosystem of North Central Florida
Physical Description: Mixed Material
Copyright Date: 2008

Record Information

Source Institution: University of Florida
Holding Location: University of Florida
Rights Management: All rights reserved by the source institution and holding location.
System ID: UFE0012041:00001


This item has the following downloads:


Full Text












SPATIAL AND DEMOGRAPHIC MODELING TECHNIQUES APPLIED TO THE
LONGLEAF PINE (Pinus Palustris) ECOSYSTEM OF NORTH CENTRAL FLORIDA















By

EVA LOUISE LOUDERMILK


A THESIS PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
MASTER OF SCIENCE

UNIVERSITY OF FLORIDA


2005














Copyright 2005

by

Eva Louise Loudermilk

































This document is dedicated to my family and friends.















ACKNOWLEDGMENTS

I foremost thank my two advisors, Wendell Cropper and Loukas Arvanitis, for

there support, dedication, and belief in my graduate career. My other committee

members, Michael Binford and Jane Southworth, provided invaluable help in all aspects

of my research. Jack Putz, Eric Jokela, and Alan Long have given their expertise in

various details of my project.

I thank the School of Forest Resources and Conservation for financial support as

well as George Blakeslee and Tim White for their dependable guidance. The

opportunities and amenities at the Forest Information Systems laboratory have been

immeasurable. Mike Rasser, Balaji Ramachandran, and Valentina Boycheva gave me

guidance early in my graduate research, while I have been able to confide in my

trustworthy peer and lab partner, Doug Shoemaker, throughout the years.

Special thanks go to Libby Zimmerman, Mike Penn, Amy Copeland, and Bobby

Cahal of the Goethe State Forest staff and Robert Mitchell and Stephen Pecot from the

Joseph W. Jones Ecological Center who have provided valuable data for my research.

This experience would not have been complete without the encouragement and support

from my family and friends, especially my husband, Erick, and my devoted parents, Eva

and Staffan Lundberg.

















TABLE OF CONTENTS


page

A C K N O W L E D G M E N T S ......... .................................................................................... iv

LIST OF TABLES ............. ............. ........ ........... .................... vii

LIST OF FIGURES ............. ................... ............ .......... ............... .. viii

A B STR A C T ................................................. ..................................... .. x

CHAPTER

1 IN T R O D U C T IO N ................................................................................. .............. 1

2 SPATIAL MODELING OF FOREST SUCCESSION AND FIRE DISTURBANCE
IN A LONGLEAF PINE (Pinus Palustris) MANAGED ECOSYSTEM IN
N ORTH-CEN TRAL FLORIDA ............................................ .......................... 3

In tro d u ctio n .................................................................................. 3
M odeling Forest Succession ............................................................. ............... 3
Longleaf Pine H habitat: Sandhill ........................................ ........................ 4
Objectives .............................................................................. .... ....... ................... 8
LA N D IS M odel D escription............................................................................. 8
General Characteristics.................... .......... ............................. 8
Fire Disturbance Characteristics.................... ...... .......................... 10
M e th o d s ........................................................................................................1 1
S tu d y A rea ......................................................... ................ 1 1
Parameterization of LAND IS ............... ...... ....................... ................. 13
Fire Simulation Scenarios and Parameters................... ........... .............. 17
Calibration and Sensitivity A nalyses........................................ ............... 18
R esu lts an d D iscu ssion ....... .. .............. .......... ........ ........................................ .. 19
Calibration and Sensitivity Analyses........................................ ............... 19
Fire Scenario Effects on Species Abundance.......................... ................23
M anagem ent Implications ...........................................................................25
Future M odel D evelopm ent........................................... .......................... 29
C o n c lu sio n s..................................................... ................ 3 0









3 DEVELOPMENT OF A DEMOGRAPHIC MATRIX MODEL OF LONGLEAF
P IN E S (P inus P alustris) ..................... ......... ......................................................... 32

In tro du ctio n ...............32..............................................
Longleaf Pine .......................................................................................................32
Population M atrix M odeling ........................................ .......... ............... 33
O b j e ctiv e s ..............................................................................................3 4
M e th o d s ..............................................................................3 4
D ata Sources ............. .............................. ..................... 34
Transition M atrix D evelopm ent ........................................ ....... ............... 35
Results and Discussion .................................... ...... .. ...... ............. 40
E rror A naly sis ...................................................40
Elasticity Analysis .............. .... ....................... ......41
Fire Frequency A nalysis........................................................... ............... 42
Future Model Development........................ ....... .......................... 45
Implications of Using Multiple Data Sources ...............................................47
C o n c lu sio n s........................................................................................................... 4 7

4 HYBRID APPROACH TO SPATIAL AND DEMOGRAPHIC MODELING OF
LONGLEAF PINES (Pinus Palustris) ............. ............................................... 49

In tro d u ctio n .......................................................................................4 9
L o n g le af P in e ................................................................................ 4 9
H ybrid M modeling .............................. ......................... ... ...... .... ..... ...... 50
O b j e ctiv es .....................................................................................5 2
M e th o d s ..............................................................................5 2
Study A rea ..................................................................................................52
Outside Data Sources .............................. ........... ...............53
Hybrid Modeling Techniques: The Spatial and Demographic Components.......54
Results and Discussion ................................................. 57
Fire Effects on Individual Populations ...... ............................................57
Implications of Hybrid Approach ................................................................60
Future M odel D evelopm ent............................................ 61
C o n c lu sio n s........................................................................................................... 6 2

5 C O N C L U S IO N S ....................................................................................................... 6 4

G en eral C on clu sion s .................................................................................. 64
Limitations, Implications, and Future W ork................................. .......... .....65
Broad Im plications of Research ......................................................................... .....67

APPENDIX: LANDIS PARAMETER FILES ...................................... ...............69

LIST O F R EFER EN CE S ...............................74.............................

B IO G R A PH IC A L SK E T C H ....................................................................................... 80
















LIST OF TABLES


Table page

2-1. Attributes of the four species studied using LANDIS .............................................19

2-2. Calibration analysis of mean fire size (MS) in each fire regime scenario..................20

2-3. Calibration analysis of mean fire return intervals (MI) in each fire regime
sc e n a rio ....................................................................... 2 0

2-4. Statistics on species abundance within each fire scenario.................. ...............29

3-1. Transition matrices developed from multiple data sources including the base
transition matrix created by Platt et al. (1988).............. ........ .... ............... 39

3-2. Error analysis comparing one simulated time step of the fire matrix model to real
five year growth data from the Ichwauy Preserve, FL.........................................41

3-3. Descriptive statistics of the various fire return intervals .......................................44
















LIST OF FIGURES


page

2-1. Map of the study site: Watermelon Pond Unit of the Goethe State Forest, Florida,
U S A ...................................... ................................................. 1 3

2-2. Fire fuel accumulation curve for all three active habitats at the WPU....................15

2-3. Proportion of area damaged by fire and bum frequency in each habitat using both
fire scenarios in LA N D IS ......... .................................................... ............... 22

2-4. Sensitivity analysis: Mean area damaged by fire each model year as the mean fire
size (MS) changed by 20% in both fire scenarios .................................................23

2-5. Sensitivity analysis: Mean area damaged by fire each model year with change in
overall mean fire return interval (MI) in the IFM. ................................................24

2-6. Fire scenario effects on species abundance in a single LANDIS model run.............26

2-7. P. palustris (PIPA) behavior following fire disturbance in a southeastern portion
of the W PU ..................................... .......................... .... .... ......... 27

2-8. Comparison of dynamics of tree species dominance between the two fire
scenarios in the east portion of the WPU ......................................... ...........28

2-9. Map of all fires, their distribution, and severity at the WPU throughout the 500
y ear sim u latio n ..................................................... ................ 2 9

3-1. Example of the function used to determine the values of the matrix elements
(trees >50cm dbh) assuming a maximum growth rate (k=1.05)..............................38

3-2. Negative exponential function of seedling establishment after the onset of fire
suppression ..................................... ................................ ........... 40

3-3. Elasticity of various lambda values ()) ........... ............................... ...............43

3-4. Total number of individuals of P. palustris during various fire intervals ...............44

3-5. Total number of P. palustris seedlings during various fire intervals ......................45

4-1. Map of the study site: Watermelon Pond Unit of the Goethe State Forest, Florida,
U S A ........................................ ......................... .............. 5 3









4-2. The three areas within the study site with P. palustris populations used in the
hybrid m odel analysis .................. ............ .............. ...... .............. .. 56

4-3. Diagram of the time step discrepancy between the matrix population model (4-
year) and LANDIS (10-year). ............................................................................57

4-4. Total abundance changes within the small population in relation to the seedling
abundance changes ......................................................... .. ............ 58

4-5. Total population changes of the matrix population model assuming fires occur
either once or every iteration (2 to 3 times) within a LANDIS time step ...............59

4-6. Total population changes of the matrix population model assuming fires occur
every iteration (2 to 3 times) within a LANDIS time step.................................63















Abstract of Thesis Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Master of Science

SPATIAL AND DEMOGRAPHIC MODELING TECHNIQUES APPLIED TO THE
LONGLEAF PINE (Pinus Palustris) ECOSYSTEM OF NORTH CENTRAL FLORIDA

By

Eva Louise Loudermilk

December 2005

Chair: Loukas G. Arvanitis
Cochair: Wendell P. Cropper
Major Department: Forest Resources and Conservation

Modeling longleaf pine (Pinuspalustris) response to fire requires addressing

processes at both the landscape and population scale. The LANDIS (LANdscape

DIsturbance and Succession) model simulates tree behavioral traits and stochastic fire

disturbances on a large but coarse temporal and spatial scale. A matrix population model

simulates growth and survival of a species through transition probabilities. Using

specific elements from each model provides an opportunity to examine the population

dynamics of P. palustris in a large managed forest.

The purpose of this research was to a) spatially model forest successional patterns

of a P. palustris ecosystem in a north-central Florida forest using LANDIS; b) develop a

density dependent matrix population model of P. palustris; and c) utilize the linked

outputs of LANDIS and the matrix population model to examine individual P. palustris

populations at the Watermelon Pond Unit (WPU) of the Goethe State Forest, FL. Remote









and ground reference data aggregated with spatial and image analysis techniques were

explicit elements used to develop the LANDIS model. An existing four year P. palustris

transition matrix model was extended to include a seedling size class, fecundity

estimates, and a density dependent function to develop the demographic model. Two fire

scenarios were applied to both LANDIS and the demographic model, simulating a fire-

maintained P. palustris ecosystem and a fire suppressed one. The fire regimes simulated

in LANDIS were used to determine which one of the two transition matrices was

employed for several individual P. palustris populations at the WPU through time.

LANDIS simulated certain traits of P. palustris dynamics well, but the results were

not entirely realistic: Fire frequency was still much lower than what is required in a fire-

maintained ecosystem, especially within individual populations. Land managers of the

WPU may use LANDIS to visualize the adverse effects that hardwoods may have on P.

palustris. Further fine-tuning of parameters and smaller time steps may improve this

model for visualizing P. palustris dynamics. At the scale of individual populations, the

matrix model may be used for management ofP. palustris. Elasticity and model

behavior analyses have demonstrated that the matrix population model plausibly

represented P. palustris dynamics in this ecosystem. A dual transition matrix technique

(with and without fire) was used to simulate the response of this conifer to fire scenarios.

The hybrid technique provided for the missing elements of one model to be supplied by

the other, but comparing outputs from each model was not possible due to their

incompatible output formats. Research on P. palustris population dynamics, coupled

with advanced modeling techniques, provides for opportunities to continue understanding

this complex conifer and its habitat.














CHAPTER 1
INTRODUCTION

Seemingly endless longleaf pine (P. palustris) savannahs once dominated the

southeastern United States. Logging, development, and fire suppression have depleted

this ecosystem, and now only a few remaining natural old growth longleaf pine areas are

left (Platt et al., 1988; Hartnett and Krofta, 1989; Myers, 1990). One of these areas is

found in the Goethe State Forest (Remnants, 2002) located in Levy and Alachua counties

of Florida, USA. One portion, known as the Watermelon Pond Unit (WPU), was once

predominantly longleaf pine habitat. It has since been severely degraded or has

succeeded into other biological communities, such as scrub and hardwood forests

(Hardin, 2004). Information on patterns and ecological functions of this system on

various scales are increasingly needed for designing sound forest management plans (He

et al., 2002), including prescribed burning, which is essential for this ecosystem's

maintenance and restoration requirements (Andrews, 1917; Platt et al. 1988; Boyer, 1990;

Hardin, 2004). In this research, we intended to model this ecosystem on two different

levels; a landscape and population scale, to determine the advantages that each has for

long-term land management plans.

Modeling successional patterns and the fire regime on a spatially explicit landscape

scale and over long time periods is an effective way to gain knowledge of complex

processes utilizing existing ecosystem-level inputs (He et al., 2002). The LANDIS

(LANdscape DIsturbance and Succession; Mladenoff et al., 1996) model was used in this

study to develop a management tool for visualizing longleaf pine dynamics at the WPU









and the adverse affects that hardwoods have on its community. Two model fire regimes

were developed to simulate a fire maintained longleaf pine ecosystem and a fire

suppressed one.

Modeling longleaf pine dynamics on a population scale, such as with a density-

dependent transition matrix approach, may provide managers with more detailed

information about the system, specifically size-class rates of survival, mortality, and

fecundity (Caswell, 2001). The population model was parameterized based on literature

estimates of this conifer's response to fire occurrence and fire suppression. The model

framework included two separate transition matrices representing each fire condition.

These two modeling techniques provide managers with two methods for examining

the long-term dynamics of this system on two very different scales. These two models

may now be brought together: Information from LANDIS, such as simulated fire

frequency, may be used to employ the transition matrix model to analyze the individual

longleaf pine populations at the WPU during various fire regimes over time.














CHAPTER 2
SPATIAL MODELING OF FOREST SUCCESSION AND FIRE DISTURBANCE IN A
LONGLEAF PINE (Pinus Palustris) MANAGED ECOSYSTEM IN NORTH-
CENTRAL FLORIDA

Introduction

Modeling Forest Succession

Modeling forest succession is essential in understanding the relatively slow

changes of large heterogeneous forests for long-term land management plans. Attempts

have been made to physically measure forest succession quantitatively. However, this is

limited in practice due to for labor-intensive sampling over long time periods.

Individual-based models of tree competition and succession have been a common

approach for this issue (e.g., Botkin, 1993). Forest simulation models are effective over

various spatial and temporal scales and useful when comparing the affects of different

land management practices, predicting land-use, climate change, and forest production

over time (Peng, 2000).

Managing fire-dependent ecosystems in Florida, such as longleaf pines (Pinus

palustris) savannahs, has become increasingly important to reduce wildfire outbreaks and

ensure future establishment of this sensitive tree species. Mechanistic and stochastic

models of fire and fuel dynamics have aided in this type of management. Mechanistic

models include FARSITE (Finney, 1998) and BEHAVE (Andrews, 1986) that require

specific physical and chemical properties of the landscape to determine the rate and shape

of fire spread. The stochastic model DISPATCH (Baker et al., 1991) simulates the

effects of changing climate or shifts in fire disturbance regimes on relatively large









landscapes. LANDIS (LANdscape DIsturbance and Succession, Mladenoff et al., 1996)

models landscapes over large spatial and temporal scales requiring relatively coarse input

data of dominant tree species and fire characteristics of the land (He and Mladenoff,

1999). This model is spatially interactive, implementing variables such as tree species,

seeding characteristics, and cohort-based growth capabilities, as well as fire ignition,

spread, and fuel traits. These features enable users to gain insight into the overall

dynamics of the forest system visually and quantitatively over decades to several

centuries (Mladenoff and He, 1999). The focus of this study was to spatially model long-

term forest succession in a longleaf pine (Pinuspalustris) ecosystem through two fire

disturbance regimes utilizing the LANDIS model and assessing its practicality as a

management tool in the Goethe State Forest, Florida.

Longleaf Pine Habitat: Sandhill

Description

The sandhill community is found on rolling hills of sandy infertile soils of north-

central Florida. Sparse longleaf pines (Pinuspalustris) dominate the canopy with a low

understory of wiregrass (Aristida strict) as well as other grasses and forbs. The mid-

canopy consists mainly of turkey oak (Quercus laevis) and blue-jack oak (Quercus

incana) in varying densities (Myers, 1990; Greenberg and Simons, 1999). Some oaks

produce "oak domes," dense thickets of sand-live oaks (Quercus geminata) with little or

no vegetative ground cover (Guerin, 1993).

History and Current State

Sandhill was once one of the largest forest types in the southeastern coastal plain of

the United States, but has diminished by over 98% due to harvesting and fire suppression

(Myers, 1990). The highly dispersed pine and low understory provided for an easily









traveled and nicely shaded route for early European settlers (Hall, 1829). Overtime, this

ecosystem became fragmented by clear-cutting for timber, development, introduction of

livestock, and agricultural practices. Many of these fragments have suffered from fire

suppression, resulting in encroachment by faster growing, shade tolerant tree species

(Frost, 1993; Rasser, 2003).

Restoration and preservation of P. palustris dominated ecosystems has become a

well known practice in the southeast which contributes to recreation, aesthetics, reducing

the onset of pest infestation, and conservation for endangered and often endemic species

that thrive in this habitat (Boyer, 1993; Neel, 1993). Land managers plan for frequent

low-intensity fires (-3-5 years) that keep encroaching hardwoods in check (Hartnett and

Krofta, 1989; Platt and Rathbun, 1993). These fires also create an environment that

supports P. palustris recruitment and open savannah like parks for recreation including

hunting and easily traveled trails (Myers, 1990). Several endemic species thrive in these

habitats, including the Sherman's fox squirrel (Sciurus niger), red-cockaded woodpecker

(Picoides borealis), brown-headed nuthatch (Sittapusilla), Bachman's sparrow

(Aimophila aestivalis), and the gopher tortoise (Gopherus polyphemus) (Engstrom, 1993).

Unique Traits

P. palustris is the characteristic fire-dependent conifer of sandhill communities.

Frequent low-intensity fires remove accumulated leaf litter, overgrown grasses, and

encroaching oaks exposing bare-mineral soil required for seed germination and

establishment (Croker, 1975; Boyer, 1990; Myers, 1990). Another distinctive feature

about this pine is its seedling's slow growing "grass stage." Here it develops a long thick

tap root (Myers, 1990), increases in diameter (Platt et al., 1988), and is protected from

fire through dense tufts of long, moisture rich needles. It may stay in this stage for 5-15









years. Within one growing season, P. palustris can grow over Im in height avoiding

lethal temperatures of fire near the ground surface while protecting the terminal bud. A

maturing tree develops fire-resilient scaly plates, which define it as the most fire-resistant

pine. Furthermore, the ecosystem has vegetation with pyrogenic traits that facilitate the

ignition and spread of fire (Fonda, 2001; Myers, 1990). Without frequent fire, the stable

P. palustris dominated forest disappears due to failure of seedling establishment and

incursion of hardwood upland species (Myers, 1990).

Competitive Oaks

If fire is suppressed on sandhill sites for less than a decade, oaks (Quercus spp.) or

other hardwood species (Prunus serotina, P. caroliniana) from adjacent communities

may encroach and often out-compete the shade intolerant P. palustris (Myers, 1990).

Southern hardwood species often resprout, even if they appear to be killed by fire or other

disturbances. Most flourish under dense canopy cover and have chemical and physical

traits that reduce the onset of fire (Platt and Schwartz, 1990). Three southeastern United

States oak species (Q. geminata, Q. hemisphaerica, Q. laevis) are focused on in this study

as competitors with P. palustris.

Turkey oak (Q. laevis) is a deciduous species uniquely adapted to fire-maintained

communities of the southeastern U.S. They are found in dry sandy areas and well

drained ridges of the coastal plain (Berg and Hamrick, 1994). Sandhill areas are their

favored habitat, sometimes out-competing longleaf pines eventually forming "turkey oak

barrens." Q. laevis life history traits, such as vegetative propagation, fast growth, and

vigorous resprouting subsequent to fire, are important factors in the interaction with P.

palustris. It does not contend well with other more shade tolerant oaks (e.g., Q.

geminata, Q. hemisphaerica), which are also examined in this study, increasing the risk









of upland forest encroachment during fire inhibition (McGinty and Christy, 1977; Putz,

personal communication, 2005).

The commonly dominant laurel oak (Q. hemisphaerica) is found in a diversity of

north and central Florida habitats which exhibit a broad range of moisture and fertility

gradients (Platt and Schwartz, 1990). The moisture rich environments created by

hardwood habitats with Q. hemisphaerica decrease the likelihood and intensity of fire,

although fires during dry periods may be detrimental to this tree and habitat (Wade et al.,

1980). It is frequently considered a facultative seral species on sandhill sites, with a high

tolerance of shade, clonal traits and resprouting capabilities subsequent to moderately

intense burns (Putz, personal communication, 2005). Boundaries between hardwood

forests and sandhills are often kept in check by low-intensity prescribed burns ignited in

sandhill areas and allowed to creep into the edges of hardwood forests (Platt and

Schwartz, 1990; Hardin, 2004).

Sand-live oak (Q. geminata) tolerates highly infertile sandy soils typical of

Florida's scrub ecosystem as well as xeric hardwood forests of north-central Florida. Its

traits are similar to Q. virginiana (live oak), but are smaller in size and produces acorns in

pairs. Q. geminata is not quite as tolerant of shade as Q. hemisphaerica and is slightly

more fire tolerant, but thrives as a clonal propagator and strong resprouter after fire. Q.

geminata form the aforementioned clonal 'oak domes' within sandhill communities that

can spread and dominate the landscape as scrub or xeric hammock if not regulated by

frequent burning (Guerin, 1993).

Sandhill, scrub, and hardwood forests are mutually exclusive ecosystems that

respond very differently to fire. Sandhill habitats have vegetation with pyrogenic traits









that facilitate the ignition and spread of fire. Scrub and sandhill and are closely linked

ecologically but the fire-frequency is much lower in scrub and its principal components

are invasive (Myers, 1990). Hardwood forests create moisture rich environments

containing less flammable leaf litter deterring the onset of fire and reducing its intensity

(Platt and Schwartz, 1990).

Objectives

The purpose of this research was to use the LANDIS model to study the spatial

relationships of these four species and their habitats in response to various fuel

treatments. Two scenarios were developed for this study: (I) the Frequent Fire Model

(FFM), simulating a fire-maintained P. palustris ecosystem; and (II) the Infrequent Fire

Model (IFM), simulating a historically fire suppressed landscape. Following are

questions proposed for this research: 1) Can the unique fire regime of the P. palustris

habitat be modeled using LANDIS? 2) What are the implications of using LANDIS as a

long-term management tool for the P. palustris habitat? 3) Are the fire and fuel

characteristics of LANDIS sufficient to simulate the short and long mean fire return

intervals of various habitats in the study site?

LANDIS Model Description

General Characteristics

LANDIS (LANdscape DIsturbance and Succession) was introduced in 1996 and

has been used extensively for spatial analysis and modeling of large heterogeneous

landscapes and long time frames (Mladenoff et al., 1996, Mladenoff and He, 1999).

Originally, LANDIS was designed to simulate the forests of the northern lake states

(Mladenoff et al., 1996, He and Mladenoff, 1999). Since then, it has been used for forest

succession modeling in many parts of the world. Successional patterns of an undisturbed









boreal forest reserve in northeast China were compared to surrounding disturbed areas

using dominant tree species in LANDIS (He et al., 2002). In Finland, Pennanen and

Kuuluvainen (2002) manipulated the simulated fire behavior and successional processes

of LANDIS to form FIN-LANDIS. Instead of using specific tree species of a forest,

Franklin et al. (2001) used plant functional groups. These groups were ascertained from

Noble and Slayter's (1980) vital attribute fundamentals describing the successional role

of plant species. He et al. (2004) created various fuel treatments to assess the risk of

wildfire occurrence in the Missouri Ozark Highlands. Overall, LANDIS is used to

simulate large forested areas that have limited inputs. The presence/absence nature and

coarse scale of LANDIS avoids the need for precise measurements and prediction of

species abundance (Mladenoff and He, 1999). The spatial scale is user defined but is

often in 10's to 100's of square meter pixels. The temporal scale is restricted to 10-year

time intervals.

LANDIS combines the ability to simulate landscape-sized forest successional

patterns while implementing specific tree species attributes and probabilistic disturbance

regimes, such as fire, wind, and harvesting. Forest succession is simulated in LANDIS

based on seed dispersal, seedling establishment, sprouting traits, competition (fire and

shade tolerance), growth, and mortality characteristics of each tree species modeled

(Gustafson et al., 2004). For instance, a shade-intolerant species cannot establish easily

on a site where shade tolerant species are well established. These species will remain

dominant unless other attributes, disturbances or environmental factors alter the

conditions throughout the simulation (Mladenoff and He, 1999).









Fire Disturbance Characteristics

Fire disturbance in LANDIS affects tree establishment, mortality, and inter-specific

competition. The probability of establishment for each tree species is based on fire

tolerance, age, and site characteristics. The stochastic and spatially explicit nature of fire

is represented through mathematically defined distributions and various algorithms. Fire

disturbance parameters influence the size, probability of fire occurrence on the land,

return interval, spread, and severity of the fire regime. The fire tolerance levels (1-5) of

each tree species are categorical representations of their relational tolerances to fire.

Fire probability (P) in LANDIS follows a negative exponential distribution and is

calculated in LANDIS with the following equation:

P bxlfxMI (+2) (2-1)

MI is the mean fire return interval (years) for each landtype. The parameter If

represents the year since the last fire on each landtype. A larger If represents a higher

probability that a fire will occur on that landtype. The fire probability coefficient (b)

represents the chance a cell randomly selected for ignition check will ignite and is used

during the calibration process (He et al., 2003; Mladenoff and He, 1999).

The fire probability function works directly with ignition and spread of fire in

LANDIS. Fire is ignited based on the percentage of cells checked for ignition (chosen

randomly) and on the probability that a fire will occur. It is also a function of map size

(Ignition = ignition coefficient x total number of cells). LANDIS is able to ignite and

therefore simulate more than one fire disturbance within an iteration, but the chance is

exponentially decreased through time (He and Mladenoff, 1999). Fires spread in

LANDIS as a function of fire size, fire probability, species attributes (e.g., fire tolerance

and age), and spatial arrangement of adjacent cells. A fire will continue to spread until









the pre-determined maximum fire size is reached, the adjacent cells cannot burn based on

probability estimates, or it reaches the edge of 'forested' land (He and Mladenoff, 1999;

Mladenoff and He, 1999).

Fuel accumulation for each landtype is established from a conceptual fire curve

representing a relationship between fuel quantities, based on detritus accumulation and

decomposition characteristics (Gustafson et al., 2004). It is dependent on the time since a

burn last occurred (He et al., 2003) and the landtype severity and species susceptibility to

fire act together to determine which species and age-cohorts are killed during a fire

disturbance (Mladenoff et al.,1996). As such, the fire disturbance in LANDIS is not

entirely stochastic as some areas within a landscape may be more susceptible to fire than

others (He and Mladenoff, 1999).

The fire disturbance file is exclusively designed for simulating fire size on the

entire landscape and is a determining factor of fire spread, fire probability, fire and forest

susceptibility, and spatial configuration of forest conditions (Gustafson et al., 2004). Fire

size also follows the notion that large disturbances are less frequent than small

disturbances through a log-normal distribution (Mladenoff and He, 1999). These fire

disturbance parameters and their calculations are described in detail elsewhere (He and

Mladenoff, 1999; Mladenoff and He, 1999; He et al., 2003).

Methods

Study Area

The Watermelon Pond Unit (WPU) of the Goethe State Forest is a 1938-ha tract

located in Levy and Alachua counties of Florida, USA (latititude 290 32' 42 ", longitude

82 36' 20"; Figure 2-1). Its checker-board structure makes management difficult,

especially for prescribed burning. Agricultural land owners, state and county roads, and









two small towns divide the forest, further complicating management. Before European

settlement, the land was a homogeneous landscape of the sandhill community type

dominated by P. palustris. During the last century, the area was fragmented by

anthropogenic effects, specifically phosphate mining and non-sustainable timber

harvesting. Many of these areas were unmanaged causing dramatic successional changes

in this ecosystem with few P. palustris trees left today (Rasser, 2003).

A main goal of land managers is the immediate implementation of prescribed fire

plans for restoration and preservation of sandhill areas (Hardin, 2004). Sandhills of the

WPU are classified into four general groups, ranked in order of quality: 1) areas with

scattered mature P. palustris (>=30 ft2 basal area) with subdominant Q. laevis and

abundant natural P. palustris regeneration, 2) stands where Q. laevis predominate with P.

palustris being a minor component (10-20 ft2 ba) and sparse P. palustris regeneration, 3)

pure Q. laevis and scrub oak stands, and 4) areas newly planted with P. palustris

seedlings. Wiregrass (Aristata strict) and other natural herbaceous groundcover are

found in these areas, which are considered strong indicators of community health

(Hardin, 2004). Note that Q. laevis is an abundant tree species found at the WPU, and is

considered a dominant component of some historic sandhill sites (Greenberg and Simons,

1999). As such, it provides a rare opportunity for the WPU staff to manage Q. laevis

dominated sandhills.

To maintain a level of biodiversity, it is important to keep intact the recently

developed scrub and hardwood communities while restraining further encroachment into

sandhill sites with prescribed burning plans. Scrub is Florida's most endangered

ecosystem with many rare and threatened endemic species. Prescribed fire initiated in










sandhill areas will be allowed to bum into the edges of these scrub and hardwood forests

as part of the land management plans (Hardin, 2004).







Archer, FL
















0 1 2 4 Kilometers
Legend
Goethe State Forest Boundary
State of Florida
Alachua County
SLevy County

Figure 2-1. Map of the study site: Watermelon Pond Unit of the Goethe State
Forest, Florida, USA.

Parameterization of LANDIS

Spatial inputs

Spatial data specific to the WPU used to develop both the landtype and the forest

composition map and their attributes are the: a) biological communities delineated from

high-resolution digital aerial photography taken in February 2002, b) geo-referenced

ground data on tree species from 2002 (Rasser, 2003), c) geo-referenced ground data on

biological communities from 2003 and d) temporal aerial photography. The high-

resolution digital aerial photography (0.3m ground sampling distance) was donated from









Mpower3/Emerge (Andover, MA). These datasets provided baseline information

representing the area's vegetative structure and composition. ArcGIS 8.x, 9.0 and

IMAGINE 8.x were used to develop the final Erdas GIS file format datasets with a spatial

resolution of 20m. All input parameter files for LANDIS, except the spatial files, are

included in the Appendix.

Landtypes. The WPU was stratified into four distinct habitats unique in their fire

and fuel traits: a) current P. palustris sandhill habitat, b) hardwoods, c) scrub, and d) wet

areas. These landtypes were created from digitally delineated biological communities on

habitat type and their characteristic tree species. The ground reference data was used to

perform an accuracy assessment on the landtype map. The fire accumulation and severity

curves for each landtype were qualitatively assessed based on literature and current site

conditions of each habitat and used for both fire scenarios (Figure 2-2). Qualities, such

as highly flammable longleaf pine needles (Fonda, 2001) and wiregrass, dry sandy soils,

and low decomposition rates create the characteristic sandhill environment with highly

frequent (<10 years) low-intensity fires. Scrub is considered a pyrogenic community that

accumulates fuels slowly and often has a high heat of ignition, but produces catastrophic

fires at a rather low MI (50 years) (Myers, 1990). Furthermore, the abundance of the

shrub Ceratiola ericoides, a highly flammable and characteristic scrub species,

contributes to the pyrogenic nature of this habitat. Hardwood forests have rather long

MIs (80 years) with a moisture rich environment and fast decomposition rates that greatly

reduce the intensity of these periodic fires (Platt and Schwartz, 1990). The wet areas are

defined as portions of the WPU with hydric soils, currently abundant with water, and/or

prairie type habitat with fluctuating water levels. Since the wet areas have little or no










trees, they were not actively modeled in LANDIS. The approximate areas for each

landtype are as follows: current P. palustris habitat (1050 ha), scrub (375 ha), hardwoods

(400 ha), and wet areas (165 ha). Even though the WPU has distinct habitats, it is small

and relatively flat, which does not restrict the establishment of any of these tree species.

Therefore, the establishment coefficients were set to one for each tree species on all

landtypes in the FFM and instead their successional behavior was purely designed based

on their attribute parameters and response to fire. The establishment coefficients in the

IFM were also set to one, except P. palustris was reduced to 0.25 because of its weak

competitive and establishment abilities during fire suppression.


5-

4 -




W Sandhill
a 1 Hardwood
Scrub
01
0 10 30 50 70 90 100
Years of Fuel Accumulation

Figure 2-2. Fire fuel accumulation curve for all three active habitats at the WPU.

Forest composition. The forest composition map and associated attributes were

derived from ground reference data, temporal aerial photography, and managerial records

and correspondence. Forest classes required for LANDIS were distributed using the

biological community spatial dataset as a foundation. The vegetative data designated the

tree species found in each forest class. The ground reference data was used to perform an

accuracy assessment on the forest composition map. For simplification, only the four

most abundant tree species at the WPU were used in the LANDIS model, namely









longleaf pine (P. palustris), turkey oak (Q. laevis), sand-live oak (Q. geminata), and

laurel oak (Q. hemisphaerica). Historic aerial photography and forest management

records were used to determine tree age-class distributions. Approximately one set of

aerial photographs of the study site per decade back to 1949 were used. The final map

consisted of 17 classes with various combinations of species and their age distributions

within the study site (see Appendix).

Species attributes

Demographic parameters for the four tree species modeled were obtained from

literature (Duncan and Duncan, 1988; Boyer, 1990; Young and Young, 1992; Preston and

Braham, 2002) and experts opinion (Long, personal communication, 2004; Jokela,

personal communication, 2004; Putz, personal communication, 2005; Table 2-1.). Unlike

other studies using LANDIS (Pennanen and Kuuluvainen, 2002; Wimberly, 2004),

longevity was input as the typical lifespan of a given species in this geographic region

rather than its maximum known lifespan. This creates a more realistic mortality

inference of each species. A seeding distance was input to represent the range that a tree

species can effectively seed out, with the assumption that other natural effects, e.g., wild

animals, influence where a seed may fall and germinate. The Quercus spp. modeled are

readily able to resprout after a disturbance at a very young age. With a long seeding

range and early propagation, some Quercus spp. establish and spread effortlessly in

LANDIS. The propagation probability coefficient was developed by understanding its

sensitivity (see below) within LANDIS and qualitatively shaping this variable for each

Quercus spp.









Fire Simulation Scenarios and Parameters

Two fire scenarios were created to simulate a fire-maintained P. palustris

ecosystem (Frequent Fire Model: FFM) and a fire suppressed one (Infrequent Fire Model:

IFM), each with unique fire return intervals and fire size parameters. All fire parameters

for both models (Table 2-2., Table 2-3.) were obtained from the Goethe State Forest

management plans (Hardin, 2004) and communication with management personnel

(Cahal, personal communication, 2005; Copeland, personal communication, 2005; Penn,

personal communication, 2005; Zimmerman, personal communication, 2005). Because

the P. palustris ecosystem requires frequent low intensity fires, the mean fire return

interval (MI) of this habitat in the FFM was set to <10 years to ensure fires every

LANDIS time step. The P. palustris ecosystem in the IFM followed historic records

noting the fire suppressed period dating back to 1890 with a MI of 45 years (Zimmerman,

personal communication, 2005). The scrub and hardwood habitats followed the same

landtype parameterization methods mentioned in the 'Spatial Input: Landtypes' section.

Fire size variables for both models were estimated from delineated prescribed burns for

the WPU, their management plans, and known wildfire outbreaks. The bum size

parameters of the FFM followed that of known sizes for prescribed bums and future

management plans. The burn size parameters of the IFM were directly related to the only

known wildfire that occurred in 1980. Each simulation was run for 50 LANDIS time

steps (500 years) to ensure reaching model stabilization. Species abundance was

recorded and used to compare the effectiveness of each fire scenario focusing on P.

palustris and its establishing abilities.









Calibration and Sensitivity Analyses

The fire parameters of LANDIS were calibrated by adjusting the associated fire

coefficients interactively to achieve a desired mean fire return interval for all landtypes

and mean fire sizes for each of the two fire scenarios within a 500 year simulation.

Default values were initially used for the fire size (50) and fire probability [b = MI of

each landtype (see Equation 2-1)] coefficients and manipulated in intervals of 10

depending on the scale and direction of error (He et al., 2003). For each treatment, these

values were systematically varied until the parameters met the simulated values or until

no further improvements could be made.

Many of the tree species attributes were interactively manipulated to develop

relational species behavioral traits. For instance, Q. hemisphaerica and Q. geminata were

set to have rather similar dominance traits if no fire was introduced and the former with

stronger seeding capabilities, while P. palustris thrived and established itself subsequent

to a burn. Q. laevis was targeted as a transitional species that dominates areas with P.

palustris if fire is somewhat infrequent or recently ceased, but may be replaced by

invading Q. hemisphaerica and Q. geminata. P. palustris was also set to not have the

ability to vegetatively propagate or establish well during fire suppression.

The vegetation propagation probability coefficients of the Q. spp. were set equal to

one another because of their similar re-sprouting abilities after a burn. Only a small (1%

to <6%) change in the abundance of any of the four species was recorded from the

simultaneously adjusted vegetation propagation coefficients over a range from 0.1 to 0.9

of the three Q. spp. To avoid bias, the median coefficient (0.5) was chosen for all three Q.

spp.









Sensitivity analyses were performed on the mean fire size and mean fire return

interval to measure their contribution to the model response. Each parameter was

adjusted 20% from the base run and P. palustris abundance was recorded during 500

simulation years as well as the simulated mean fire size on the landscape in each interval

(Mladenoff and He, 1999). The MI analysis was only performed on the IFM to illustrate

the effects of a 20% change in the MI of all landtypes simultaneously, which could not be

done with the already minimal fire return interval (<10 yrs.) of the FFM.

Table 2-1. Attributes of the four species studied using LANDIS.
PIPA QUGE QULA QUHE
Longevity (yrs.) 200 130 100 100
Age of Maturity (yrs.) 20 20 20 20
Shade Tolerance (1-5) 1 4 3 4
Fire Tolerance (1-5) 5 1 4 1
Effective Seeding Distance (m) 30 50 50 70
Maximum Seeding Distance (m) 50 70 70 90
Propagation Probability After Fire (0-1) 0 0.5 0.5 0.5
Minimum Resprout Age (yrs.) 0 5 5 5
PIPA: Pinuspalustris; QUGE: Quercus geminata; QULA: Quercus laevis; QUHE:
Quercus hemisphaerica

Results and Discussion

Calibration and Sensitivity Analyses

Mean fire size was calibrated for both the FFM and IFM, with a -12% and -16%

error, respectively (Table 2-2.). Error is calculated by the dividing the difference

between observed fire size and the expected fire size by the expected fire size. The

simulated MS was accurate to within <0.7% between both models. The IFM was more

accurately calibrated than the FFM in terms of expected mean fire return intervals (Table

2-3.). The low MIs of the hardwood and scrub habitats could not be calibrated accurately

in the FFM due to the high MI (<10 yrs.) of the P. palustris sandhill habitat. This result









appeared realistic as fires may ignite within sandhill areas and spread into adjacent

habitats. The total amount of area burned in hardwood and scrub sites were relatively

low in the FFM while sandhill areas burned readily (Figure 2-3.). The hardwood sites

were the most difficult to correctly calibrate for either scenario because the low intensity

fire after long durations of fire suppression increased the probability that fire could occur

again in LANDIS. As such, the increase MI of the sandhill areas in the IFM contributed

to the decrease in fire frequency of both the hardwood and scrub habitats. The hardwood

habitat MI became more accurately simulated while scrub habitat resulted in an

exceedingly larger MI. Ultimately, the fire and fuel characteristics in LANDIS were

sufficient to simulate the diverse fire regimes of theses three habitats during longer fire

return intervals.

Table 2-2. Calibration analysis of mean fire size (MS) in each fire regime scenario.
Frequent Fire Model Infrequent Fire Model
Predetermined MS, ha (% area) 120 (5.9%) 106 (5.3%)
Simulated MS, ha (% area) 105 (5.2%) 101 (5.0%)
Standard Deviation (% area) 59 (2.9%) 28 (1.4%)
Error -12% -16%

Table 2-3. Calibration analysis of mean fire return intervals (MI) in each fire regime
scenario.
Frequent Fire Model Infrequent Fire Model
Sandhill Hardwoods Scrub Sandhill Hardwoods Scrub
Predetermined MI 10 80 50 45 80 50
Simulated MI 13 15 25 38 46 77
Standard Deviation 6 6 16 19 27 71
Error 31% -81% -50% -15% -43% 53%

The mean fire size (MS) parameters were generally more sensitive in the IFM than

the FFM (Figure 2-4.). In the IFM, there was a 17% increase and 7% decrease in MS

with the direct 20% change in MS. The strong increase in fire size was directly related to

the average increase of 36% in P. palustris abundance throughout the simulation. This









simulation reflects a management goal of fire re-introduction at the WPU. The FFM

resulted in a 2% and 4% increase in MS with a 20% increase and decrease in the MS,

respectively. These small changes correlated with the small average changes in P.

palustris abundance, ranging from 0.5% to 3%. The lack of change in the simulated MS

of the FFM is the direct result of the influence of the time since the last fire in LANDIS.

Once a fire has occurred, fire probability and severity diminish on that site. Since the MS

showed little change in the FFM, the P. palustris population was only slightly influenced.

The MIs, MS, and P. palustris abundance changes in the IFM were more sensitive

to a 20% increase in MI within all landtypes rather than a decrease (Figure 2-5). There

was a 13% increase in overall simulated MI with an increase in overall MI. The increase

in MI resulted in two fires dropped in years 90 and 320. MS did not show much change

(-2%) with increased MI, but the total amount of area burned increased by 5%. P.

palustris abundance reacted with a 16% increase in average abundance with increased

MI. A large fire on the sandhill habitat in year 400 brought on new P. palustris growth

and influenced the calculated increase in average abundance greatly. There was no

change in simulated MIs, MS, or P. palustris abundance throughout the simulation when

the MIs were reduced by 20%. These changes are complex (Mladenoff and He, 1999)

and largely due to fire probability estimates calculated for each landtype as a function of

MI, time since last fire, and a calibrated fire probability coefficient.









0.300
0.250
0.200
0.150
0.100 -
0.050 -
0.000




0.300
0.250
0.200
0.150
0.100
0.050
n nnn


Frequent Fire Model


0 100 2


300


400


Infrequent Fire Model


II I


I, IH II


500


* Sandhill
O Scrub
o Hardwoods


I Il


0 100 200 300 400 500
Model Year
Figure 2-3. Proportion of area damaged by fire and burn frequency in each habitat using
both fire scenarios in LANDIS.


Ill IIII1N I












Frequent Fire Model


600

500

400

300

200

100

0


Infrequent Fire Model


600

500

400

300

200

100

0


O +20% MS
* Base MS
O -20% MS


0 100 200 300 400 500
Model Year

Figure 2-4. Sensitivity analysis: Mean area damaged by fire each model year as the mean
fire size (MS) changed by 20% in both fire scenarios.

Fire Scenario Effects on Species Abundance

Species abundance changes throughout the LANDIS simulations illustrate the

effects of the two fire regimes on each species. The results were analyzed after model

stabilization (-yr. 200), unless otherwise stated. The most dramatic effects between the

two fire regimes was the change in abundance ofP. palustris (Figure 2-6.). In the FFM,

the population was able to recover and gradually increase in number through regular

recruitment. Gaps were formed from trees that died of old age or killed by fires (Figure

2-7.). As fires occurred, mature P. palustris were able to seed into these gaps and within


1111 11111111111111111111111111111111 11111111 1111111111111111111llllllllllllllln11










a few decades become a dominant species. The moderately fire resistant Q. laevis

becomes more dominant on the land as well during a frequent fire regime, while Q.

hemisphaerica and Q. geminata are suppressed. At year 500, the higher mean population

abundance (34%) of P. palustris in the FFM demonstrates the ability of this species to

thrive in a fire-maintained community (Figure 2-8.). Over 500 years, the mean

abundance ofP. palustris and Q. laevis, the two more fire tolerant tree species increased

by 28% and 22%, respectively, when using the FFM compared to the IFM. The other

two oaks, Q. hemisphaerica and Q. geminata, only revealed a slight increase in mean

abundance (4% and 2%, respectively) in the FFM. Fires do not entirely extirpate any of

these three oak populations because of their ability to vegetatively propagate subsequent

to fire at a relatively young age (Boyer, 1990; Long, personal communication, 2004,

Table 2-1.). Their quick re-establishment capabilities make P. palustris ecosystems

difficult to restore after long periods of successional changes caused by fire suppression

(Hartnett and Krofta, 1989).


600 -
SDO +20% MI
500 U Base MI
O -20% MI
M 400
E
a 300
S200
C 100
0
0 100 200 300 400 500
Model Year

Figure 2-5. Sensitivity analysis: Mean area damaged by fire each model year with
change in overall mean fire return interval (MI) in the IFM.









The FFM demonstrated more stability in the tree populations over time than the

IFM, illustrated by the overall lower variation of species abundance throughout the 500

year simulations (Table 2-4.) This is caused by the highly intermittent fires in the IFM

(Figure 2-3.). Even though the FFM maintained a relatively stable population of P.

palustris, the outcomes were not entirely realistic. The northern portions of the WPU

with P. palustris were burned only 1 to 10 times over the 50 LANDIS time steps. A

typical sandhill community requires prescribed fires every 3-5 years (Hardin, 2004), or at

least a 10 year return time (Platt et al., 1991). As such, the entire community at the WPU

should burn every LANDIS time step. The model may have been affected by the

fragmented structure of the WPU. Fires burned more readily (20 fires) within the non-

restrictive boundaries of the southeastern portion of the WPU (Figure 2-9.) suggesting

fire spread constraints in the smaller confined areas of the northern WPU.

Management Implications

Using this LANDIS model as a management tool for the WPU is plausible for

understanding the adverse effects that some of Florida's oak species have on P. palustris

ecosystems within just a few decades. The strong seeding and re-sprouting abilities,

shade tolerance, fast growth, and diverse site characteristics of Q. hemisphaerica make it

a difficult tree to control on fire suppressed sandhill sites and may require other methods

of control (Myers, 1990; Putz, personal communication, 2005). Q. geminata may also be

difficult to control, is tolerant of the typical xeric conditions of the sandhill and scrub

communities, and therefore is a strong competitor with P. palustris (Myers, 1990). Both

of these Q. spp. are regularly kept in check by frequent fires that are allowed to creep into

their borders (Hardin, 2004), but are opportunistic during fire suppression.












Frequent Fire Model


0 100 200


Infrequent Fire Model


* PIPA
------ QUGE
-- QULA
QUHE


-' -. P*rC'


0 100 200 300 400 500
Model Year

Figure 2-6. Fire scenario effects on species abundance in a single LANDIS model run.
PIPA: Pinus palustris; QUGE: Quercus geminata; QULA: Quercus laevis;
QUHE: Quercus hemisphaerica

Q. laevis tolerates fires well and re-sprouts vigorously following fire (McGinty and

Christy, 1977), but was the least abundant species on the landscape during the model

runs. In LANDIS, this species was replaced by the other hardwoods due to its lack of

shade tolerance and short longevity. P. palustris was able to out-compete Q. laevis with

frequent fire in spite of the oak's re-sprouting abilities. As such, the species is complex


2000



1500



1000



500


2000



1500



1000



500



0









and difficult to model. Otherwise, LANDIS is known to be capable of adequately

simulating forest dynamics when only a few species are included (Franklin et al., 2001;

Pennanen and Kuulavainen, 2002).

y380 y390 y400



















SYounger Age Cohorts
Older Age Cohorts .Fire between Fres between
PIPA iUGE QUHE QULA Open Area Iyes 30 & 4 I
0 500 0 250 500 1,000 Meters


I I.













Figure 2-7. P. palustris (PIPA) behavior following fire disturbance in a southeastern
portion of the WPU. The top row represents the dominant species of each
cell. The bottom row illustrates the recruitment of young PIPA in openings
created by fire induced tree mortality. PIPA: Pinuspalustris; QUGE:
Quercus geminata; QULA: Quercus laevis; QUHE: Quercus hemisphaerica

Even though the P. palustris population did stabilize and continued to increase in

abundance in the FFM, this version (3.7) of LANDIS is applicable for only a few of the

management plans for the WPU. LANDIS maintained an aesthetically pleasing mosaic

type landscape pattern and illustrated the invading capabilities of particular hardwood

species. It also provided the spatial and temporal aspects of fire and tree species











behavior with minimal inputs (Mladenoff et al., 1996; Mladenoff and He, 1999). The


seeding characteristics of all four species were effectively modeled: They all took


advantage of gaps created by fire or age induced mortality and the oaks readily re-


sprouted subsequent to fire disturbance. P. palustris typically became dominant shortly


after a fire through tolerance and recruitment (Figure 2-7.). The short fire return interval


of 3-5 years (Hardin, 2004) required to maintain and/or restore sandhill communities


could not be modeled with LANDIS 3.7 because of the 10 year time step. Furthermore,


the fragmented structure of the WPU restricted fire spread across the landscape.


y400


y450



, -.





4L
**i :1M


y500
y< -


SP
." .. ':-. ,

.' ".. '



Dominant Tree Species
SPIPA I QUGE QUHE | QULA | Open Area


IJ .




..*l V
t



I I I I I I I I I t
0 05 1 2 Kilometers


Figure 2-8. Comparison of dynamics of tree species dominance between the two fire
scenarios in the east portion of the WPU. PIPA: Pinuspalustris; QUGE:
Quercus geminata; QULA: Quercus laevis; QUHE: Quercus hemisphaerica










Table 2-4. Statistics on species abundance within each fire scenario.
Frequent Fire Model Infrequent Fire Model
Mean Std Mean Std
PIPA 924 125 664 279
QUGE 1,319 108 1,285 128
QULA 745 357 583 413
QUHE 876 81 837 103


Frequent Fire Model








-
l


UpperScenes
0 05 1 2KNlometers
Lower Scenes
0 05 1 2 lometers


Infrequent Fire Model


k


t


Fire Severity

13
4
i5
[ No Fires


Figure 2-9. Map of all fires, their distribution, and severity at the WPU throughout the
500 year simulation. Darker reds are more severe fires. The most recent fire
is displayed on top of older ones.

Future Model Development

Modeling techniques that may further advance this study's use of LANDIS

includes implementing smaller time steps, further calibration, and fine-tuning of the

spatial inputs. The next version ofLANDIS (LANDIS II) will have one to several year

time steps, which may enhance the realism ofP. palustris behavior to fire and its


WU.-.









competitive tree species. Since the fragmented structure of the WPU restricted fire

spread in this LANDIS model, simulating each section separately may induce more

frequent fires on each individual population ofP. palustris. The fire and fuel

characteristics of LANDIS were sufficient to simulate the various fire regimes desired in

this study, especially during higher fire return intervals and severe fires. Although,

additional calibration may increase accuracy of the fire disturbance parameters. Other

tree species present at the WPU may be added to the model to study more complex

interactions within the landscape. Wiregrass (Aristida Stricta), a grass species and an

indicator of P. palustris sandhill community health (Hardin, 2004) may be included as a

relational species (Gustafson et al., 2004). Additional historical data or tree cores from

the WPU may provide for more accurate age distributions of species. Although long-

term validation procedures are difficult to establish, they would support the quality and

viability of this model.

Conclusions

Several of this study's objectives were met. Two fire regimes were created to

assess the landscape over time in relation to P. palustris and its competitive hardwoods in

an ideal fire-maintained habitat and a fire suppressed one. Even though this model did

demonstrate certain characteristics of P. palustris dynamics well, the results were not

entirely realistic: Fire frequency was still considerably lower in the FFM, especially

within individual populations, than what is required to maintain a viable P. palustris

ecosystem. Q. hemisphaerica and Q. geminata were quite accurately depicted on the

landscape, while some of the simulated dynamics of Q. laevis lacked realism. LANDIS

II may provide the details to more precisely simulate the fire regime desired for this

study, although the fragmented structure of the WPU may still adversely affect the






31


model's fire spread processes. The WPU may be able to use LANDIS as a management

tool to visualize the adverse effects hardwoods may have on P. palustris sandhill

communities within just a few decades. With some fine-tuning of parameters and

implementation of smaller time steps, this model may dramatically improve and be

applicable for management to visualize P. palustris population dynamics through time.














CHAPTER 3
DEVELOPMENT OF A DEMOGRAPHIC MATRIX MODEL OF LONGLEAF PINES
(Pinus Palustris)

Introduction

Longleaf Pine

P. palustris is a unique fire-dependent tree species of the southeastern United

States. Frequent low-intensity fires remove accumulated leaf litter, overgrown grasses,

and encroaching oaks exposing bare-mineral soil required for seed germination and

establishment (Croker, 1975; Boyer, 1990; Myers, 1990). Another distinctive feature

about this pine is its seedling's slow growing 'grass stage'. Here it develops a long thick

tap root (Myers, 1990), increases in diameter (Platt et al., 1988), and is protected from

fire through dense tuft of long, moisture rich needles. It may stay in this stage for 5-15

years. Within one growing season, P. palustris can grow over Im in height avoiding

lethal temperatures of fire near the ground surface while protecting the terminal bud. A

maturing tree develops fire-resilient scaly plates, which define it as the most fire-resistant

pine. Furthermore, the ecosystem has vegetation with pyrogenic traits that facilitate the

ignition and spread of fire (Fonda, 2001; Myers, 1990). Without frequent fire, the stable

P. palustris dominated forest disappears due to failure of seedling establishment and

incursion of hardwood upland species (Myers, 1990).

Sandhill, a P. palustris dominated ecosystem, was once one of the largest forest

types in the southeastern coastal plain of the United States, but has diminished by over

90% due to harvesting and fire suppression. The highly dispersed pine and low









understory provided for an easily traveled and nicely shaded route for early European

settlers (Hall, 1829). The ecosystem became fragmented by clear-cutting for timber,

development, introduction of livestock, and agricultural practices. Many of these

fragments have suffered from fire suppression resulting in encroachment by more fast

growing, shade tolerant tree species (Frost, 1993; Rasser, 2003).

Restoration and preservation of P. palustris dominated ecosystems has become a

well known practice in the southeast which contributes to recreation, aesthetics, reducing

the onset of pest infestation, and conservation for endangered and often endemic species

that thrive in this habitat (Boyer, 1993; Neel, 1993). Land managers plan for frequent

low-intensity fires (-3-5 years) that keep encroaching hardwoods in check (Hartnett and

Krofta, 1989; Platt and Rathbun, 1993). These fires also create an environment that

supports P. palustris recruitment and open savannah like parks for recreation including

hunting and easily traveled trails (Myers, 1990). Several endemic species thrive in these

habitats, including the Sherman's fox squirrel (Sciurus niger), red-cockaded woodpecker

(Picoides borealis), brown-headed nuthatch (Sittapusilla), Bachman's sparrow

(Aimophila aestivalis), and the gopher tortoise (Gopherus polyphemus) (Engstrom, 1993).

Population Matrix Modeling

Population matrix models are an ideal starting point for analyzing tree population

dynamics (Escalante et al., 2004; Zuidema and Zagt, 2000). Population matrix models

often use size-based classes and transition probabilities developed from individual tree

measurements over time. They are mechanistically designed to simulate survival,

growth, death, and fecundity in user defined time steps assuming that the environment in

which their data was collected is constant over time (Caswell, 2001). The labor required

to census and map the data needed to estimate these probabilities is intense (Silvertown et









al., 1996). Estimation of size-specific fecundity is especially difficult because the

process of seedling establishment depends on seed production, seed viability, dispersal,

and site quality on multiple scales.

Objectives

The purpose of this research was to study P. palustris demographics in order to

develop comprehensive density-dependent transition matrix models using multiple data

sources. Two transition matrices were developed: (I) A P. palustris population

transition matrix for a fire-maintained ecosystem and; and (II) A P. palustris population

transition matrix for an ecosystem during fire suppression. These matrices were

combined into one model to assess its response to various fire frequencies.

Methods

Data Sources

Data were compiled on longleaf pine demography from several sources for the

development of the fire-maintained transition matrix model. A population transition

matrix built by Platt et al. (1988) for longleaf pines (P. palustris) at the Wade Tract,

Thomas County, Georgia was used as a base for our model. Longleaf pine seedling and

mature tree growth data recorded by the Joseph W. Jones Ecological Center at the

Ichwauy Preserve, Baker County, Georgia (Jones et al., 2003; Palik et al., 2003) were

used to extend the Platt model. Longleaf pine seedling survivorship data were extracted

from the Jones Center data and a seedbed preparation study performed at the Escambia

Experimental Forest in southwest Alabama (Croker, 1975). Simulated fire suppression

effects on longleaf pine recruitment were estimated from a study in Chinsegut Hill

McCarty-Woods in northern Glades County, Florida (Hartnett and Krofta, 1989).









Transition Matrix Development

We extended an existing four year transition matrix model (Platt et al., 1988) with a

new 0-2cm diameter size class (Table 3-la.). The model was implemented in the python

programming language. The transition matrix created by Platt et al. was specifically

designed for a frequently burned P. palustris population (1-4 years), and therefore an

ideal starting point. The four year transition probabilities were in eight size classes,

starting with 2-10cm and continuing in 10cm increments (e.g., 10-20cm) until reaching

the 70cm+ size class. The Platt model lacked seedling (0-2cm) size class transition

probabilities and fecundity estimates for a fire maintained P. palustris ecosystem. An

intensive gap analysis on P. palustris seedling survivorship at the Ichwauy Preserve from

1998 through 2003 provided a complete population state vector (Jones et al., 2003; Palik

et al., 2003). In this study, ground measurements were collected on adult longleaf pines

[>4cm dbh (diameter at breast height)] and randomly planted plots of seedlings. Seedling

survivorship (34%) was estimated in two frequently burned P. palustris sites (Croker,

1975; Jones et al., 2003; Palik et al., 2003) which was similar to Boyer's (1963) seedling

survival rate (38%). Although the seedling mortality rate (including significant fire

scorch) seems high, the effects of fire suppression on seedling establishment are profound

(Boyer, 1963; Boyer, 1990; Myers, 1990).

The current population of P. palustris at the Ichwauy Preserve is essentially at an

ideal maximum density or K (basal area of 11.58 m2/ha) according to Boyer's (1999)

estimates for maintaining a healthy and naturally regenerating P. palustris population and

therefore applied as such to the model. We assumed that fecundity and seedling survival

were functions of tree density. Size-specific relative cone production from Platt et al.

(1988) was used to scale fecundity estimates. Fecundity values in the matrix represent









the size-specific per capital number of newly established seedlings that enter the

population. Because seedling establishment depends on a complex process that includes

cone production, seed viability, seed dispersal, and germination safe sites, these

parameters are difficult to estimate. In order to estimate fecundity parameters, we

assumed a maximum population growth rate (k) of 1.05. The maximum growth rate

would only occur in a population with no density limitation. Size-specific fecundity

values were estimated with an iterative procedure (Cropper and DiResta, 1999) using the

Bisection method. We assumed that fecundity scaled with size according to the ratio of

1:1:3:3:5:5:5 for the seven largest size classes. Model behavior is not sensitive to the

assumption of how fecundity scales with size (Cropper and DiResta, 1999). This is also

consistent with the relatively low elasticity (asymptotic growth rate sensitivity) for

fecundity values in the matrix model (see below).

Longleaf pine seedling growth (McGuire et al., 2001; Gagnon et al., 2004) and

distribution pattern in gaps (Platt et al., 1988; Brockway and Outcalt, 1998) indicate that

seedlings are competitively inhibited by larger trees. Although the functional relationship

between adult density and seedling establishment and survival cannot be quantified, long-

term simulations require density-dependence for realism. A population characterized by

our matrix model growing at k = 1.05 could increase from 200 trees per ha to 1,662 trees

per ha in 80 years. For comparison, the density of a virgin longleaf pine stand in

Louisiana in 1935 was reported to be between 191 and 233 trees per ha (Penfound and

Watkins, 1937).

In order to simulate seedling density-dependence, we assumed that the fecundity

and seedling survival matrix elements were functions of the number of trees greater than









10cm diameter. We also assumed a local population carrying capacity (K) of 500 trees

greater than 10cm diameter (100m radius plot). Calibrating the density function starts

with two sets of matrix parameters; one, the matrix for the maximum growth rate (X =

1.05, or other assumed values); the other being a matrix that represents stable growth at K

( = 1.00). Fecundity and seedling survival values are calculated as:

f(N) = Max e (-kd* N) (3-1)

where f(N) is the value of the matrix element at population size N, Max is the value

of the matrix element in the maximum growth matrix, and kd is a constant (e.g., a

fecundity value of 2.63 for trees greater than 50 cm at k = 1.05; Figure 3-1.). The value of

kd is determined by solving the equation for a value of f(N) equal to the k = 1 matrix

element and a value of N set at K, which is at a stable equilibrium in this model. This

density-dependent model has been used to model fisheries (Caswell, 2001) and a Poa

population in the UK (Case, 2000) based on measured fecundity and seedling survival at

different densities.

A second transition matrix model was created to represent the effects of fire

suppression on P. palustris establishment (Table 3-lb.). It has been determined that after

just 15 years without fire disturbance, P. palustris recruitment ceases (Hartnett and

Krofta, 1989) due to lack of bare mineral soil, low shade tolerance, and poor competitive

abilities of seedlings (Myers, 1990). Even just a few years of fire suppression can

dramatically reduce the amount of recruitment and vigorous herbaceous encroachment

may adversely affect growth and survival of young longleaf pines (Hartnett and Krofta,

1989; Platt and Rathbun, 1993). As such, we assumed that seedling number would

decline exponentially as fires are suppressed (Figure 3-2.). We were able to achieve this











by simply using the same transition probabilities as the first model and setting the size-

specific fecundity parameters to zero.


3


2.5

C
E 2
LU

- 1.5

o
CO
S1
>

0.5


0


0 200 400 600 800


1000


Population Density

Figure 3-1. Example of the function used to determine the values of the matrix elements
(trees >50cm dbh) assuming a maximum growth rate (X=1.05). The dotted
line represents the population at K, which is at a stable equilibrium.
















Table 3-1. Transition matrices developed from multiple data sources including the base transition matrix created by Platt et al. (1988).
The measurements refer to the diameter of the trees (at breast height for those applicable).
a.) Fire Transition Matrix


0 to 2
0.114
0.228


2 to 10
0
0.756
0.082


10 to 20
0.123

0.834
0.138


Year 0 (cm)
20 to 30 30 to 40
0.123 0.369


0.865
0.127


0.834
0.152


0.658 0.162 0.028 0.008 0.014


40 to 50
0.369





0.899
0.077


50 to 60
0.615







0.875
0.084


0.054
0.024 0.041 0.055


b.) No Fire Transition Matrix
Year 0 (cm)
Year 4 (cm) 0to2 2 to 10 10 to 20 20 to 30 30 to 40 40 to 50 50 to 60 60 to 70 70+
0to2 0.114 0 0 0 0 0 0 0 0
2 to 10 0.228 0.756
10 to 20 0.082 0.834
20 to 30 0.138 0.865
30 to 40 0.127 0.834
40 to 50 0.152 0.899
50 to 60 0.077 0.875
60 to 70 0.084 0.891
70 + 0.054 0.882
Deaths 0.658 0.162 0.028 0.008 0.014 0.024 0.041 0.055 0.118


Year 4 (cm)
0 to 2
2 to 10
10 to 20
20 to 30
30 to 40
40 to 50
50 to 60
60 to 70


Deaths


60 to 70 70 +
0.615 0.615


0.891


0.882
0.118












2500

2000

c 1500

S1000

Co 500 -

0
0 4 8 12 16
Model Year

Figure 3-2. Negative exponential function of seedling establishment after the onset of fire
suppression. Seedling numbers are below zero by year 16.

Results and Discussion

Error Analysis

A simulated population with one fire in each four year time step was compared to

real data from the Ichwauy Preserve collected on a five year time interval (1998-2003)

for an error analysis (Table 3-2.). Only the size-classes from 2cm on were used in this

analysis because the same individuals were measured in 1998 and 2003. This does not

account for additional individuals entering the population, which is otherwise the case in

the fire matrix model using fecundity and seedling transition probabilities. As a result,

there was a high percentage of error (118%) in the 2-10cm size class. The 60-70cm size

classes resulted in the most percentage error (303%) due to the small population and

associated changes in that size class. Otherwise, the % error ranged from -11 to 6%.

Determining P. palustris growth from diameter measurements within five years is

difficult because of the slow and variable growth characteristics of P. palustris. As the

habitat is burned, the pine's scaly plates protect the tree, but may stunt its growth and

allow the bark to slough off causing growth patterns to seem absent or even reversed

(e.g., lower dbh) during short periods (Myers, 1990; Putz, personal communication,









2005). As such, it is also relatively safe to compare P. palustris growth measurements

with a one year discrepancy.

Table 3-2. Error analysis comparing one simulated time step of the fire matrix model to
real five year growth data from the Ichwauy Preserve, FL.
Size-Class (cm) % Error Difference
2-10 118 474
10-20 -11 -105
20-30 6 55
30 40 -7 -88
40-50 4 32
50 60 -5 -7
60 70 303 6
70+ 0 0

Elasticity Analysis

An elasticity analysis of the transition matrix was performed to evaluate the relative

sensitivity of k to changes in the matrix elements (Caswell, 2001). Elasticity (e,j) is

calculated using the following equation:

a,, vwj
e *- (3-2)
A

Where a,j are elements of the stage population matrix, Wis the right eigenvector of
the dominant eigenvalue (A), Vis the left eigenvector, and is the scalar
product of the two eigenvectors.

The population matrix elements were classified as P, the probability of surviving

and remaining in the same size class; G, the probability of growing into the next largest

size class; and F, the per capital production of established seedlings. Overall, the

elasticity values (Figure 3-3.) demonstrated that P. palustris population growth was far

more sensitive to P (range: 0.003 0.17), while G and F were much less sensitive (range:

0.001 0.04). This suggests that accurate estimates for P are necessary for representing

the population growth rate (Doak et al., 1994). G values for smaller size class trees were

higher, signifying the relative importance of P. palustris growth of younger trees. As V's









increase, G and F begin to play a stronger role in the model as P weakens. In conclusion,

this elasticity analysis follows similar analyses of other natural populations by

summarizing the comparative effects of small changes in demographic rates (Doak et al.,

1994; Escalante et al., 2004).

Fire Frequency Analysis

Simulated fire frequency was varied to assess model behavior (Figure 3-4.; Figure

3-5.). With constant fires (1-4 year fire return interval) the population remained steady

and at K. Lower standard deviation and standard error of the total population with

frequent fires suggested a more stable population through time (Ott and Longnecker,

2001; Table 3-3.). As the fire return intervals became larger, spikes in species abundance

became more dramatic with each onset of fire but the population remained well below K

and continued to deplete with proliferation of fire suppression (Figure 3-4.). These

spikes in population increase are the direct result of seedling recruitment when fires are

brought back into the system (Figure 3-5.). It is not clear how well this simulates reality.

A healthy P. palustris population that has experienced fire suppression lasting 40 years

may not be able to re-establish itself from a newly introduced fire regime due to the large

fuel load that has most likely accumulated and other vegetative encroachment (Hartnett

and Krofta, 1989; Gilliam and Platt, 1999). However, the model does assume that fire

occurs 1-4 times in each time step. As such, some recruitment is likely especially if cone

production is in a mast year and conditions are favorable (Platt et al., 1988; Boyer, 1999).

The shorter fire return intervals then depict the more realistic behavior ofP. palustris in a

fire maintained community.







43


Lambda = 1.05


1 2 3 4 5 6 7 8 9


72


Lambda = 1.03
mF
DP
DG









2 3 4 5 6 7 8 9


Lambda = 1.00


1 2 3 4 5 6 7 8 9
Size Class
Figure 3-3. Elasticity of various lambda values (X). F: fecundity; P: probability of
remaining in the same size class; G: probability of growing to the next size
class. Size classes are in order from smallest to largest.










Table 3-3. Descriptive statistics of the various fire return intervals. Every four year fire
return interval represents 1-4 years of low-intensity burns in one matrix time
step.
Fire Return Intervals (yrs.) 4 8 12 20 40 No Fire
Mean 5,940 4,994 4,620 4,226 3,957 3,537
Standard Error 35 125 153 181 192 199
Standard Deviation 177 639 779 920 980 1,015
Sample Variance 31,447 407,997 606,372 847,299 960,623 1,029,834
Minimum 5,776 4,027 3,444 2,909 2,715 2,247
Maximum 6,756 6,756 6,756 6,756 6,756 6,756


8000

7000

6000

S5000

-1 4000 -"" ". : -
.0 -- 4yr. "",
< 3000 4 y-r.
-- 12yr. ---...
2000 ------- 40 yr.
1000 --- No Fire
1000

0
0 20 40 60 80 100
Model Year
Figure 3-4. Total number of individuals of P. palustris during various fire intervals. The
peaks in population growth are from seedlings establishing during years of
fire occurrence.

When the population was indefinitely subjected to fire suppression, the seedling

size class effectively disappears within four time steps (16 years) in a negatively

exponential fashion (Figure 3-5.), while the total population continued to decline until it

reached below one at year 536. With pines reaching maximum and mean ages of 400-

500 (Boyer, 1999) and 200-250 (Putz, personal communication, 2005), respectively, and

assuming no anthropogenic (e.g., logging) or catastrophic disturbances (e.g., hurricanes),

this estimate of the youngest tree reaching 536 years of age is unlikely. Although, the










population is under 100 by year 316, suggesting a struggling population in route to its

demise.


3000
3000 4 yr.
-- 12 yr.
2500...... 40 yr.
S--x-- No Fire
S2000

< 1500

i 1000 -

500 -


0 20 40 60 80 100
Model Year
Figure 3-5. Total number ofP. palustris seedlings during various fire intervals. The
peaks coincide with population growth rates in Figure 3-4.

An advantage to the frequent fire regime is the consistent uneven age structure of

the population. Infrequent fires that stimulate new P. palustris growth tend to create gaps

of missing age-classes in the population. This may influence cone production and seed

crops with the lack of necessary mature trees at particular periods in time. Gap sizes or

densities of adult trees may therefore fluctuate dramatically through time as a result.

Unnecessarily large gaps allow some P. palustris recruitment, but also allow for faster

growing competing vegetation to take over the landscape. If gaps are not large enough,

seedling establishment and success may be inadequate for a sustainable population

(Grace and Platt, 1995; Palik et al., 2003).

Future Model Development

Several additions to the transition matrix and more in-depth validation procedures

may be used to further develop this model. Since most P. palustris forests are managed









to some degree and are often in a restoration stage, there are many variations in

population structure and composition, understory and site conditions, and fuel traits,

which may influence population development. The model should include validation with

site specific management strategies, such as burning initiatives, seedbed preparation, and

thinning. These managed forests spanning from east Texas to the east coast of Virginia

(Frost, 1993) provide for an opportunity to use and compare this model with their P.

palustris population dynamics. More specifically, K may vary between sites and between

size and even age classes within a population. In our model, fecundity estimates were

assumed based on a hypothetical growth rate. Applying empirical fecundity values in a

stage based distribution may improve these estimates. Seedling parameters are difficult

to estimate due to their complex growth and survival patterns related to environmental

influences, such as fire and nutrient competition from adult P. palustris or encroaching

vegetation (Grace and Platt, 1995; Gagnon et al., 2004). Their growth rate in diameter

was especially difficult. For instance, Grace and Platt (1995) showed that many P.

palustris seedlings do not even reach 2cm in diameter in four years. As such, it may be

required to separate this size class into smaller size classes for particular P. palustris

populations.

This modeling technique may be further improved by acquiring more long-term

data on this slow growing and long-lived species. Unfortunately, funding and research

opportunities are often rather short-term and therefore may influence the model's utility.

For instance, the data sources used in this study did not have measurements recorded

beyond a six year period. Other studies (e.g., Veno, 1976) used past research and data









collected to continue understanding forest dynamics of a site on a larger temporal scale

(>20 years).

Implications of Using Multiple Data Sources

It would have been ideal to collect all of the data for model development and

testing within the Goethe State Forest. This was not practical because of the incredible

time, money, and effort required to collect and analyze the raw data needed to create such

a comprehensive demographic model on P. palustris. On the other hand, using multiple

data sources may produce large errors associated with environmental and anthropogenic

factors. P. palustris populations may vary by site condition, such as vegetative structure

and composition, landscape history, climatic conditions, soil and available nutrients, and

high genetic variation (Boyer, 1990). Management and research objectives impact these

populations as well, through variation in the fire regime, physical manipulation of the

landscape (e.g., planting, thinning), and having different data collection methods and

goals.

Conclusions

This model has the potential to become practical and useful for further

understanding of the population dynamics of this long-lived conifer. The essential

elements of a complete transition matrix have been estimated from reliable resources. A

K was implemented into the model as a spatial restriction or density-dependent function.

The unique survivorship characteristics ofP. palustris seedlings and fecundity estimates

were applied to the transition matrix to develop a more comprehensive modeling scheme.

Elasticity and model behavior analyses provided for model responsiveness to its inherent

elements and validation procedures resulted in relatively low error. Using the dual

transition matrix technique applied the behavioral response of this conifer to fire. The






48


abundant research done on P. palustris population dynamics (e.g., Boyer, 1990) coupled

with population modeling techniques provide opportunities to continue understanding

this complex conifer and its habitat.














CHAPTER 4
HYBRID APPROACH TO SPATIAL AND DEMOGRAPHIC MODELING OF
LONGLEAF PINES (Pinus Palustris)

Introduction

Longleaf Pine

P. palustris is a unique fire-dependent tree species of the southeastern United

States. Frequent low-intensity fires remove accumulated leaf litter, overgrown grasses,

and encroaching oaks exposing bare-mineral soil required for seed germination and

establishment (Croker, 1975; Boyer, 1990; Myers, 1990). Another distinctive feature

about this pine is its seedling's slow growing 'grass stage'. Here it develops a long thick

tap root (Myers, 1990), increases in diameter (Platt et al., 1988), and is protected from

fire through dense tuft of long, moisture rich needles. It may stay in this stage for 5-15

years. Within one growing season, P. palustris can grow over Im in height avoiding

lethal temperatures of fire near the ground surface while protecting the terminal bud. A

maturing tree develops fire-resilient scaly plates, which define it as the most fire-resistant

pine. Furthermore, the ecosystem has vegetation with pyrogenic traits that facilitate the

ignition and spread of fire (Fonda, 2001; Myers, 1990). Without frequent fire, the stable

P. palustris dominated forest disappears due to failure of seedling establishment and

incursion of hardwood upland species (Myers, 1990).

Sandhill, a P. palustris dominated ecosystem, was once one of the largest forest

types in the southeastern coastal plain of the US, but has diminished by over 98% due to

harvesting and fire suppression (Myers, 1990). The highly dispersed pine and low









understory provided for an easily traveled and nicely shaded route for early European

settlers (Hall, 1829). The ecosystem became fragmented by clear-cutting for timber,

development, introduction of livestock, and agricultural practices. Many of these

fragments have suffered from fire suppression resulting in encroachment by more fast

growing, shade tolerant tree species (Frost, 1993; Rasser, 2003).

Restoration and preservation of P. palustris dominated ecosystems has become a

well known practice in the southeast which contributes to recreation, aesthetics, reducing

the onset of pest infestation, and conservation for endangered and often endemic species

that thrive in this habitat (Boyer, 1993; Neel, 1993). Land managers plan for frequent

low-intensity fires (-3-5 years) that keep encroaching hardwoods in check (Hartnett and

Krofta, 1989; Platt and Rathbun, 1993). These fires also create an environment that

supports P. palustris recruitment and open savannah like parks for recreation including

hunting and easily traveled trails (Myers, 1990). Several endemic species thrive in these

habitats, including the Sherman's fox squirrel (Sciurus niger), red-cockaded woodpecker

(Picoides borealis), brown-headed nuthatch (Sittapusilla), Bachman's sparrow

(Aimophila aestivalis), and the gopher tortoise (Gopherus polyphemus) (Engstrom, 1993).

Hybrid Modeling

Creating hybrid ecosystem process models provide an opportunity to employ

important aspects of two or more models to create a new more robust model. Hybrid

models can be constructed with components that include simulation models, statistical

models, and empirical spatial models [Geographic Information System (GIS) models].

Robinson and Ek (2003) utilized at least six various empirical and physiological based

modules to develop an individual-tree growth and dynamics model. Logofet and

Korotkov (2002) incorporated spatially explicit forest inventory data to apply a









geographical component within a Markov chain model measuring natural succession of

forest types. Lamar and McGraw (2005) compared two matrix population models of

eastern hemlock (Tsuga canadensis L.), one created from field data and one created from

remote sensing techniques using low elevation aerial photography. Each of these studies

used the inherent strengths from each model to establish a more viable model of forest

dynamics. In this study, density-dependent matrix population models use particular

outputs from the spatially explicit LANDIS (LANdscape DIsturbance and Succession;

Mladenoff et al., 1996) model to assess P. palustris population dynamics.

Population matrix models are an ideal starting point for analyzing tree population

dynamics (Escalante et al., 2004; Zuidema and Zagt, 2000). They have become powerful

tools used in the natural sciences and have gone on to be used in hybrid model research

(Logofet and Korotkov, 2002; Lamar and McGraw, 2005). Population matrix models

often use size-based classes and transition probabilities developed from individual tree

measurements over time. They are mechanistically designed to simulate survival,

growth, death, and fecundity in user defined time steps assuming that the environment in

which their data were collected is constant over time (Caswell, 2001). The labor required

to census and map the data needed to estimate these probabilities is intense (Silvertown et

al., 1996). Estimation of size-specific fecundity is especially difficult because the process

of seedling establishment depends on seed production, seed viability, dispersal, and site

quality on multiple scales.

LANDIS on the other hand, models landscapes over large spatial and temporal

scales requiring relatively coarse input data of dominant tree species and fire

characteristics of the land (He and Mladenoff, 1999). The model is spatially interactive,









implementing variables such as tree species seeding characteristics and cohort-based

growth capabilities, as well as fire ignition, spread, and fuel traits. These features enable

the user to gain insight into the overall dynamics of the forest system visually and

quantitatively over decades to several centuries (Mladenoff and He, 1999).

Objectives

The purpose of this research was to utilize the linked outputs of a spatially explicit

forest succession model (LANDIS), and a density dependent transition matrix model to

examine P. palustris population dynamics over time. Temporal fire data from LANDIS

was used to select between two transition matrices: (I) A P. palustris population

transition matrix for a fire-maintained ecosystem and; (II) A P. palustris population

transition matrix during fire suppression. The ultimate goal was to tie P. palustris

population numbers to specific management units or independent populations.

Methods

Study Area

The Watermelon Pond Unit (WPU) of the Goethe State Forest is a 1938-ha tract

located in Levy and Alachua counties of Florida, USA (latititude 290 32' 42 ", longitude

82 o 36' 20"; Figure 4-1.). It has a checker-board structure that makes land management

difficult, especially for prescribed burning. Agricultural land owners, state and county

roads, and two small towns divide the forest, further complicating management. Before

European settlement, the land was a homogeneous landscape of the sandhill community

type dominated by P. palustris. During the last century, the area was fragmented by

anthropogenic effects, specifically phosphate mining and non-sustainable timber

harvesting. Many of these areas were unmanaged causing dramatic successional changes

in this ecosystem with few P. palustris trees left today (Rasser, 2003). Now, one of the










main management goals is the immediate implementation of prescribed burning

initiatives for restoration and preservation of the remaining sandhill areas (Hardin, 2004).






Archer, FL

















Legend
i Goethe State Forest Boundary
State of Florida
Alachua County
SLevy County

Figure 4-1. Map of the study site: Watermelon Pond Unit of the Goethe State Forest,
Florida, USA.

Outside Data Sources

Data were collected on longleaf pine demographics from several outside sources

for the development of the fire-maintained transition matrix model used in this study. A

population transition matrix built by Platt et al. (1988) for longleaf pines (Pinus palustris)

at the Wade Tract, Thomas County, Georgia was used as a base for our model. Longleaf

pine seedling and mature tree growth data recorded by the Joseph W. Jones Ecological

Center at the Ichwauy Preserve, Baker County, Georgia were used for assessing density

parameters (Jones et al., 2003; Palik et al., 2003). Longleaf pine seedling survivorship









data were extracted from the Jones Center data and a seedbed preparation study

performed at the Escambia Experimental Forest in southwest Alabama (Croker, 1975).

Simulated fire suppression effects on longleaf pine recruitment were estimated from a

study in Chinsegut Hill McCarty-Woods in northern Glades County, Florida (Hartnett

and Krofta, 1989).

Hybrid Modeling Techniques: The Spatial and Demographic Components

Modeling P. palustris response to fire requires addressing processes at both the

landscape scale and at the population scale. It is certainly possible to construct a single

model that simulates the landscape fire regime as well as seedling establishment and tree

growth and mortality. A hybrid approach, however, has the advantage of using models

that have already been developed (LANDIS, P. palustris matrix population model),

which are operating at different time steps, to meet the model objective with significantly

less model development effort. The principal problem of the hybrid model approach is

passing information between the individual models.

LANDIS outputs, and other data sources, were used to estimate the: a) initial P.

palustris size-classes and fire distributions; b) P. palustris carrying capacity; and c) fire

disturbance frequencies, all within individual P. palustris populations of the WPU.

These populations were chosen because of their size and relative fire disturbance

frequency and distribution simulated within LANDIS (Figure 4-2.). A small isolated

population in the northern territory (270ha) of the WPU was used as well as the much

larger southeastern section (WPU East; 792ha). For comparison, the entire population

(1874ha) at the WPU was also simulated assuming each fire disturbance in LANDIS

affects the entire site's P. palustris population. From a graphical analysis of age and size

relationships in Platt et al. (1998), the initial age-class distributions created for LANDIS









(see Chapter 1) were used to determine the size-class distributions found in each of the

three P. palustris populations. The number of individuals within each population was

calculated from: a) the known area ofP. palustris sandhill habitat within each section; b)

the size-class distribution within each section; and c) the P. palustris population density

of the Ichwauy Preserve (basal area of 11.58 m2/ha). This population was essentially at

an ideal maximum density or carrying capacity (K) according to Boyer's (1999) estimates

for maintaining a healthy and naturally regenerating P. palustris population. We

assumed that fecundity and seedling survival were functions of tree density. Size-

specific relative cone production from Platt et al. (1988) was used to scale fecundity

estimates (see Chapter 2). Fire frequency distributions were purely based on the

occurrence of fire within each of the three populations for each model created in

LANDIS. The Frequent Fire Model (FFM) simulated fires every one to two LANDIS

time steps (10-20 years) and the Infrequent Fire Model (IFM) simulated a more fire

suppressed ecosystem with a fire return interval of 1-7 LANDIS time steps (10-70 years).

The size, intensity, and frequency of bums varied across the landscape which were

influenced by land characteristics and tree species inter- and intra- specific behavior.

Fire was simulated with the matrix population model through application of two

size class transition matrices representing population dynamics with fire or with no fire

during each ten year time step in LANDIS. The development and analysis of the two

matrices are described in detail elsewhere (see Chapter 2). Essentially, one four year

transition matrix represents a P. palustris population subjected to a regular low-intensity

burning regime (1-4 years). The other four year transition matrix represents a P. palustris

population subjected to fire suppression with fecundity set to zero. When the fire










transition matrix is chosen, fecundity and seedling survival rates are density-dependent,

with equilibrium at the carrying capacity (K). Density-dependence is not applied in the

fire suppression transition matrix because of the rapid decline in recruitment in the

absence of frequent burning (Hartnett and Kofta, 1989).

















0 05 1 2 Kilometers


"Small" Isolated
Population




I I I I 1 I I I
0 05 1 2 Kilometers


Figure 4-2. The three areas within the study site with P. palustris populations used in the
hybrid model analysis: 1) The entire WPU (1874ha); 2) The southeastern
portion of the WPU, namely WPU East (792ha); 3) A small isolated
population in the northern section of the WPU (270ha). Refer to Figure 4-1.
for the layout of the WPU.

Two application methods were proposed to handle the time step discrepancy

between LANDIS and the matrix model. The hybrid matrix model runs an alternate of 2

to 3 times per LANDIS time step, so that five 4-year time steps equal two LANDIS time

steps (Figure 4-3.). When a fire occurs in LANDIS, it only disturbs an area once and

therefore, the fire matrix model should only run once per LANDIS time step. The









positioning of whether the fire matrix runs first, second, or third, is randomly selected.

For comparison, a second matrix model was designed to employ the fire transition matrix

2 or 3 times when a fire occurs in LANDIS.






0 10 20
Time Steps
Time (yrs)
LANDIS

Matrix Model

Overlap

Figure 4-3. Diagram of the time step discrepancy between the matrix population model
(4-year) and LANDIS (10-year).

Results and Discussion

Fire Effects on Individual Populations

The overall simulated longleaf pine population abundance changes were a direct

result of seedling recruitment fluctuations caused by fire disturbance and fire suppression

(Figure 4-4.) As fires occur on the landscape, the accumulated fuel (e.g., leaf litter,

herbaceous growth) burns to expose bare mineral soil for the establishment and

germination of P. palustris seedlings, assuming good seed production and site conditions

(Gilliam and Platt, 1999). Without fires, seedling recruitment decreases exponentially

until they cease to establish 15 years (- 4 times steps) from the onset of fire suppression

(Hartnett and Krofta, 1989).











50000
-Total
Seedlings
40000


30000


20000


10000


0
0 100 200 300 400 500
Model Year
Figure 4-4. Total abundance changes within the small population in relation to the
seedling abundance changes. The spikes relate directly between total
population numbers and the recruitment of seedlings.

The two application methods for using the fire transition matrix (e.g., 1 or 2/3 times

per LANDIS time step) were compared (Figure 4-5.). Having the fire transition matrix

run 2 to 3 times more results in a clear percent increase in mean (41-59%) and abundance

at year 500 (74-84%) in all three populations. The entire WPU and WPU East had higher

variation (31% and 28%, respectively) in population numbers using the transition matrix

2/3 times per LANDIS time step, while the isolated portion of the WPU did not show

much difference (0.03%). In the isolated population, there was a time lag that allowed

for fire suppression to take effect for both application methods. In the two larger

populations, the fires were frequent enough to allow the 2/3 fire method to keep a steady

increase or maintained P. palustris population (e.g., smooth curve).













Entire WPU


1000000


800000


600000


400000


200000


0







100000


80000


S60000


S40000


20000


0


-- 1 Fire
- 2/3 Fires


0 40 80 120 160 200



WPU East


0 40 80 120 160 200



Isolated Population


50000

40000


30000


20000


10000


0


0 40 80 120 160 200
Model Year


Figure 4-5. Total population changes of the matrix population model assuming fires
occur either once or every iteration (2 to 3 times) within a LANDIS time step.
Scale may be different between graphs.









P. palustris population dynamics were compared over 500 years assuming fires

occurred in all transition matrices (all 2/3 times steps) during fire disturbance in LANDIS

(Figure 4-6.). The fire frequencies of the FFM resulted in relatively stable numbers in the

two larger populations, while the isolated population continued to decline in both the

FFM and the IFM. The 62-80% increase in fire frequency in all populations resulted in

90-96% larger numbers at year 500 using the FFM. The high fluctuations in the WPU

East numbers was due to the large number of fires from the FFM that were spaced far

enough apart for multiple reactions to fire suppression and seedling recruitment following

fires.

Comparing P. palustris simulation outputs of LANDIS and the matrix model was

not possible due to their incompatible output formats. LANDIS outputs the presence of a

species in particular locations (cells) across a landscape and through time, lacking

population numbers. The matrix model outputs a species population in size-classes

through time, lacking a spatial component. However, utilizing the output fire frequencies

of LANDIS in the matrix model to provide P. palustris numbers in individual populations

at the WPU offered an opportunity for the missing elements of one model to be supplied

by the other.

Implications of Hybrid Approach

This modeling approach demonstrates the ability to combine strengths of

ecosystem models and compensate for their weaknesses (Kimmins et al., 1999). Here we

utilized the P. palustris transition probabilities derived from field work done by Platt et

al. (1988) and others (Croker, 1975; Hartnett and Krofta, 1989; Jones et al., 2003, Palik et

al., 2003) and added a density dependent function (see Chapter 2). A spatial component









was built-in to specifically look at individual populations within the WPU as well as the

fire frequencies and distributions created in LANDIS.

Although this was an opportunity to use the elements of two models at two

different scales, each model simulates species dynamics quite differently, especially

related to fecundity and population changes. LANDIS simulates fecundity from tree age-

cohorts reaching maturity and dispersing 'seeds' into adjacent cells (He et al., 2003),

while the fire matrix model utilizes fecundity transition probabilities (Caswell, 2001).

LANDIS simulates tree growth dynamics through a presence/absence approach with

interacting tree age-cohorts and influential disturbances and site conditions (He et al.,

2003), while matrix models use size-based transition probabilities of one population of

species assuming no competition, disturbance, or site variability (Caswell, 2001).

Future Model Development

The matrix population model approach is not spatially explicit, which may be a

major limitation for longleaf pine simulation. Seedling establishment is clearly sensitive

to overstory density (Boyer, 1993; Grace and Platt, 1995; McGuire et al,. 2001). The

hybrid modeling approach, using LANDIS to provide some of the missing processes, is

not a complete solution. Although long-term data is often difficult to obtain, utilizing

GIS data may provide the long-term information needed for calibration of such

demographic models (Logogfet and Korotkov, 2002). Acquiring and comparing

demographic data collected from field-work and remotely sensed imagery (Lamar and

McGraw, 2005) may be another calibration and validation option. Obtaining site-specific

demographic data is an ideal approach to enhancing this modeling scheme. Since the

data sources used to create these models are from a variety of projects and sites,

demographic data from the WPU would be critical for model improvement. Other details









that are essential to the integration of these two types of models include understanding

size and age relationships of trees such as P. palustris (Platt et al., 1988) and measuring

fecundity and seedling growth in sites with different fire histories.

The time step discrepancy may be solved by using the next version of LANDIS

(LANDIS II), which incorporates one to several year time steps. This may enhance the

realism ofP. palustris behavior to fire and its competitive tree species in LANDIS while

providing analogous time steps between the matrix models and LANDIS. As a healthy P.

palustris population requires frequent (3-5 years) low-intensity fires (Hardin, 2004), the

fire return interval may decrease dramatically using LANDIS II, especially in the many

isolated populations of P. palustris within the WPU.

Conclusions

The hybrid techniques used in this study were successful. The temporal fire

frequency and spatial data from LANDIS provided for inputs to simulate the P. palustris

matrix models using various fire regimes. Individual populations, or management units

within the WPU, were simulated and analyzed. The outputs of these two models were

not compared because of their incompatible output formats [e.g., presence of species

(LANDIS) vs. population numbers (matrix model)]. Using LANDIS II, spatial

technologies, such as GIS and remote sensing, and more site-specific long-term ground

data on growth and fecundity ofP. palustris may improve justification procedures. The

abundant research done on P. palustris population dynamics (e.g., Boyer, 1990) together

with demographic and hybrid modeling techniques provide opportunities to continue

understanding this complex conifer and its habitat.













1000000


Entire WPU


--FFM
- IFM


800000


600000


400000


200000


0
0 100 200 300 400 500



WPU East

100000 / A
80000 ,. A A


60000


40000


20000


50000


40000


30000


20000


10000


0


0 100 200 300 400 500



Isolated Population


0 100 200 300
Model Year


400 500


Figure 4-6. Total population changes of the matrix population model assuming fires
occur every iteration (2 to 3 times) within a LANDIS time step. Scale is
different between graphs. FFM: Frequent Fire Model; IFM: Infrequent Fire
Model.














CHAPTER 5
CONCLUSIONS

General Conclusions

Results, conclusions, implications, and overall applicability for each portion of this

research are discussed here. Conclusions from the spatial modeling portion of the

research include:

Two fire regimes were created in relation to P. palustris and its competitive

hardwoods in an ideal fire-maintained habitat and a fire suppressed one.

LANDIS demonstrated certain characteristics of P. palustris dynamics well:

recruitment of young after fire, seeding distance, and fire tolerance.

Results were not entirely realistic: Fire frequency was lower in the FFM than

what is required to maintain a viable P. palustris ecosystem.

Q. hemisphaerica and Q. geminata were accurately depicted, while dynamics of

Q. laevis lacked realism.

The demographic modeling led to the following conclusions:

The essential elements of a complete P. palustris transition matrix were

estimated from data and literature.

Additions to the P. palustris matrix population model developed by Platt et al.

(1988) included:

o A density-dependent function

o Survivorship and growth rates of P. palustris seedlings (<2cm dbh)

o Fecundity estimates









Model sensitivity and dynamics were analyzed through elasticity and scenario

analyses.

Validation procedures resulted in relatively low error.

A dual transition matrix technique (with and without fire) applied the

behavioral response of this conifer to fire.

The hybrid modeling approach, consisting of the combined application of LANDIS

and the demographic model, led to the following conclusions:

The temporal fire frequency and spatial data from LANDIS provided inputs to

simulate the P. palustris matrix models using various fire regimes. General

responses to fire suppression were realistic.

Individual populations within the WPU were simulated and analyzed.

The incompatible output formats of the two models [e.g., presence of species

(LANDIS) vs. population numbers (matrix model)] did not allow for model

comparison.

Limitations, Implications, and Future Work

Spatial Modeling

There are several limitations with the spatial model developed using LANDIS. Q.

laevis tolerates fires well and re-sprouts vigorously following fire (McGinty and Christy,

1977), but was the least abundant species on the landscape during the model runs. The

short fire return interval of 3-5 years (Hardin, 2004) required to maintain and/or restore

sandhill communities could not be modeled with LANDIS 3.7 because of the 10 year

time step. Furthermore, the fragmented structure of the WPU restricted fire spread across

the landscape.









With some fine-tuning of parameters and implementation of smaller time steps, this

model may dramatically improve and be applicable for management to visualize P.

palustris population dynamics through time. Species behavior for Q. laevis could be

improved and additional calibration may fine-tune the model. LANDIS II may provide a

superior framework to simulate the fire regime desired for this study, although the

fragmented structure of the WPU may still adversely affect the model's fire spread

processes. Simulating each section separately may induce more frequent fires on each

individual population of P. palustris. Future work may also include adding more trees

species found at the WPU to LANDIS, obtaining more historical data, collecting tree

cores for ground referenced species age distributions, and applying long-term validation

procedures to support the quality and viability of the model.

Demographic Modeling

There are several limitations to the development of the P. palustris demographic

model. The model should include site specific validation where management strategies

vary. More specifically, density-dependent dynamics may vary between sites and

between size and even age classes within a population. In our model, fecundity estimates

were assumed based on a hypothetical growth rate. Applying empirical fecundity values

in a stage based distribution may improve these estimates. Obtaining data on seedling

growth rates in diameter was difficult and seemed to vary highly within and between sites

(Grace and Platt, 1995). It may be required to separate this size class into smaller size

classes for particular P. palustris populations.

This modeling technique may be further improved by acquiring more long-term

data on this slow growing and long-lived species and collecting all of the data for model

development and testing within the Goethe State Forest would have been ideal. This









deemed impractical because of the incredible time, money, and effort required to collect

and analyze the raw data needed to create such a comprehensive demographic model on

P. palustris.

Hybrid Modeling Approach

There are several limitations to the hybrid modeling approach. Although this

method demonstrates the ability to combine strengths of ecosystem models, each model

simulates species dynamics quite differently, especially related to fecundity and

population changes. Furthermore, using LANDIS to provide some of the missing

processes is not a complete solution. The time step discrepancy may be solved by using

LANDIS II which incorporates one to several year time steps. Using LANDIS II, spatial

technologies, such as GIS and remote sensing, and more site-specific long-term ground

data on growth and fecundity ofP. palustris may improve justification procedures.

Broad Implications of Research

Forest managers should understand the long-term implications of their

management plans on forests as a whole. Using a broad-scaled model, such as LANDIS,

provides managers with a glimpse of how their forests are changing over time in respect

to tree behaviors and interactions, site conditions, and disturbances without exhaustive

input parameters. They may easily change parameters, such as fire disturbances or

species attributes in LANDIS using its graphical user interface. To study the finer details

of a forest, a comprehensive demographic model aids to assess one species over time

within specific populations of a forest. Although these inputs may be difficult to collect,

using outside data sources on a commonly researched species may provide for the

information needed.









Much of this research and data collected may be used for studies in other natural

areas. Although the spatial inputs and site conditions used in LANDIS were specifically

designed for the WPU, the attributes for each tree species may be used elsewhere. As the

four tree species and their habitats are very common in north-central Florida forests, the

methodology may be utilized to study landscape level processes in other managed

landscapes. As mentioned earlier, because the P. palustris demographic model used

multiple data sources, the model may be tested in various areas in the southeast to assess

its viability. The site conditions and management regimes within these areas may refine

the demographic model for use in a specific area.

















APPENDIX
LANDIS PARAMETER FILES

Main Parameter File


# Base parameter file #
# This file created by Landis Interface version 1.01 #
myspecies.dat # species life history attribute file #
myLandTypes.dat # Attributes describing each land type class #
lm_20m_2nd_gis.gis # 8-bit Erdas GIS map of land type or ecoregion #
fcm_20m_gis.gis # 8-bit Erdas GIS maps of species and age cohort classes #
myMapAttrib. dat # Attributes describing each map class #
myMapIndex.dat # Map names for output #
myAgeIndex.dat # Age map names for output #
C:\LANDIS\Project\OutputFFM\ # Path for LANDIS output #
myDisturb.dat # Disturbance input file #
myDefaultplt.dat # color information file #
myOutputFreq.dat # output frequency file #
50 # Number of iterations (10 year time steps) #
1 # Random seed: 0 generate randomly; other fixed seed #
20 # spatial resolution in meters #
DISPERSAL # seed dispersal routine #
0 # wind switch: 0-none; 1-standard; 2-mean; 3-strong; 4-light #
1 # Fire switch: 0-none; 1-standard; 2-mean; 3-strong; 4-light #
0 # old harvesting switch currently disabled #
0 # Harvesting: 0-off 1-on #
0 # Adjacency: 0-not considered 1-considered #
0 # Number of decades to consider adjacency #
1.0 # Harvest proportion #
myHarvEvents.dat # Harvest prescription #
fcm_20m_gis.i9s # 16-bit Erdas GIS map for stand boundaries #
lm_0m_2nd_gis.gis # 16-bit Erdas GIS map for management areas #
stands.log # Harvest history log summarized by stands #
ma.log # Harvest history log summarized by management area #



Species Attribute File


# LANDIS species attribute input file::version 3.2 #
# This file created by Landis Interface version 1.01 #
# name longevity maturity shade fire eff_seed max_seed veg_prop sprout reclass #
PIPA 200 20 1 5 30 50 0.0 0 0.5
QUGE 130 20 4 1 50 70 0.5 5 0.5
QULA 100 20 3 4 50 70 0.5 5 0.5
QUHE 100 20 4 1 70 90 0.5 5 0.5











Landtype Attribute File: FFM



" Landtype file c:\LANDIs\ProC ect\FreqFire\myLandTypes.dat #
# This file created by Landis Interface version 1.01 #
empty 0 0 0.0 10 0 0
0.0
0.0
0.0
0.0
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
CurrentPIPAhabitat 40 5 1.0 5 0 100


50 70 90 1 2 3 4 4 0
ood 40 80 0.02 80 0 1C




50 70 90 1 2 3 3 3 0
40 50 0.02 50 0 100




50 70 90 2 4 5 5 5 0
0 0.0 10 0 0


0 0 0 00 0 0 0 0
I0




0 0 0 0 0 0 0 0 0





0 0 0 0 0 0 0 0 0
000000000


1.0
1.0
1.0
1.0
10 30
Hardwc
1.0
1.0
1.0
1.0
10 30
scrub
1.0
1.0
1.0
1.0
10 30
Wet 0
0.0
0.0
0.0
0.0
0 0 0


0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0











Landtype Attribute File: IFM



# Landtype file c:\LANDIS\Pro ect\InFreqFire\myLandTypes. dat #
# This file created by Landis Interface version 1.01 #
empty 0 0 0.0 10 0 0
0.0
0.0
0.0
0.0
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
CurrentPIPAhabitat 40 45 0.02 10 0 22


50 70 90 1 2 3 4 4
od 40 80 0.02 20 0




50 70 90 1 2 3 3 3
40 50 0.02 10 0 22




50 70 90 2 4 5 5 5
0 0.0 10 0 0


0 0 0 0 0 0 0 0 0 0
22




0 0 0 0 0 0 0 0 0 0





0 0 0 0 0 0 0 0 0 0


0.5
1.0
1.0
1.0
10 30
Hardwo
0. 5
1.0
1.0
1.0
10 30
scrub
0.5
1.0
1.0
1.0
10 30
wet 0
0.0
0.0
0.0
0.0
00 0


0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0









72




Map Attribute File




# Map attribute class file c:\LANDIS\Project\FreqFire\myMapAttrib.dat #
# This file created by Landis Interface version 1.01 #
# Map attribute class 0 #
0 00000000000000000000
0 0000000000000
0 0000000000
0 0000000000
# Map attribute class 1 I
0 00000000000000000000
0 0000000000000
0 0000000000
0 0000000000
# Map attribute class 2 #
0 11100000000000000000
0 0000000000000
1 1110000000
0 0000000000
# Map attribute class 3 #
0 11111100000000000000
0 0000000000000
1 1111110000
0 0000000000
# Map attribute class 4 #
0 00000000000000000000
0 0000000000000
1 1110000000
0 0000000000
# Map attribute class 5 #
0 00000000000000000000
0 0000000000000
1 1111110000
0 0000000000
# Map attribute class 6 #
0 11111100000000000000
1 1111110000000
1 1111110000
1 1000000000
# Map attribute class 7 #
0 11100000000000000000
1 1110000000000
1 1110000000
1 1110000000
# Map attribute class 8 #
0 11100000000000000000
0 0000000000000
1 1110000000
1 1110000000
# Map attribute class 9 #
0 00000000000000000000
1 1111110000000
1 1000000000
1 1000000000
# Map attribute class 10 #
0 00000000000000000000
1 1111110000000
0 0000000000
1 1111110000
# Map attribute class 11 #
0 00000000000000000000
1 1110000000000
0 0000000000
1 1110000000
# Map attribute class 12 #
0 00000000000000000000
0 0000000000000
0 0000000000
1 1111110000
# Map attribute class 13 #
0 00000000000000000000
0 0000000000000
0 0000000000
1 1100000000
# Map attribute class 14 #
0 00000000000000000000
0 0000000000000
0 0000000000
1 1111000000
# Map attribute class 15 #
0 00000000000000000000
1 1110000000000
1 1110000000
0 0000000000
# Map attribute class 16 #
0 00000000000000000000
1 1111110000000
1 1111110000
0 0000000000







73



Map Index File




# Map index file C:\LANDIs\Project\FreqFire\myMapIndex.dat #
# This file created by Landis Interface version 1.01 #
DOM PIPA QUGE QULA QUHE sandhill Hammock SH-HW




Age Index File


# Age reclass file C:\LANDIS\Project\FreqFire\myAgelndex.dat #
# This file created by Landis interface version 1.01 #
'pipa yquge yqula yquhe





Disturbance File: FFM


# Disturbance File C:\LANDIS\Project\FreqFire\myDisturb.dat #
# This file created by Landis Interface version 1.01 #
[WIND] # default values #
[FIRE]
600000
2600000
1200000
40
3
7
10
10
fire
firefin
fire.log
[HARVEST] # default values #



Disturbance File: IFM


# Disturbance File c:\LANDIS\Project\InFreqFire\myDisturb.dat #
# This file created by Landis Interface version 1.01 #
[WIND] # default values #
[FIRE]
883000
1272000
1060000
80
3
7
10
10
fire
firefin
fire.log
[HARVEST] # default values #
















LIST OF REFERENCES


Andrews, E.F., 1917. Agency of fire in propagation of longleaf pines. Botanical Gazette.
64, 497-508.

Andrews, P.L., 1986. BEHAVE: Fire behavior prediction and fuel modeling system -
BURN subsystem, part 1. In: Mladenoff, D.J., Baker, W.L. (Eds.), Spatial
Modeling of Forest Landscape Change: Approaches and Applications. Cambridge
University Press, Cambridge, UK, pp. 172-174.

Baker, W.L., Egbert, S.L., Frazier, G.F., 1991. A spatial model for studying the effects of
climatic change on the structure of landscapes subject to large disturbances.
Ecological Modelling. 56, 109-125.

Berg, E.E., Hamrick, J.L., 1994. Spatial and genetic structure of two sandhills oaks:
Quercus laevis and Quercus margaretta (Fagaceae). American Journal of Botany.
81, 7-14.

Botkin, D.B., 1993. Forest Dynamics: An Ecological Model. Oxford University Press,
New York, NY, USA, 309 pp.

Boyer, W.D., 1963. Development of Longleaf Pine Seedlings under Parent Trees. USDA
Forest Service Research Paper SO-4. Southern Forest Experiment Station, New
Orleans, LA, USA, 7 pp.

Boyer, W.D., 1990. Pinuspalustris Mill. longleaf pine. In: Burns, R, M., Honkala, B.H.
(technical coordinators), Silvics of North America, Volume 1, Conifers. USDA
Forest Service Agricultural Handbook 654. Washington D.C., USA, pp. 405-412.

Boyer, W.D., 1993. Regenerating longleaf pine with natural seeding. In: Herman, S.M.
(Ed.), The Longleaf Pine Ecosystem: Ecology, Restoration, and Management.
Proceedings, 18th of the Tall Timbers Fire Ecology Conference. Tall Timbers, Inc.,
Tallahassee, FL, USA, pp. 299-310.

Boyer, W.D., 1999. Longleaf pine: natural regeneration and management. Alabama
Treasured Forests. 18, 7-9.

Brockway, D.G., Outcalt, K.W., 1998. Gap-phase regeneration in longleaf pine wiregrass
ecosystems. Forest Ecology and Management. 160, 125-139.

Case, T.D., 2000. An Illustrated Guide to Theoretical Ecology. Oxford press, Oxford,
NY, USA, 449 pp.









Caswell, H., 2001. Matrix Population Models: Construction, Analysis, and Interpretation.
2nd Ed. Sinauer Associates, Inc., Sunderland, MA, USA, 722 pp.

Croker, T.C., 1975. Seedbed Preparation Aids Natural Regeneration of Longleaf Pine.
USDA Forest Service Research Paper SO-112. Southern Forest Experiment
Station, Ashville, NC, USA, 6 pp.

Cropper, Jr. W.P., DiResta, D., 1999. Simulation of a commercial sponge population
in Biscayne Bay, Florida: recovery following Hurricane Andrew and
management implications. Ecological Modelling. 118, 1-15.

Doak, D., Kareiva, P., Klepetka, B., 1994. Modeling population viability for the desert
tortoise in the Western Mojave Desert. Ecological Applications. 4, 446-460.

Duncan, W.H., Duncan, M.B., 1988. Trees of Southeastern United States. University of
Georgia Press, Athens, GA, USA, 322 pp.

Engstrom, R.T., 1993. Characteristic mammals and birds of longleaf pine forests. In:
Herman, S.M. (Ed.), The Longleaf Pine Ecosystem: Ecology, Restoration, and
Management. Proceedings, 18th of the Tall Timbers Fire Ecology Conference. Tall
Timbers, Inc., Tallahassee, FL, USA, pp. 127-138.

Escalante, S., Montana, C., Orellana, R., 2004. Demography and potential extractive use
of the liana palm, Desmoncus oi th, ,i1itih,, Martius (Arecaceae), in southern
Quintana Roo, Mexico. Forest Ecology and Management. 187, 3-18.

Finney, M.A., 1998. FARSITE: Fire Area Simulator- Model Development and
Evaluation. USDA Forest Service Research Paper RMRS-RP-4. Rocky Mountain
Research Station, Ogden, UT, USA, 47 pp.

Fonda, R. W., 2001. Burning characteristics of needles from eight pine species. Forest
Science. 47, 390-396.

Franklin, J., Syphard, A.D., Mladenoff, D.J., He, H.S., Simons, D.K., Martin, R.P.,
Deutschman, D., O'Leary, J.F., 2001. Simulating the effects of different fire
regimes on plant functional groups in Southern California. Ecological Modelling.
142, 261-283.

Frost, C.C., 1993. Four centuries of changing landscape patterns in the longleaf pine
ecosystem. In: Herman, S.M. (Ed.), The Longleaf Pine Ecosystem: Ecology,
Restoration, and Management. Proceedings, 18th of the Tall Timbers Fire Ecology
Conference. Tall Timbers, Inc., Tallahassee, FL, USA, pp. 17-44.

Gagnon, J. L., Jokela, E.J., Moser, W.K., Huber, D. A., 2004. Characteristics of gaps and
natural regeneration in mature longleaf pine flatwoods ecosystems. Forest Ecology
and Management. 187, 373-380.









Gilliam, F.S., Platt, W.J., 1999. Effects of long-term fire exclusion on tree species
composition and stand structure in an old-growth Pinuspalustris (longleaf pine)
forest. Plant Ecology. 140, 15-26.

Grace, S.L., Platt, W.J., 1995. Effects of adult tree density and fire on the demography of
pregrass stage juvenile longleaf pine (Pinus palustris Mill.). Journal of Ecology.
83, 75-86.

Greenberg, C.H., Simons, R.W., 1999. Age, composition, and stand structure of old-
growth oak sites in the Florida High Pine Landscape: implications for ecosystem
management and restoration. Natural Areas Journal. 19, 30-40.

Guerin, D.N., 1993. Oak dome clonal structure and fire ecology in a Florida longleaf pine
dominated community. Bulletin of the Torrey Botanical Club. 120, 107-144.

Gustafson, E.J., Zollner, P.A., Sturtevant, B.R., He, H.S., Mladenoff, D.J., 2004.
Influence of forest management alternatives and land type on susceptibility to fire
in northern Wisconsin, USA. Landscape Ecology. 19, 327-341.

Hall, B., 1829. Travels in North America, in the Years 1827 and 1828. In: Myers, R.L.,
Ewel, J.J. (Eds.), 1990. Ecosystems of Florida. University of Central Florida Press,
Orlando, FL, USA, pp. 153.

Hardin, D., 2004. Watermelon Pond Restoration Plan: the First Five Years. Division of
Forestry, Forest Management Division, Tallahassee, FL, USA, 27 pp.

Hartnett, D.C., Krofta, D.M., 1989. Fifty-five years of post-fire succession in a southern
mixed hardwood forest. Bulletin of the Torrey Botanical Club. 116, 107-113.

He, H.S., Hao, Z., Larsen, D.R., Dai, L., Hu, Y., Chang, Y., 2002. A simulation study of
landscape scale forest succession in northeastern China. Ecological Modelling.
156, 153-166.

He, H.S., Mladenoff, D.J., 1999. Spatially explicit and stochastic simulation of forest-
landscape fire disturbance and succession. Ecology. 80, 81-99.

He, H.S., Mladenoff, D.J., Nimerfro, K.K., Gustafson, E.J., 2003. LANDIS 3.7 User's
Guide. School of Natural Resources, University of Missouri-Columbia, Columbia,
MO, USA, 63 pp.

He, H.S., Shang, B.Z., Crow, T.R., Gustafson, E.J., Shifley, S.R., 2004. Simulating forest
fuel and fire risk dynamics across landscapes LANDIS fuel module design.
Ecological Modelling. 180, 135-151.

Jones, R.H., Mitchell, R.J., Stevens, G.N., Pecot, S.D., 2003. Controls of fine root
dynamics across a gradient of gap sizes in a pine woodland. Oecologia. 134, 132-
143.









Kimmins, J.P., Mailly, D., Seel, B., 1999. Modelling forest ecosystem net primary
production: the hybrid simulation approach used in FORECAST. Ecological
Modelling. 122, 195-224.

Lamar, W.R., McGraw, J.B., 2005. Evaluating the use of remotely sensed data in matrix
population modeling for eastern hemlock (Tsuga canadensis L.). Forest Ecology
and Management. 212, 50-64.

Logofet D.O., Korotkov V.N., 2002. 'Hybrid' optimization: a heuristic solution to the
Markov-chain calibration problem. Ecological Modelling. 151, 51-61.

McGinty, D.T., Christy, E.J., 1977. Turkey oak ecology on a Georgia sandhill. American
Midland Naturalist. 98, 487-491.

McGuire, J.P., Mithcell, R.J., Moser, E.B., Pecot, S.D., Gjerstad, D.H., Hedman, C.W.,
2001. Gaps in a gappy forest: plant resources, longleaf pine regeneration, and
understory response to tree removal in longleaf pine savannas. Canadian Journal of
Forest Research. 31, 765-778.

Mladenoff, D.J., He, H.S., 1999. Design and behavior of LANDIS, an object-oriented
model of forest landscape disturbance and succession. In: Mladenoff, D.J., Baker,
W.L. (Eds.), Spatial Modeling of Forest Landscape Change: Approaches and
Applications. Cambridge University Press, Cambridge, UK, pp. 125-162.

Mladenoff, D.J., Host, G.E., Boeder, J., Crow, T.R., 1996. LANDIS: A spatial model of
forest landscape disturbance, succession, and management. In: Goodchild, M.F.,
Steyaert, L.T., Parks, B.O. (Eds.), GIS and Environmental Modeling: Progress and
Research Issues. GIS World Books, Fort Collins, CO, USA, pp. 175-180.

Myers, R.L., 1990. Scrub and high pine. In: Myers, R.L., Ewel, J.J. (Eds.), Ecosystems of
Florida. University of Central Florida Press, Orlando, FL, USA, pp. 150-193.

Neel, W.L., 1993. An ecological approach to longleaf pine forestry. In: Herman, S.M.
(Ed.), The Longleaf Pine Ecosystem: Ecology, Restoration, and Management.
Proceedings, 18th of the Tall Timbers Fire Ecology Conference. Tall Timbers, Inc.,
Tallahassee, FL, USA, pp. 337-338.

Noble, I.R., Slayter, R.O., 1980. The use of vital attributes to predict succession changes
in plant communities subject to recurrent disturbances. Vegetatio. 43, 5-21.

Ott, R.L., Longnecker, M., 2001. An Introduction to Statistical Methods and Data
Analysis. 5th Ed. Duxbury, Pacific Grove, CA, USA, 1152 pp.

Palik, B., Mitchell, R.J., Pecot, S., Battaglia, M., Pu, M., 2003. Spatial distribution of
overstory retention influences resources and growth of longleaf pine seedlings.
Ecological Applications. 13, 674-686.









Penfound, W.T., Watkins, A.G., 1937. Phytosociological studies in the pinelands of
southeastern Louisiana. American Midland Naturalist. 18, 661-682.

Peng C., 2000. Quantitative methods: understanding the role of forest simulation models
in sustainable forest management. Environmental Impact Assessment Review. 20,
481-501.

Pennanen, J., Kuuluvainen, T., 2002. A spatial simulation approach to natural forest
landscape dynamics in boreal Fennoscandia. Forest Ecology and Management. 164,
157-175.

Platt, W.J., Glitzenstein, J.S., Streng, D. R., 1991. Evaluating pyrogenicity and its effects
on vegetation in longleaf pine savannas. In: High Intensity Fire in Wildlands:
Management, Challenges, and Options. Proceedings, 17th of the Tall Timbers Fire
Ecology Conference. Tall Timbers, Inc., Tallahassee, FL, USA, pp. 143-161.

Platt, W.J., Gregory, W.E., Rathbun, S.L., 1988. The population dynamics of a long-lived
conifer (Pinuspalustris). The American Naturalist. 131, 491-525.

Platt, W.J., Rathbun, S.L., 1993. Dynamics of an old-growth longleaf pine population. In:
Herman, S.M. (Ed.), The Longleaf Pine Ecosystem: Ecology, Restoration, and
Management. Proceedings, 18th of the Tall Timbers Fire Ecology Conference. Tall
Timbers, Inc., Tallahassee, FL, USA, pp. 275-298.

Platt, W.J., Schwartz, M.W., 1990. Temperate hardwood forests. In: Myers, R.L., Ewel,
J.J. (Eds.), 1990. Ecosystems of Florida. University of Central Florida Press,
Orlando, FL, USA, pp. 150-193.

Preston, R.J., Braham, R.R., 2002. North American Trees. Iowa State Press, Ames, IA,
USA, 520 pp.

Rasser, M.K., 2003. Monitoring and Assessing Longleaf Pine Ecosystem Restoration: A
Case Study in North-Central Florida [Theses]. University of Florida. University
Microfilms, Gainesville, FL. Available from: University library electronic archives
or [http://purl.fcla.edu/fcla/etd/UFE0001256]. 60 pp.

Remnants of Florida's Longleaf Forest, 2002. The Longleaf Alliance.
Data accessed: July,
2005.

Robinson, A.P., Ek, A.R., 2003. Description and validation of a hybrid model of forest
growth and stand dynamics for the Great Lakes region. Ecological Modelling. 170,
73-104.









Silvertown, J., Franco, M., Menges, E., 1996. Interpretation of elasticity matrices as an
aid to the management of plant populations for conservation. In: Lamar, W.R.,
McGraw, J.B., Evaluating the use of remotely sensed data in matrix population
modeling for eastern hemlock (Tsuga canadensis L.). Forest Ecology and
Management. pp. 51.

Veno, P.A., 1976. Successional relationships of five Florida plant communities. Ecology.
57, 498-508.

Wade, D., Ewel, J., Hofstetter, R., 1980. Fire in South Florida Ecosystems. USDA Forest
Service General Technical Report SE-17. Southeastern Forest Experiment Station,
Asheville, NC, USA, 125 pp.

Wimberly, M.C., 2004. Fire and forest landscapes in the Georgia Piedmont: an
assessment of spatial modeling assumptions. Ecological Modelling. 180, 41-56.

Young, C.G., Young, J.A., 1992. Seeds of Woody Plants in North America. Dioscorides
Press, Portland, OR, USA, 407 pp.

Zuidema, P. A., Zagt, R. J., 2000. Using population matrices for long-lived species: a
review of published models for 35 woody plants. In: Zuidema, P.A. Demography
of Exploited Tree Species in the Bolivian Amazon [Dissertation]. Utrecht
University, Utrecht, NL, pp. 159-182.















BIOGRAPHICAL SKETCH

Eva Louise Loudermilk was born a first generation American in 1976 with a strong

Swedish heritage. With a B.S. degree from the University of Florida (UF) in animal

science in 2001, she worked as an assistant curator of ornithology at the Florida Museum

of Natural History as well as a field assistant monitoring several animal and plant species

throughout the state. She became affiliated with the Forest Information Systems

Laboratory at the School of Forest Resources and Conservation at UF and began her

graduate career in 2003 with an emphasis in Geographical Information Systems. She

plans to continue her education within the interdisciplinary ecology PhD program at the

School of Natural Resources and Environment at UF.