<%BANNER%>

Changes in Ecosystem Carbon Balance Where Permafrost Is Thawing

Permanent Link: http://ufdc.ufl.edu/UFE0024627/00001

Material Information

Title: Changes in Ecosystem Carbon Balance Where Permafrost Is Thawing
Physical Description: 1 online resource (107 p.)
Language: english
Creator: Lee, Hanna
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2009

Subjects

Subjects / Keywords: carbon, climate, global, methane, net, permafrost, soil, thermokarst
Botany -- Dissertations, Academic -- UF
Genre: Botany thesis, Ph.D.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract: One of the biggest potential feedbacks to global climate change from high latitude ecosystems may come from thawing of permafrost, which stores more than 50% of the total global terrestrial soil carbon. Thawing of permafrost may accelerate decomposition of soil organic matter and increase carbon dioxide (CO2) emissions. When permafrost thaws in ice-rich areas, it creates localized topographical surface subsidence called thermokarst, which can induce variations in soil abiotic properties. Also, depending on the location of thermokarst formation it can create anaerobic conditions to soil. By altering multiple resources in soil, thermokarst can change C cycling in high latitude ecosystems beyond simple increases in temperature alone. My research was conducted at a subarctic tundra site near Healy, Alaska (Latitude: 63.7masculine ordinalN), where permafrost thaw and thermokarst development have been observed and monitored for two decades. I established soil gas wells to explain how thermokarst affects soil respiration and to determine which depth in the soil profile has the greatest soil CO2 flux. I also established plot scale study to determine how thermokarst affects ecosystem C exchange measuring above ground CO2 fluxes using clear chambers. Then I collected permafrost soils with different substrate quality and incubated them at 15degreeC under aerobic and anaerobic conditions to observe C loss and climate forcing under different environment after permafrost thaw. I found that there was an increase in soil CO2 production where thermokarst development progressed; however, this was mostly driven by surface soil layer CO2 production rather than deeper soil layer. This was likely as a result of changes in environment such as soil temperature, moisture, and vegetation and was also shown in ecosystem carbon exchange in plot scale. I was able to estimate annual ecosystem carbon balance using surface subsidence created by thermokarst development, thaw depth, and plant biomass. Permafrost soil incubation showed that carbon loss was 3 times greater under aerobic conditions, but climate forcing was 1.15 times greater under anaerobic conditions due to methane emissions. Therefore, permafrost thaw and thermokarst development may stimulate soil CO2 production, ecosystem carbon exchange, but will affect climate warming more under anaerobic conditions.
General Note: In the series University of Florida Digital Collections.
General Note: Includes vita.
Bibliography: Includes bibliographical references.
Source of Description: Description based on online resource; title from PDF title page.
Source of Description: This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility: by Hanna Lee.
Thesis: Thesis (Ph.D.)--University of Florida, 2009.
Local: Adviser: Schuur, Edward A.

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2009
System ID: UFE0024627:00001

Permanent Link: http://ufdc.ufl.edu/UFE0024627/00001

Material Information

Title: Changes in Ecosystem Carbon Balance Where Permafrost Is Thawing
Physical Description: 1 online resource (107 p.)
Language: english
Creator: Lee, Hanna
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2009

Subjects

Subjects / Keywords: carbon, climate, global, methane, net, permafrost, soil, thermokarst
Botany -- Dissertations, Academic -- UF
Genre: Botany thesis, Ph.D.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract: One of the biggest potential feedbacks to global climate change from high latitude ecosystems may come from thawing of permafrost, which stores more than 50% of the total global terrestrial soil carbon. Thawing of permafrost may accelerate decomposition of soil organic matter and increase carbon dioxide (CO2) emissions. When permafrost thaws in ice-rich areas, it creates localized topographical surface subsidence called thermokarst, which can induce variations in soil abiotic properties. Also, depending on the location of thermokarst formation it can create anaerobic conditions to soil. By altering multiple resources in soil, thermokarst can change C cycling in high latitude ecosystems beyond simple increases in temperature alone. My research was conducted at a subarctic tundra site near Healy, Alaska (Latitude: 63.7masculine ordinalN), where permafrost thaw and thermokarst development have been observed and monitored for two decades. I established soil gas wells to explain how thermokarst affects soil respiration and to determine which depth in the soil profile has the greatest soil CO2 flux. I also established plot scale study to determine how thermokarst affects ecosystem C exchange measuring above ground CO2 fluxes using clear chambers. Then I collected permafrost soils with different substrate quality and incubated them at 15degreeC under aerobic and anaerobic conditions to observe C loss and climate forcing under different environment after permafrost thaw. I found that there was an increase in soil CO2 production where thermokarst development progressed; however, this was mostly driven by surface soil layer CO2 production rather than deeper soil layer. This was likely as a result of changes in environment such as soil temperature, moisture, and vegetation and was also shown in ecosystem carbon exchange in plot scale. I was able to estimate annual ecosystem carbon balance using surface subsidence created by thermokarst development, thaw depth, and plant biomass. Permafrost soil incubation showed that carbon loss was 3 times greater under aerobic conditions, but climate forcing was 1.15 times greater under anaerobic conditions due to methane emissions. Therefore, permafrost thaw and thermokarst development may stimulate soil CO2 production, ecosystem carbon exchange, but will affect climate warming more under anaerobic conditions.
General Note: In the series University of Florida Digital Collections.
General Note: Includes vita.
Bibliography: Includes bibliographical references.
Source of Description: Description based on online resource; title from PDF title page.
Source of Description: This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility: by Hanna Lee.
Thesis: Thesis (Ph.D.)--University of Florida, 2009.
Local: Adviser: Schuur, Edward A.

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2009
System ID: UFE0024627:00001


This item has the following downloads:


Full Text

PAGE 1

CHANGES IN ECOSYSTEM CARBON BALANCE WHERE PERMAFROST IS THAWING By HANNA LEE A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2009 1

PAGE 2

2009 Hanna Lee 2

PAGE 3

TABLE OF CONTENTS page LIST OF TABLES...........................................................................................................................5 LIST OF FIGURES.........................................................................................................................6 ABSTRACT.....................................................................................................................................8 CHAPTER 1. INTRODUCTION..................................................................................................................10 2. SOIL CO2 PRODUCTION IN UPLAND TUNDRA WHERE PERMAFROST IS THAWING.............................................................................................................................14 Introduction.............................................................................................................................14 Methods..................................................................................................................................15 Site Description...............................................................................................................15 Gas Well Measurements..................................................................................................16 Soil Properties.................................................................................................................17 Estimating Soil Profile CO2 Flux and Production...........................................................18 Surface Microtopography................................................................................................19 Statistical Analysis..........................................................................................................20 Results.....................................................................................................................................20 Soil Profile CO2 Concentration.......................................................................................20 VWC and Soil Gas Diffusion Coefficient.......................................................................21 Soil Profile CO2 Flux and Production.............................................................................22 Modeling Growing Season Soil CO2 Production............................................................23 Discussion...............................................................................................................................24 Conclusion..............................................................................................................................28 3. A SPATIALLY EXPLICIT ANALYSIS TO MODEL CARBON FLUXES AT A LARGER SCALE IN UPLAND TUNDRA WHERE PERMAFROST IS THAWING.......34 Introduction.............................................................................................................................34 Methods..................................................................................................................................37 Site Description...............................................................................................................37 Land Surface Survey.......................................................................................................37 Soil Properties.................................................................................................................39 Aboveground Biomass and Vegetation Index.................................................................40 Ecosystem Respiration, Gross Primary Production, and Net Ecosystem Exchange of CO2..........................................................................................................................40 Spatial Autocorrelation....................................................................................................41 Statistical Analysis..........................................................................................................42 New Sites to Verify the Model........................................................................................43 Estimating Annual Carbon Fluxes at a Plot Scale...........................................................45 3

PAGE 4

Results.....................................................................................................................................46 Correlations and Spatial Patterns of Explanatory and Response Variables....................46 C fluxes Model Construction...........................................................................................47 Verifying the Model Using New Dataset........................................................................47 Estimating Annual C fluxes Using Environmental Variables.........................................48 Discussion...............................................................................................................................50 Conclusion..............................................................................................................................57 4. THE RATE OF PERMAFROST CARBON RELEASE UNDER AEROBIC AND ANAEROBIC CONDITIONS................................................................................................69 Introduction.............................................................................................................................69 Methods..................................................................................................................................71 Soil Sampling and Preparation........................................................................................71 Laboratory Incubation Experiment and CO2/CH4 Fluxes...............................................72 Soil Enzyme Assay..........................................................................................................73 Soil C and N Analysis and Stable Isotope Measurements..............................................74 Soil pH.............................................................................................................................75 Analyses..........................................................................................................................75 Results.....................................................................................................................................76 Substrate Quality and Soil Enzyme Activities................................................................76 Carbon Fluxes under Aerobic and Anaerobic Conditions...............................................77 Soil -glucosidase Activity..............................................................................................77 Cumulative Carbon Mineralization and Climate Forcing...............................................78 The Relationship between Carbon Mineralization and Substrate Quality......................79 Discussion...............................................................................................................................80 Conclusion..............................................................................................................................84 5. CONCLUSION.......................................................................................................................95 LIST OF REFERENCES...............................................................................................................97 BIOGRAPHICAL SKETCH.......................................................................................................107 4

PAGE 5

LIST OF TABLES Table page 2-1. Pearsons pairwise correlation coefficients for the environmental variables used in the soil CO2 production model...........................................................................................29 2-2. Soil CO2 production model using single and multiple variables.......................................29 3-1. Pearsons pairwise correlation coefficients of the explanatory variables at the three EML gradient sites.............................................................................................................58 3-2. Model parameters of the curves fitted through each semivariogram in environmental variables.............................................................................................................................58 3-3. Site combined C fluxes model using measured environmental variables. Reco, GPP, and NEE were in log scale to seek for normal distribution...............................................59 3-4. The AIC and BIC values of error terms in spatial covariance structure of the growing season C flux models (growing season Reco, GPP, and NEE).........................................60 3-5. Annual C flux models constructed using same group of environmental variables chosen from the plot scale C fluxes model........................................................................60 3-6. Model parameters of the surces fitted through each semivariogram in surface subsidence (MT) measured each site in 2005....................................................................61 3-7. Datasets used in each analysis...........................................................................................62 3-8. Growing season C flux models using measured environmental variables to seek for spatial pattern in the error terms of the mixed model........................................................63 3-9. Model parameters of the curves fitted through each semivariogram.................................64 3-10. Summary statistics of the explanatory variables used in the growing season carbon flux models for the three EML gradient sites....................................................................64 4-1. Substrate quality of soil samples used in the incubation study..........................................86 4-2. Cumulative C mineralization per gram of dry soil after 250 days of incubation at 15C under aerobic conditions and anaerobic conditions..................................................87 4-3. Areas of lake and non-lake in continuous, discontinuous, and sporadic permafrost zone....................................................................................................................................88 5

PAGE 6

LIST OF FIGURES Figure page 2-1. Soil CO2 concentrations measured within the active layer during the growing season from 2005 to 2007..............................................................................................................30 2-2. The distribution of volumetric water content (VWC) measured in 2006 and 2007 during the growing season at all gas wells for 10, 20, and 30cm depth............................31 2-3. Soil profile CO2 fluxes for the three sites during the growing season (JuneAug) from 2005 to 2007 calculated for 10 cm intervals.............................................................32 2-4. Mean (SE) growing season soil CO2 production by year................................................33 2-5. Relationship between relative microtopographic position of individual gas wells and soil CO2 production............................................................................................................33 3-1. Spatial pattern of surface subsidence in meter scale created by permafrost thaw and thermokarst........................................................................................................................65 3-2. Semivariograms of growing season C fluxes measured at plot scale for ecosystem respiration (Reco), gross primary production (GPP), and net ecosystem exchange of CO2 (NEE) at Minimal, Moderate, and Extensive Thaw...................................................66 3-3. Semivariograms of the error terms from multiple regression models for ecosystem respiration (Reco), gross primary production (GPP), and net ecosystem exchange of CO2 (NEE) at Minimal, Moderate, and Extensive Thaw...................................................67 3-4. A 1:1 fit of measured and model predicted Reco, GPP, and NEE. Different symbols represent different sites near the EML gradient sites.........................................................68 3-5. The spatial patterns in predicted annual Reco and GPP at the three EML gradient sites using ordinary Kriging...............................................................................................68 4-1. A map of soil sampling locations indicated by the closed circles.....................................89 4-2. The CO2 fluxes during 250 days of soil incubation at 15C under aerobic conditions.....90 4-3. The CO2 (left) and CH4 (right) fluxes during 250 days of incubation at 15C under anaerobic conditions. ........................................................................................................91 4-4. Soil -glucosidase activity before incubation and after 125 days of incubation at 15C under aerobic and anaerobic conditions...................................................................92 4-5. Cumulative C mineralization via CO2 emissions under aerobic conditions (A) and via CO2 and CH4 emissions under anaerobic conditions (B) per grams of dry soil................93 6

PAGE 7

4-6. Cumulative C mineralization via CO2 emissions under aerobic conditions (A) and by CO2 and CH4 emissions under anaerobic conditions (B) per grams of soil carbon...........93 4-7. Relationship between %C, %N, and C to N ratio in soils and cumulative C mineralized per soil dry weight after 250 days of incubation at 15C under aerobic and anaerobic conditions, and relative climate forcing.....................................................94 4-8. Relationship between soil pH and cumulative C mineralized per grams of soil after 250 days of incubation via CO2 emissions under aerobic conditions, and via CO2 and CH4 emissions under anaerobic conditions........................................................................94 7

PAGE 8

Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy CHANGES IN ECOSYSTEM CARBON BALANCE WHERE PERMAFROST IS THAWING By Hanna Lee August 2009 Chair: Edward A. G. Schuur Major: Botany One of the biggest potential feedbacks to global climate change from high latitude ecosystems may come from thawing of permafrost, which stores more than 50% of the total global terrestrial soil carbon. Thawing of permafrost may accelerate decomposition of soil organic matter and increase carbon dioxide (CO2) emissions. When permafrost thaws in ice-rich areas, it creates localized topographical surface subsidence called thermokarst, which can induce variations in soil abiotic properties. Also, depending on the location of thermokarst formation it can create anaerobic conditions in soil. By altering multiple resources in soil, thermokarst can change C-cycling in high latitude ecosystems beyond simple increases in temperature alone. My research was conducted at a subarctic tundra site near Healy, Alaska (Latitude: 63.7N), where permafrost thaw and thermokarst development have been observed and monitored for two decades. I established soil gas wells to explain how thermokarst affects soil respiration and to determine which depth in the soil profile has the greatest soil CO2 flux. I also established plotscale studies to determine how thermokarst affects ecosystem C exchange by measuring above ground CO2 fluxes using clear chambers. Then I collected permafrost soils with different substrate qualities and incubated them at 15C under aerobic and anaerobic conditions to observe C loss and climate forcing in different environments after permafrost thaw. I found that there was 8

PAGE 9

an increase in soil CO2 production where thermokarst development progressed; this was mostly driven by surface soil layer CO2 production rather than deeper soil layer CO2 production This was likely due to changes in the environment such as soil temperature, moisture, and vegetation. I was able to estimate the annual ecosystem carbon balance using surface subsidence created by thermokarst development, thaw depth, and plant biomass. Permafrost soil incubation showed that carbon loss was 3 times greater under aerobic conditions, but climate forcing was 1.15 times greater under anaerobic conditions due to methane emissions. Therefore, permafrost thaw and thermokarst development may stimulate soil CO2 production, ecosystem carbon exchange, but will affect climate warming more under anaerobic conditions. 9

PAGE 10

CHAPTER 1 INTRODUCTION Over the past century, high latitude ecosystems have undergone drastic changes due to global scale warming [ACIA, 2005; Serreze et al., 2000]. One consequence of warming in high latitude ecosystems is permafrost thaw. Permafrost soil is distributed across 14% of the global land surface [Tarnocai et al., in press], and accounts for more than half of global terrestrial soil carbon (C) [Schuur et al., 2008]. The frozen conditions of permafrost slow down decomposition of annual plant litter input, storing them as new soil organic matter every year. Recent studies have shown increased permafrost temperatures as deep as 50m [Osterkamp and Romanovsky, 1999] due to increased mean annual temperature and changes in precipitation. Thawing of permafrost and thermokarst development may stimulate organic matter decomposition, releasing greenhouse gases such as carbon dioxide (CO2) and methane (CH4) to the atmosphere. These emissions may create a positive feedback cycle between global warming and permafrost thaw. When ice-rich permafrost thaws, it may cause ground surface subsidence called thermokarst terrain. The scale and magnitude of thermokarst alter local hydrology [Jorgenson et al., 2006] and may change soil properties such as increased belowground temperature, moisture, and nutrient availability [Fortier et al., 2007; Jorgenson et al., 2001; Osterkamp, 2007b]. On sloped areas or uplands that are relatively well drained, thermokarst creates patchy microtopographic depressions that result in well drained soils [Schuur et al., 2008]. When thermokarst soils are well drained, they create optimum soil moisture conditions for decomposition and may increase CO2 emissions by stimulating decomposition of SOM [Zimov et al., 2006]. Depending on the landscape scale topography and degrees of erosion, large excavations and lakes may form [Jorgenson and Osterkamp, 2005; Jorgenson and Shur, 2007; Smith et al., 2007] and make the soils anaerobic. A recent study showed that CH4 emissions from 10

PAGE 11

water-logged soil conditions created by thermokarst may play a major role in climate change [Walter et al., 2007]. Very rarely, upland thermokarst creates water-logged conditions depending on the degree of depression, which may contribute to CH4 emissions through anaerobic SOM decomposition. As a result of thaw effects on hydrology, soil organic matter from permafrost may be deposited in an oxic or an anoxic environment after permafrost thaw. A subsided ground surface may expose deeper permafrost to further thaw and increase active layer thickness (ALT), the seasonally-thawed soil layer found in the permafrost zone, which in turn may stimulate decomposition of soil organic matter stored deep within. Some researchers have observed increased net ecosystem exchange of CO2 (NEE) from high latitude ecosystems as a result of increased air temperature and soil temperature [Chapin et al., 2000; Oechel et al., 2000], while others have found that increased soil moisture and temperature corresponded to increased NEE as a direct effect [Oberbauer et al., 1992]. Indirectly, increased temperature may stimulate soil organic matter decomposition and release nutrients to soil, which in turn positively feed back to faster decomposition of soil organic matter [Mack et al., 2004]. Alternatively, changes in soil properties caused by permafrost thaw and thermokarst may stimulate carbon uptake by increased primary productivity in the permafrost zone. Many studies have observed the effects of climate change in carbon uptake in high latitude ecosystems, such as increased primary production [Schuur et al., 2007], invasion and enlargement of shrubs, and invasion of trees to tundra [Macdonald et al., 1993; Sturm et al., 2005]. Increased primary production and changes in species composition affect ecosystem carbon uptake, since woody parts in plants store more carbon than herbaceous parts. Therefore, changes in tundra carbon balance as a result of permafrost thaw and thermokarst development will vary based processes 11

PAGE 12

that simulate decomposition (soil temperature and moisture availability) and those that stimulate nutrient availability and a change in species composition. In an upland tundra ecosystem, Schuur et al. [2007] observed increased primary productivity and changes in plant species composition from tussock forming sedge species to shrub species as a result of permafrost thaw and thermokarst development as it progressed over several decades. Vogel et al. [in press] found that at the initial stage of permafrost thaw and thermokarst, the ecosystem became a net carbon sink because of increased primary productivity but at the later stage of permafrost thaw and thermokarst (about 50 years) the ecosystem became a net carbon source because of increased carbon emissions especially from decomposition of old carbon stored in the deeper permafrost layer. These patterns in net ecosystem exchange (NEE) corresponded to greater permafrost thaw and vegetation productivity. Supporting these results, Lee et al. [submitted] observed greater belowground CO2 production in areas with increased permafrost thaw, likely due to increased soil organic matter decomposition. Typically, organic matter decomposition occurs at a much faster rate under aerobic conditions than anaerobic conditions. Even under anaerobic conditions, CO2 emissions are greater than CH4 emissions. However, CH4 has a much higher global warming potential (GWPCH4 = 25, GWPCO2 = 1) than CO2 in a 100-year time frame [IPCC, 2007]. Even though total C emissions are much smaller under anaerobic conditions, more reactive greenhouse gas emissions such as CH4 may contribute more to global warming in the long term. Therefore, the magnitude of permafrost carbon emissions to global warming may depend on the soil conditions, especially aerobic or anaerobic conditions after thawing of permafrost, such as newly formed or enlargement of thermokarst lakes or drainage of thermokarst lakes in the northern ecosystems. 12

PAGE 13

Three experiments were established to observe changes in the ecosystem carbon balance under thawing of permafrost. First, soil CO2 production was estimated from a natural permafrost thaw gradient to examine how permafrost thaw and thermokarst development over decadal time scales affects CO2 emissions from an upland tundra ecosystem. I hypothesized that soil CO2 production would increase with the degree of permafrost thaw and thermokarst development, especially in the deeper soil layers, because increased soil temperature and moisture would stimulate microbial decomposition of soil organic matter. Second, ecosystem C exchange was scaled up in time and space. A plot scale carbon balance study was established in an Alaskan tundra where permafrost thaw and thermokarst had been observed for several decades. I hypothesized that there would be a positive relationship between depth of microtopography and carbon emissions and carbon uptake because increased resource availability stimulates plant growth as well as microbial decomposition of soil organic matter. Lastly, I tested how the oxygen status and soil substrate quality affect CO2 and CH4 emissions from permafrost soil by conducting a laboratory soil incubation experiment. Permafrost soils were collected from Alaska and Siberia, which showed different soil characteristics such as organic and mineral, acidic and non-acidic to observe the fate of permafrost carbon when it thaws. I hypothesized that total C loss would be higher under aerobic conditions, but the global warming effect would be higher under anaerobic conditions because of CH4 emissions. I also hypothesized that C loss would be positively correlated to soil substrate quality such as %C, %N, or C to N ratio, because the rate of microbial decomposition is controlled by substrate quality of soil. 13

PAGE 14

CHAPTER 2 SOIL CO2 PRODUCTION IN UPLAND TUNDRA WHERE PERMAFROST IS THAWING Introduction Over the past century, high latitude ecosystems have undergone drastic changes due to global scale warming [ACIA, 2005; Serreze et al., 2000]. One consequence of warming in high latitude ecosystems is permafrost thaw. Permafrost soil is distributed across 14% of the global land surface [Tarnocai et al., in press], and accounts for more than half of global terrestrial soil carbon (C) [Schuur et al., 2008]. Recent studies have shown increased permafrost temperatures as deep as 50m [Osterkamp and Romanovsky, 1999] due to increased mean annual temperature and changes in precipitation. This may expose soil organic matter stored in permafrost to increased microbial decomposition, releasing greenhouse gases such as carbon dioxide (CO2) and methane (CH4) to the atmosphere. These emissions may create a positive feedback cycle between global warming and permafrost thaw. When ice-rich permafrost thaws, it may cause ground surface subsidence called thermokarst terrain. Thermokarst may alter the soil environment by changing soil hydrology, which in turn contributes to changes in microsite soil temperature, moisture, and nutrient availability [Fortier et al., 2007; Jorgenson et al., 2001; Osterkamp, 2007b]. A subsided ground surface may expose deeper permafrost to further thaw and increase active layer thickness (ALT), the seasonally-thawed soil layer found in the permafrost zone, which in turn may stimulate decomposition of soil organic matter stored deep within. Several studies have observed increased whole-ecosystem C emissions from climatic changes in northern ecosystems [Oechel et al., 2000; Shaver et al., 1998; Vourlitis and Oechel, 1999; Welker et al., 2000]. However, chamber measurements and eddy covariance methods of measuring CO2 exchange cannot distinguish between belowground and aboveground C 14

PAGE 15

emissions. Some researchers have observed belowground CO2 emissions directly by detecting CO2 gas from different soil profiles using gas wells or other similar instruments [Davidson and Trumbore, 1995; Elberling and Ladegaard-Pedersen, 2005; Hirsch et al., 2002; Oh et al., 2005; Risk et al., 2002; Takahashi et al., 2004; Tang et al., 2003]. Using Ficks law of gaseous diffusion, observations of soil CO2 concentrations in soil profile can be used to quantify CO2 production from different soil layers and to estimate total belowground CO2 production as soil CO2 production [Davidson and Trumbore, 1995]. In this study, I estimated soil CO2 production from a natural permafrost thaw gradient to examine how permafrost thaw and thermokarst development over decadal time scales affects CO2 emissions from an upland tundra ecosystem. I hypothesized that soil CO2 production, would increase with the degree of permafrost thaw and thermokarst development especially in the deeper soil layers, because increased soil temperature and moisture would stimulate microbial decomposition of soil organic matter. Methods Site Description This study was conducted in upland tundra near Healy, Alaska, just outside of Denali National Park (Eight Mile Lake site: 63o52'42"N, 149o15'12"W). Ground temperature and deep permafrost temperatures to 30 m have been monitored since 1985 in this area [Osterkamp and Romanovsky, 1999] and ground subsidence as a result of permafrost thaw and thermokarst has also been observed [Osterkamp, 2005]. The area is a gentle north-facing slope (< 5) with discharge water draining into the adjacent Eight Mile Lake. Three sites were established in 2003 based on the degree of permafrost thaw and resulting ground subsidence [Schuur et al., 2007]: Minimal Thaw is the least disturbed typical moist acidic tussock tundra site, with little ground subsidence. Moderate Thaw is located adjacent to the permafrost monitoring borehole where 15

PAGE 16

patchy areas of ground subsidence have started to occur. Extensive Thaw contains large-scale ground subsidence and shrubs that have become the dominant vegetation at the expense of tussock-forming sedges, likely as a result of permafrost thaw and thermokarst. The changes in ground topography as a result of subsidence has resulted in a complex microtopography where, subsided areas collect water and become wetter, while nearby within the same site, relatively elevated areas become drier [Lee et al., 2007]. These different moisture conditions are coupled with soil temperature, microsite differences in plant community composition, and increased active layer thickness (ALT), with wetter areas having deeper thaw depth. Gas Well Measurements I used soil gas wells to collect air from the soil profile in order to estimate CO2 production from different soil layers. Gas wells were constructed from 1/8 diameter stainless steel tubing. The tubing was the length of the designated depth plus an additional 40 cm, which was bent into an L-shape and pushed horizontally into the soil at the designated depth in the soil profile. At the soil surface, an airtight stopcock was glued to the end of the stainless steel tubing and was kept closed, except during sampling, to minimize ambient air exchange with soil gas. Gas wells had three holes near the tip inside the soil to decrease the chance of clogging from soil organic matter or silt. Five gas well locations were established at each site, with each location having wells at four depths in the soil profile (10, 20, 30, and 40 cm), to measure soil profile CO2 flux in 2004. In 2005, I installed gas wells at 6 additional locations at each site, with each site having shorter depth intervals (5, 10, 15, 20, 25, and 30 cm), to obtain finer scale estimates of soil profile CO2 flux. I used airtight plastic syringes to sample CO2 from each gas well and analyzed it using an injection loop system attached to an infrared gas analyzer (Li-820, LICOR Corp., Lincoln, Nebraska). Samples were taken weekly throughout the growing season (May to September) from 2005 to 2007 on calm mornings, to minimize pressure-driven soil gas exchange between ambient 16

PAGE 17

air and soil that occurs as a result of wind [Hirsch et al., 2004]. I was able to sample CO2 from deeper soil layers in the beginning of the growing season, which was taken from air pockets within frozen soil. These air pockets trapped CO2 produced over the winter and often contained high CO2 concentrations up to 100,000 ppmv. However, since these air pockets were not subject to the standard assumptions made when applying the normal diffusion law, I did not include these measurements in the flux estimates. On average, across years and sites, I was able to sample gas from ice-free soil starting at day of year 150 for the 10 cm depth, starting at day of year 165 for the 20 cm depth, and starting at day of year 185 for the 30 cm depth, as the seasonal thaw front descended within the active layer. At the EML gradient sites, deeper soil layers were saturated with water most of the time due to a water table that was perched on the ground ice surface. When it rained, even surface layers (10 cm) could become saturated with water. Therefore, when soil was waterlogged, I collected 10ml of soil water and mixed it with 10ml of ambient air, shook it for 1 minute to equilibrate, then measured CO2 from the headspace of the mixed gas sample. In this way, I was able to estimate the CO2 dissolved in soil water. Soil Properties Soil temperature was monitored at 10 cm of depth using IButton Thermochron temperature loggers (Maxim Inc., Dallas, TX) buried adjacent to each gas well and data-logged for three days. Mean soil temperatures at 10 cm were obtained for the three days of measurement and they were normalized to a mean of 0C. These were used to describe the spatial variability of soil temperature at each gas well. Active layer thickness (ALT) was measured at the end of growing season by pushing a thin rod into the ground until it hit the permafrost surface. Soil moisture content was measured at a location close to gas wells during each gas sampling period. I measured soil moisture as volumetric water content (VWC) using a water content reflectometer (CS616, Campbell Scientific Inc.) and a CR10X data logger. The CS616 was calibrated for the 17

PAGE 18

EML gradient sites by comparing CS616 values measured at nearby points to direct water content measurements that were determined through destructive soil sampling from samples taken at the same nearby points. To calibrate the CS616, I vertically installed the probe at 10, 20, and 30 cm depths in the soil at 60 random locations at the EML gradient sites and obtained measurements. Directly afterwards, a 10 cm diameter soil core was removed at each location. Soil cores were weighed wet, then dried at 60C and reweighed to obtain VWC of the soil. I used linear regressions for each depth to obtain the best fit between CS616 values and direct measure of VWC. Estimating Soil Profile CO2 Flux and Production I calculated soil profile CO2 flux using Ficks first law from one-dimensional measurements of CO2 concentration at each soil profile. CO2 Flux = -Deff dC/dz (2-1) In this equation, Deff is gas diffusion coefficient in soil, dC is the difference of CO2 concentrations between two adjacent soil horizons, and dz is thickness of each soil horizon that was sampled. The gas diffusion coefficient (Deff) was estimated based on air-filled porosity in soil using Millingtons equation, as well as the CO2 gradient [Davidson and Trumbore, 1995; Gaudinski et al., 2000; Hirsch et al., 2002; Millington, 1959]. Deff = D0 e4/3 (e/a)2 (T/273)1.75 (2-2) Here, D0 is diffusivity in air, a is total porosity (1 Bulk density/Particle density), and e is air-filled porosity (Total porosity VWC). The diffusion coefficient for CO2 in free air was 1.39 10-5 m2 s-1 [Gaudinski et al., 2000], and 1.7 10-9 m2 s-1 in water [Jahne et al., 1987]. 18

PAGE 19

Soil profile CO2 flux was estimated from the difference between CO2 concentrations at two different soil depths and the diffusion coefficient (Equation 2-1). The CO2 production for each layer was calculated as the difference between CO2 fluxes at two adjacent layers (Equation 2-3). CO2 production = Fi Fi+1 (2-3) Where Fi is CO2 flux for one horizon and Fi+1 is CO2 flux in the horizon below. Soil CO2 production was estimated as a sum of CO2 production for each soil layer across the entire soil profile. Surface Microtopography I used a differential Global Positioning System (GPS) [Little et al., 2003] to determine the fine scale (1 cm vertical resolution) topographical features of all gas well locations and the degree of surface subsidence created by permafrost thaw and thermokarst. I installed one GPS antenna (Trimble 5400) at a nearby USGS marker (63'16.56"N, 149'17.92"W) and used another GPS antenna to survey gas well coordinates, so that sampling coordinates were always corrected relative to the marker to obtain better accuracy. The x, y, z (Longitude, Latitude, Altitude) coordinates were then normalized for hillslope trends along the z-axis to minimize the effect of hillslope. From this, I excluded the top 5% of the z-axis measurements as outliers and obtained means of the remaining highest 10% of the points per site. All of the z-axis values were normalized according to the mean of the highest 10% of the values as a measure of surface subsidence, which isolated microtopography as a result of thermokarst. This procedure was done separately for each site. Therefore, the surfaces with lower relative elevation largely represent surface subsidence created by permafrost thaw and thermokarst, whereas the surfaces with higher relative elevation represent elevated surfaces that did not subside [Lee et al., 2007]. This relative elevation measurement representing surface subsidence created by permafrost thaw and thermokarst development will be referred to as Microtopography throughout this chapter. 19

PAGE 20

Statistical Analysis Soil CO2 production was transformed using natural log to achieve a normal distribution of the data for regression analysis. I used repeated measures analysis to compare soil CO2 production at each site over multiple time periods. Using Mixed Model in SAS, sites were fixed with individual wells nested within. I compared site, year, and a site year interaction. I used simple regression analysis to model mean growing season soil CO2 production using measured soil variables: soil temperature at 10 cm (T), soil moisture at 10 cm (VWC), active layer thickness (ALT), and surface subsidence created by permafrost thaw and thermokarst (Microtopography). I used measurements at 10 cm for temperature and moisture measurements since soil CO2 production was largely driven by production in the surface layer and PCA of soil temperature and moisture at 10, 20, and 30 cm was the most correlated to 10cm measurements. I also used the three sites as dummy variables with a slope of 0 to test whether the response curves were different. Dummy variables indicated differences among sites (Site1 and Site2). The analyses were done using SAS 9.0 and JMP 5.1. Results Soil Profile CO2 Concentration Mean CO2 concentrations at different depths in the soil profile ranged from 680 ppmv near the surface to 13000 ppmv at deeper depths, with median values ranging from 640 to 5400 ppmv (Figure 2-1a). Soil CO2 concentrations varied by sites, wells, time of the growing season, and with temperature, and precipitation. At every depth profile and year sampled, the mean CO2 concentration was higher than the median, illustrating the influence of some very high observed concentrations on the mean. The coefficient of variance for all concentrations was low (CV < 1) for the 5 and 10 cm depths and was higher (CV > 1) for all deeper depths. At this study area, CO2 concentrations increased with depth down to 30 cm, but then decreased below 30 cm, 20

PAGE 21

possibly due to the presence of permafrost and waterlogged conditions. Mean growing season thaw depth was 32 cm across all wells sampled. Generally, the seasonal trend in CO2 concentrations between day of year 150 and 260 was for peak values to occur in July (Figure 2-1b), which follows the trend of seasonal increase in soil temperature and plant productivity during the growing season. However, much higher CO2 concentrations, as high as 1 104 ppmv, were observed even in surface soil layers in the beginning of the growing season (DOY <150) in 2006 and 2007. These concentrations were likely a result of winter respiration CO2 trapped in air pockets within the frozen soil. VWC and Soil Gas Diffusion Coefficient VWC at 10 cm ranged from 0.1 to 0.3 mL cm-3, whereas at 30 cm it remained close to 1.0 mL cm-3 throughout the growing season (Figure 2-2). In contrast, VWC at 20 cm showed large variation between 0.1 to 1.0 mL cm-3 due to the timing of rainfall events and/or the microtopographical location of individual wells, whether they were located in lower, wetter or higher, drier areas. The diffusion coefficient showed an opposite trend from VWC by definition (Equation 2-2), because air-filled porosity decreases as VWC increases and water slows CO2 diffusion rates. The estimates of the diffusion coefficient indicated that at 30 cm and deeper, the diffusion coefficient was generally the same as that in water, and this corresponded to the fact that a water table was perched on the ice surface throughout the growing season. The ratio of an estimated diffusion coefficient in soil at 10 cm and that of the atmosphere (Deff/Da) ranged from 0.1 to 0.5 across wells and through time, whereas the diffusion coefficient ratio for 20 cm depth varied by 100 times due to periods of water saturation alternating with drier periods at the mid-profile depth. Therefore, the high CO2 concentrations observed at deeper soil layers did not lead to high CO2 production for those layers because they were largely a consequence of lower diffusivity, by 2 orders of magnitude, rather than higher production. 21

PAGE 22

Soil Profile CO2 Flux and Production Soil profile CO2 flux for the 0 cm layer ranged from 0.72 to 2.64 g CO2-C m-2 day-1 across all three years of growing season, and CO2 flux was highest here among all soil layers (Figure 2-3). Even though CO2 concentrations were high in the deepest soil layer (20 cm), CO2 flux was low because the diffusion coefficient in this layer decreased due to water saturation. As a result of lower diffusivity, mean CO2 flux in the 10 cm layer was only 1020% that of the layer above, while mean CO2 flux of 20 cm layer was 1 to 2 orders of magnitude lower than that of 10-20 cm layer. When the mean CO2 flux at 0 cm were compared by sites, Extensive Thaw was higher than Moderate and Minimal Thaw (p < 0.001 for both pairwise comparisons), but Moderate Thaw and Minimal Thaw were not significantly different from each other. This trend of the highest mean CO2 flux being in the Extensive Thaw site was the same for the 10-20 cm (p = 0.01 for both pairwise comparisons) layer. In the deepest layer (20 cm), mean CO2 flux at Extensive Thaw and Moderate Thaw was about 10% higher than Minimal Thaw; however, they were not statistically different from one another. Fluxes from soil deeper than 30 cm were not considered here, because those layers were either frozen or saturated during most of the growing season, and flux rates were very low. Aggregated across the growing season, the mean soil CO2 production ranged from 0.96 to 2.52 g CO2-C m-2 day-1, and was highest at Extensive Thaw and lowest at Minimal Thaw (Figure 2-4). Mean soil CO2 production at Extensive Thaw was significantly different from Moderate Thaw and Minimal Thaw (p < 0.001); however, Moderate Thaw was not significantly different from Minimal Thaw. Over the three years of observation, soil CO2 production in 2007 was the highest and 2005 was the lowest, which may be due to higher frequency of rainfall and increased soil moisture especially at 5 and 10 cm in 2005. When soil CO2 production was analyzed using repeated measures analysis, there were differences among sites (p = 0.02) and across years (p < 22

PAGE 23

0.001). However, the site year interaction was not significant (p = 0.23), implying that the trend across sites was consistent throughout the three sampled years. When the mean soil CO2 production was extrapolated to whole growing season CO2 production (May 1 to September 31, 150 days), estimates of growing season CO2 production from three sites were 270.4 g CO2-C m-2 at Extensive Thaw, 229.1 g CO2-C m-2 at Moderate Thaw, and 177.3 g CO2-C m-2 at Minimal Thaw. These estimates explained approximately 43% of total ecosystem respiration when they were compared with surface CO2 flux measurements from this site [Vogel et al., in press]. Modeling Growing Season Soil CO2 Production Pearsons partial correlations showed that Microtopography was negatively correlated to VWC (-0.5780), positively correlated with both dummy site variables (Table 2-1), and only weakly correlated with ALT (0.0676). When all of the variables were used to model mean growing season soil CO2 production, 54% of the variance in soil CO2 production was explained in a full model (adjusted R2 = 0.38). However, none of the parameters were significant at a p = 0.10 level in this model, except for Site 1 (the dummy variable for Extensive Thaw site), which implies that there is a multicoliniarity embedded within all of the parameters in the model. Variables ALT and VWC were not added to the regression model because adding those variables from Sites+T+Microtopography did not increase the R2 of the model. Adjusted R2 values account for increasing the number of variables. Variables were manually put in the model in addition to the single best variables (Microtopography, Sites) to maximize R2 while maintaining significance for the rest of the variables. Therefore, I separated each variable in the multiple regression model to find the best predictor variable for soil CO2 production. From the analysis, Microtopography was the best single predictor variable (R2 = 0.33) for soil CO2 production (Figure 2-5) followed by Site 1 and 2 (R2 = 0.29), and soil temperature (R2 = 0.15) (Table 2-2). Soil temperature was positively correlated to soil CO2 production, whereas 23

PAGE 24

Microtopography was negatively correlated to soil CO2 production. Neither VWC nor ALT was significantly correlated to soil CO2 production by themselves. When the Site variables were combined with microtopography, the R2 increased only 0.15, which implies that Site variables are not correlated to Microtopography. When I used Microtopography and soil temperature in the model, 41% of soil CO2 production was explained by these two variables (adjusted R2 = 0.36). When site variables were added to this model, R2 increased up to 0.55, making it the best overall model for soil CO2 production given the measured variables. Discussion Soil profile CO2 measurements provide detailed information on CO2 production across the soil profile [Davidson and Trumbore, 1995]. I used this method to understand how permafrost thaw and thermokarst affects soil C emissions from different soil layers. Soil CO2 production in this upland tundra site showed both temporal and spatial variation; soil CO2 production generally increased with increasing temperature during the growing season. Production also increased with the degree of permafrost thaw and thermokarst development across sites, likely as a result of changes in soil properties apart from seasonal variations in soil temperature. Higher total soil CO2 production at Extensive Thaw (Figure 2-4) reflected warmer temperatures, deeper thaw depth, and more ground subsidence at this site. Soil CO2 flux estimated for each 10 cm soil layer showed that the 0 cm surface layer contributed the most to total soil CO2 production, with the 10-20 cm layer in general having CO2 production an order of magnitude less (Figure 2-3). Even though deeper soil layers had 100 times higher CO2 concentrations (Figure 2-1), CO2 flux from these deeper layers was two orders of magnitude lower than surface soil CO2 flux (Figure 2-3) due to low diffusivity in these water saturated soils. 24

PAGE 25

High correlations between microtopography and other environmental variables (Table 2-1) shows that surface subsidence may be a good predictor of changes in soil environment such as increased moisture and surface temperature. Among all the variables, soil moisture was correlated the highest with surface subsidence. When thermokarst develops, subsided areas are more likely to collect water and increase soil moisture content [Jorgenson et al., 2006], while leaving other nearby areas drier. I observed that soil temperature in subsided areas was higher than in nearby less subsided soil [Osterkamp et al., 2000; Schuur et al., 2007]. Soil temperatures are higher in thermokarst depressions than in the surroundings because subsided ground surfaces trap snow during winter as wind redistributes it across the landscape, which then further insulates soil during winter [Osterkamp, 2007a]. Warm winter soil positively feeds back to soil temperature during summer [Stieglitz et al., 2003]. Sites represented as dummy variables consisted of 0 and 1 in the regression model, and I speculate that correlations between sites and surface subsidence represent changes in the environment that were not detected through the measurements, such as shifts in vegetation from tussock-forming sedges to shrub dominating tundra, or changes in total plant biomass [Schuur et al., 2007]. These changes in the plant community could alter belowground CO2 production independently of the observed environmental changes, and there were no overall differences in soil C pools among sites [Schuur et al., 2008]. I suggest that surface subsidence created by permafrost thaw and thermokarst development in upland tundra may be a driver of changes in soil environment by redistributing water and snow that secondarily increase soil temperature and moisture that then affect decomposition of soil organic matter and the production of soil CO2. A strong relationship between soil CO2 production and microtopography, temperature, and sites (Table 2-3, Figure 2-5) shows that surface subsidence induces change in soil properties that 25

PAGE 26

stimulates CO2 emissions from soil organic matter decomposition and root respiration. Several field studies have shown relationships between ecosystem respiration and topography driven by factors such as soil moisture, soil texture, and vegetation [Epron et al., 2006; Hanson et al., 1993; Kang et al., 2003]. Soil CO2 production increases with temperature because both root respiration and microbial decomposition of soil organic matter respond to the direct effect of temperature. Most studies observed increased soil respiration and ecosystem respiration as a result of increased air and soil temperature [Davidson and Janssens, 2006; Davidson et al., 2000; Fang and Moncrieff, 2001], and several others observed a positive relationship between soil CO2 emissions and soil moisture in tundra ecosystems [Chapin et al., 1988; Christensen et al., 1998; Illeris et al., 2004a; Illeris and Jonasson, 1999; Oberbauer et al., 1992; Oberbauer et al., 1991], suggesting that either plant or microbial respiration may be limited by soil moisture in tundra ecosystems. The results are consistent with both moisture and temperature having a positive effect on soil CO2 production, with the main effect of microtopography acting to redistribute soil moisture. While thickening of the active layer may expose larger amounts of soil organic matter to microbial decomposition [Romanovsky et al., 1997], I found that ALT was not significantly correlated to soil CO2 production at these sites (Table 2-3). Therefore, I suggest that surface subsidence created by permafrost thaw and thermokarst development stimulated soil organic matter decomposition and root respiration likely as a result of changes in soil properties such as increased soil temperature, moisture, and plant biomass. Soil CO2 flux from the deepest soil layers (200 cm, Figure 2-3) was not different among sites, suggesting that permafrost thaw and thermokarst development did not stimulate deeper soil organic matter decomposition. This was contrary to the original hypothesis that deeper soil layers may contribute more to soil CO2 production following permafrost thaw. In contrast to deep CO2 26

PAGE 27

fluxes, CO2 fluxes in the surface layers (both 00 cm and 10 cm) were different among sites, suggesting that permafrost thaw and thermokarst development stimulated CO2 fluxes in these layers. Even though CO2 concentrations collected from deeper depths (30 and 40 cm) gas samples were, on average, 10 times higher than at 10 cm depth, the deepest soil layer contributed less than 10% to soil CO2 production; high concentrations were due to low diffusivity, which was 1/100 times that of surface layers. I observed increased ALT at the Extensive and Moderate Thaw sites, which extended down to a meter with thermokarst development, as well as increased soil temperatures at 30 and 40 cm [Vogel et al., in press]. However, water saturation throughout the growing season in many subsided places may have prevented faster decomposition of soil organic matter. Therefore, water saturation in the deepest thawed areas may have offset the effect of higher temperature where the most ground subsidence occurred. Finally, I observed high CO2 concentrations early in the growing season (Figure 2-1b), which does not agree with normal CO2 temperature response curves because soil temperatures are still quite low at this time of year and the seasonal thaw depth was shallow. High CO2 concentrations in gas wells below the frozen soil suggest a different interpretation for winter and early spring soil CO2 concentrations beyond the normal flux diffusion model. Assuming the diffusion was near zero for these trapped air pockets, I estimated the winter soil CO2 storage in pore spaces within frozen soils to better understand the potential magnitude of this C storage. To estimate potential soil CO2 storage, I first assumed that the air volume in frozen soils was similar to air-filled porosity in unfrozen soil (measured during the growing season), and that the sampled CO2 concentrations in the early spring represented the storage of CO2 during the winter. Using sampled CO2 concentrations, estimated pore-space volume, and the ideal gas law, I calculate that this winter soil CO2 storage could be between 0.01 g CO2-C m-2 and 1.48 g CO2-C m-2 for a 27

PAGE 28

given 10cm layer of soil. This likely represents a maximum estimate because it assumes that the trapped air pocket is continuous for a given layer. This amount is small compared to CO2 flux from annual respiration, which is on the order of 300-400 g CO2-C m-2 from these sites [Vogel et al., in press], but pulses of high surface CO2 flux during early spring has been observed in several other tundra flux studies [Elberling and Brandt, 2003; Grogan et al., 2004], which may be from release of trapped CO2 in these air pockets when soil starts to thaw. I suggest that these observations may be caused by physical storage and release of winter respiration [Monson et al., 2006; Schimel et al., 2006] within frozen soils due to limited diffusion through ice layers [Albert and Perron, 2000], rather than a stimulation of biological CO2 production during spring thaw. Conclusion In summary, I supported the projection that permafrost thaw and thermokarst development will increase permafrost carbon emissions by stimulating soil organic matter decomposition, but show that the surface and middle of the soil profile are the layers most affected by thaw. This stimulation results from increased soil temperature, moisture, and thickening of the active layer associated with surface subsidence created by permafrost thaw and thermokarst development. Ground surface subsidence created by permafrost thaw and thermokarst development was the best predictor of soil CO2 production but was also correlated to other environmental variables such as soil moisture, indicating that ground subsidence induces changes in soil properties and can be used as an integrated metric for other environmental variables. 28

PAGE 29

Table 2-1. Pearsons pairwise correlation coefficients for the environmental variables used in the soil CO2 production model. Values with star are correlation coefficients and the rest of the values are p-values. Microtopography T VWC ALT Site1 Site2 Microtopography -0.1059* -0.5752* -0.3442* 0.4362* 0.4835* T 0.4679 0.1912* -0.0179* 0.0288* 0.0217* VWC 0.0014 0.3298 0.0258* -0.2657* -0.3709* ALT 0.0675 0.9252 0.8965 -0.1946* -0.3190* Site1 0.0180 0.8798 0.1717 0.2859 0.5000* Site2 0.0079 0.9095 0.0520 0.0752 0.0030 T is soil temperature at 10 cm. VWC is volumetric water content at 10 cm. ALT is active layer thickness. Site1, and 2 are dummy variables of sites represented by 0 and 1. Table 2-2. Soil CO2 production model using single and multiple variables. Values in Model p column represent the p-values of the full model and values in T and Microtopography column represent p-values of each variable in the model. Model R2 R2adj. Model p Microtopography T ALT 0.00 0.9637 VWC 0.11 0.1062 T 0.15 0.0508 Sites 0.29 0.0189 Microtopography 0.33 0.0021 T+Microtopography 0.41 0.36 0.0022 0.0040 0.0898 Sites+T 0.42 0.34 0.0069 0.0401 Sites+Microtopography 0.48 0.41 0.0021 0.0097 Sites+T+Microtopography 0.55 0.46 0.0017 0.0241 0.0985 29

PAGE 30

Figure 2-1. Soil CO2 concentrations measured within the active layer during the growing season from 2005 to 2007. a) Soil CO2 concentrations sampled at 5 cm increments in soil profile across three sites varying in the degrees of permafrost thaw. The moss layer (0 cm) consists of live mosses. The wet layer (5.5 cm) below the moss is moist, and the moisture content varies seasonally with rainfall. The water-logged layer (>17.5 cm) is typically saturated with water for most of the growing season as a result of rain and melt water perched on the ground ice. Mean values are shown with open diamonds and median values with gray diamonds for all soil CO2 concentrations. b) The seasonal trend of soil CO2 concentration over the three-year study. Each point represents a mean value across all three sites for a given sampling period. A single depth (10 cm) is shown for clarity; this depth reflects similar seasonal trends that occurred at all depths. 30

PAGE 31

Figure 2-2. The distribution of volumetric water content (VWC) measured in 2006 and 2007 during the growing season at all gas wells for 10, 20, and 30cm depth. The plots show means with the 75 percentile within the box and the 95 percentile within the error bars. 31

PAGE 32

Figure 2-3. Soil profile CO2 fluxes for the three sites during the growing season (JuneAug) from 2005 to 2007 calculated for 10 cm intervals. Each symbol of soil CO2 fluxes represents a weekly mean averaged by site. Note the scale of the y-axis changes for each depth interval. 32

PAGE 33

Figure 2-4. Mean (SE) growing season soil CO2 production by year. Different letters represent significant differences in means between sites within a study year at < 0.05 using Tukeys paired test adjusted for multiple comparisons. Figure 2-5. Relationship between relative microtopographic position of individual gas wells and soil CO2 production. Microtopography was the best single predictor variable in the soil CO2 production model (R2 = 0.33, p = 0.0021, y = -3.35 0.59x). The symbols are means of log-transformed soil CO2 production from 2005 to 2007 for each gas well. 33

PAGE 34

CHAPTER 3 A SPATIALLY EXPLICIT ANALYSIS TO MODEL CARBON FLUXES AT A LARGER SCALE IN UPLAND TUNDRA WHERE PERMAFROST IS THAWING Introduction Climate change has led to an increase in mean annual air temperature on a global scale [ACIA, 2005]. High latitude ecosystems (boreal forest and tundra) have undergone the most drastic changes over the last 100 years globally [Overpeck et al., 1997; Serreze et al., 2000]. One of the important changes in high latitude ecosystems is the increased temperature of permafrost soil, which contributes to the thawing of permafrost [Hinzman et al., 2005; Osterkamp and Romanovsky, 1999]. In the upland tundra, thawing of ice-rich permafrost can create localized surface subsidence called thermokarst [Davis, 2001; Jorgenson et al., 2006] due to draining water that formerly sustained the ground as ice-wedges of permafrost. Even though thermokarst is caused by thawing of permafrost, it is different from active layer thickening because thermokarst results in permanent landscape depressions [Schuur et al., 2008]. The scale and magnitude of thermokarst affects local hydrology [Jorgenson et al., 2006] and may change soil properties such as increased belowground temperature, moisture, and nutrient availability. Permafrost stores more than 50% of terrestrial carbon (C, 1672Pg of soil C) as soil organic matter [Schuur et al., 2008; Tarnocai et al., in press]. The frozen conditions of permafrost slow down decomposition of annual plant litter inputs, storing them as new soil organic matter every year. Thawing of permafrost and thermokarst development may stimulate organic matter decomposition as a result of altered soil properties, and increased rates of decomposition may in turn exacerbate global scale warming. Some researchers have observed increased net ecosystem exchange of CO2 (NEE) from high latitude ecosystems as a result of increased air temperature and soil temperature [Chapin et al., 2000; Oechel et al., 2000], while others have found that increased soil moisture and temperature corresponded to increased NEE as direct effect 34

PAGE 35

[Oberbauer et al., 1992]. Indirectly, increased temperature may stimulate soil organic matter decomposition and the release of nutrients to soil, which in turn positively feed back to faster decomposition of soil organic matter [Mack et al., 2004]. Alternatively, changes in soil properties caused by permafrost thaw and thermokarst may stimulate carbon uptake by increased primary productivity in the permafrost zone. Many studies have observed the effects of climate change on carbon uptake in high latitude ecosystems such as increased primary production [Schuur et al., 2007] and invasion and enlargement of shrubs and invasion of trees in tundra [Macdonald et al., 1993; Sturm et al., 2005]. Increased primary production and changes in species composition affect ecosystem carbon uptake since woody parts in plants store more carbon than herbaceous parts. Therefore, changes in tundra carbon balance as a result of permafrost thaw and thermokarst development will vary based on processes that stimulate decomposition (soil temperature and moisture availability) and those that stimulate nutrient availability and a change in aboveground productivity. In an upland tundra ecosystem, Schuur et al. [2007] observed an increase in primary productivity and changes in plant species composition from tussock forming sedge species to shrub species as a result of permafrost thaw and thermokarst development as it progressed over several decades. Vogel et al. [in press] found that at the initial stage of permafrost thaw and thermokarst, the ecosystem was a net carbon sink because of increased primary productivity. However, at the later stage (about 50 years of permafrost thaw and thermokarst) the ecosystem was a net carbon source because of increased carbon emissions especially from decomposition of old carbon stored in the deeper permafrost layer. These patterns in net ecosystem exchange (NEE) corresponded to greater permafrost thaw and vegetation productivity. Supporting these 35

PAGE 36

results, Lee et al. [submitted] observed greater belowground CO2 production in areas with increased permafrost thaw, likely as a result of increased soil organic matter decomposition. The previous research conducted a well replicated study over time; however, the measurements were taken at relatively few points (six points per site) throughout the study area. Therefore, it remains unclear whether the factors that affect ecosystem C exchange for these few points also regulate ecosystem C exchange in the larger research area. The first objective of this study was to scale up ecosystem C flux estimates from point scale to plot scale. Intensive sampling at a point scale over time describes large variability (i.e. diurnal C cycles and seasonal C cycles); however, sampling intensively at the landscape scale takes great amount of time and effort. The second objective of this study was to find environmental variables such as soil temperature, moisture, thaw depth, and aboveground biomass that best explain ecosystem C fluxes on an annual time scale. I established a plot scale (50mm) ecosystem C flux study in an Alaskan tundra where permafrost thaw and thermokarst have been observed for several decades. From a previous land surface survey (Figure 3-1), I observed the existence of spatial structures in surface topography created as a result of thermokarst formation and permafrost thaw across the three field sites. Therefore, I was interested in characterizing spatial heterogeneity of environmental properties in the landscape created by permafrost thaw and thermokarst development using geostatistical analysis. I hypothesized that there would be a positive relationship between depth of surface subsidence created by permafrost thaw and thermokarst development and carbon emissions and carbon uptake because increased resource availability would stimulate plant growth as well as microbial decomposition of soil organic matter. 36

PAGE 37

Methods Site Description This study was conducted in an upland tundra site near Healy, Alaska (Eight Mile Lake site: 63o52'42"N, 149o15'12"W, hereafter: EML), just outside of Denali National Park. Increased permafrost temperatures down to 30m belowground have been monitored in this area since the 1980s [Osterkamp and Romanovsky, 1999], and ground subsidence has been observed as a result of permafrost thawing [Osterkamp, 2005]. There is a gentle north-facing slope (<5) throughout the site, such that the flows from water discharging from thawing permafrost drain into nearby EML. Three study sites were established in 2003, according to the estimated age and degree of permafrost thawing and resulting ground subsidence known as thermokarst: Minimal Thaw (least disturbed typical tussock tundra), Moderate Thaw (15 years of thaw), and Extensive Thaw (over 50 years of thaw) [Schuur et al., 2007]. Each site varies in the dominant vegetation type and productivity, likely as a result of permafrost thawing and thermokarst development. Land Surface Survey As a preliminary study, I surveyed the landscape to quantify surface subsidence created by permafrost thaw and thermokarst at the three EML gradient sites (Figure 3-1) in 2004. Fine scale ground subsidence was estimated using Theodolite Total Station surveying equipment (Leica TPS400, Leica Geosystems, St. Gallen, Switzerland). About 12 transects were established within a 50 m 50 m area at each site and 600 () points were surveyed along transects. Sampling points along a single transect were spaced approximately 50 cm apart from each other. Tussock vegetation was carefully avoided to minimize variations in surface topography not directly due to thermokarst. The data points were normalized within a site to remove the overall slope effect; therefore, the only remaining variable was vertical topography, most of which was formed as a result of permafrost thaw and thermokarst development [Schuur et al., in press.]. Lower surfaces 37

PAGE 38

represented depressions created by melting ice wedges, whereas higher surfaces remained at the original elevation and were not likely underlain by ice wedges. This relative elevation measurement representing surface subsidence created by permafrost thaw and thermokarst development will be referred to as microtopography (MT) throughout this chapter. As thermokarst develops, microtopographic depressions become saturated with water, while other elevated patches spots become drier. The EML gradient sites showed different patterns of surface subsidence (Figure 3-1). Minimal Thaw showed weak spatial dependence, or structural variance (see Spatial Autocorrelation in Methods) and minimal variation in surface topography, Moderate Thaw showed moderate spatial dependence and patchy distribution of surface subsidence, and Extensive Thaw showed the most spatial dependence with more widespread surface subsidence throughout the survey area. These patterns observed at the EML gradient sites are often shown as classic examples of different spatial patterns [Fortin and Dale, 2007; Grunwald, 2008]. In 2006, a subset of the surveyed area was further measured at each gradient site, to quantify variables from soil as well as ecosystem CO2 exchange. The plot was an equally spaced 5m grid (hereafter plot or plot scale) within a 50 m 25 m area including 50 sampling points at each plot. I used differential Global Positioning Systems (GPS) to determine the location and topographical surface features of the 50 sampling points in each plot, because the landscape survey was done prior to plot set up. Also, I was interested in fine scale topographical features, which the differential GPS is capable of measuring, since the method provides precision up to 1cm vertically [Little et al., 2003]. One GPS unit (Trimble 5400) was installed at a nearby USGS marker (63'16.56"N, 149'17.92"W) and the other GPS unit was used to measure the coordinates of interest so that the measurements were always corrected relative to the marker to 38

PAGE 39

obtain better accuracy. The longitude and latitude information collected from the differential GPS and the Theodolite Total Station was converted to the Universal Transverse Mercator (UTM) coordinate system using ArcGIS 9.3 (ESRI, 2008). Soil Properties Soil properties were measured at each grid point within the plot at the three EML gradient sites. Soil temperature was measured at 10 cm of depth in 2008 using Ibutton Thermochron temperature loggers (Maxim Inc., Dallas, TX). The soil temperatures were normalized at the mean of 0 to capture relative differences in soil temperatures at the microtopographic variations. Soil thaw depth (TD) was measured during the peak of the growing season (mid-July) in 2006 and 2007 using a 1/16 rod, which was pushed into the ground until it hit the frozen layer and the depth from the surface was recorded. Soil moisture content was measured as volumetric water content (VWC) using a Campbell Scientific CS616 water content reflectometer and a hand-held voltmeter. The VWC was measured at 10 cm of depth at each sampling grid every time CO2 flux measurements were taken. I measured soil moisture as volumetric water content (VWC) using a water content reflectometer (CS616, Campbell Scientific Inc.) and a CR10X data logger. The CS616 was calibrated for the EML gradient sites at nearby points using direct sampling of soil water content made with destructive soil sampling to compare with CS616 values made at those same nearby points. To calibrate the CS616, I vertically installed the probe at 10, 20, and 30 cm depths in the soil at 60 random locations at the EML gradient sites and obtained measurements. Directly afterwards, a 10 cm diameter soil core was removed at each location. Soil cores were weighed wet, then dried at 60C and reweighed to obtain VWC of the soil. I used linear regressions for each depth to obtain the best fit between CS616 values and direct measure of VWC. The means of 2006 and 2007 measurements were used for further analysis. 39

PAGE 40

Aboveground Biomass and Vegetation Index Aboveground plant biomass was estimated for each grid at Minimal, Moderate, and Extensive Thaw sites using the point-framing method at the most productive time of the year (mid-July to early-August) in 2006. The point-framing method quantifies productivity as well as species composition. I used the same method presented in [Schuur et al., 2007] and used the allometric equations developed for the EML gradient sites to estimate total aboveground biomass (g m-2) of a small quadrat (40 cm 40 cm) used in point-framing. I indirectly assessed vegetation biomass using the normalized difference vegetation index (NDVI), which was measured at the same time as aboveground biomass was estimated with point-framing. NDVI is an estimate of vegetation cover measured by the differences between near infrared radiation (NIR) and visible wavelength (VIS) (NDVI = (NIRVIS) / (NIR+VIS)). A hand-held ADC camera (Tetracam Inc., Chatsworth CA) was used to take photographs of the vegetation sampling implemented in the point-framing and later used the TetracamPixelWrench2 software (Tetracam Inc., Chatsworth CA) to process the images to estimate NDVI. Ecosystem Respiration, Gross Primary Production, and Net Ecosystem Exchange of CO2 Ecosystem respiration (Reco), gross primary production (GPP), and net ecosystem exchange (NEE) of CO2 were measured in each plot at Minimal, Moderate, and Extensive Thaw sites. Net ecosystem exchange of CO2 was estimated from measurements of CO2 flux under ambient light and absolute dark (NEElight = GPP Reco; NEEdark = () Reco). The CO2 flux was measured three times at the peak of the growing season (mid to late July) during 2006 and 2007. I used a closed static chamber system to measure CO2 flux [Vogel et al., in press] using an infra-red gas analyzer (Li-820, LI-COR Biosciences, Lincoln NB), as air circulated through attached to a 40 cm 40 cm 40 cm plastic chamber. The flow of air was adjusted as 1.5 L min-1 continuously. Light intensity was measured as photosynthetically active radiation (PAR, mol m40

PAGE 41

2 sec-1) using a LICOR quantum sensor placed inside the chamber. To take the dark CO2 flux measurement, I blocked the light by covering the chamber with a reflective cloth designed to fit the chamber exactly. Reco was corrected by air temperature at 25C. GPP and NEE were corrected by the light intensity. The mean of the three measurements taken in 2006 and 2007 were used for further analyses of Reco, GPP, and NEE. Spatial Autocorrelation I produced semivariograms to evaluate spatial dependence and spatial autocorrelation of the variables measured at each grid within the plots at the three EML gradient sites. These variables include microtopography (MT), soil temperature (T), soil moisture (VWC), active layer thaw depth measured in July (TD), normalized difference vegetation index (NDVI), and total aboveground biomass (B). I also produced semivariograms to evaluate spatial dependence and spatial autocorrelation of ecosystem C fluxes (Reco, GPP, and NEE) at each grid within the plots. Semivariance is a measure of dissimilarity, which therefore increases with increasing distance; if it shows a flat line instead of any form of increase, then the data are randomly distributed in space [Rossi et al., 1992]. Semivariance increases over distance because in nature dissimilarity increases when the things are farther apart from one another [Tobler, 1970]. Over a certain distance, semivariance becomes flat, indicating that the variable is spatially independent and is randomly distributed, but until then it is considered to be autocorrelated. The y-intercept from the model is called Nugget (C0), the y-value of where the semivariogram becomes a plateau is called Sill (C), and the distance of the presence of spatial pattern and autocorrelation is called Autocorrelation Distance (m). Spatial dependence (%) was calculated from Equation 3-1 [Jackson and Caldwell, 1993]. Spatial Dependence (%) = (C-C0) / C 100 (3-1) 41

PAGE 42

The data for spatial autocorrelation were analyzed within each site due to the distances among different sites (200 to 500 m apart). I minimized the total range of analysis to 30 m and assigned each bin as 5 m wide to insure the number of independent samples in each bin. Therefore, in each semivariogram and correlogram, there were at least 200 pairs of data points per distance interval and significance. Semivariograms, ordinary block Kriging, and contour plots were generated with GS+ Version 9.0 (Gamma Design Software, 2008) using the ten nearest neighbors and a distance of 0.5 meter between block centers. Statistical Analysis I used Pearsons pairwise correlations to observe correlations within the variables. The response variables were transformed with natural log to obtain normality. I constructed simple regression models for Reco, GPP, and NEE (response variables) using multiple environmental variables (MT, T, TD, VWC, NDVI, and B) measured at each grid within the plots (explanatory variables) to find best-fit models that describes Reco, GPP, and NEE. This procedure was done for sites separately, instead of using sites as a block, to find any spatial pattern from the model error terms. The sites were more than 200 m apart from one another, which made it difficult to analyze the data at the same time. Also, just by using environmental variables that capture characteristics in different sites, I intended to construct a generalized model of C fluxes for tundra where permafrost is thawing, without being constrained by sites. From this analysis, the spatial mixed-effect ANOVAs are represented by Yk = + kXk + k (3-2) k ~ Normal (0, 2kk) (3-3) where the subscript k is one of ANOVA from multiple regression models used to predict the set of response variables (Y) [MacKenzie et al., 2008]. X is the matrix of data, is the vector of parameters, is the residual error, and 2 is the variance term for the random effect and error 42

PAGE 43

terms (k). I used different covariance models to account for spatial structure, as reflected by the covariance matrix, [Legendre et al., 2002; Littell et al., 2006]. To find best fit models for the response variables, I used manual backward stepwise fit with MT as a fixed variable, because I considered MT as the proxy of change in permafrost thaw and thermokarst development. This procedure was conducted for each regression model from each site. I then constructed regression models with the three sites combined to find sets of environmental variables that best explain Reco, GPP, and NEE regardless of site, because I was interested in finding a generalized model that explained these three variables in the upland tundra outside of the site boundary. I used manual backward stepwise fit with MT as a fixed variable to construct a best fit model that explains ecosystem C fluxes in larger scale. An F-test was used to test the significance of the model at the cost of increasing the number of explanatory variables. The best-fit model was driven by the least number of variables that are significant in the model, when the model was also significant. I used Kenward-Rogers adjustment for the degrees of freedom to eliminate potential sample size bias. The spatial patterns observed from the semivariograms were put in each model as the error term to verify whether it enhanced the model. Akaike Information Criterion (AIC) and BIC were observed to test the significance of using the spatial pattern in the model as well as changes p-value of the model. The statistical analyses were done using SAS 9.0 and R. New Sites to Verify the Model Four additional sites were selected nearby Eight Mile Lake (EML) in 2006 and 2008 to verify the growing season C fluxes model constructed from EML gradient sites. Three of the four sites were located on thaw-slump thermokarst features similar to the thermokarst features at the EML gradient sites; one of these sites was dominated by Ericaceous shrubs (hereafter Shrub). Two of the thaw-slump thermokarst sites (Thaw-slump1 and 2) were dominated by tussock43

PAGE 44

forming sedges; these sites were similar in vegetation and microtopography to the Moderate Thaw site. The fourth site was characterized by a collapse-scar fen thermokarst feature (Sedge-fen), which was distinctive from any other EML gradient sites. This site was dominated by sedges that did not form tussocks, had a thick layer of Sphagnum, and was waterlogged most of the year, which was shallower than 5 cm above the soil surface. Even though this site was waterlogged most of the year, it did not show anoxic conditions in surface soil to 30 cm. The same variables were measured at these four sites as at the EML gradient sites; data were collected from Shrub and Sedge-fen in 2006 and Thaw-slump 1 and 2 in 2008. At each site, three 50 m transects were established, and only 30 grid points that are 5m apart from one another were established. Environmental properties and ecosystem C fluxes were measured at Shrub and Sedge-fen sites (MT, TD, T, VWC, NDVI, and Biomass as environmental properties and Reco, GPP, and NEE as ecosystem C fluxes) in 2006 and at Thaw-slump1 and 2 sites in 2008 (all of the above except Biomass). Microtopography was measured at the four new sites using differential GPS (Shrub and Sedge-fen sites) in 2006 and the Theodolite Total Station (Thaw-slump1 and 2 sites) in 2008. Soil temperature at 10 cm and thaw depth was measured using the same method described above in mid-July 2008. VWC at 10 cm was measured once at Shrub and Sedge-fen sites in mid-July 2006 and once at Thaw-slump1 and 2 sites in mid-July 2008. NDVI was measured using the Tetracam ADC camera once at Shrub and Sedge-fen sites during the peak of the growing season in 2006 and at Thaw-slump1 and 2 sites during the peak of the growing season in 2008. Aboveground biomass was estimated using point-framing method at Shrub and Sedge-fen sites during the peak of the growing season in 2006, but was not measured in Thaw-slump1 and 2 sites. Reco, GPP, and NEE were measured using the same method as above at 44

PAGE 45

Shrub and Sedge-fen sites twice in the growing season of 2006 (June and mid-July), and at Thaw-slump1 and 2 sites once in mid-July 2008. The explanatory variables from these four sites were used to fit into the parameters estimated from the growing season C fluxes model constructed from the EML gradient sites to predict growing season C fluxes. The predicted C fluxes were then compared with measured C fluxes from the four new sites. Pearsons Chi-squared test was conducted for each site using predicted and measured C fluxes to observe goodness-of-fit. Estimating Annual Carbon Fluxes at a Plot Scale The data from Vogel et al. [in press] was used to estimate annual C fluxes from the study site using the explanatory variables that best explained Reco, GPP, and NEE at the plot scale. Vogel et al. [in press] established 6 replicate chamber collars at each of the three sites described in this study (Minimal Thaw, Moderate Thaw, and Extensive Thaw) at a point scale. MT was collected using Theodolite Total Station surveying equipment and normalized with the same method as this study. The three year means (2004-2006) of TD measured at the peak of the growing season (mid-July) was used for this analysis. Aboveground biomass was estimated using point-framing at the peak of the growing season in 2004. Reco, GPP, and NEE were estimated intensively throughout the growing season (May to September) from 2004 to 2006 using static chamber and automated chamber systems, simultaneously. Several measurements of Reco were taken during the winter using the static chamber system to estimate annual Reco, GPP, and NEE of the three sites. Growing season C fluxes (Reco, GPP, and NEE) are highly correlated to annual C fluxes, therefore, I assumed that the environmental variables that explained growing season C fluxes would also explain large variability in annual C fluxes. A simple regression model was constructed using MT, TD, and Biomass for annual Reco, GPP, and NEE to find the model that 45

PAGE 46

would explain C fluxes the best. Later, I used the parameters estimated from annual Reco, GPP, and NEE models at the point scale to estimate annual Reco, GPP, and NEE for the plot scale measurements in this study. Results Correlations and Spatial Patterns of Explanatory and Response Variables Pearsons pairwise correlations (Table 3-1) showed that MT was significantly correlated at = 0.05 level to four of the five measured variables such as T (Pearsons correlation coefficient = -0.2315), VWC (-0.4696), TD (-0.4887), and NDVI (-0.3070), which indicates that surface subsidence created by permafrost thaw and thermokarst development influenced the soil and ecosystem environment. Other variables showed significant but moderate correlations: TD was correlated to T (0.3911), VWC (0.3144), and NDVI (0.3003), and Biomass was correlated to T (0.2158). Semivariograms of variables measured at the plot scale in Extensive Thaw showed more explanatory variables with spatial dependence and greater range of autocorrelation (MT, TD, VWC, NDVI) than those of Moderate (T and TD) and Minimal Thaw (TD and VWC) (Table 3-2). A best fit linear model indicates a lack of significant autocorrelation (nugget only model), and these data are assumed to have random spatial distributions. Biomass did not show any spatial structure at any of the three sites. Contrary to Biomass, NDVI showed moderate spatial dependence at Extensive Thaw (Spatial dependence = 29.7%). Spatial dependence of the variables ranged widely from 29.7 to 91.7%. Semivariograms of response variables (Reco, GPP, and NEE), on the other hand, did not show strong spatial patterns compared to explanatory variables (Figure 3-2). None of the variables from Extensive Thaw showed spatial patterns, but Reco from Minimal Thaw showed a linear pattern as well as did GPP from Minimal Thaw, and NEE from Moderate Thaw. 46

PAGE 47

Contrary to the spatial patterns shown from the response variables, spatial patterns in the error terms (Figure 3-3; Table 3-9) from plot scale linear models of the growing season C fluxes (Table 3-8) showed more spatial patterns. The spatial pattern in the error terms of growing season C flux models at Extensive Thaw site was weak at Extensive Thaw; only NEE showed spatial pattern and spatial dependence was 46%. On the other hand, the error terms in Moderate and Minimal Thaw showed moderate to high spatial dependence from 60 to 86%. Error terms of all three response variable models showed spatial patterns in Moderate Thaw, and error terms of Reco and GPP showed spatial patterns in Minimal Thaw. C fluxes Model Construction I combined the three sites to construct a regression model to explain growing season ecosystem C fluxes (Reco, GPP, and NEE) measured at the plot scale using sets of environmental variables. When all the variables were put in the model, none of the variables were significant in the model (Table 3-3). MT, TD, and Biomass best explained Log(-Reco) (adj. R2 = 0.1986), MT, TD, and Biomass best explained Log(GPP) (adj. R2 = 0.2404), and MT and TD best explained Log(NEE) (adj. R2 = 0.1093). I then used the spatial structure to better explain the model since spatial patterns were observed from the error terms of C fluxes models created at different sites. The fitted covariance models were Gaussian, Exponential, Spherical, Linear, Linear-log, Power and Anisotropic Exponential. AIC and BIC values were lower for Exponential, Spherical, Power, Linear, Linear-log, and Gaussian; however, Anisotropic covariance structure consistently showed slightly higher AIC and BIC values (Table 3-4), indicating that spatial structures were very weak and unnecessary. Verifying the Model Using New Dataset I verified the growing season C flux models developed from EML gradient sites at the four different nearby tundra sites (Figure 3-4). I compared measured growing season C fluxes at the 47

PAGE 48

four new sites and predicted growing season C fluxes at the four new sites using parameters derived from EML gradient sites. The Chi-square goodness-of-fit test showed that the measured and predicted growing season C fluxes at the four new sites were not different (the p-values ranged from 0.99 to 1). This implies that even though measured and predicted values did not fit well at a 1:1 relationship, they were not statistically different. Each site showed a different trend in measured values of C fluxes and predicted growing season C fluxes using the environmental variables measured at the grid from the four sites and regression model parameters derived from EML sites. The Shrub site showed a significant relationship between measured and predicted growing season C fluxes; R2 of growing season Reco was 0.2237 (p = 0.0096), GPP was 0.2822 (p = 0.0107), and NEE was 0.2179 (p = 0.0030). However, the Sedge-fen site only showed a significant relationship in NEE between measured and predicted (R2 = 0.3041, p = 0.0011). I was not able to estimate growing season Reco and GPP for the two Thaw-slump sites, because I did not measure plant biomass at those sites. But measured and predicted growing season NEE did not show a significant relationship at either of the Thaw-slump sites. Estimating Annual C fluxes Using Environmental Variables The same three variables (MT, TD, and Biomass) as growing season C flux models were picked from stepwise regression to construct annual C flux models (Table 3-5). Annual Reco was explained by MT and TD with an adjusted R2 = 0.6630, annual GPP was explained by MT, TD, and Biomass with an adjusted R2 = 0.7519, and annual NEE was explained by MT and Biomass with an adjusted R2 = 0.2775. MT was not significant in any of the C flux models, which may be due to the strong correlation shown between MT and TD (Pearsons correlation coefficient = -0.8302, p < 0.0001). Among these models, annual Reco and GPP models estimated the same group of variables as the model constructed using growing season measurements taken at the plot scale (Table 3-4). 48

PAGE 49

When annual Reco, GPP, and NEE were estimated for the plot scale, the mean annual Reco was -379.5 (.7), -413.5 (.6), and -432.4 (.1) g CO2-C m-2 at Minimal, Moderate, and Extensive Thaw, respectively. The mean annual Reco at Moderate and Extensive Thaw were significantly greater than Minimal Thaw (p = 0.0006), and Extensive Thaw was different from Moderate Thaw at = 0.10 level (p = 0.0812). The mean annual GPP was 349.4 (.7), 421.5 (.0), and 426.4 (.7) g CO2-C m-2 at Minimal, Moderate, and Extensive Thaw, respectively. The mean annual GPP at Moderate and Extensive Thaw were significantly greater than Minimal Thaw (p = 0.0043), but Moderate and Extensive Thaw sites were not statistically different. The mean annual NEE was -43.4 (.8), 0.1 (.6), and -8.8 (.7) g CO2-C m-2 at Minimal, Moderate, and Extensive Thaw, respectively. The mean annual NEE at Moderate and Extensive Thaw were significantly great than Minimal Thaw (p = 0.0076), but Moderate and Extensive Thaw were not statistically different. There was a difference between modeled annual NEE using environmental variables and annual NEE driven by the difference between annual Reco and annual GPP. When the annual NEE was estimated by the difference between annual Reco and annual GPP, it was -30.1 (.3), 8.1 (.5), and -6.0 (.1) g CO2-C m-2 at Minimal, Moderate, and Extensive Thaw, respectively. This may be due to a low R2 in annual NEE regression model (Table 3-5). The Kriging maps of predicted annual Reco and GPP at the three EML gradient sites showed spatial patterns on a larger scale (Figure 3-5). It is interesting that the spatial patterns of predicted annual C fluxes were similar to those of microtopography at the three EML gradient sites presented in Figure 3-1. At Minimal Thaw, predicted annual GPP showed similar spatial patterns as microtopography, whereas at Moderate Thaw, both predicted annual Reco and GPP showed similar patterns as microtopography, and at Extensive Thaw, predicted annual Reco 49

PAGE 50

showed similar patterns as microtopography. This implies that not only the autocorrelation distance of the variables but also the spatial patterns of the variables were similar to microtopography created by permafrost thaw and thermokarst development. This also explains that the spatial pattern in the annual C fluxes at Minimal Thaw was regulated by GPP, Moderate Thaw was regulated by a combination of Reco and GPP, and Extensive Thaw was regulated by Reco. Discussion Recent ecological findings emphasize the importance of the relationship between spatial and temporal scales of variation in predicting ecological phenomena [Wiens et al., 1993]. In this study, I observed spatial patterns in the environment as well as C fluxes at the three levels of permafrost thaw and thermokarst development. I also observed that the spatial patterns in the growing season C fluxes showed similar patterns and autocorrelation distances as those in the environment (Figure 3-1, Figure 3-2, Table 3-2, and Table 3-6). Most carbon flux studies were done either on a point scale, or a much larger landscape scale using eddy covariance or remote sensing method. It is ideal to conduct a large scale study to observe changes in ecological phenomena; however, it takes much time and effort to make observations on larger scales with as much repetition as would be done on a point scale. I showed possibilities of estimating growing season and annual C fluxes in tundra, where permafrost is thawing, on a plot scale (the scale in between point scale and the eddy covariance tower footprint) using the three simple variables collected in the environment such as microtopography, thaw depth, and aboveground biomass (Table 3-3). These three variables also predicted 75% of the variability in the annual C fluxes measured at the same site intensively for three years on a point scale (Table 3-5). The spatial pattern in microtopography was different at each of the three EML gradient sites that corresponded with the different degrees of permafrost thaw and thermokarst 50

PAGE 51

development (Figure 3-1 and Table 3-6). These patterns also corresponded with the changes in soil environment (Table 3-2) as well as the growing season C fluxes (Figure 3-2). Extensive Thaw showed the most variables with spatial patterns. Thaw depth was the only variable that showed a spatial pattern at all three EML gradient sites. At Moderate and Extensive Thaw, thaw depth showed a similar range of autocorrelation as soil temperature at 10cm. Also, Pearsons correlation coefficient showed that both thaw depth and soil temperature were the two variables other than microtopography that were significantly correlated to each other (Table 3-1). The variables with spatial pattern, at both explanatory (Table 3-2) and error terms of response variable models (Table 3-9), showed the autocorrelation distance of 2.9 to 5.7m, which overlapped with that of microtopography (3.5 to 8.1 m) from the preliminary surveying (Table 3-6). This clearly shows that the variables are correlated to one another and also that the spatial pattern follows those of surface subsidence created by permafrost thaw and thermokarst development. These results supported the assumption that permafrost thaw and thermokarst change soil environment, which in turn stimulate C fluxes in the upland tundra. However, the sampling distance (5 m) in this study overlapped with the autocorrelation distance of the variables and the error terms in the growing season C flux models, which suggest that the variables that did not show spatial pattern may show spatial patterns if they were sampled at a shorter distance [Bellier et al., 2007; Legendre and Legendre, 1998]. This was also supported by semivariograms shown in microtopography measured at a finer scale than (50 cm within each transect) at the three sites previously (Figure 3-1). Using the spatial patterns in the model error terms did not change the parameter estimates or significance of the variables in the model even with the presence of spatial pattern in the error terms shown through semivariogram. This suggests that the spatial component did exist in the 51

PAGE 52

models, but it did not play a significant role because the variables used to predict the models have strong spatial structures (Figure 3-3, Table 3-2). Five out of six environmental variables measured showed spatial patterns at a mean distance of 4.5m (Table 3-2) with moderate to high spatial dependence. I postulate that the lack of spatial patterns in Reco, GPP, and NEE models may be due to multicollinearity shown among all of the explanatory variables used in the model as well as correlations shown among explanatory variables and response variables. Spatial autocorrelation was seen in various soil properties such as temperature, moisture, pH, soil organic matter, and nutrients in different ecosystems [Jackson and Caldwell, 1993; Saetre, 1999; Smithwick et al., 2005] at a shorter distance than the sampling distance used in this study. This may be because soil properties at the EML gradient sites showed more gradual changes over larger scales as shown in the distribution of land surface subsidence from the preliminary survey (Figure 3-1) than the systems shown in other studies that often include more homogeneous agricultural systems. The growing season C flux models constructed at the plot scale showed that the combination of microtopography, thaw depth, and aboveground biomass explained a large portion of variability of growing season C fluxes (Table 3-3). These results reflect the importance of microtopography, thaw depth, and biomass out of six variables measured including soil temperature, moisture, and NDVI in predicting growing season C fluxes. This result, however, does not imply that the variables excluded in the models were not important in explaining growing season ecosystem C fluxes. Instead, it implies that these variables may be correlated to other variables such as microtopography, which was fixed in the C flux models and was correlated to the three variables excluded from the growing season C flux models such as soil temperature, moisture, and NDVI (Table 3-1). High correlations between microtopography 52

PAGE 53

and other environmental variables measured support that the changes in the environment such as increased soil temperature, moisture, and aboveground biomass may likely be caused as a result of permafrost thaw and thermokarst development. When thermokarst develops, subsided areas are more likely to collect water and increase soil moisture content [Jorgenson et al., 2006], while leaving other nearby areas drier. Osterkamp et al., [2000] observed that soil temperatures in subsided areas were higher than nearby less subsided soil, likely because subsided ground surfaces trap snow during winter as wind redistributes it across the landscape, which then further insulates soil during winter [Osterkamp, 2007a]. Also, warm winter soil positively feeds back to soil temperatures in summer [Stieglitz et al., 2003]. Increased resource availability such as soil temperature and nutrients may also result in increased shrub density and biomass under permafrost thaw and thermokarst [Schuur et al., 2007]. Overall, my results showed that growing season C fluxes at the plot scale can be estimated by the three environmental variables that represented the variability in the C fluxes. Ecosystem C fluxes change as a function of C emissions and uptake, which are also governed by the environmental factors. Some researchers observed increased Reco as a function of increased soil temperature [Chapin et al., 2000; Oechel et al., 2000; Shaver et al., 1992], increased GPP as a function of increased soil temperature [Oberbauer et al., 2007; Starr et al., 2004], increased Reco as a function of increased soil moisture [Chapin et al., 1988; Illeris et al., 2004b; Oberbauer et al., 1992], increased Reco and GPP as a function of increased NDVI [La Puma et al., 2007], differences in Reco and GPP as a function of different vegetation types [Sullivan et al., 2008], and increased NEE as a function of increased active layer thickness [Weller et al., 1995] in tundra ecosystem. This study focused on the relationship between C fluxes and microtopography created by permafrost thaw and thermokarst development; because I 53

PAGE 54

hypothesized that thermokarst stimulate changes in the environment, which in turn affect ecosystem C fluxes. Several researchers observed relationship between topographic patterns and carbon emissions [Epron et al., 2006; Hanson et al., 1993; Kang et al., 2003] in various ecosystems. Most of the studies that observed relationships between carbon emissions and topographic patterns concluded that this relationship was not a direct effect of topography itself, but due to factors such as changes in soil moisture, soil texture, and vegetation driven by topographic patterns. For example, Riveros-Iregui and McGlynn [in press] showed that different patterns of slope in the topography changed patterns of watershed and soil moisture that in turn affected ecosystem C fluxes. Their results support that topographic patterns are useful predictors of ecosystem C fluxes because they often represent larger scale in the landscape covering the entire watershed area [Riveros-Iregui and McGlynn, in press]. A hydrologic modeling study using DEM at an arctic tundra system [Stieglitz et al., 1999] also showed that using DEM incorporates larger scale watersheds, which fit well with other model parameters. My results show that topographic patterns created by permafrost thaw and thermokarst development, thaw depth, and aboveground biomass are good predictors of changes in growing season ecosystem C fluxes as a result of permafrost thaw and thermokarst in the upland tundra. The variability in annual C fluxes, on the other hand, was explained much more by the same three variables. In tundra ecosystems, C fluxes measured in several studies show that CO2 efflux is much too variable throughout the day and year, and weather conditions that it needs in order to be measured continuously [Heikkinen et al., 2002; Kwon et al., 2006]. Also, several other studies show that aboveground biomass and NDVI was much too variable year to year in tundra ecosystems [Boelman et al., 2005], which in turn affect high variability in ecosystem gas exchange in tundra ecosystems [Boelman et al., 2003]. Whereas, surface subsidence and thaw 54

PAGE 55

depth do not show much variability in daily time scale and are well correlated to inter-annual comparison [Vogel et al., in press.]. How many measurements are needed to accurately represent natural phenomena has always been question to scientists. The previous study [Lee et al., submitted] showed an R2 value of 0.33 for soil CO2 production only by microtopography (with all the variables, an adjusted R2 = 0.48), which indicates the power of repeated measurements in explaining the variability of ecosystem C fluxes. The growing season C flux models (Reco, GPP, and NEE) developed using environmental variables (MT, TD, and Biomass) in this study were constructed with the means of three C fluxes measurements taken at the peak of the growing season in 2005 and 2006. If the response variables were measured repeatedly throughout the growing season in several years, the model would have shown higher R2 value than the current estimates. When the same sets of environmental variables were used to model annual C fluxes measured intensively through out the growing season for three years using both static chamber and automated chamber systems, it explained much more variability of annual C fluxes than growing season C fluxes measured three times during the growing season for two years. However, this study showed the variability that could be explained by taking more measurements throughout the year and the possibilities of explaining annual C fluxes using three environmental variables that could be easily measured in the field. The annual C flux models presented in this study showed possibilities of estimating annual C fluxes at the plot scale using three variables measured in the field (Table 3-5), which also implies that there is a possibility of estimating annual C fluxes with only one time measurement taken in the peak of the growing season. The results are encouraging with regards to scaling up point measurements to a larger scale, such as plots, sites, or landscape. Shaver et al. [2007] showed that NEE could be modeled using base Reco measurements, soil temperature, and leaf 55

PAGE 56

area index with R2 values ranging from 0.62 to 0.98 in tundra ecosystems regardless of the vegetation type. A similar approach as was used in this study of up-scaling was used in Bubier et al. [1993] and Heikkinen et al. [2002], where environmental measurements such as soil temperature and water table depth were used to model CO2 and CH4 emissions in tundra. Also, several studies used point measurements inside the eddy covariance tower footprint or satellite image to verify the scaling [Oechel et al., 2000; Riveros-Iregui et al., 2008]. My study showed scaling of areal and temporal scaling, when other studies only showed either one of areal or temporal scaling. The regression analysis used in this study supports the possibility of making global extrapolations for C fluxes outside of these interior Alaska sites, as a function of microtopography created by permafrost thaw and thermokarst development; however, still there are limits present, as the verifying model only fit with certain vegetation types indicating that it might be site specific (Figure 3-4). The spatial patterns shown in predicted C fluxes at a plot scale (Figure 3-5) coincide with the trends of ecosystem C fluxes shown in previous studies with point scale measurements [Vogel et al., in press; Schuur et al., in press]. After intensively measuring C fluxes at the EML gradient sites, they concluded that the Minimal Thaw site is carbon neutral, the Moderate Thaw site is carbon sink, and the Extensive Thaw site is carbon source. From the ordinary-Kriging map, it seems that the spatial pattern of the annual C fluxes were not affected by Reco nor GPP at Minimal Thaw, by both Reco and GPP in Moderate Thaw, and by Reco in Extensive Thaw. However, there are ways to correct for different sites representing various types of vegetation [Shaver et al., 2007]. Even without correcting for vegetation differences, these results still would be a useful source of verifying the assumptions of a larger scale study using remote sensing, aerial photography or satellite imagery in the future. 56

PAGE 57

Conclusion In this study, I showed spatial patterns of changes in the environment as well as C fluxes created as a result of permafrost thaw and thermokarst development. The correlation coefficients and the autocorrelation distances of surface subsidence were correlated to soil temperature, moisture, and thaw depth, confirming that permafrost thaw and thermokarst development may affect changes in the environment, which in turn may stimulate changes in ecosystem C fluxes. I showed a scaling of growing season C fluxes at a plot scale from a point scale and found that surface subsidence, thaw depth, and biomass explains large portion of the variability in the growing season C fluxes. The spatial patterns shown in the error terms of growing season C flux models were not strong enough to enhance C fluxes model, which suggests that it was not necessary to use spatial pattern in modeling C fluxes in the upland tundra where permafrost thaw and thermokarst development prevail. Surface subsidence, thaw depth, and biomass were also good explanatory variables to estimate annual C fluxes in this study by over 70%, which shows the connection between these measurements over greater space and longer time. This study showed the possibilities of scaling and verifying of the measurements taken at a smaller areal scale and time and successfully extrapolating to a greater area and time. 57

PAGE 58

Table 3-1. Pearsons pairwise correlation coefficients of the explanatory variables at the three EML gradient sites. The values with star are the correlation coefficients and the rest of the values are p-values. MT T VWC TD B NDVI MT -0.2315* -0.4696* -0.4887* -0.0896* -0.3070* T 0.0067 0.0170* 0.3911* 0.2158* 0.0295* VWC 0.0000 0.8362 0.3144* -0.0770* 0.1297* TD 0.0000 0.0000 0.0001 0.0998* 0.3003* B 0.2993 0.0080 0.3488 0.2244 0.1339* NDVI 0.0003 0.7213 0.1148 0.0002 0.1036 MT is surface subsidence measured by relative elevation noted as microtopography T is soil temperature at 10 cm. VWC is volumetric water content at 10 cm. TD is soil thaw depth measured in late July. B is aboveground biomass measured by point-framing method. NDVI is normalized difference vegetation index. Table 3-2. Model parameters of the curves fitted through each semivariogram in environmental variables. The nugget (C0) is intercept of graph, sill (C) is semivariance of each graph where it plateaus, range is the distance where the plateau begins in meters, and spatial dependence ((C-C0) / C 100) is the ratio of structural to population variance. Site Variable Model Nugget Sill Autocorrelation distance Spatial dependence Minimal MT LIN 0.00453 Thaw T LIN 1.668 TD GAU 13.203 23.814 2.9 44.6 VWC EXP 0.00132 0.00498 71.1 73.5 NDVI LIN 0.00155 B LIN 26105.7 Moderate MT LIN 0.00614 Thaw T GAU 0.191 2.292 5.2 91.7 TD GAU 22.997 41.812 5.2 45.0 VWC LIN 0.00187 NDVI LIN 0.00728 B LIN 19144.8 Extensive MT EXP 0.00821 0.03202 71.0 74.4 Thaw T EXP 1.562 2.357 3.5 33.7 TD GAU 86.427 127.500 4.3 32.2 VWC EXP 0.00164 0.00329 39.6 50.2 NDVI EXP 0.00417 0.00593 3.4 29.7 B LIN 17290.9 EXP: Exponential, GAU: Gaussian, LIN: Linear. 58

PAGE 59

Table 3-3. Site combined C fluxes model using measured environmental variables. Reco, GPP, and NEE were in log scale to seek for normal distribution. Values in MT, TD, and B columns are parameter estimates and the values in parenthesis are p-values of each variable in the model. A log transformation of the response variables was used to meet the assumption of normally distributed data. Model R2 Adj. R2 p-value MT TD B Reco MT 0.0956 0.0002 -0.1144 (0.0002) TD 0.1175 <0.0001 0.0074 (<0.0001) B 0.0929 0.0003 0.0004 (0.0003) MT+TD 0.1439 0.1310 <0.0001 -0.0868 (0.0450) 0.0054 (0.0070) MT+B 0.1729 0.1605 <0.0001 -0.1326 (0.0005) 0.0003 (0.0006) MT+TD+B 0.2158 0.1980 <0.0001 -0.0788 (0.0584) 0.0051 (0.0082) 0.0003 (0.0007) GPP MT 0.1224 <0.0001 -0.1301 (<0.0001) TD 0.1924 <0.0001 0.0075 (<0.0001) B 0.0607 0.0038 0.0002 (0.0038) MT+TD 0.2165 0.2047 <0.0001 -0.0662 (0.0450) 0.0060 (0.0001) MT+B 0.1690 0.1565 <0.0001 -0.1229 (<0.0001) 0.0002 (0.0072) MT+TD+B 0.2573 0.2404 <0.0001 -0.0614 (0.0574) 0.0059 (0.0001) 0.0002 (0.0080) NEE MT 0.0806 0.0008 -0.1341 (0.0008) TD 0.1006 0.0002 0.0069 (0.0002) MT+TD 0.1225 0.1093 0.0002 -0.0800 (0.0711) 0.0051 (0.0130) 59

PAGE 60

Table 3-4. The AIC and BIC values of error terms in spatial covariance structure of the growing season C flux models (growing season Reco, GPP, and NEE) shown in Table 3-3. Response Covariance Structure AIC BIC EXP, SPH, POW, LIN, LINL, GAU 177.1 183.1 Reco EXPA 183.1 198.1 EXP, SPH, POW, LIN, LINL, GAU 112.9 119 GPP EXPA 118.9 134 EXP, SPH, POW, LIN, LINL, GAU 458 464.1 NEE EXPA 464 479.1 EXP: Exponential, SPH: Spherical, GAU: Gaussian, LIN: Linear, LINL: Linear-log, POW: Power, EXPA: Anisotropic covariance structure. Table 3-5. Annual C flux models using same group of environmental variables chosen from the plot scale C fluxes model. The p-value column represents p-value of the whole model. MT, T, TD, and B columns represent p-value of the variables in the model. [The data were used with permission from Vogel, J.G., E.A.G. Schuur, C. Trucco, and H. Lee (in press.), The carbon cycling response of tussock tundra to permafrost thaw and thermokarst development, Journal of Geophysical Research-Biogeosciences.] Model R2 adj. R2 p-value MT T TD B Annual Reco MT T TD B MT+TD T+TD MT+T+TD 0.4505 0.2636 0.6540 0.2790 0.7051 0.7532 0.7646 0.6630 0.7203 0.7103 0.0032 0.0293 <0.0001 0.0242 0.0002 <0.0001 0.0002 0.0032 0.7625 0.9340 0.0293 0.0268 0.0930 <0.0001 0.0037 <0.0001 0.0086 0.0242 Annual GPP MT T TD B TD+B MT+TD+B 0.5030 0.1104 0.6525 0.4361 0.8011 0.7984 0.7747 0.7519 0.0014 0.1778 <0.0001 0.0029 <0.0001 0.0001 0.0014 0.9976 0.1778 <0.0001 <0.0001 0.0132 0.0029 0.0044 0.0173 Annual NEE MT T TD B MT+B T+B 0.1447 0.0258 0.1227 0.2610 0.3678 0.3953 0.2775 0.3147 0.1320 0.5240 0.1542 0.0303 0.0403 0.0230 0.1320 0.5407 0.5240 0.0879 0.1542 0.0303 0.0432 0.0085 60

PAGE 61

Table 3-6. Model parameters of the surces fitted through each semivariogram in surface subsidence (MT) measured each site in 2005 shown in Figure 3-1. The sampling area was approximately 50m and 600 points were sampled. Sites Model Nugget Sill Range (m) Minimal Thaw EXP 0.0189 0.0575 6.3 Moderate Thaw EXP 0.0294 0.1641 7.3 Extensive Thaw EXP 0.0006 0.2102 3.5 61

PAGE 62

Table 3-7. Datasets used in each analysis. Analysis ExplanatoryVariables Response Variables Scale Data source Table / Figure no. Kriging UTM MT 50m50m at Minimal, Moderate, and Extensive Thaw sites Survey using Total Station in 2005 Figure 3-1 Table 3-6 Pearsons pairwise correlation All 50m25m plot (Minimal, Moderate, and Extensive) with 5m grid This study Table 3-1 Semivariogram UTM All Reco GPP NEE 50m25m plot (Minimal, Moderate, and Extensive) with 5m grid This study Figure 3-2 Table 3-2 Table 3-9 Growing season C fluxes model All Reco GPP NEE 50m25m plot (Minimal, Moderate, and Extensive) with 5m grid This study Table 3-3 Table 3-3 Verifying growing season C fluxes model MT TD B Reco GPP NEE Four new sites (Shrub, Sedge-fen, Thaw-slump1 and 2) with 5m grid This study Figure 3-4 Annual C fluxes model MT TD B Reco GPP NEE Point measure of 6 replicates at each EML gradient sites Vogel et al., in press Table 3-5 Annual C fluxes estimate MT TD B Reco GPP NEE 50m25m plot (Minimal, Moderate, and Extensive) with 5m grid This study Figure 3-5 All: MT, T, TD, VWC, NDVI, and B 62

PAGE 63

Table 3-8. Growing season C flux models using measured environmental variables to seek for spatial pattern in the error terms of the mixed model. A log transformation of the response variables was used to meet the assumption of normally distributed data. Best fit model was found to reduce any pattern other than remaining spatial pattern. Site Response Var. Model R2 (adj. R2) Model p-value Explanatory Var. p-value Minimal Thaw Reco MT+TD+B 0.155 (0.086) 0.0973 MT (0.3922) TD (0.1352) B (0.0248) GPP MT+T+VWC+NDVI 0.215 (0.125) 0.0688 MT (0.1537) T (0.1305) VWC (0.0582) NDVI (0.0691) NEE MT+T+VWC 0.124 (0.053) 0.1757 MT (0.5046) T (0.1398) VWC (0.0462) Moderate Thaw Reco MT+T+B 0.160 (0.104) 0.0476 MT (0.0623) T (0.0937) B (0.1588) GPP MT+T 0.169 (0.133) 0.0140 MT (0.0266) T (0.0244) NEE MT+T+VWC+B 0.126 (0.046) 0.1953 MT (0.0467) T (0.1304) VWC (0.1192) B (0.2329) Extensive Thaw Reco MT+TD 0.245 (0.210) 0.0011 MT (0.9647) TD (0.0144) GPP MT+TD+B 0.375 (0.330) 0.0001 MT (0.5749) TD (0.0017) B (0.0095) NEE MT+TD+VWC 0.262 (0.208) 0.0513 MT (0.8096) TD (0.0099) VWC (0.1071) 63

PAGE 64

Table 3-9. Model parameters of the curves fitted through each semivariogram in Figure 3-3. The models were constructed for the sites separately to observe spatial patterns of the model. Sites C balance Model Nugget Sill Autocorrelation distance (m) Spatial dependence (%) Minimal Reco GAU 0.00339 0.02338 4.8 85 Thaw GPP EXP 0.00175 0.01170 5.7 85 NEE LIN 0.01394 Moderate Reco GAU 0.00470 0.02914 4.3 84 Thaw GPP EXP 0.00270 0.01920 4.1 86 NEE EXP 0.02510 0.06220 91.0 60 Extensive Reco LIN 0.01743 Thaw GPP LIN 0.01486 NEE EXP 0.01608 0.02971 5.5 46 EXP: Exponential, GAU: Gaussian, LIN: Linear. Table 3-10. Summary statistics of the explanatory variables used in the growing season carbon flux models for the three EML gradient sites. Variables Statistics Minimal Thaw Moderate Thaw Extensive Thaw MT Average -0.28 -0.39 -0.63 Minimum -0.81 -1.04 -1.58 Maximum 0.25 0.39 0.48 Standard deviation 0.27 0.31 0.46 T Average 3.82 5.26 5.24 Minimum 0.28 1.33 1.83 Maximum 6.88 10.30 8.25 Standard deviation 1.31 1.70 1.57 TD Average 44.2 45.3 48.9 Minimum 37.0 38.5 36.8 Maximum 58.4 65.9 82.5 Standard deviation 4.9 6.6 11.1 VWC Average 0.18 0.19 0.19 Minimum 0.13 0.12 0.13 Maximum 0.36 0.31 0.35 Standard deviation 0.05 0.04 0.05 NDVI Average 0.74 0.69 0.74 Minimum 0.65 0.47 0.51 Maximum 0.84 0.82 0.87 Standard deviation 0.04 0.09 0.08 B Average 393.9 485.3 465.2 Minimum 101.3 191.1 161.2 Maximum 750.6 780.3 704.7 Standard deviation 161.5 152.5 130.0 64

PAGE 65

Figure 3-1. Spatial pattern of surface subsidence in meter scale created by permafrost thaw and thermokarst. The spatial patterns were generated by using semivariogram and ordinary Kriging. The negative values indicate subsided surface, whereas the positive values indicate the elevated or non-subsided surface. 65

PAGE 66

Figure 3-2. Semivariograms of growing season C fluxes measured at plot scale for ecosystem respiration (Reco), gross primary production (GPP), and net ecosystem exchange of CO2 (NEE) at Minimal, Moderate, and Extensive Thaw. 66

PAGE 67

Figure 3-3. Semivariograms of the error terms from multiple regression models for ecosystem respiration (Reco), gross primary production (GPP), and net ecosystem exchange of CO2 (NEE) at Minimal, Moderate, and Extensive Thaw. 67

PAGE 68

Figure 3-4. A 1:1 fit of measured and model predicted Reco, GPP, and NEE. Different symbols represent different sites near the EML gradient sites. Figure 3-5. The spatial patterns in predicted annual Reco and GPP at the three EML gradient sites using ordinary Kriging. The values are in gCO2-C m-2. The negative values indicate CO2 emissions, and the positive values indicate CO2 uptake from atmosphere to the ecosystem. 68

PAGE 69

CHAPTER 4 THE RATE OF PERMAFROST CARBON RELEASE UNDER AEROBIC AND ANAEROBIC CONDITIONS Introduction Permafrost soil is distributed across 14% of the global land surface over boreal forest, tundra, and arctic desert [Tarnocai et al., in press], which is estimated to store over 50% of global terrestrial carbon [Schuur et al., 2008]. Frozen conditions of permafrost have prevented plant litter decomposition and thus stabilized soil organic matter (SOM) from decomposition since the last glacial period [Harden et al., 1992]. In the past century, high latitude ecosystems have undergone the most drastic changes due to global scale warming [ACIA, 2005; Serreze et al., 2000]. Increased air and soil temperatures contributed to thawing of permafrost and as a result stimulated microbial decomposition of SOM stored within, releasing greenhouse gases such as carbon dioxide (CO2) and methane (CH4) to the atmosphere. Emissions of these greenhouse gases will positively feedback to warming on a global scale. In ice-rich permafrost areas, permafrost thaw is often followed by land surface subsidence as a result of draining water from the ice-wedges that formerly sustained the ground [Davis, 2001; Jorgenson et al., 2006]. As a result of thaw effects on hydrology, SOM from permafrost can be deposited in an aerobic or an anaerobic environment. On sloped areas or uplands underlain by relatively well-drained soils, thermokarst creates patchy microtopographic depressions [Schuur et al., 2008]. Well drained soils in the thermokarst zone may create optimum soil moisture conditions for microbial decomposition of SOM, which will result in increased CO2 emissions [Zimov et al., 2006]. Alternatively, depending on landscape level topography and degrees of erosion, large excavations or lakes can form as a result of thermokarst development [Jorgenson and Osterkamp, 2005; Jorgenson and Shur, 2007; Smith et al., 2007]. This type of thermokarst development will result in anaerobic conditions in the soil. A recent 69

PAGE 70

study suggested that the CH4 emissions from thermokarst lakes may have played a major role in climate change during the last glacial maximum [Walter et al., 2007]. Depending on the degrees of the surface subsidence, upland thermokarst can result in water-logged conditions and contribute to CH4 emissions through anaerobic SOM decomposition. Recent studies showed that increases in CO2 emissions as a result of permafrost thaw and thermokarst development at well-drained upland tundra was likely due to changes in the environment such as increased soil temperature, moisture, and thaw depth that stimulated microbial decomposition of SOM [Vogel et al., in press; Lee et al., submitted]. Typically, organic matter decomposition occurs at a much faster rate under aerobic conditions than anaerobic conditions. Even under anaerobic conditions, carbon mineralization via CO2 emissions is greater than CH4 emissions. However, CH4 has a much higher global warming potential (GWPCH4 = 25, GWPCO2 = 1) than CO2 in a 100 year time frame [IPCC, 2007]. Even though total greenhouse gas emissions can be much smaller under anaerobic conditions, more reactive greenhouse gas emissions such as CH4 may contribute more to global warming in the long term than CO2. Therefore, the magnitude of climate forcing may depend on the soil conditions, especially aerobic or anaerobic conditions after thawing of permafrost. Even under the same physical conditions, soils with different substrate quality can be expected to differ in the rates of carbon mineralization and thus differ in climate forcing from the greenhouse gas emissions. Generally, organic soils decompose at a faster rate than mineral soils because organic soils contain more carbon compounds available for microorganisms to consume. Soil substrate quality such as the ratio of carbon to nitrogen is often used as a predictor of the rate of SOM decomposition, because SOM with a low C to N ratio decomposes quickly [Gholz et al., 2000]. Microbial community composition, microbial biomass, and the nature of the soil 70

PAGE 71

matrix also affect the rate of decomposition [Chapin et al., 2002]. The soil conditions that affect the rate of decomposition have been studied much under aerobic conditions. Several studies have shown that soil pH affects CH4 emissions [Goodwin and Zeikus, 1987], but less is known about the controls on the rate of decomposition under anaerobic conditions. I tested how the aerobic and anaerobic conditions in soils and soil substrate quality affect CO2 and CH4 emissions from permafrost soil by conducting laboratory soil incubation experiments. I used permafrost soils collected from Alaska and Siberia, which have varying soil characteristics such as the organic and mineral soils, soil acidity, and a wide range of carbon to nitrogen ratios. I hypothesized that total C mineralization would be greater under aerobic conditions, but the global warming effect would be greater under anaerobic conditions because microbial activity is higher under aerobic conditions than anaerobic conditions. Furthermore, I hypothesized that C mineralization will be positively correlated to soil substrate qualities such as %C, %N, and the C to N ratio, because the rate of microbial decomposition is controlled by the substrate quality of soil. Methods Soil Sampling and Preparation Frozen soils were collected from 12 different locations in Alaska and Siberia (Figure 4-1). All of the soil samples were collected in the permafrost zone; however, they varied in C and N content, type of vegetation, parent material, and parent material age. Three soil cores were collected from four different areas in north-slope of the Brooks Range in Alaska near Toolik Lake Biological Station (68.63N, -149.72E). The moist acidic tussock tundra sites were located on the Itkillik I glacial drift (hereafter, Itkillik I), whereas the non-acidic tundra sites were on the Itkillik II glacial drift (hereafter, Itkillik II) [Hobbie et al., 2002], and Sagavanirktok glacial drift (hereafter, Sag), which is much older in parent material age [Walker and Everett, 1991]. The 71

PAGE 72

fourth soil samples were collected from a recently formed thermokarst cave near Toolik Lake (ToolikKarst). Three replicates of soil cores were collected from the three moist acidic tundra sites at Eight Mile Lake (EML) research site located in central Alaska, on the foothills of the Alaska Range outside of Denali National Park (63.88N, -149.25E). The EML site is a natural gradient site containing three different intensities of permafrost thaw and thermokarst development (Minimal Thaw, Moderate Thaw, and Extensive Thaw; Schuur et al., submitted; Vogel et al., in press; hereafter, Ext, Ext-o, Mod, Mod-o, Min, and Min-o for mineral soils and o for organic soils). Three soil cores were collected from Fox, Alaska (64.95N, -149.72E) along the Steese Highway, where the soil layer was exposed vertically due to a land development (hereafter, Fox). Three soil cores were collected from Zelenyi Mys (68.80N, 161.38E) along Kolyma River in Siberia (hereafter, Siberia), which consists of windblown loess called Yedoma [Zimov et al., 2006]. The soil samples were collected within 1m from the surface, except for Fox and Siberia samples that were collected around 10m depth from the surface. All soil samples were kept frozen continuously after collection and shipped frozen to the University of Florida for further analysis. The soils were thawed and separated into organic and mineral layers in the laboratory. For all other sites, only mineral soil layers were used. The organic soils from the three sites from EML were used as a reference organic layer. Laboratory Incubation Experiment and CO2/CH4 Fluxes I prepared 1 L jars with airtight lids and attached luerlock stopcocks on the lid for sampling. Approximately 50 to 100 grams of soil with field moisture content were placed in each jar and kept at 15C for aerobic soil incubation. The incubation temperature was chosen at 15C with the intention of comparing the results from the previous soil incubation study that was conducted at 5 and 15C using the same subset of Siberian soils. Also, it is likely that these soils may be exposed to soil temperatures of 15C in summer when permafrost thaws. For anaerobic 72

PAGE 73

soil incubation, I added 150 ml of deionized water to each jar to completely submerge the soil and flushed them with helium (He) to remove any dissolved oxygen in the water and head space of the jar. I used anaerobic indicator strips (BBL GasPak Anaerobic Indicator Strips) to confirm that the environment in the jar was anaerobic. The incubation jars remained closed throughout the experiment to maintain anaerobic conditions for anaerobic incubation jars and to maintain humidity for aerobic incubation jars. Soils were incubated for over 250 days. Aerobic jars were flushed with humidified CO2-free air when CO2 concentrations in the headspace air exceeded 1%. I sampled 10 ml from the headspace in the 1L jars using an air-tight syringe and measured CO2 with an infrared gas analyzer (LI-COR 6252, Li-COR Biosciences, Lincoln, NE, USA), calibrated with certified CO2-in-air standards with concentrations of 608, 1,000, and 10,000 ppmv. I used a gas chromatograph (Shimadzu GC-FID, Column 1/8 SS 45/60 Carboxen 1000) to measure CH4 from anaerobic incubation. I used 1% CH4 as a standard. For aerobic incubations, I collected gas samples daily for two weeks, then every other day for a month, then weekly for a month, then biweekly for two months, and then monthly. For anaerobic incubation, I collected gas samples weekly for two months, then biweekly for two months, and then monthly, thereafter. I injected the same amount of He gas into the anaerobic incubation jars when I took the gas samples for measurement to prevent negative pressure in the jars from sampling. The same procedure was used for aerobic incubation jars, but humidified CO2-free air was used instead of He gas. Replacing the sampled amount of headspace gas by CO2-free air or He gas resulted in dilution of the CO2 and CH4 concentrations in the jars. This dilution factor was used to calculate CO2 and CH4 concentrations in the jars as well as the gas fluxes. Soil Enzyme Assay I analyzed potential activity of -glucosidase to better understand microbial activity related to carbon mineralization. One subsample of soil (approximately 3 g of field moist soil) was taken 73

PAGE 74

at the initial step of the soil incubation and analyzed for soil enzyme activity. The other subsamples (approximately 3 grams of soil in field moisture content) were incubated under the same conditions as soil incubation study for CO2 and CH4 measurement in 20 mL vials under aerobic and anaerobic conditions. The aerobic vials were kept in field moisture conditions and the anaerobic vials were filled with 5ml of water that submerged the soil sample and were purged with He gas in the beginning of the incubation to maintain the same conditions as in the main soil incubation jars for carbon mineralization. Soil enzyme activities were assayed using the fluorescent model substrate 4-methylumbelliferone (MUF) and 7-amino-4-methyl coumarin (AMC) in acetate buffer at pH 5.0. The soil samples were placed in 10 mL of acetate buffer and homogenized using brief agitation with Tissue Tearor Model 398 (Biospec Products, Bartlesville, OK). A 1/10 dilution of soil slurry was transferred to a microtiter plate and 50 L of enzyme substrate solution was added. Samples were incubated for 2 h in the dark at room temperature. The assay was stopped by adding 10 L of 0.1M NaOH and read on a Bio-Tek Model FL600 fluorometric plate reader (Bio-Tek Instruments, Inc., Winooski, VT) [Prenger and Reddy, 2004]. Standard curves were generated using 0, 0.5, 1.0, 3.0, 5.0, and 10.0 mol substrates and corrected by the quenching ability of the soil using one of each of the soil samples for the same standard curves [Freeman et al., 1995]. After 125 days of incubation, the soils were analyze for changes in potential -glucosidase activity under aerobic and anaerobic conditions using the same methods. Soil C and N Analysis and Stable Isotope Measurements Soil % C and N were measured using a Costech ECS 4010 Elemental Analyzer (Valencia, CA). The 13C/12C isotope ratios and 15N/14N isotope ratios of soils used in the incubation were measured with a Finnigan Delta Plus XL continuous flow mass spectrometer (Finnigan MAT GmbH, Bremen, Germany). The carbon and nitrogen stable isotope 13C and 15N data of soil 74

PAGE 75

samples were obtained by combustion with the Costecht CN Analyzer (Costech Instruments, Milan, Italy), which was coupled to the Finnigan Delta XL Plust mass spectrometer through a ConfloIIt control unit. Results were expressed as delta () values and deviations from standard reference materials were calculated from the following equations (Equation 4-1, 4-2). 15N or 13C = [(Rsample/Rstandard) 1] 1000 (4-1) R = 13C/12C or 15N/14N (4-2) I used NIST certified sucrose standard and peach as working standard (SRM 1547; 13C = -26.06 15N/14N = 1.91 ). Average 1 precision for the isotopes was 0.05 for 13C and 0.07 for 15N. Soil pH Soil pH was measured with oven dried (60C) soil samples. The soils were diluted 1:5 using deionized water and stabilized. I measured the stabilized soil solution using an Orion pH meter (ThermoOrion model 250, Orion Instruments, Beverly, Mass.). The pH of deionized water was 5.71. Analyses The C mineralization from the soil incubation experiment was calculated from multiplying the number of days by mean fluxes of those days. Total C mineralization after 250 days of soil incubation was expressed in two different ways: C mineralization from grams of dry soil and C mineralization from grams of soil carbon. Climate forcing of cumulative C mineralization under anaerobic conditions was calculated by adding cumulative C mineralization via CO2 emissions with cumulative C mineralization via CH4 emissions multiplied by 25 (GWP of CH4). Relative climate forcing was estimated by the ratio between cumulative C mineralized via CO2 emissions under aerobic conditions and CO2 equivalent climate forcing in cumulative C mineralized under anaerobic conditions via CO2 and CH4 emissions. This was expressed as climate forcing of CO2 75

PAGE 76

equivalent. Simple regression analysis was used to obtain the relationships between soil substrate quality, soil -glucosidase activity, and C mineralization. Repeated measures analysis was used to assess differences in means among samples in CO2 and CH4 fluxes per day both in aerobic conditions and anaerobic conditions. Paired t-tests were used to compare the initial -glucosidase activity and after 125 days of incubation. Repeated measures analysis was used to compare the differences in CO2 fluxes for organic soils and mineral soils. Statistical analyses were done using JMP 7.0.2. (SAS Institute Inc. 2007). Results Substrate Quality and Soil Enzyme Activities The soils used in the incubation study showed a wide range of soil % C; the lowest soil %C (1.10%) was found in soils from Fox, Alaska and the highest (41.43%) was found in organic soils from the EML tundra sites in interior Alaska (Moderate Thaw organic layer). The highest mean %C in the mineral soil was 16.46%, which was found in the Extensive Thaw site at EML at a depth of 70 to 83 cm from the surface. Mean % N in soils ranged from 0.15 to 1.01%, which resulted in the range of C to N ratio of 2.5 to 45.8 (Table 4-1). The mean sample C to N ratio of mineral soils ranged from 2.5 to 26.3 and that of organic soils ranged from 40.7 to 45.8. Mean soil pH ranged from 3.32 to 7.23. The surface organic soils were more acidic than deeper mineral soils. The soils with low %C had higher pH levels ranging from slightly acidic to neutral (5.27 to 7.23). The 13C signatures in soil ranged from -29.52 to -24.36 in mineral soils and from -26.81 to -25.60 in organic soils, which were not statistically different. Mineral soils were more enriched in 15N (0.42 2.86) than organic soils (-0.68 to 0.04), and soil 15N was negatively correlated with soil %N (R2 = 0.5719, p = 0.0226). 76

PAGE 77

Carbon Fluxes under Aerobic and Anaerobic Conditions Organic soils showed sharp peaks of CO2 fluxes in the beginning of the incubation, whereas mineral soils showed a gradual increase of CO2 fluxes over the first 3 to 5 days of the incubation study and then gradually decreased. This trend is not common for most soils; however, it is not uncommon in frozen soils after thawing. The freeze-thaw cycle in soils may release nutrients from microbial cell breakdown and show a pulse of CO2 emissions as a result. CO2 fluxes were highest in the beginning of the incubation period under both aerobic and anaerobic conditions (Figure 4-2 and Figure 4-3). The CO2 fluxes from organic soils were an order of magnitude greater than mineral soils under both aerobic and anaerobic conditions. The CO2 fluxes under aerobic conditions reached steady-state by 30 days of incubation, whereas CO2 fluxes under anaerobic conditions did not reach steady-state until 150 days of incubation. The magnitude of CO2 fluxes were different in mineral and organic soils analyzed by repeated measures analysis (p = 0.0535). On the other hand, the CH4 fluxes from some soil samples, such as Itkillik I and II and organic layers from EML tundra sites, did not reach steady-state even after 250 days of incubation. The CH4 fluxes were very low in mineral soils from EML tundra sites and in mineral soils with low %C content. Even though the CH4 fluxes were low in some of the soil samples, the CH4 concentrations in the incubation jars were higher than those in the atmosphere; the lowest CH4 concentrations in the incubation jars were 5ppm. Soil -glucosidase Activity Corresponding to the drop in the rate of CO2 fluxes under aerobic and anaerobic conditions, soil -glucosidase activity decreased significantly in aerobic and anaerobic incubations after 125 days. Initial soil -glucosidase activity ranged from 0.035 to 2.123 mmol gdw-1 hr-1 (Figure 4-4), and was an order of magnitude greater in organic soils than in mineral soils. The range of soil -glucosidase activity in mineral soils was 0.035 to 0.109 mmol gdw-1 hr-1, whereas that of organic 77

PAGE 78

soils was 1.360 to 2.123 mmol gdw-1 hr-1. After 125 days of incubation under aerobic and anaerobic conditions, the soil -glucosidase activity was significantly lower than initial activity in both aerobic (p = 0.0126) and anaerobic conditions (p = 0.0020) using t-tests. Also, soil -glucosidase activity was significantly lower in the soils incubated under anaerobic conditions (p = 0.0294) than under aerobic conditions. Cumulative Carbon Mineralization and Climate Forcing Cumulative C mineralization was calculated based on 250 days of soil incubation. The range of cumulative C mineralized under aerobic conditions was 0.42 to 3.06 mg C gdw-1 in mineral soils and 10.82 to 31.49 mg C gdw-1 in organic soils (Figure 4-5 and Table 4-2). Cumulative C mineralization in organic soils showed greater variations among and within samples than in mineral soils. Under anaerobic conditions, cumulative C mineralization via CO2 and CH4 ranged from 0.09 to 0.86 mg C gdw-1 in mineral soils and from 3.80 to 5.21 mg C gdw-1 in organic soils (Figure 4-5 and Table 4-2). Among the mineral soils, Itkillik I had the highest cumulative C mineralization via CH4 emissions. In most mineral soils, cumulative C mineralization via CH4 emissions was small (0.2% at the lowest) compared to cumulative C mineralization via CO2. In organic soils, however, cumulative C mineralization via CH4 constituted up to 27% of total C mineralization. The ratio of cumulative C mineralization from aerobic to anaerobic incubation ranged from 1.23 to 9.83, showing that the rate of C mineralization was on average 4.2 times faster under aerobic than anaerobic conditions. The mean ratio between cumulative C mineralization under aerobic and anaerobic was 3 after 100 days of incubation, but continuously increased to 4.2 after 100 days of incubation. The trend of cumulative C mineralization per gram of soil C (mgC gC-1) was similar to C mineralization per gram of soil for organic soils, but was very different in mineral soils (Figure 4-6), because the C content in mineral soils ranged from 1.10 to 16.30 % (Table 4-1). The range 78

PAGE 79

of cumulative C mineralization under aerobic conditions was 6.47 to 81.75 mg C g C-1 in mineral and organic soils. In fact, cumulative C mineralization per gram of soil C in Siberia and in the Extensive Thaw organic layer was not statistically different from each other at = 0.05 level, and both Siberia and Fox soil samples showed higher cumulative C mineralization than in the Moderate Thaw organic layer. The range of C mineralization under anaerobic conditions was 1.96 to 22.61 mg C gC-1 in mineral and organic soils. Soils with low C content (Fox and Siberia) also showed greater C mineralization under anaerobic conditions than the organic soil samples. Itkillik I sample had the most amount of C mineralized via CH4 emissions per gram of soil C (mgC gC-1) than any other soil sample. Climate forcing from C mineralized under anaerobic conditions ranged from 0.10 to 12.66 CO2 equivalent per gram of soil (Table 4-2), which resulted in relative climate forcing from 0.2 to 7.5. The lowest relative climate forcing was found in Itkillik I and Minimal Thaw mineral soil samples due to unusually high CH4 emissions. The highest relative climate forcing was found in Toolik Karst soil due to the very low emissions of both CO2 and CH4 under anaerobic conditions. Relative climate forcing in mineral and organic soils were not different. The Relationship between Carbon Mineralization and Substrate Quality Overall, there was an exponential increase in cumulative C mineralization under aerobic and anaerobic conditions in relation to %C, %N, or C to N ratios of the soil samples (Figure 4-7). Cumulative C mineralization under aerobic conditions and %C were exponentially correlated (R2 = 0.8098, p < 0.0001). On the other hand, relative climate forcing estimated from these soil samples were not statistically correlated to %C, %N, and C to N ratio. The log transformed cumulative C mineralized under aerobic and anaerobic conditions were negatively correlated to soil pH (Figure 4-8; p = 0.0002 and R2 = 0.37 for aerobic C mineralization, and p < 0.0001 and 79

PAGE 80

R2 = 0.39 and p < 0.0001 and R2 = 0.50 for anaerobic C mineralization via CO2 and CH4, respectively). Discussion I conducted permafrost soil incubation study under aerobic and anaerobic conditions to estimate C mineralization and the greenhouse gas (CO2 and CH4) emissions in aerobic and anaerobic conditions that may be caused after thawing of permafrost. Many studies were done in controlled environment to investigate the rate of permafrost carbon mineralization; however, this study compared the climate forcing from the greenhouse gas emissions under aerobic and anaerobic conditions. This study showed that cumulative C mineralized during 250 days of soil incubation was 4.2 times greater under aerobic conditions than anaerobic conditions (Table 4-3). The relative climate forcing from these emissions ranged from 0.2 to 7.5 in the wide variety of soil samples used in this incubation study (Table 4-3), when global warming potential of CH4 was considered as 25 times that of CO2 on a 100-year time scale. There was an exponential relationship between carbon mineralization and soil substrate quality such as %C, %N, and C to N ratio (Figure 4-6). The cumulative C mineralized under aerobic conditions was greater than anaerobic conditions (Table 4-3; Figure 4-5), which is consistent with the general understanding that typically, aerobic decomposition of soil organic matter occurs at a faster rate than anaerobic decomposition. The ratio of C mineralized under aerobic and anaerobic conditions may vary depending on how long the soils were incubated and under what conditions the soils were incubated. For example, Updegraff et al. [1995] observed that the rate of C mineralization in Sphagnum and sedge dominated peatland soils under aerobic conditions were 2.0 to 4.3 times higher than that under anaerobic conditions during a 80 week of laboratory incubation study at 15C. On the other hand, Bridgham et al. [1998] observed that the rate of C mineralization was 80

PAGE 81

4.5 to 7.7 higher under aerobic conditions than anaerobic conditions using same soils at 30C after a 59 week of laboratory incubation study. Closer observations in daily CO2 and CH4 fluxes in this study (Figure 4-2 and 4-3) showed that the rate of CO2 emissions under aerobic conditions reached steady-state within 30 days of the incubation, but that under anaerobic conditions did not reach steady-state until 150 days of incubation. Both under aerobic and anaerobic conditions, the CO2 fluxes peaked in the beginning of the incubation; however, the CH4 fluxes started increasing after 50 days of the incubation and did not reach steady state even after 250 days of incubation. While, most soil incubation studies were carried out for short periods of time, one soil incubation study that was conducted for 80 weeks under anaerobic conditions, showed a similar increasing trend in CH4 emissions [Updegraff et al., 1995] as this study. The climate forcing under anaerobic conditions seemed to depend on the ratio of CO2 to CH4 emissions (Table 4-2). Therefore, it is important to understand the soil conditions stimulating CH4 fluxes even though the carbon mineralized via CH4 fluxes was minute in the total carbon mineralized. The CH4 fluxes in this study were much higher than those presented in several other studies [Bridgham et al., 1998; Rodionow et al., 2006]. The two pathways of CH4 production are through decarboxylation of acetate and reduction of CO2 [Paul and Clark, 1996]. In nature, some proportion of the CH4 production is consumed by methanotrophs and is released as CO2 to the atmosphere. Recent studies showed that in certain types of Sphagnum moss communities, methanotrophs form a symbiotic relationship with Sphagnum and consume CH4 at the Sphagnum surface that was produced in the deeper layer [Raghoebarsing et al., 2006; Raghoebarsing et al., 2005]. However, this study was conducted under a controlled environment, such that it was assumed that the methanotrophic activity did not exist, because the incubation jars were kept closed throughout the entire incubation study. In my study, I assumed that the CO2 81

PAGE 82

from anaerobic conditions were produced from the acetotrophic methanogen activity. Many studies show that CO2 reduction is a major pathway of CH4 production in northern peatland environments [Horn et al., 2003; Hornibrook et al., 1997]. Recent findings show that in some fen environments, acetate decarboxylation was predominant in the upper layers and the nutrient rich conditions [Chasar et al., 2000; Kelley et al., 1992]. It is important to point out that there were possible CH4 productions under anaerobic conditions, which increased climate forcing from the gas emitted; however, there was also possible CH4 uptake under aerobic conditions, which may decrease climate forcing. I did not observe N2O emissions in the anaerobic incubation study; however, some other studies have observed N2O emissions under anaerobic conditions, and N2O uptake under aerobic conditions. In permafrost soils, I would expect that N2O emissions to be low, because these northern systems are generally limited by nitrogen availability. The magnitude of C mineralization depended on the substrate quality of the soils (Figure 4-6 and Figure 4-7). When C mineralization was expressed by the fraction of soil C content instead of soil dry weight, soils with relatively less C content lost more C during the same incubation period (Figure 4-6). This indicates that the rate of SOM decomposition varies by the substrate quality of organic matter and not due to the quantity of C in the soil. A positive relationship between %C and cumulative C mineralization was observed in Siberian yedoma soils incubated under aerobic conditions [Dutta et al., 2006] previously. Generally, SOM decomposition in permafrost soils is limited by the nitrogen availability. A long term fertilization study conducted on the north-slope of Alaska showed an increased microbial response to soil available N [Mack et al., 2004]. Also, several freeze-thaw cycle studies concluded that sudden high pulse of CO2 emissions after soil thaw may be due to nutrient leaching from the dead microbial cells after thawing [Lipson et al., 1999; Rivkina et al., 2000]. Soil pH is shown to 82

PAGE 83

affect SOM decomposition [Fierer and Jackson, 2006]; generally, decomposition occurs at a faster rate at a neutral pH than in acidic soils [Chapin et al., 2002]. Also, the greatest rate of methanogenesis was observed under slightly acidic to neutral conditions, rather than in acidic soils [Goodwin and Zeikus, 1987], which is contrasting from the results shown in this study. The trends shown in this study may have been caused by the differences in the characteristics of mineral versus organic soils. For example, the soils with high %C were organic soils, and these soils also had a low pH. Organic soils not only have more complex compositions of soil organic matter, but also contain more diverse microbial community than the mineral soils [Schimel and Chapin, 2006]. When physical and chemical properties in soils are similar, the microbial community composition will be the controlling factor of SOM decomposition [Chapin et al., 2002]. Therefore, my results showed that mineral soils and organic soils have different characteristics and also have different microbial communities that largely affect the decomposition process. Relative climate forcing of the soils ranged from 0.2 to 7.5 (Table 4-2), which was caused as a result of differences in CH4 emissions under anaerobic conditions. The relative climate forcing estimated from these soils contradicted to those found in Schuur et al. [2008] (summarized from Bridgham et al. [1998]), likely due to the wide variety of mineral soils used in this study compared to the study conducted by Bridgham et al. [1998] that only used different types of peat soils. I estimated how changes in the land surface, represented by aerobic and anaerobic conditions created as a result of permafrost thaw and thermokarst development, may affect climate forcing. Northern wetlands are reported as a long-term sink of atmospheric CO2, but a source of CH4 [Smith et al., 2005]. These wetlands and lakes drain several mechanisms, such as stream piracy, tapping, bank overflow, and ice-wedge erosion [Mackay and Slaymaker, 83

PAGE 84

1989; Marsh and Neumann, 2001] and form drain thaw lake basin. When these wetlands and lakes drain, the soil conditions may shift away from being a sink of CO2 and source of CH4,to being a source of CO2 and sink of CH4. Recent studies that reported expansion of thaw lakes in continuous permafrost zone as a result of permafrost thaw [Smith et al., 2005; Walter et al., 2006], observed shrinking of thermokarst ponds and lakes in the discontinuous permafrost zone [Marsh et al., 2009; Riordan et al., 2006; Smith et al., 2005; Yoshikawa and Hinzman, 2003]. Due to thaw lake expansion, Walter et al. [2006] reported a 58% increase in CH4 emissions. I used the estimates of a 6% decrease in northern thermokarst lakes from Smith et al. [2005] and multiplied the land area estimated by Smith et al. [2007] to estimate the total land area and lake area in the northern ecosystem excluding Greenland (Table 4-4). I used relative climate forcing estimated from my study for aerobic and anaerobic conditions as 1 and 0.65 (Table 4-3) and multiplied this value by total area without lakes (aerobic) and total lake area (anaerobic). My calculations show that if lake areas decrease by 6%, climate forcing will increase 0.04% from current conditions. While an increase in climate forcing of 0.04% may seem like a small amount, this value does not take into account 1) the amount of climate forcing due to uptake of methane under aerobic conditions, nor 2) bursts of greenhouse gases, such as a CH4 ebullition observed in thermokarst lakes, while CH4 ebullition could be a big source of changes in climate forcing. My results thus far, support that changes in the arctic landscape as a result of permafrost thaw and thermokarst development may result in increased climate forcing. Conclusion I conducted soil incubation studies to determine the fate of permafrost carbon under aerobic and anaerobic conditions. My results show that C mineralization was 4.2 times greater under aerobic conditions than anaerobic conditions after 250 days of soil incubation. However, when the global warming potential of CH4 was considered to be 25 times that of CO2 in a 10084

PAGE 85

year time scale, then the relative climate forcing ranged from 0.2 to 7.5 depending on the type of soil and how much CH4 was produced. Cumulative C mineralized under aerobic and anaerobic conditions were positively correlated to soil substrate quality such as %C, %N, and C to N ratio. However, relative climate forcing was statistically not correlated to any of the substrate qualities observed. Soil pH, on the other hand, showed negative correlations with cumulative C mineralized under aerobic conditions and C mineralized via CO2 and CH4 emissions under anaerobic conditions, which contrasted existing general knowledge of soil pH and SOM decomposition. I postulate that this may be due to the wide variety of mineral soils observed in this incubation study, and the associated microbial communities that are affecting C mineralization in this study. When I used the recent estimate of overall 6% decrease in thermokarst lake areas in the permafrost zone to estimate changes in climate forcing due to changes in soil aerobic and anaerobic conditions, I calculated that climate forcing will increase by 0.04% from current conditions. My findings suggest that changes in the arctic landscape as a result of permafrost thaw and thermokarst development may result in increased climate forcing. 85

PAGE 86

Table 4-1. Substrate quality of soil samples used in the incubation study. Values in parentheses are standard errors. 86 Sites Sampling locations (N, E) Soil type Depth (cm) %C %N C:N pH 13C 15N Itkillik I (68.6, -149.7) Mineral 42-66 6.06 (0.16) 0.31 (0.01) 19.34 (0.41) 4.14 (0.07) -26.28 (0.03) 1.90 (0.07) Itkillik II (68.6, -149.7) Mineral 40-55.5 9.34 (4.05) 0.65 (0.27) 14.32 (0.77) 4.78 (0.18) -26.27 (1.03) 0.79 (0.11) Sag (68.6, -149.7) Mineral 42.5-66 5.04(0.84) 0.25 (0.01) 19.76 (2.46) 4.27 (0.02) -29.52 (2.42) 0.42 (0.37) Minimal Thaw (63.9, -149.3) Mineral 56-76 13.34 (11.13) 0.56 (0.50) 23.79 (3.70) 4.60 (0.20) -26.71 (0.16) 0.76 (0.09) Moderate Thaw (63.9, -149.3) Mineral 62-90 9.44 (5.81) 0.35 (0.18) 26.25 (4.12) 4.66 (0.12) -27.59 (0.49) 0.62 (0.14) Extensive Thaw (63.9, -149.3) Mineral 70-83 16.46 (0.33) 0.75 (0.05) 21.89 (0.88) 4.01 (0.74) -26.96 (0.12) 0.84 (0.06) ToolikKarst (68.6, -149.7) Mineral 100 2.71(0.57) 0.19 (0.04) 14.31 (0.04) 5.27 (0.01) -25.40 (0.19) 1.13 (0.45) Fox (65.0, -149.7) Mineral 10m 1.10(0.63) 0.15 (0.04) 6.84 (2.31) 7.23 (0.18) -25.87 (0.55) 2.19 (1.04) Siberia (68.8, 161.4) Mineral 10m 1.15(0.05) 0.46 (0.01) 2.48 (0.04) 6.55 (0.42) -24.36 (0.33) 2.86 (0.04) Minimal Thaw (63.9, -149.3) Organic 5-15 40.64 (1.55) 1.01 (0.12) 40.66 (6.36) 3.36 (0.06) -25.60 (0.70) 0.04 (0.64) Moderate Thaw (63.9, -149.3) Organic 5-15 41.43 (0.76) 0.97 (0.19) 45.53 (9.89) 3.32 (0.03) -26.40 (0.21) 0.02 (0.31) Extensive Thaw (63.9, -149.3) Organic 5-15 40.12 (3.30) 0.88 (0.01) 45.76 (3.68) 3.64 (0.14) -26.81 (0.68) -0.68 (0.26)

PAGE 87

Table 4-2. Cumulative C mineralization per gram of dry soil after 250 days of incubation at 15C under aerobic conditions and anaerobic conditions. Relative climate forcing was estimated as the ratio between aerobic cumulative C mineralization and CO2 equivalent climate forcing (CF) of cumulative C mineralization under anaerobic conditions. The values represent means of 3 samples and values in parentheses are standard errors. Aerobic C mineralization Anaerobic CO2-C Anaerobic CH4-C CH4 to CO2 ratio Anaerobic CF CO2 eq. Aerobic / Anaerobic Sites mg gdw-1 mg gdw-1 mg gdw-1 % mgCO2 eq. gdw-1 Relative climate forcing Itkillik I 1.03 (0.63) 0.60 (0.32) 0.25 (0.23) 30.9 (21.4) 6.97 (6.04) 2.9 (2.6) 0.2 Itkillik II 1.81 (0.65) 0.30 (0.05) 0.04 (0.02) 13.9 (9.7) 1.30 (0.65) 5.3 (1.5) 1.4 Sag 0.43 (0.19) 0.14 (0.01) 0.03 (0.00) 19.6 (6.6) 0.79 (0.16) 2.6 (1.1) 0.5 Minimal Thaw 1.25 (0.52) 0.89 (0.02) 0.18 (0.00) 2.0 (1.4) 5.48 (0.06) 3.8 (0.5) 0.2 Moderate Thaw 0.59 (0.18) 0.14 (0.03) 0.01 (0.00) 3.6 (1.4) 0.29 (0.10) 3.8 (0.5) 2.1 Extensive Thaw 3.08 (1.61) 0.48 (0.17) 0.01 (0.01) 2.0 (1.3) 0.81 (0.43) 4.8 (0.6) 3.8 ToolikKarst 0.78 (0.46) 0.10 (0.01) 0.00 (0.00) 0.3 (0.1) 0.10 (0.01) 7.5 (3.7) 7.5 Fox 0.54 (0.22) 0.18 (0.01) 0.00 (0.00) 0.2 (0.1) 0.19 (0.02) 2.8 (1.0) 2.8 Siberia 0.80 (0.24) 0.25 (0.03) 0.00 (0.00) 0.2 (0.1) 0.26 (0.03) 3.1 (0.5) 3.0 Minimal Thaw-O 32.47 (9.35) 3.27 (0.68) 0.38 (0.18) 13.2 (9.1) 12.66 (4.65) 5.8 (2.3) 1.7 Moderate Thaw-O 11.15 (3.04) 3.92 (0.88) 0.29 (0.23) 8.3 (7.4) 11.14 (5.52) 2.8 (0.7) 1.0 Extensive Thaw-O 21.45 (2.77) 4.38 (0.62) 0.19 (0.13) 4.1 (2.9) 9.09 (3.45) 7.6 (1.8) 3.6 87

PAGE 88

Table 4-3. Areas of lake and non-lake in continuous, discontinuous, and sporadic permafrost zone and changes in relative climate forcing when the lake area decreases by 6%. When CO2 emissions under aerobic conditions in this study were set at 1, the ratio of CO2 emissions under anaerobic was 0.194 and CH4 was 0.018. Non-lake Lake Aerobic climate forcing Anaerobic climate forcing Total climate forcing Climate forcing ratio Current 26,693,400 560,200 26,693,400 364,789 27,058,189 1 6% decrease 26,727,012 526,588 26,727,012 342,901 27,069,913 1.000433 88

PAGE 89

Figure 4-1. A map of soil sampling locations indicated by the closed circles. 89

PAGE 90

Figure 4-2. The CO2 fluxes during 250 days of soil incubation at 15C under aerobic conditions. Error bars indicate standard errors. Note the different scale for organic soils. 90

PAGE 91

Figure 4-3. The CO2 (left) and CH4 (right) fluxes during 250 days of incubation at 15C under anaerobic conditions. Error bars indicate standard error. Note the different scale for organic soils. Legends are the same for left and right panels. 91

PAGE 92

Figure 4-4. Soil -glucosidase activity before incubation and after 125 days of incubation at 15C under aerobic and anaerobic conditions. Error bars indicate standard error. (Min = Minimal Thaw, Mod = Moderate Thaw, and Ext = Extensive Thaw) 92

PAGE 93

Figure 4-5. Cumulative C mineralization via CO2 emissions under aerobic conditions (A) and via CO2 and CH4 emissions under anaerobic conditions (B) per grams of dry soil. Error bars indicate standard errors. Note the difference in y-axis between organic and mineral soils. The negative error bars in panel B are standard error for CO2-C and the positive error bars are standard error for CH4-C mineralization. (Min = Minimal Thaw, Mod = Moderate Thaw, Ext = Extensive Thaw, and TK = ToolikKarst) Figure 4-6. Cumulative C mineralization via CO2 emissions under aerobic conditions (A) and by CO2 and CH4 emissions under anaerobic conditions (B) per grams of soil carbon. Error bars indicate standard errors. The negative error bars in panel B are standard error for CO2-C and the positive error bars are standard error for CH4-C mineralization. Abbreviations are the same as Figure 4-5. 93

PAGE 94

Figure 4-7. Relationship between %C, %N, and C to N ratio in soils and cumulative C mineralized per soil dry weight after 250 days of incubation at 15C under aerobic and anaerobic conditions, and relative climate forcing. Figure 4-8. Relationship between soil pH and cumulative C mineralized per grams of soil after 250 days of incubation via CO2 emissions under aerobic conditions, and via CO2 and CH4 emissions under anaerobic conditions. 94

PAGE 95

CHAPTER 5 CONCLUSION In summary, I supported the projection that permafrost thaw and thermokarst development will increase permafrost carbon emissions by stimulating soil organic matter decomposition, but show that the surface and middle of the soil profile are the layers most affected by thaw. This stimulation results from increased soil temperature, moisture, and thickening of the active layer associated with surface subsidence created by permafrost thaw and thermokarst development. Ground surface subsidence created by permafrost thaw and thermokarst development was the best predictor of soil CO2 production but was also correlated to other environmental variables such as soil moisture, indicating that ground subsidence induces changes in soil properties and can be used as an integrated metric for other environmental variables. I showed spatial patterns of changes in the environment as well as C fluxes created as a result of permafrost thaw and thermokarst development. The correlation coefficients and the autocorrelation distances of surface subsidence were correlated to soil temperature, moisture, and thaw depth, confirming that permafrost thaw and thermokarst development may affect changes in the environment, which in turn may stimulate changes in ecosystem C fluxes. I showed a scaling of growing season C fluxes at a plot scale from a point scale and found that surface subsidence, thaw depth, and biomass explains large portion of the variability in the growing season C fluxes. The spatial patterns shown in the error terms of growing season C flux models were not strong enough to enhance C fluxes model, which suggests that it was not necessary to use spatial pattern in modeling C fluxes in the upland tundra where permafrost thaw and thermokarst development prevail. Surface subsidence, thaw depth, and biomass were also good explanatory variables to estimate annual C fluxes in this study by over 70%, which shows the connection between these measurements over greater space and longer time. This study showed 95

PAGE 96

the possibilities of scaling and verifying of the measurements taken at a smaller areal scale and time and successfully extrapolating to a greater area and time. I conducted soil incubation studies to determine the fate of permafrost carbon under aerobic and anaerobic conditions. My results show that C mineralization was 4.2 times greater under aerobic conditions than anaerobic conditions after 250 days of soil incubation. However, when the global warming potential of CH4 was considered to be 25 times that of CO2 in a 100-year time scale, then the relative climate forcing ranged from 0.2 to 7.5 depending on the type of soil and how much CH4 was produced. Cumulative C mineralized under aerobic and anaerobic conditions were positively correlated to soil substrate quality such as %C, %N, and C to N ratio. However, relative climate forcing was statistically not correlated to any of the substrate qualities observed. Soil pH, on the other hand, showed negative correlations with cumulative C mineralized under aerobic conditions and C mineralized via CO2 and CH4 emissions under anaerobic conditions, which contrasted existing general knowledge of soil pH and SOM decomposition. I postulate that this may be due to the wide variety of mineral soils observed in this incubation study, and the associated microbial communities that are affecting C mineralization in this study. When I used the recent estimate of overall 6% decrease in thermokarst lake areas in the permafrost zone to estimate changes in climate forcing due to changes in soil aerobic and anaerobic conditions, I calculated that climate forcing will increase by 0.04% from current conditions. My findings suggest that changes in the arctic landscape as a result of permafrost thaw and thermokarst development may result in increased climate forcing. 96

PAGE 97

LIST OF REFERENCES ACIA (2005), Arctic Climate Impact Assessment Scientific Report, Cambridge University Press, New York, NY. Albert, M.R., and F.E. Perron (2000), Ice layer and surface crust permeability in a seasonal snow pack, Hydrological Processes, 14 (18), 3207-3214. Bellier, E., P. Monestiez, J.P. Durbec, and J.N. Candau (2007), Identifying spatial relationships at multiple scales: principal coordinates of neighbour matrices (PCNM) and geostatistical approaches, Ecography, 30 (3), 385-399. Boelman, N.T., M. Stieglitz, K.L. Griffin, and G.R. Shaver (2005), Inter-annual variability of NDVI in response to long-term warming and fertilization in wet sedge and tussock tundra, Oecologia, 143 (4), 588-597. Boelman, N.T., M. Stieglitz, H.M. Rueth, M. Sommerkorn, K.L. Griffin, G.R. Shaver, and J.A. Gamon (2003), Response of NDVI, biomass, and ecosystem gas exchange to long-term warming and fertilization in wet sedge tundra, Oecologia, 135 (3), 414-421. Bridgham, S.D., K. Updegraff, and J. Pastor (1998), Carbon, nitrogen, and phosphorus mineralization in northern wetlands (vol 79, pg 1545, 1998), Ecology, 79 (7), 2571-2571. Chapin, F.S., N. Fetcher, K. Kielland, K.R. Everett, and A.E. Linkins (1988), Productivity and Nutrient Cycling of Alaskan Tundra Enhancement by Flowing Soil-Water, Ecology, 69 (3), 693-702. Chapin, F.S., P.A. Matson, and H.A. Mooney (2002), Principles of Terrestrial Ecosystem Ecology, 436 pp., Springer-Verlag, New York, USA. Chapin, F.S., A.D. McGuire, J. Randerson, R. Pielke, D. Baldocchi, S.E. Hobbie, N. Roulet, W. Eugster, E. Kasischke, E.B. Rastetter, S.A. Zimov, and S.W. Running (2000), Arctic and boreal ecosystems of western North America as components of the climate system, Global Change Biology, 6, 211-223. Chasar, L.S., J.P. Chanton, P.H. Glaser, D.I. Siegel, and J.S. Rivers (2000), Radiocarbon and stable carbon isotopic evidence for transport and transformation of dissolved organic carbon, dissolved inorganic carbon, and CH4 in a northern Minnesota peatland, Global Biogeochemical Cycles, 14 (4), 1095-1108. Christensen, T.R., S. Jonasson, A. Michelsen, T.V. Callaghan, and M. Havstrom (1998), Environmental controls on soil respiration in the Eurasian and Greenlandic Arctic, Journal of Geophysical Research-Atmospheres, 103 (D22), 29015-29021. Davidson, E.A., and I.A. Janssens (2006), Temperature sensitivity of soil carbon decomposition and feedbacks to climate change, Nature, 440 (7081), 165-173. 97

PAGE 98

Davidson, E.A., and S.E. Trumbore (1995), Gas Diffusivity and Production of Co2 in Deep Soils of the Eastern Amazon, Tellus Series B-Chemical and Physical Meteorology, 47 (5), 550-565. Davidson, E.A., L.V. Verchot, J.H. Cattanio, I.L. Ackerman, and J.E.M. Carvalho (2000), Effects of soil water content on soil respiration in forests and cattle pastures of eastern Amazonia, Biogeochemistry, 48 (1), 53-69. Davis, T.N. (2001), Permafrost: a guide to frozen ground in transition, 351 pp., University of Alaska Press, Fairbanks. Dutta, K., E.A.G. Schuur, J.C. Neff, and S.A. Zimov (2006), Potential carbon release from permafrost soils of Northeastern Siberia, Global Change Biology, 12 (12), 2336-2351. Elberling, B., and K.K. Brandt (2003), Uncoupling of microbial CO2 production and release in frozen soil and its implications for field studies of arctic C cycling, Soil Biology & Biochemistry, 35 (2), 263-272. Elberling, B., and P. Ladegaard-Pedersen (2005), Subsurface CO2 dynamics in temperate beech and spruce forest stands, Biogeochemistry, 75 (3), 479-506. Epron, D., A. Bosc, D. Bonal, and V. Freycon (2006), Spatial variation of soil respiration across a topographic gradient in a tropical rain forest in French Guiana, Journal of Tropical Ecology, 22, 565-574. Fang, C., and J.B. Moncrieff (2001), The dependence of soil CO2 efflux on temperature, Soil Biology & Biochemistry, 33 (2), 155-165. Fierer, N., and R.B. Jackson (2006), The diversity and biogeography of soil bacterial communities, Proceedings of the National Academy of Sciences of the United States of America, 103 (3), 626-631. Fortier, D., M. Allard, and Y. Shur (2007), Observation of rapid drainage system development by thermal erosion of ice wedges on Bylot island, Canadian Arctic Archipelago, Permafrost and Periglacial Processes, 18 (3), 229-243. Fortin, M.J., and M. Dale (2007), Spatial Analysis: a guide for ecologists, 365 pp., Cambridge University Press, Cambridge, UK. Freeman, C., G. Liska, N.J. Ostle, S.E. Jones, and M.A. Lock (1995), The Use of Fluorogenic Substrates for Measuring Enzyme-Activity in Peatlands, Plant and Soil, 175 (1), 147-152. Gaudinski, J.B., S.E. Trumbore, E.A. Davidson, and S.H. Zheng (2000), Soil carbon cycling in a temperate forest: radiocarbon-based estimates of residence times, sequestration rates and partitioning of fluxes, Biogeochemistry, 51 (1), 33-69. 98

PAGE 99

Gholz, H.L., D.A. Wedin, S.M. Smitherman, M.E. Harmon, and W.J. Parton (2000), Long-term dynamics of pine and hardwood litter in contrasting environments: toward a global model of decomposition, Global Change Biology, 6 (7), 751-765. Goodwin, S., and J.G. Zeikus (1987), Physiological Adaptations of Anaerobic-Bacteria to Low Ph Metabolic Control of Proton Motive Force in Sarcina-Ventriculi, Journal of Bacteriology, 169 (5), 2150-2157. Grogan, P., A. Michelsen, P. Ambus, and S. Jonasson (2004), Freeze-thaw regime effects on carbon and nitrogen dynamics in sub-arctic heath tundra mesocosms, Soil Biology & Biochemistry, 36 (4), 641-654. Grunwald, S. (2008), Disaggregation and scientific visualization of earthscapes considering trends and spatial dependence structures, New Journal of Physics, 10. Hanson, P.J., S.D. Wullschleger, S.A. Bohlman, and D.E. Todd (1993), Seasonal and Topographic Patterns of Forest Floor Co2 Efflux from an Upland Oak Forest, Tree Physiology, 13 (1), 1-15. Harden, J.W., E.T. Sundquist, R.F. Stallard, and R.K. Mark (1992), Dynamics of Soil Carbon During Deglaciation of the Laurentide Ice-Sheet, Science, 258 (5090), 1921-1924. Heikkinen, J.E.P., V. Elsakov, and P.J. Martikainen (2002), Carbon dioxide and methane dynamics and annual carbon balance in tundra wetland in NE Europe, Russia, Global Biogeochemical Cycles, 16 (4). Hinzman, L.D., N.D. Bettez, W.R. Bolton, F.S. Chapin, M.B. Dyurgerov, C.L. Fastie, B. Griffith, R.D. Hollister, A. Hope, H.P. Huntington, A.M. Jensen, G.J. Jia, T. Jorgenson, D.L. Kane, D.R. Klein, G. Kofinas, A.H. Lynch, A.H. Lloyd, A.D. McGuire, F.E. Nelson, W.C. Oechel, T.E. Osterkamp, C.H. Racine, V.E. Romanovsky, R.S. Stone, D.A. Stow, M. Sturm, C.E. Tweedie, G.L. Vourlitis, M.D. Walker, D.A. Walker, P.J. Webber, J.M. Welker, K. Winker, and K. Yoshikawa (2005), Evidence and implications of recent climate change in northern Alaska and other arctic regions, Climatic Change, 72 (3), 251-298. Hirsch, A.I., S.E. Trumbore, and M.L. Goulden (2002), Direct measurement of the deep soil respiration accompanying seasonal thawing of a boreal forest soil, Journal of Geophysical Research-Atmospheres, 108 (D3), doi:10.1029/2001JD000921. Hirsch, A.I., S.E. Trumbore, and M.L. Goulden (2004), The surface CO2 gradient and pore-space storage flux in a high-porosity litter layer, Tellus Series B-Chemical and Physical Meteorology, 56 (4), 312-321. Hobbie, S.E., T.A. Miley, and M.S. Weiss (2002), Carbon and nitrogen cycling in soils from acidic and nonacidic tundra with different glacial histories in Northern Alaska, Ecosystems, 5 (8), 761-774. 99

PAGE 100

Horn, M.A., C. Matthies, K. Kusel, A. Schramm, and H.L. Drake (2003), Hydrogenotrophic methanogenesis by moderately acid-tolerant methanogens of a methane-emitting acidic peat, Applied and Environmental Microbiology, 69 (1), 74-83. Hornibrook, E.R.C., F.J. Longstaffe, and W.S. Fyfe (1997), Spatial distribution of microbial methane production pathways in temperate zone wetland soils: Stable carbon and hydrogen isotope evidence, Geochimica Et Cosmochimica Acta, 61 (4), 745-753. Illeris, L., T.R. Christensen, and M. Mastepanov (2004a), Moisture effects on temperature sensitivity of CO2 exchange in a subarctic heath ecosystem, Biogeochemistry, 70 (3), 315-330. Illeris, L., and S. Jonasson (1999), Soil and plant CO2 emission in response to variations in soil moisture and temperature and to amendment with nitrogen, phosphorus, and carbon in northern Scandinavia, Arctic Antarctic and Alpine Research, 31 (3), 264-271. Illeris, L., S.M. Konig, P. Grogan, S. Jonasson, A. Michelsen, and H. Ro-Poulsen (2004b), Growing-season carbon dioxide flux in a dry subarctic heath: Responses to long-term manipulations, Arctic Antarctic and Alpine Research, 36 (4), 456-463. IPCC (2007), Climate Change 2007: The Physical Science Basis. Contribution of Working Group I to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change, Cambridge University Press, New York. Jackson, R.B., and M.M. Caldwell (1993), Geostatistical Patterns of Soil Heterogeneity around Individual Perennial Plants, Journal of Ecology, 81 (4), 683-692. Jahne, B., G. Heinz, and W. Dietrich (1987), Measurement of the Diffusion-Coefficients of Sparingly Soluble Gases in Water, Journal of Geophysical Research-Oceans, 92 (C10), 10767-10776. Jorgenson, M.T., and T.E. Osterkamp (2005), Response of boreal ecosystems to varying modes of permafrost degradation, Canadian Journal of Forest Research-Revue Canadienne De Recherche Forestiere, 35 (9), 2100-2111. Jorgenson, M.T., C.H. Racine, J.C. Walters, and T.E. Osterkamp (2001), Permafrost degradation and ecological changes associated with a warming climate in central Alaska, Climatic Change, 48 (4), 551-579. Jorgenson, M.T., and Y. Shur (2007), Evolution of lakes and basins in northern Alaska and discussion of the thaw lake cycle, Journal of Geophysical Research-Earth Surface, 112 (F2). Jorgenson, M.T., Y.L. Shur, and E.R. Pullman (2006), Abrupt increase in permafrost degradation in Arctic Alaska, Geophysical Research Letters, 33 (2), doi:10.1029/2005GL024960. 100

PAGE 101

Kang, S.Y., S. Doh, D. Lee, V.L. Jin, and J.S. Kimball (2003), Topographic and climatic controls on soil respiration in six temperate mixed-hardwood forest slopes, Korea, Global Change Biology, 9 (10), 1427-1437. Kelley, C.A., N.B. Dise, and C.S. Martens (1992), Correction of BA 95025957. Temporal variations in the stable carbon isotopic composition of methane emitted from Minnesota peatlands. Correction of author name from Cheryl A. Kelly. Erratum published in GLOBAL BIOGEOCHEM CYCLES Vol. 6. Iss. 4. 1992. p. 347, Global Biogeochemical Cycles, 6 (3), 263-269. Kwon, H.J., W.C. Oechel, R.C. Zulueta, and S.J. Hastings (2006), Effects of climate variability on carbon sequestration among adjacent wet sedge tundra and moist tussock tundra ecosystems, Journal of Geophysical Research-Biogeosciences, 111 (G3). La Puma, I.P., T.E. Philippi, and S.F. Oberbauer (2007), Relating NDVI to ecosystem CO2 exchange patterns in response to season length and soil warming manipulations in arctic Alaska, Remote Sensing of Environment, 109 (2), 225-236. Lee, H., E.A.G. Schuur, and J.G. Vogel (submitted), Soil CO2 production in upland tundra where permafrost is thawing. Legendre, P., M.R.T. Dale, M.J. Fortin, J. Gurevitch, M. Hohn, and D. Myers (2002), The consequences of spatial structure for the design and analysis of ecological field surveys, Ecography, 25 (5), 601-615. Legendre, P., and L. Legendre (1998), Numerical Ecology, 870 pp., Elsevier Science, Amsterdam. Lipson, D.A., S.K. Schmidt, and R.K. Monson (1999), Links between microbial population dynamics and nitrogen availability in an alpine ecosystem, Ecology, 80 (5), 1623-1631. Littell, R.C., G.A. Milliken, W.W. Stroup, R.D. Wolfinger, and O. Shabenberger (2006), SAS for Mixed Models, 437-478 pp., SAS Institute Inc., Cary, NC, USA. Little, J.D., H. Sandall, M.T. Walegur, and F.E. Nelson (2003), Application of Differential Global Positioning Systems to monitor frost heave and thaw settlement in tundra environments, Permafrost and Periglacial Processes, 14 (4), 349-357. Macdonald, G.M., T.W.D. Edwards, K.A. Moser, R. Pienitz, and J.P. Smol (1993), Rapid Response of Treeline Vegetation and Lakes to Past Climate Warming, Nature, 361 (6409), 243-246. Mack, M.C., E.A.G. Schuur, M.S. Bret-Harte, G.R. Shaver, and F.S. Chapin (2004), Ecosystem carbon storage in arctic tundra reduced by long-term nutrient fertilization, Nature, 431 (7007), 440-443. 101

PAGE 102

Mackay, J.R., and O. Slaymaker (1989), The Horton River Breakthrough and Resulting Geomorphic Changes in a Permafrost Environment, Western Arctic Coast, Canada, Geografiska Annaler Series a-Physical Geography, 71 (3-4), 171-184. MacKenzie, M.D., E.J.B. McIntire, S.A. Quideau, and R.C. Graham (2008), Charcoal Distribution Affects Carbon and Nitrogen Contents in Forest Soils of California, Soil Science Society of America Journal, 72 (6), 1774-1785. Marsh, P., and N.N. Neumann (2001), Processes controlling the rapid drainage of two ice-rich permafrost-dammed lakes in NW Canada, Hydrological Processes, 15 (18), 3433-3446. Marsh, P., M. Russell, S. Pohl, H. Haywood, and C. Onclin (2009), Changes in thaw lake drainage in the Western Canadian Arctic from 1950 to 2000, Hydrological Processes, 23 (1), 145-158. Millington, R.J. (1959), Gas Diffusion in Porous Media, Science, 130 (3367), 100-102. Monson, R.K., D.L. Lipson, S.P. Burns, A.A. Turnipseed, A.C. Delany, M.W. Williams, and S.K. Schmidt (2006), Winter forest soil respiration controlled by climate and microbial community composition, Nature, 439 (7077), 711-714. Oberbauer, S.F., C.T. Gillespie, W. Cheng, R. Gebauer, A.S. Serra, and J.D. Tenhunen (1992), Environmental-Effects on Co2 Efflux from Riparian Tundra in the Northern Foothills of the Brooks Range, Alaska, USA, Oecologia, 92 (4), 568-577. Oberbauer, S.F., J.D. Tenhunen, and J.F. Reynolds (1991), Environmental-Effects on Co-2 Efflux from Water Track and Tussock Tundra in Arctic Alaska, USA, Arctic and Alpine Research, 23 (2), 162-169. Oberbauer, S.F., C.E. Tweedie, J.M. Welker, J.T. Fahnestock, G.H.R. Henry, P.J. Webber, R.D. Hollister, M.D. Walker, A. Kuchy, E. Elmore, and G. Starr (2007), Tundra CO2 fluxes in response to experimental warming across latitudinal and moisture gradients, Ecological Monographs, 77 (2), 221-238. Oechel, W.C., G.L. Vourlitis, S.J. Hastings, R.C. Zulueta, L. Hinzman, and D. Kane (2000), Acclimation of ecosystem CO2 exchange in the Alaskan Arctic in response to decadal climate warming, Nature, 406 (6799), 978-981. Oh, N.H., H.S. Kim, and D.D. Richter (2005), What regulates soil CO2 concentrations? A modeling approach to CO2 diffusion in deep soil profiles, Environmental Engineering Science, 22 (1), 38-45. Osterkamp, T.E. (2005), The recent warming of permafrost in Alaska, Global and Planetary Change, 49 (3-4), 187-202. Osterkamp, T.E. (2007a), Causes of Warming and Thawing Permafrost in Alaska, EOS, TRANSACTIONS AMERICAN GEOPHYSICAL UNION, 88 (48), 522. 102

PAGE 103

Osterkamp, T.E. (2007b), Characteristics of the recent warming of permafrost in Alaska, Journal of Geophysical Research-Earth Surface, 112 (F2), doi:10.1029/2006JF000578. Osterkamp, T.E., and V.E. Romanovsky (1999), Evidence for warming and thawing of discontinuous permafrost in Alaska, Permafrost and Periglacial Processes, 10 (1), 17-37. Osterkamp, T.E., L. Viereck, Y. Shur, M.T. Jorgenson, C. Racine, A. Doyle, and R.D. Boone (2000), Observations of thermokarst and its impact on boreal forests in Alaska, USA, Arctic Antarctic and Alpine Research, 32 (3), 303-315. Overpeck, J., K. Hughen, D. Hardy, R. Bradley, R. Case, M. Douglas, B. Finney, K. Gajewski, G. Jacoby, A. Jennings, S. Lamoureux, A. Lasca, G. MacDonald, J. Moore, M. Retelle, S. Smith, A. Wolfe, and G. Zielinski (1997), Arctic environmental change of the last four centuries, Science, 278 (5341), 1251-1256. Paul, E.A., and F.E. Clark (1996), Soil Microbiology and Biochemistry, Academic Press, Inc., San Diego, CA. Prenger, J.P., and K.R. Reddy (2004), Microbial enzyme activities in a freshwater marsh after cessation of nutrient loading, Soil Science Society of America Journal, 68 (5), 1796-1804. Raghoebarsing, A.A., A. Pol, K.T. van de Pas-Schoonen, A.J.P. Smolders, K.F. Ettwig, W.I.C. Rijpstra, S. Schouten, J.S.S. Damste, H.J.M. Op den Camp, M.S.M. Jetten, and M. Strous (2006), A microbial consortium couples anaerobic methane oxidation to denitrification, Nature, 440 (7086), 918-921. Raghoebarsing, A.A., A.J.P. Smolders, M.C. Schmid, W.I.C. Rijpstra, M. Wolters-Arts, J. Derksen, M.S.M. Jetten, S. Schouten, J.S.S. Damste, L.P.M. Lamers, J.G.M. Roelofs, H. den Camp, and M. Strous (2005), Methanotrophic symbionts provide carbon for photosynthesis in peat bogs, Nature, 436 (7054), 1153-1156. Riordan, B., D. Verbyla, and A.D. McGuire (2006), Shrinking ponds in subarctic Alaska based on 1950-2002 remotely sensed images, Journal of Geophysical Research-Biogeosciences, 111 (G4). Risk, D., L. Kellman, and H. Beltrami (2002), Soil CO2 production and surface flux at four climate observatories in eastern Canada, Global Biogeochemical Cycles, 16 (4), doi:10.1029/2001GB001831. Riveros-Iregui, D.A., and B.L. McGlynn. 2009. Landscape Structure Controls Soil CO2 Efflux Variability in Complex Terrain: Scaling from Point Observations to Watershed-Scale Fluxes. Journal of Geophysical Research Biogeosciences. doi: 10.1029/2008JG000885 (in press). Riveros-Iregui, D.A., B.L. McGlynn, H.E. Epstein, and D.L. Welsch (2008), Interpretation and evaluation of combined measurement techniques for soil CO2 efflux: Discrete surface chambers and continuous soil CO2 concentration probes, Journal of Geophysical Research-Biogeosciences, 113 (G4). 103

PAGE 104

Rivkina, E.M., E.I. Friedmann, C.P. McKay, and D.A. Gilichinsky (2000), Metabolic activity of permafrost bacteria below the freezing point, Applied and Environmental Microbiology, 66 (8), 3230-3233. Rodionow, A., H. Flessa, O. Kazansky, and G. Guggenberger (2006), Organic matter composition and potential trace gas production of permafrost soils in the forest tundra in northern Siberia, Geoderma, 135, 49-62. Romanovsky, V.E., T.E. Osterkamp, and N.S. Duxbury (1997), An evaluation of three numerical models used in simulations of the active layer and permafrost temperature regimes, Cold Regions Science and Technology, 26 (3), 195-203. Rossi, R.E., D.J. Mulla, A.G. Journel, and E.H. Franz (1992), Geostatistical Tools for Modeling and Interpreting Ecological Spatial Dependence, Ecological Monographs, 62 (2), 277-314. Saetre, P. (1999), Spatial patterns of ground vegetation, soil microbial biomass and activity in a mixed spruce-birch stand, Ecography, 22 (2), 183-192. Schimel, J.P., and F.S. Chapin (2006), Microbial processes in the Alaskan boreal forest, in Alaskas Changing Boreal Forest., edited by F.S.I. Chapin, M.W. Oswood, K. van Cleve, L.A. Viereck, and D.L. Verbyla, pp. 227-240, Oxford University Press, Oxford, UK. Schimel, J.P., J. Fahnestock, G. Michaelson, C. Mikan, C.L. Ping, V.E. Romanovsky, and J. Welker (2006), Cold-season production of CO2 in arctic soils: Can laboratory and field estimates be reconciled through a simple modeling approach?, Arctic Antarctic and Alpine Research, 38 (2), 249-256. Schuur, E.A.G., J. Bockheim, J.G. Canadell, E. Euskirchen, C.B. Field, S.V. Goryachkin, S. Hagemann, P. Kuhry, P.M. Lafleur, H. Lee, G. Mazhitova, F.E. Nelson, A. Rinke, V.E. Romanovsky, N. Shiklomanov, C. Tarnocai, S. Venevsky, J.G. Vogel, and S.A. Zimov (2008), Vulnerability of permafrost carbon to climate change: Implications for the global carbon cycle, Bioscience, 58 (8), 701-714. Schuur, E.A.G., K.G. Crummer, J.G. Vogel, and M.C. Mack (2007), Plant species composition and productivity following permafrost thaw and thermokarst in alaskan tundra, Ecosystems, 10 (2), 280-292. Schuur, E.A.G., J.G. Vogel, K.G. Crummer, H. Lee, J.O. Sickman, and T.E. Osterkamp (in press.), The impact of permafrost thaw on old carbon release and net carbon exchange from tundra. Serreze, M.C., J.E. Walsh, F.S. Chapin, T. Osterkamp, M. Dyurgerov, V. Romanovsky, W.C. Oechel, J. Morison, T. Zhang, and R.G. Barry (2000), Observational evidence of recent change in the northern high-latitude environment, Climatic Change, 46 (1-2), 159-207. Shaver, G.R., W.D. Billings, F.S. Chapin, A.E. Giblin, K.J. Nadelhoffer, W.C. Oechel, and E.B. Rastetter (1992), Global Change and the Carbon Balance of Arctic Ecosystems, Bioscience, 42 (6), 433-441. 104

PAGE 105

Shaver, G.R., L.C. Johnson, D.H. Cades, G. Murray, J.A. Laundre, E.B. Rastetter, K.J. Nadelhoffer, and A.E. Giblin (1998), Biomass and CO2 flux in wet sedge tundras: Responses to nutrients, temperature, and light, Ecological Monographs, 68 (1), 75-97. Shaver, G.R., L.E. Street, E.B. Rastetter, M.T. Van Wijk, and M. Williams (2007), Functional convergence in regulation of net CO2 flux in heterogeneous tundra landscapes in Alaska and Sweden, Journal of Ecology, 95 (4), 802-817. Smith, L.C., Y. Sheng, G.M. MacDonald, and L.D. Hinzman (2005), Disappearing Arctic lakes, Science, 308 (5727), 1429-1429. Smith, L.C., Y.W. Sheng, and G.M. MacDonald (2007), A first pan-Arctic assessment of the influence of glaciation, permafrost, topography and peatlands on northern hemisphere lake distribution, Permafrost and Periglacial Processes, 18 (2), 201-208. Smithwick, E.A.H., M.C. Mack, M.G. Turner, F.S. Chapin, J. Zhu, and T.C. Balser (2005), Spatial heterogeneity and soil nitrogen dynamics in a burned black spruce forest stand: distinct controls at different scales, Biogeochemistry, 76 (3), 517-537. Starr, G., D.S. Neuman, and S.F. Oberbauer (2004), Ecophysiological analysis of two arctic sedges under reduced root temperatures, Physiologia Plantarum, 120 (3), 458-464. Stieglitz, M., S.J. Dery, V.E. Romanovsky, and T.E. Osterkamp (2003), The role of snow cover in the warming of arctic permafrost, Geophysical Research Letters, 30 (13). Stieglitz, M., J. Hobbie, A. Giblin, and G. Kling (1999), Hydrologic modeling of an arctic tundra watershed: Toward Pan-Arctic predictions, Journal of Geophysical Research-Atmospheres, 104 (D22), 27507-27518. Sturm, M., J. Schimel, G. Michaelson, J.M. Welker, S.F. Oberbauer, G.E. Liston, J. Fahnestock, and V.E. Romanovsky (2005), Winter biological processes could help convert arctic tundra to shrubland, Bioscience, 55 (1), 17-26. Sullivan, P.F., S.J.T. Arens, R.A. Chimner, and J.M. Welker (2008), Temperature and microtopography interact to control carbon cycling in a high arctic fen, Ecosystems, 11 (1), 61-76. Takahashi, A., T. Hiyama, H.A. Takahashi, and Y. Fukushima (2004), Analytical estimation of the vertical distribution of CO2 production within soil: application to a Japanese temperate forest, Agricultural and Forest Meteorology, 126 (3-4), 223-235. Tang, J.W., D.D. Baldocchi, Y. Qi, and L.K. Xu (2003), Assessing soil CO2 efflux using continuous measurements of CO2 profiles in soils with small solid-state sensors, Agricultural and Forest Meteorology, 118 (3-4), 207-220. Tobler, W.R. (1970), Computer Movie Simulating Urban Growth in Detroit Region, Economic Geography, 46 (2), 234-240. 105

PAGE 106

Updegraff, K., J. Pastor, S.D. Bridgham, and C.A. Johnston (1995), Environmental and Substrate Controls over Carbon and Nitrogen Mineralization in Northern Wetlands, Ecological Applications, 5 (1), 151-163. Vogel, J.G., E.A.G. Schuur, C. Trucco, and H. Lee (in press.), The carbon cycling response of tussock tundra to permafrost thaw and thermokarst development, Journal of Geophysical Research-Biogeosciences. Vourlitis, G.L., and W.C. Oechel (1999), Eddy covariance measurements of CO2 and energy fluxes of an Alaskan tussock tundra ecosystem, Ecology, 80 (2), 686-701. Walker, D.A., and K.R. Everett (1991), Loess Ecosystems of Northern Alaska Regional Gradient and Toposequence at Prudhoe Bay, Ecological Monographs, 61 (4), 437-464. Walter, K.M., M.E. Edwards, G. Grosse, S.A. Zimov, and F.S. Chapin (2007), Thermokarst lakes as a source of atmospheric CH4 during the last deglaciation, Science, 318 (5850), 633-636. Walter, K.M., S.A. Zimov, J.P. Chanton, D. Verbyla, and F.S. Chapin (2006), Methane bubbling from Siberian thaw lakes as a positive feedback to climate warming, Nature, 443 (7107), 71-75. Welker, J.M., J.T. Fahnestock, and M.H. Jones (2000), Annual CO2 flux in dry and moist arctic tundra: Field responses to increases in summer temperatures and winter snow depth, Climatic Change, 44 (1-2), 139-150. Weller, G., F.S. Chapin, K.R. Everett, J.E. Hobbie, D. Kane, W.C. Oechel, C.L. Ping, W.S. Reeburgh, D. Walker, and J. Walsh (1995), The arctic flux study: A regional view of trace gas release, Journal of Biogeography, 22 (2-3), 365-374. Wiens, J.A., N.C. Stenseth, B. Vanhorne, and R.A. Ims (1993), Ecological Mechanisms and Landscape Ecology, Oikos, 66 (3), 369-380. Yoshikawa, K., and L.D. Hinzman (2003), Shrinking thermokarst ponds and groundwater dynamics in discontinuous permafrost near Council, Alaska, Permafrost and Periglacial Processes, 14 (2), 151-160. Zimov, S.A., E.A.G. Schuur, and F.S. Chapin (2006), Permafrost and the global carbon budget, Science, 312 (5780), 1612-1613. 106

PAGE 107

107 BIOGRAPHICAL SKETCH Hanna Lee received Master of Agriculture fr om Korea University in 2004 on Cadmium phytoremediation of agricultural fields at the Division of Environmental and Ecological Engineering under the supervision of Dr. JeongGyu Kim. She received her Ph.D. from the University of Florida in the summer of 2009 under the supervision of Dr. Ted Schuur on Changes in carbon cycling under thawing permafro st from the Department of Botany. She will be working with Dr. Heather Throop at New Mexico State University and Dr. Thom Rahn at Los Alamos National Laboratory as a post-doctoral fellow on Decomposition study in the dry lands starting in May 2009.