Citation
Detection of Turfgrass Stress Using Ground Based Remote Sensing

Material Information

Title:
Detection of Turfgrass Stress Using Ground Based Remote Sensing
Creator:
Frank, Jason Hamilton
Place of Publication:
[Gainesville, Fla.]
Publisher:
University of Florida
Publication Date:
Language:
english
Physical Description:
1 online resource (96 p.)

Thesis/Dissertation Information

Degree:
Master's ( M.S.)
Degree Grantor:
University of Florida
Degree Disciplines:
Horticultural Sciences
Environmental Horticulture
Committee Chair:
Unruh, Joseph B.
Committee Members:
Lee, Won Suk
Trenholm, Laurie E.
Brecke, Barry J.
Judge, Jasmeet
Graduation Date:
5/1/2008

Subjects

Subjects / Keywords:
Atrial heart septal defects ( jstor )
Biomass ( jstor )
Crop circles ( jstor )
Crops ( jstor )
Dehydration ( jstor )
Moisture content ( jstor )
Reflectance ( jstor )
Remote sensing ( jstor )
Spectral reflectance ( jstor )
Turf grasses ( jstor )
Environmental Horticulture -- Dissertations, Academic -- UF
Genre:
Electronic Thesis or Dissertation
bibliography ( marcgt )
theses ( marcgt )
Horticultural Science thesis, M.S.

Notes

Abstract:
Many turf managers and researchers still use time and labor intensive techniques to manage and assess turf that originate from decades ago. Many of these management practices have been proven repeatedly to work in a variety of situations to assess turfgrass stress, but may be time consuming and inconsistent. However, there are new developments in ways of assessing and mapping stress that increase the efficiency by which one manages and assesses turf. With increasing water and other environmental restrictions turf managers and researchers need be aware of technological advances in their industry. Two field experiments were conducted at the University of Florida, West Florida Research and Education Center near Jay, FL. The first was the detection of leaf nitrogen concentration and soil volumetric water content using remote sensing technology. Plots were arranged in a randomized complete block design and treatments were arranged in a split plot factorial design with four N levels split across four irrigation regimes with three replications per treatment. The second experiment used remote sensing technology to assess biomass production created by applications of plant growth regulators. Plots were arranged in a factorial design observing three different PGRs, each at four incremental levels with three replications per treatment. Spectral measurements were taken from two remote sensing devices. A ground based vehicle mounted optical sensor Crop Circle (Model ACS-210) (Holland Scientific) which produces its own light to compute a normalized difference vegetation index (NDVI) which is calculated as (880nm-650nm) / (880nm+650nm). A hand-held hyperspectral spectroradiometer (FieldSpec Pro; Analytical Spectral Devices, Inc., Boulder, CO) with a spectral range of 300 to 2500 nm was also used to collect reflectance data. All reflectance readings were taken in full sunlight between the hours of 1100 and 1400 Central Standard Time (CST) to minimize variance caused by solar radiation. Plots were ground-truthed using time domain reflectometry for soil moisture content (VWC) and clipping samples obtained from plots were dried, weighed for biomass, and tested for Total Kjeldahl N content. Visual ratings were also taken. The results from the nitrogen and irrigation experiment indicate reflectance data best modeled soil VWC, visual ratings, and some treatment effects from N. Highest correlations for VWC were achieved from PLS regression on ASD data (r2=0.71) and the Crop Circle device (r2=0.70). The results from the PGR experiment modeled the biomass reduction expected from increasing rates of flurprimidol and trinexapac-ethyl compared to the untreated control, but not ethephon. Similar trends were observed in NDVI and leaf area index (LAI) indices computed from Crop Circle and ASD instruments, as well as several stress indices computed from the ASD device. If remote sensing technology could be used to adequately assess turf parameters, and was coupled with the global positioning system (GPS) and geographic information systems (GIS) technologies, irrigations and chemical applications could be applied on a site specific basis. In this setting, turf managers could potentially reduce inputs thereby reducing cost and environmental impact. ( en )
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.
Thesis:
Thesis (M.S.)--University of Florida, 2008.
Local:
Adviser: Unruh, Joseph B.
Statement of Responsibility:
by Jason Hamilton Frank

Record Information

Source Institution:
UFRGP
Rights Management:
Copyright by Jason Hamilton Frank. Permission granted to the University of Florida to digitize, archive and distribute this item for non-profit research and educational purposes. Any reuse of this item in excess of fair use or other copyright exemptions requires permission of the copyright holder.
Embargo Date:
7/11/2008
Classification:
LD1780 2008 ( lcc )

Downloads

This item has the following downloads:


Full Text





validated through a single sample cross-validation. For cross-validation, ten percent of the

sample was omitted for prediction purposes so that the number of factors chosen creates the

minimal predicted residual sum of squares (PRESS). This process was repeated so that every

observation was used exactly once for cross-validation.

Results and Discussion

Influence of Irrigation Treatments

ANOVA (Tables 2-1, 2-2, 2-3) indicated no irrigation treatment effects were significant.

This is likely due to rainfall received during the experiment which eliminated differential soil

VWC. As stated previously in the materials and methods, irrigation treatments were only

applied once during the experiment due to the fact that the weekly cumulative ET calculations

(Appendix C) based on daily rainfall and ET values (Appendix B) were only in deficit of more

than 1.27cm once during the experiment. In review of the average rainfall and ET data from the

last 3 years, the experiment should be run earlier in the year when cumulative ET minus rainfall

differences are not as great (Appendix D).

Nitrogen Effects

ANOVA indicated a large source of variation in bermudagrass response to N treatments.

Leaf tissue N concentration, color, quality, NDVI and LAI from both optical sensing devices, as

well as many individual reflectance wavelengths and reflectance ratios indicated a significant

response to N treatments (Tables 2-1, 2-2, 2-3). This can be attributed to the ability of N to

increase overall turf health since it is the single most important macronutrient to the plant

(Hopkins, 1999).

NDVI and LAI from the Crop Circle device indicated a sampling date effect (Table 2-1).

Additionally, reflectance at wavelength 813 nm, LAI, and 915/975 nm and 865/725 nm ratios

quantified from the ASD device also indicated a sampling date effect (Tables 2-1, 2-2, 2-3).









basis. In this setting, turf managers could potentially reduce chemical input thereby reducing the

high cost of broadcast PGR applications.









935/661 nm, Stress1 computed as 706/760 nm, and Stress2 computed as 706/813 nm. The

results indicated that reflectance data was not successful in explaining variability in the models

for clipping yield. However, Kruse et al. (2006) found conducting Partial Least Squares

regression on hyperspectral data (400-1,100nm) was an adequate predictor of biomass in

creeping bentgrass (Agrostis stolinifera L.)

With the introduction of hyperspectral data, which typically measures more than 200

wavelengths, there has also been investigation of different statistical methods to quantify and

interpret remotely sensed data. Traditionally, linear regression is the simplest way to understand

variance. However, in interpreting reflectance data, it has been found that other regression

techniques better assess plant status. Partial Least Squares Regression (PLS) is a method used to

predict variables when there are a large number of factors, typically more than the number of

observations (Tobias, 1997). The use of multiple variables is generally tested through Multiple

Linear Regression (MLR) (Tobias, 1997). However, the limitation with MLR is when the

number of factors becomes too large the model may become over-fitting (Tobias, 1997). This is

when it uses a large number of variables but only a few account for most of the variation, also

called latent factors (Tobias, 1997). Almost perfect models may be achieved; however they will

not be highly predictive of new data. When prediction is the obj ective then PLS regression is a

highly useful tool (Tobias, 1997).

Currently, the maj ority of remote sensing research is done on agronomic food crops and

forested areas. There is only limited research available in the area of remote sensing and

turfgrass management (Trenholm et al., 1999b). If conditions across the entire golf course were

uniform in microclimates, soil types, turf varieties, pest densities, nutrient status, and irrigation

performance there would be little need for advanced sensor systems to measure the variability of









research is available in the area of turfgrass management (Trenholm et al., 1999). Further

research is needed to incorporate remote sensing into turfgrass management (Trenholm et al.,

1999). If conditions across the entire golf course were uniform in microclimates, soil types, turf

varieties, pest densities, nutrient status, and irrigation performance there would be little need for

advanced sensor systems to measure the variability of these parameters to better meet the needs

on a locational basis (Stowell and Gelerntner, 2006). However, that is not the case.

Optimistically, Bell et al. (2002a) concludes "If an optical sensing (In reference to remote

sensing) system and software can be economically produced, reasonably priced, and mounted

effectively on normal maintenance equipment, a turf practitioner could save enough money in

fertilizers and pesticides to pay for the equipment. This approach would be useful for reducing

the amount of fertilizers and pesticides needed to adequately manage large turf areas. The use of

optical sensing to determine fertilizer rate before or during application could increase turf

uniformity and possibly, turf health. Sensor maps of large turf areas could be used to signal turf

decline and provide an early warning system for turf mangers."









Research conducted by Xiong et al. (2007) showed seasonal NDVI and LAI effects in response

to N treatments.

There was a significant improvement in visual color between no N applied and N applied

at all levels across all dates (Table 2-4). Visual Quality rating (Table 2-4) and 650 nm

reflectance from the Crop Circle device were both greater at the highest N rate (100 kg ha l)

compared to no N applied (Table 2-5). Similarly, reflectance wavelengths (510 nm, 535 nm, 545

nm, 550 nm, and range average 630-690 nm) (Table 2-6), ratios (605/515 nm, 915/975 nm, and

865/725 nm) (Table 2-7), and indices (NDVI, Stress1, and Stress2) derived from ASD data

detected the high rates of applied N (Table 2-7).

Nitrogen X Date Interaction

It is apparent from the N X Date interaction means that there is a significant sequential

response to increasing N treatments on percent leafN concentration the first three sampling dates

(22 June, 12 July, and 25 July 2007) (Tables 2-8, 2-9, 2-10, and 2-11). NDVI and LAI indices

calculated from the Crop Circle device (Table 2-8) also show this response as well as LAI (Table

2-8), 813 nm and 935 nm (Table 2-10), and 915/975 nm and 865/925 nm ratios (Table 2-11)

calculated from the ASD reflectance data. This response in reflectance data can be attributed to

the ability of the devices to detect overall increases in turf health due to a sequential vigor

response in reaction to varying N treatments. This agrees with the work of Bell et al. (2002a,

2004).

The last two sampling dates (22 Aug. and 4 Sept. 2007) do not show a response to

increasing N treatments for any parameter measured that was significant by date (Tables 2-8, 2-

9, 2-10, and 2-11). This is likely due to a waning N response, since N had not been applied for

33 and 46 days respectively for the final two sampling dates.










Table 3-6. Partial Least Squares regression coefficients on hyperspectral data from ASD device to predict volumetric water content
(VWC), biomass, and visual quality (rated on NTEP 1-9 scale 9 is best and 6 is acceptable) in fairway height bermudagrass
(Cynodon dactylon X C. transvaalensis, 'Princess 77) averaged over five dates (22 June, 12 July, 25 July, 22 Aug., and 4
Sept. 2007)
Calibration No. of FactorsY r2 SE n
VWC 5 0.36 0.05 195
Biomass 2 0.03 0.05 195
Visual Quality 8 0.26 0.03 195
SNumber of factors required to achieve a minimal Predicted Residual Sum of Squares (PRESS) of prediction for the partial least
squares regression model.
z Field Spec Pro; Analytical Spectral Devices, Inc., Boulder, CO












Table 3-4. Coeffieient estimates (r2) Of ASD reflectance data at individual wavelengths, as well as range averages from 630-690nm
and 2080-23 50nm to volumetric water content (VWC) visual color, quality, density ratings, and biomass production from
fairway height hybrid bermudagrass (Cynodon dactylon X C. transvaalensis, 'Princess 77) averaged over Hyve dates (22
June, 12 July, 25 July, 22 Aug., and 4 Sept. 2007). All numbers in column headings are expressed nanometers (nm).
ASD z (nm)
r2 510 535 545 550 661 735 755 813 935 2132
VWC 0.12*** 0.06** 0.05* 0.04* 0.07*** 0.02* 0.04* 0.06* 0.11*** 0.43***
Color 0.35*** 0.25*** 0.25*** 0.25*** 0.42*** 0.05* 0.01 0.01 0.02* 0.11***
Density 0.51*** 0.39*** 0.38*** 0.39*** 0.53*** 0.07* 0.01 0.01 0.02 0.06**
Quality 0.48*** 0.35*** 0.35*** 0.35*** 0.53*** 0.06* 0.01 0.01 0.02 0.08***
Biomass 0.00 0.00 0.00 0.01 0.01 0.02 0.03* 0.04* 0.05* 0.01
*, **, *** Signifieant at the 0.05, 0.01, and 0.001 probability levels, respectively
z FieldSpec Pro; Analytical Spectral Devices, Inc., Boulder, CO
















40
ASD *
35 y = 0.7166x + 6.231
U R2 = 0.7083 4 e


20 4
a ~~***4 *~
~ zj *


10 4* *i *

10


0 5 10 15 20 25 30 35 4
-5
Actual VWC


Figure 2-1. Partial Least Squares regression on hyperspectral data from ASD device with a range Of 350-2500 nm and a spectral range
of 1 nm for prediction of volumetric water content (VWC) collected from fairway height hybrid bermudagrass (Cynodon
dactylon X C. transvaalensistr~r~r~r~r~r~ Burtt-Davy), four irrigation rates (60, 80, 100, and 120% estimated ET values), four N rates
(0, 25, 50, and 100 kg ha )~, averaged over all dates (22 June, 12 July, 25 July, 22 Aug., and 4 Sept. 2007).









CHAPTER 3
USE OF GROUND-BASED REMOTE SENSING TECHNOLOGY TO ASSESS BIOMASS
PRODUCTION IN RESPONSE TO PLANT GROWTH REGULATOR APPLICATIONS

Introduction

Remote sensing is defined as obtaining information about an obj ect, area, or phenomenon

by analyzing data acquired by a device that is not in contact with that obj ect, area, or

phenomenon (Lillesand and Keifer, 1987). For many years researchers have entertained the idea

of measuring various plant parameters using remote sensing technology. Plant light interception

significantly influences growth and physiological responses (Salisbury and Ross, 1992). When

light is intercepted by the plant it is absorbed, transmitted, or reflected (Salisbury and Ross,

1992). Using remote sensing technology to quantify the light that is reflected could help to

detect the onset of turfgrass stress (Ikemura and Leinauer, 2006). Real world applications of

remote sensing technology for use in turfgrass management are still in their infancy; however,

studies have shown that image analysis and various remote sensing devices have strong potential

to detect a variety of turfgrass stresses (Ikemura and Leinauer, 2006). Recent investigations into

this technology show that there is potential for turfgrass managers to manage and assess turf

more efficiently. The use of remote sensing technology to estimate turfgrass stress could

significantly decrease the time and labor required to assess these levels in traditional ways, thus

reducing cost as well (Osborne et al., 2002).

The many great achievements in U.S. agricultural productivity can be attributed to the use of

agricultural chemicals including fertilizer and pesticides (Lee et al., 1999). However, the

increased dependence on pesticides and fertilizers have heightened many environmental

concerns especially in Florida due to the sandy soils and heavy rainfall, which increases potential

for heavy runoff and leaching of chemicals, if applied in excess (Min and Lee, 2003). With

increasing water and environmental restrictions, turf managers and researchers must be aware of












Table 2-2. Analysis of variance of individual reflectance bands from ASD reflectance hyperspectral data with a range of 3 50-2500 nm
and spectral resolution of 1 nm collected from fairway height hybrid bermudagrass (Cynodon dactylon X C. transvaalensis,
Burtt-Davy), four irrigation rates (60, 80, 100, and 120% estimated ET values), four N rates (0, 25, 50, and 100 kg ha )~,
and five dates (22 June, 12 July, 25 July, 22 Aug., and 4 Sept. 2007) combined.
ASDz (nm)
Source DF 510 535 545 550 635 661 735 755 813 935 2132
Rep 2 0.0429 0.0002 0.0001 0.0001 0.0596 0.1303 0.0001 0.0001 0.0001 0.0001 0.0584
Date 4 0.0001 0.0001 0.0001 0.0001 0.0001 0.0001 0.0001 0.0001 0.0001 0.0001 0.0001
Irrigatio 3 0.2045 0.4189 0.4541 0.4650 0.2572 0.1908 0.0999 0.1148 0. 1094 0.1134 0.1102
Dxl 12 0.0826 0.2307 0.2667 0.2756 0.0948 0.0659 0.9956 0.9554 0.9698 0.9933 0.0774
error a 12
Nitroge 3 0.0029 0.0002 0.0001 0.0001 0.0039 0.0116 0.9039 0.0275 0.0053 0.0073 0.097
IxN 9 0.5871 0.3050 0.2683 0.2631 0.6254 0.6496 0.0874 0.4093 0.4387 0.4533 0.726
DxN 12 0.4677 0.2312 0.2048 0.2034 0.4701 0.5078 0.4558 0.7430 0.0222 0.1810 0.7521
error b 36
Probability of greater F ratio (P>F) for Date, Irrigation, and N treatment components
z FieldSpec Pro; Analytical Spectral Devices, Inc., Boulder, CO










Table 3-1. Evaluation of means of biomass (g), volumetric water content (VWC), visual ratings: color, quality, density (rated on
NTEP 1-9 scale 9 is best and 6 is acceptable), from fairway height hybrid bermudagrass (Cynodon dactylon X C.
transvaalensistr~r~r~r~r~r~ 'Princess 77) over five dates (22 June, 12 July, 25 July, 22 Aug., and 4 Sept. 2007) combined.
Biomass (g) VWC Color Quality Density
Trinexapac-ethyl
0.05 kg a.i. hal 52.47 a-c 35.73 a 8.00 ab 8.00 ab 7.87 a-c
0.1 kg a.i. ha-l 45.20 bc 34.45 a-c 8.10 ab 8.05 ab 8.00 ab
0.2 kg a.i. ha-l 47.20 a-c 32.87d 7.73 a-c 7.80 a-c 7.93 ab
0.4 kg a.i. ha-l 43.80 bc 34.20 b-d 7.83 a-c 7.98 ab 8.00 ab
Flurprimidol
0.3 kg a.i. ha-l 53.93 a-c 35.85 a 8.10 ab 8.17 a 8.10 ab
0.6 kg a.i. ha-l 49.93 a-c 34.85 a-c 8.19 a 8.11 a 8.17 a
1.1 kg a.i. ha-l 48.13 a-c 35.23 ab 8.07 ab 8.08 a 8.10 ab
2.3 kg a.i. ha-l 36.27 c 34.72 a-c 7.83 ab 7.93 ab 7.83 a-c
Ethephon
3.8 kg a.i. ha-l 45.73 bc 35.21 ab 7.96 ab 7.91 ab 7.80 a-c
7.6 kg a.i. ha-l 56.07 a-c 35.24 ab 7.63 bc 7.58 b-d 7.67 b-d
15.2 kg a.i. hal 66.13 ab 33.88 b-d 7.47 c 7.33 cd 7.40 cd
30.4 kg a.i. hal 66.27 ab 33.37 cd 7.40 c 7.25 d 7.30 d

Untreated 71.60 a 34.01 b-d 7.87 a-c 7.77 a-c 7.73 a-d
Means in the same column followed by the same letter are not significantly different (LSD; P< 0.05)









these parameters to better meet the needs on a locational basis (Stowell and Gelernter, 2006).

However highly variable agronomic conditions create the need for further research to incorporate

the use of remote sensing into turfgrass management (Trenholm et al., 1999b). The obj ective of

this research was to assess the ability of three remote sensing instruments to detect soil moisture

and leaf nitrogen levels of Tifsport bermudagrass turf.

Materials and Methods

The field experiment was conducted at the University of Florida, West Florida Research

and Education Center near Jay, FL. Plots were arranged in a randomized complete block design

and treatments were arranged in a split plot factorial design with four N levels (sub plot) split

across four irrigation regimes (whole plot) with three replications per treatment (Appendix A).

Whole plots were 4.88m x 4.88m in size and subplots within the whole plots were 1.22m x

4.88m in size. Whole plots were spaced 4.88m apart on each side to minimize the impact of drift

from irrigation from plot to plot. Ammonium sulfate was applied on 14 June 2007 and 20 July

2007 at 0, 25, 50, and 100 kg hal with a Gandy drop spreader and then watered in with 0.5 cm of

supplemental irrigation. Irrigation was applied at 60%, 80%, 100%, and 120% of weekly

estimated ET values obtained from an onsite weather station that is part of the Florida

Agricultural Weather Network (http:.//fawn.ifas.u-fl .edu/) when ET was in deficit of 1.27cm or

more. This occurred only on 24 August 2007 (Appendix B-D). Plots were mowed 2 to 3 times

per week at a height of 1.2 cm with a Toro Reelmaster 3100-D. Data were collected five times

(22 June, 12 July, 25 July, 22 Aug., and 4 Sept.) throughout the growing season in 2007. All

measurements were taken within a one hour time period and collected between 1100-1300 hrs.

Plots were rated visually for color, quality, and density based on the standard 1-9 National

Turfgrass Evaluation Program (NTEP) rating scale where 9 is the highest possible rating, 6 is the

lowest acceptable rating, and 1 indicates dead turf. Volumetric water content (VWC) readings












Table 2-11i. Evaluation of means of ratios 915 nm/975 nm and 865 nm/925 nm computed from analytical spectral device (ASD) with a
range of 350-2500 nm and a spectral resolution of 1 nm, collected from fairway height hybrid bermudagrass (Cynodon
dactylon X C. transvaalensistr~r~r~r~r~r~ Burtt-Davy), four irrigation rates (60, 80, 100, and 120% estimated ET values), four N rates
(0, 25, 50, and 100 kg ha- ), and five dates (22 June, 12 July, 25 July, 22 Aug., and 4 Sept. 2007) combined, which had N
treatment X date interactions.


915/975 nm_ (ASDz)
25 kg ha-l 50 kg ha
1.03a2 1.03a2
1.02bl2 1.02bl2
1.02b23 1.02b2
0.95dl2 0.94dl2
0.98c1 0.9801


865/925 nm (ASDz)


0 kg ha-
1.03a2
1.01b2
1.01b3
0.95dl
0.9901


100 kg ha-
1.04al
1.02b1
1.03b1
0.94d2
0.9901


0 kg ha-
2.12a3
2.00b2
1.96b3
1.66c1
1.68c


25 kg ha-
2.20a23
2.02b2
2.03b23
1.63cl2
1.60cl


50 kg ha-' 100 kg ha'
2.23a2 2.35al
2.04b2 2.13b1
2.04b2 2.12b1
1.61cl2 1.63c2
1.57c1 1.63c1
0.05); means in the same


Date
6/22
7/12
7/25
8/22
9/4


Means in the same column and category followed by the same letter are not significantly different (LSD; P,
row and category followed by the same number are not significantly different (LSD; P, 0.05).
z FieldSpec Pro; Analytical Spectral Devices, Inc., Boulder, CO










Table 3-3. Coefficient estimates (r2) Of Crop Circle reflectance data, Normalized Difference Vegetation Index (NDVI), Leaf Area
Index (LAI), 880nm, and 650nm, and, ASD reflectance data computed into NDVI, LAI, Stress1, and Stress2 indices to
volumetric water content (VWC) visual color quality, density ratings, and biomass production from fairway height hybrid
bermudagrass (Cynodon dactylon X C. transvaalensis, 'Princess 77) averaged over five dates (22 June, 12 July, 25 July, 22
Aug., and 4 Sept. 2007).
Crop Circlez ASDY
r2 NDVIx LAI" 880nm 650nm NDVIx LAI" Stress 1 Stress2u
VWC 0.10*** 0.12*** 0.43*** 0.00 0.11*** 0.10*** 0.09*** 0.10***
Color 0.18*** 0.16*** 0.00 0.28*** 0.23*** 0.22*** 0.19*** 0.21***
Density 0.22*** 0.19*** 0.03* 0.21*** 0.28*** 0.23*** 0.24*** 0.25***
Quality 0.22*** 0.18*** 0.02 0.24*** 0.29*** 0.24*** 0.23*** 0.26***
Biomass 0.06** 0.10*** 0.00 0.09*** 0.03* 0.08*** 0.05* 0.05***
*, **,** Significant at the 0.05, 0.01, and 0.001 probability levels, respectively
z Model ACS-210, Holland Scientific
YFieldSpec Pro; Analytical Spectral Devices, Inc., Boulder, CO
xNDVI= (880-650nm)/(880+650nm)
~" LAI= 880/650nm
SStressl= 706/760nm
a Stress2= 706/813nm












Table 2-13. Coefficients estimates (r2) Of ASD reflectance readings of individual reflectance wavelengths taken from an Analytical
spectral device (ASD) with a range of 3 50- 2500 nm and a spectral resolution of 1 nm to %N concentration leaf tissue, soil
volumetric water content (VWC) visual color, quality, density ratings, and biomass production of fairway height hybrid
bermudagrass (Cynodon dactolon X C. transvaalensistr~r~r~r~r~r~ Burtt-Davy), four irrigation rates (60, 80, 100, and 120% estimated
ET values), four N rates (0, 25, 50, and 100 kg ha )~, and averaged over all dates (22 June, 12 July, 25 July, 22 Aug., and 4
Sept. 2007).
ASDz Reflectance (nm)

Parameter
510 535 545 550 635 661 735 755 813 935 2132
%/N 0.00 0.02* 0.02* 0.02* 0.01 0.02* 0.25*** 0.29*** 0.29*** 0.28*** 0.02*
VWC 0.08*** 0.01 0.00 0.36 0.19*** 0.23*** 0.38*** 0.53*** 0.52*** 0.49*** 0.29***
Color 0.28*** 0.15*** 0.13*** 0.13*** 0.36*** 0.38*** 0.09*** 0.23*** 0.24*** 0.19*** 0.40***
Quality 0.52*** 0.32*** 0.31*** 0.31*** 0.60*** 0.62*** 0.06*** 0.22*** 0.23*** 0.17*** 0.69***
ulDensity 0.55*** 0.37*** 0.34*** 0.35*** 0.62*** 0.65*** 0.17** 0.18*** 0.19*** 0.13*** 0.71***
Biomass 0.25*** 0.19*** 0.18*** 0.18*** 0.26*** 0.25*** 0.01 0.06*** 0.07*** 0.05** 0.25***
*, **, *** Significant at the 0.05, 0.01, and 0.001 probability levels, respectively
z FieldSpec Pro; Analytical Spectral Devices, Inc., Boulder, CO













Table 3-5. Coefficient estimates (r2) Of range averages from 630-690 nm and 2080-23 50 nm and ratios 605/5 15 nm, 915/975 nm, and
865/725 nm to volumetric water content (VWC) visual color, quality, density ratings, and biomass production from fairway
height hybrid bermudagrass (Cynodon dactylon X C. transvaalensistr~r~r~r~r~r~ 'Princess 77) averaged over five dates (22 June, 12
July, 25 July, 22 Aug., and 4 Sept. 2007). All numbers in column headings are expressed nanometers (nm).
ASD z (nm)
r2 630-690 2080-2350 605/515 nm 915/975 nm 865/725 nm
VWC 0.19*** 0.40*** 0.12** 0.14** 0.03
Color 0.19*** 0.12*** 0.41*** 0.32*** 0.24***
Density 0.30*** 0.08*** 0.32*** 0.39*** 0.22***
Quality 0.29*** 0.10 0.26*** 0.42*** 0.23***
Biomass 0.00 0.01 0.08* 0.06* 0.13**
*, **, *** Significant at the 0.05, 0.01, and 0.001 probability levels, respectively
z FieldSpec Pro; Analytical Spectral Devices, Inc., Boulder, CO












Table 2-9. Evaluation of Means of leaf N concentration tissue (%N) collected from fairway height hybrid bermudagrass (Cynodon
dactylon X C. transvaalensistr~r~r~r~r~r~ Burtt-Davy), four irrigation rates (60, 80, 100, and 120% estimated ET values), four N rates
(0, 25, 50, and 100 kg ha- ), and five dates (22 June, 12 July, 25 July, 22 Aug., and 4 Sept. 2007) combined, which had N
treatment X date interactions.
%N
Date 0 kg ha-' 25 kg ha-' 50 kg ha-' 100 kg ha-'
6/22 3.51b3 4.02a2 4.19a2 4.66al
7/12 3.24bc3 3.25b2 3.38b2 3.81b1
7/25 3.26bc3 3.91a2 3.91a2 4.59al
8/22 2.90cl 2.91b1 2.75cl 2.83cl
9/4 4.01al 3.98al 3.88al 4.10b1
Means in the same column and category followed by the same letter are not significantly different (LSD; P, 0.05); means in the same
row and category followed by the same number are not significantly different (LSD; P, 0.05).










greatly from one location to another. It is assumed that all the turf within a given area is

precisely the same and responds exactly the same to environmental conditions. Conversely,

tensiometers have proven to be an adequate assessment of plant water status, but require periodic

maintenance, and may have to be removed during periods of cold weather (Unruh and Elliott,

1999). In addition, the tensiometer reading is only appropriate in the area adjacent to the

placement of the ceramic tips and do not indicate water status for a large area (Unruh and Elliott,

1999). Therefore, because conservation of this limited resource is critical, it is increasingly

necessary to develop technology to help turf managers allocate irrigation water more efficiently.

Nutrient Importance

Aside from the importance of water to plants, available nutrients also drive growth and

development. Nutrient availability affects many physiological processes the plant must perform

in order to develop properly. Nitrogen (N) is the most limiting nutrient in production of non-

legumous crops (Osborne et al., 2002). It is essential to the building of several important plant

components including proteins, nucleic acids, hormones, and chlorophyll, which harvests light

energy used in photosynthesis (Hopkins, 1999). Phosphorus (P) plays an important role in the

transfer of energy within the plant in the form of compounds like adenosine triphosphate (ATP),

adenosine diphosphate (ADP), and phosphate (Pi). P also is an important factor in root growth

(Hopkins, 1999). Potassium (K) is also a very important nutrient in plant growth and is required

in large amounts. Potassium serves to activate a number of enzymes and regulates stomatal

conductance; a very important process controlling transpiration (Hopkins, 1999). All of these

nutrients, if in deficit of plant requirement, can cause a number of physiological disruptions

which are detrimental to plant growth and development. Once these stresses become visible it

may be too late to correct the damage. Therefore, methods are needed to predict plant nutrient

status that can detect physiological stress before it becomes visible to the naked eye.









data collection with the ASD unit, and again every 15 minutes depending on the length of time

needed to collect data. Usually only one white reference was needed to take data on all plots.

All reflectance readings were taken in full sunlight between the hours of 1100 and 1400 Central

Standard Time (CST) to minimize variance caused by incoming solar radiation.

Several bands were extracted from ASD reflectance data for comparison based on previous

research. These bands consisted of wavelengths 550 nm (Blackmer et al., 1994; Blackmer et al.,

1996; Yoder and Pettigree-Crosby, 1995); 510 nm, 535 nm, 635 nm, and 735 nm (Kruse et al.,

2005); 545 nm, 755 nm and 935 nm (Starks et al., 2006); and 661 nm and 813 nm (Trenholm et

al., 1999a). Spectral ranges 630-690 nm and 2080-2350 nm were averaged (Ripple 1986) and

ratios 605/515 nm, 915/975 nm, and 865/725 nm were calculated (Starks et al., 2006). Also

NDVI ((880-650 nm)/ (880+650 nm)) and LAI (880/650 nm) indices were calculated for

comparison against Crop Circle NDVI and LAI calculations, as well as Stress1 (706/760 nm)

and Stress2 (706/813 nm) indices (Trenholm et al., 2000).

Statistical Analysis

Analysis of variance (ANOVA) was performed using the PROC GLM method (SAS

Inst., 2003) to compare the differences among N and irrigation treatments at various dates to

visual ratings, Crop Circle readings, IR temperature readings calculated into a CWSI, various

individual reflectance wavelengths quantified from the ASD unit, soil VWC, and leaf N

concentration. Regression analysis was performed using the PROC REG method (SAS Inst.,

2003) to correlate relationships between visual ratings, Crop Circle readings, and IR temperature

readings, various individual reflectance wavelengths quantified from the ASD unit, soil VWC,

and leaf nitrogen concentration. Additionally, due to the large amount of data generated by the

ASD device, ASD data were rendered using PROC PLS (SAS Inst., 2003) for prediction of soil

VWC, leaf N concentration, visual quality ratings, and biomass production. Equations were









as: reflectance in the Near Infrared (NIR) region minus reflectance in the Red (R) region divided

by a sum of both (NIR-R / NIR+R) (Rouse et al., 1973) .

Human Visual Ratings

Traditionally, visual observation has been the standard for assessing turf stress for turf

managers. Researchers quantified this observation by creating a scale. The National Turfgrass

Evaluation Program (NTEP) scale is a 1-9 assessment of turf color, quality and density, where 6

is the least acceptable and has been the standard for turf research for many years (Morris, 2007).

However, recent investigations have looked beyond this scale to compare remote sensing

systems with traditional human evaluation. Trenholm et al. (1999) tested four of the current

indices and compared them to visual evaluations. These indices were:

* NDVI computed as (935-661 nm)/ (935+661 nm)
* Leaf area index (LAI) computed as 93 5/661 nm
* Stress 1 computed as 706/760 nm
* Stress 2 computed as 706/813 nm

They concluded that NDVI, LAI, and Stress 2 indices, and individual wavelength reflectance

measurements at 661 nm and 813 nm were comparable to visual evaluations (Trenholm et al.,

1999). Bell et al. (2002b) found strong correlations between NDVI and turf color, and

subsequent work by Bell et al. (2004) concluded that NDVI was a better estimator of chlorophyll

content than visual color evaluation.

Irrigation

Early research involving remote sensing to predict plant water status related plant canopy

temperature to plant water potential (Ehrler et al., 1978). This was derived from the fact that

transpiration plays a maj or role in the plant' s ability to cool itself through ET. As soil water

decreases, evaporative cooling also decreases causing the canopy temperature to rise. This is

what many researchers have used to relate canopy temperature to relative soil water content









A ground based vehicle mounted optical sensor Crop Circle (Model ACS-210) (Holland

Scientific) which produces its own light to compute a normalized difference vegetation index

(NDVI) which is calculated as (880nm-650nm) / (880nm+650nm). A hand-held hyperspectral

spectroradiometer (FieldSpec Pro; Analytical Spectral Devices, Inc., Boulder, CO) with a

spectral range of 300-2500 nm was also used to collect reflectance data. All reflectance readings

were taken in full sunlight between the hours of 1100 and 1400 Central Standard Time (CST) to

minimize variance caused by solar radiation. Plots were ground-truthed using time domain

reflectometry for soil moisture content (VWC) and clipping samples obtained from plots were

dried, weighed for biomass, and tested for Total Kj eldahl N content. Visual ratings were also

taken.

The results from the nitrogen and irrigation experiment indicate reflectance data best

modeled soil VWC, visual ratings, and some treatment effects from N. Highest correlations for

VWC were achieved from PLS regression on ASD data (r2=0.71) and the Crop Circle device

(r2=0.70). The results from the PGR experiment modeled the biomass reduction expected from

increasing rates of flurprimidol and trinexapac-ethyl compared to the untreated control, but not

ethephon. Similar trends were observed in NDVI and leaf area index (LAI) indices computed

from Crop Circle and ASD instruments, as well as several stress indices computed from the ASD

device.

If remote sensing technology could be used to adequately assess turf parameters, and was

coupled with the global positioning system (GPS) and geographic information systems (GIS)

technologies, irrigations and chemical applications could be applied on a site specific basis. In

this setting, turf managers could potentially reduce inputs thereby reducing cost and

environmental impact.











101
102
103
104


201
202
203
204


301
302
303
304


401
402
403
404


Figure A-1. Experimental layout for detection of leaf nitrogen concentration and soil volumetric water content using ground-based
remote sensing technology










APPENDIX C
CUMALITIVE WEEKLY RAINFALL AND ET DATA FOR JAY, FL

Table C-1. Cumalitive weekly rainfall and et data for Jay, FI
Cumulative


Week
Ending On Rainfall (cm)a ET (cm)a
June 15 0.38 0.89
22 6.43 3.15
29 0.38 2.74
July 6 2.67 2.95
20 3.61 2.87
27 5.56 3.07
August 3 5.31 3.02
10 0.41 3.10
17 0.86 2.90
24 0.33 3.10
31 5.97 2.44
aBased on data from http://fawn.ifas.ufl .edu/.
bIrrigation applied to water in fertilizer application


Irrigation
Applied (cm)
0.00
0.00
0.00
0.00
0.20b
0.00
0.00
0.00
0.00
2.54
0.00


Difference (cm)
(0.31)
2.97
0.61
0.33
(0.64)
2.05
4.34
1.64
(0.39)
(3.16)
2.91









correlation in the NIR region to biomass, or the energy status of the plant. Similar results were

found by Kruse et al. (2005) on creeping bentgrass (Agrostis stolinifera L.), where reflectance in

the green (5 10 nm and 53 5 nm), red (63 5 nm), and NIR (73 5 nm) were significant in predicting

biomass. Trenholm et al. (1999a) found that wear stress in several varieties of hybrid

bermudagrass (Cynodon dactylon x C. transvalensis Burtt-Davy) and seashore paspalum

(Paspalum vaginatum Swartz) is associated with the specific wavelength at 813 nm. Starke et

al., (2006) found that crude protein concentration, biomass production, and crude protein

availability of bermudagrass (C. dactylon L. Pers.) pastures is closely related with canopy

reflectance ratios of 605/515 nm, 915/975 nm, and 865/725 nm.

Researchers have been exploring the idea that plant water status can be determined by

measuring reflectance in the short wave infrared (SWIR) region (1,300-2,500 nm). Ripple (1986)

found that SWIR reflectance is strongly correlated with leaf water content and that significant

correlations exist between certain spectral bands (630-690 nm and 2,080-2,350 nm ranges), with

SWIR reflectance having the highest correlations. As reflectance decreased in the 630-690 nm

range there was an inverse linear relationship with leaf relative water content. However, Ripple

(1986) concluded that relationships between spectral reflectance and leaf water content can be

direct or indirect and are wavelength dependent. It is believed that elevated correlations with

spectral reflectance data was due to a reduction in chlorophyll as the leaves dried and not a true

relation to actual plant water status (Ripple, 1986). Hutto et al. (2006) also documented that

drought stressed creeping bentgrass had a higher amount of reflectance in the SWIR regions than

that of the non-stressed control.

Several different spectral responses to light to assess indications of plant stress, nutrient

and water status have been tested in the form of indices. One of the most common indices is the












Table 2-10. Evaluation of means of individual reflectance wavelengths at 813 nm and 935 nm collected from fairway height hybrid
bermudagrass (Cynodon dactylon X C. transvaalensistr~r~r~r~r~r~ Burtt-Davy), four irrigation rates (60, 80, 100, and 120% estimated
ET values), four N rates (0, 25, 50, and 100 kg ha- ), and five dates (22 June, 12 July, 25 July, 22 Aug., and 4 Sept. 2007)
combined, which had N treatment X date interactions.
813 nm (ASDz) 935 nm (ASDz)
Date 0 kg ha-l 25 kg ha-l 50 kg ha-l 100 kg ha-l 0 kg ha-l 25 kg ha-l 50 kg ha-l 100 kg ha-l
6/22 0.47al 0.47al 0.48al 0.49al 0.49al 0.49al 0.50al 0.51al
7/12 0.43bc1 0.43b1 0.43b1 0.44b1 0.45b1 0.45cl 0.45cl 0.46b1
7/25 0.42cl 0.43b1 0.43b1 0.43b1 0.45b1 0.46bc1 0.46bc1 0.46b1
8/22 0.26dl 0.25cl2 0.24c2 0.24c2 0.30cl 0.29dl2 0.28d2 0.28c2
9/4 0.45ab1 0.44b1 0.43b1 0.45b1 0.49al 0.48ab1 0.47b1 0.49al
Means in the same column and category followed by the same letter are not significantly different (LSD; P, 0.05); means in the same
row and category followed by the same number are not significantly different (LSD; P, 0.05).
z FieldSpec Pro; Analytical Spectral Devices, Inc., Boulder, CO










3-3 Coefficient estimates (r2) Of Crop Circle reflectance data, Normalized Difference
Vegetation Index (NDVI), Leaf Area Index (LAI), 880nm, and 650nm, and, ASD
reflectance data computed into NDVI, LAI, Stress1, and Stress2 indices to
volumetric water content (VWC) visual color quality, density ratings, and biomass
production from fairway height hybrid bermudagrass (Cynodon dactylon X C.
transvaalensistr~r~r~r~r~r~ 'Princess 77) averaged over Hyve dates (22 June, 12 July, 25 July, 22
Aug., and 4 Sept. 2007). ............. ...............76.....

3-4 Coefficient estimates (r2) Of ASD reflectance data at individual wavelengths, as well
as range averages from 630-690nm and 2080-23 50nm to volumetric water content
(VWC) visual color, quality, density ratings, and biomass production from fairway
height hybrid bermudagrass (Cynodon dactylon X C. transvaalensistr~r~r~r~r~r~ 'Princess 77)
averaged over Hyve dates (22 June, 12 July, 25 July, 22 Aug., and 4 Sept. 2007). All
numbers in column headings are expressed nanometers (nm) ................. ............... .....77

3-5 Coefficient estimates (r2) Of range averages from 630-690 nm and 2080-23 50 nm
and ratios 605/515 nm, 915/975 nm, and 865/725 nm to volumetric water content
(VWC) visual color, quality, density ratings, and biomass production from fairway
height hybrid bermudagrass (Cynodon dactylon X C. transvaalensistr~r~r~r~r~r~ 'Princess 77)
averaged over Hyve dates (22 June, 12 July, 25 July, 22 Aug., and 4 Sept. 2007). All
numbers in column headings are expressed nanometers (nm) ................. ............... .....78

3-6 Partial Least Squares regression coefficients on hyperspectral data from ASD device
to predict volumetric water content (VWC), biomass, and visual quality (rated on
NTEP 1-9 scale 9 is best and 6 is acceptable) in fairway height bermudagrass
(Cynodon dactylon X C. transvaalensis, 'Princess 77) averaged over five dates (22
June, 12 July, 25 July, 22 Aug., and 4 Sept. 2007) ................. ...............79...........

A-1 Description of treatments for Figure A-1 ................. ...............86..............

B-1 Daily Rainfall and ET data for Jay, FL............... ...............87...

C-1 Cumalitive weekly rainfall and et data for Jay, Fl ................ ...............90.............

D-1 MothlyRainfall and ET averages from 2003-2006 for Jay, FL .............. ....................91










Irrigation Importance

Water is the primary requirement for development and survival of turfgrass and many

other agronomic crops (Turgeon, 2008). Plant water potential is essential for plant growth as

well as most physiological processes the plant must perform (Idso et al., 1981). Therefore, water

availability is one of the most critical factors in plant health (Ripple, 1986). Plants are made up

of cells, which are filled with water to maintain their turgor. However, if turgor is not

maintained the plant begins to wilt (Unruh and Elliott, 1999). Many plants contain 75-85 %

water and begin to die if this percentage decreases below 65 % (Unruh and Elliott, 1999). Water

is taken up by the roots in the soil and is moved throughout the plant and eventually released

through the stomata through a process known as transpiration (Salisbury and Ross, 1992).

Stomata are responsible for gas exchange in the plant and they close when the plant is under a

water deficit or stress (Salisbury and Ross, 1992). This ultimately reduces photosynthesis and

stunts leaf growth (Unruh and Elliott, 1999). Geeske et al. (1997) found that this stunting in

growth is the plant' s defense mechanism to reduce nutrient and water requirements. The

reduction in leaf area serves as a beneficial factor that ultimately reduces transpiration and the

need for increased physiological processes (Geeske et al., 1997).

In addition, transpiration serves many different purposes such as nutrient movement

through the plant and evaporative cooling, also known as evapotranspiration (ET) (Turgeon,

2008). Evapotranspiration is the total water lost by all transpiratory movement of water from

soil, through the plant and ultimately into the atmosphere, where it evaporates (Turgeon, 2008).

Many factors affect this process such as light intensity, humidity, wind velocity and temperature

(Unruh and Elliott, 1999).

Water stress has a direct affect on the rate of photosynthesis. It first causes stomatal

closure, which causes a reduced supply of CO2. Secondly, water stress reduces water potential,









ratings, Crop Circle readings (NDVI and LAI) and ASD readings (NDVI, LAI, Stress1, and

Stress2). Regression analysis was performed using the PROC REG method (SAS Inst., 2003) to

correlate relationships between visual ratings, Crop Circle readings, various individual

reflectance wavelengths quantified from the ASD unit, and soil VWC. Additionally, due to the

large amount of data generated by the ASD device, ASD data were rendered using partial least

squares regression PROC PLS (SAS Inst., 2003) to soil VWC, visual quality ratings, and

biomass production. Equations were validated through a single sample cross-validation. For

cross-validation, ten percent of the sample was omitted for prediction purposes so that the

number of factors chosen creates the minimal predicted residual sum of squares (PRESS). This

process was repeated so that every observation was used exactly once for cross-validation.

Results and Discussion

Biomass

ANOVA reveals that there was a significant reduction in biomass in the 0.1 kg a.i. ha-l

and 0.4 kg a.i. ha-l rates of TE, the highest rate of flurprimidol (2.3 kg a.i. ha-l ) and the low rate

of ethephon (30.4 kg a.i. L ha- ) (Table 3-1). Although not always significant, a dose response

trend was observed with TE (Figure 3-1) and flurprimidol (Figure 3-2). This dose response was

not observed with Ethephon due to the fact that ethephon is primarily used for seedhead

suppression as opposed to biomass suppression in turf (Anonymous 2007).

Biomass did not correlate well with any remotely sensed parameter quantified from either

device (Tables 3-3, 3-4, 3-5, and 3-6). This data is in agreement with the research conducted by

Fenstemaker-Shaulis et al. (1997) and Trenholm et al. (1999b) who found remotely sensed data

was not an adequate predictor of biomass. The inability to adequately model biomass reduction

through increasing rates of PGR applications may be attributed to a number of factors. PGRs

have been found to limit vertical growth but increase prostrate growth, thus increasing density

















70


60


50




00 .P
F9 30


20


10



Untreated 0.3 0.6 1.1 2.3
Rate (kg a.i. ha l)


Figure 3-2. Biomass measurements collected from fairway height hybrid bermudagrass (Cynodon dactylon X C. transvaalensis~t~t~t~t~t~t~
'Princess 77) treated with flurprimidol at 0.3, 0.6, 1.1, and 2.3 kg a.i. ha l. Plots measured 1.5 X 3.0 m with three
replications. Means for biomass measurements was computed using Fishers protected LSD (p< 0.05) from data over 5
dates (22 June, 12 July, 25 July, 22 Aug., and 4 Sept. 2007) combined.





Parameter NDVIx LAI" 880 nm 650 nm
%N 0.19*** 0.21*** 0.15*** 0.15***
VWC 0.71*** 0.59*** 0.61*** 0.52***
Color 0.59*** 0.57*** 0.39*** 0.61***
Quality 0.70*** 0.67*** 0.50*** 0.71***
Density 0.67*** 0.65*** 0.48*** 0.69***
Biomass 0.26*** 0.33*** 0.14*** 0.34***
*, **,** Significant at the 0.05, 0.01, and 0.001 probability levels, respectively
z MLodel ACS-210, Holland Scientific
SIR thermometer Raytek Raynger@ STTw, TTI, Inc.
xNDVI= (880-650 nm)/(880+650 nm)
LAI= 880/650 nm
"CWSI= (To-Ta~a (To-Ta)1 / (To-Ta)u (To-Ta)I
Where Te= Canopy Temperature, Ta= Actual Temperature, a= actual, u= upper limit, and 1= lower limit


Table 2-12. Coefficient estimates (r2) Of Crop Circle reflectance data Normalized Difference Vegetation Index (NDVI), Leaf Area
Index (LAI), NIR, and Red, and crop water stress index (CWSI) to soil volumetric water content (VWC) visual color,
quality, density ratings collected from fairway height hybrid bermudagrass (Cynodon dactylon X C. transvaalensis, Burtt-
Davy), four irrigation rates (60, 80, 100, and 120% estimated ET values), four N rates (0, 25, 50, and 100 kg ha )~, and
averaged over all dates (22 June, 12 July, 25 July, 22 Aug., and 4 Sept. 2007).
Crop, Circlez IR ThermometerY


CWSI"
0.02*
0.02*
0.13***
0.33***
0.38***
0.05**











Table 2-6. Evaluation of means of individual band reflectance from the ASD device with a range of 3 50-2500 nm and a spectral
resolution of 1 nm collected from fairway height hybrid bermudagrass (Cynodon dactylon X C. transvaalensistr~r~r~r~r~r~ Burtt-
Davy), four irrigation rates (60, 80, 100, and 120% estimated ET values), four N rates (0, 25, 50, and 100 kg ha- ) and
five dates (22 June, 12 July, 25 July, 22 Aug., and 4 Sept. 2007) combined.
ASD z (nm)
N Rate 510 535 545 550 635 661 735 755 813 935 2132
0 kg hal 0.054a 0.081a 0.086a 0.088a 0.074a 0.069ab 0.29 0.366a 0.407 0.437 0.183
25 kg hal 0.055a 0.081a 0.086a 0.088a 0.075a 0.071a 0.29 0.363a 0.404 0.433 0.190
50 kg hal 0.055a 0.081a 0.086a 0.088a 0.076a 0.072a 0.29 0.361a 0.402 0.431 0.194
100 kg hal 0.051b 0.077b 0.082b 0.084b 0.068b 0.064b 0.29 0.367a 0.409 0.437 0.178
LSD 0.002 0.002 0.002 0.002 0.005 0.005 N.S. 0.007 N.S. -- N.S.
"--" denotes LSD is not valid due to a significant date by N treatment interaction
z FieldSpec Pro; Analytical Spectral Devices, Inc., Boulder, CO










(Aston and Van-Bavel, 1972). Stomatal closure of sunlit leaves results in increased leaf

temperature if factors such as wind speed and vapor pressure remain constant (Ehrler et al.,

1978). This temperature difference between canopy and air can be determined in several ways.

Originally, microthermocouples were used, but this only worked on a small scale. However,

advances in infrared thermometry have increased the ability to detect crop water stress through

the utilization of remote sensing devices (Slack et al., 1981). Infrared thermometers registered in

a narrow bandpass (8-14 Clm or 10.5 12.5 Clm) are capable of measuring crop temperatures with

0.5 oC accuracy (Ehrler et al., 1978). They can be used for measuring canopy temperature at

ground level, from aircraft, or satellite. These methods are done in a much timelier manner

compared to old methods such as tensiometers and microthermocouples (Ehrler et al., 1978).

From this relation of canopy temperature to plant water potential, many indices have been

developed to best predict plant water status. Idso et al. (1981) proposed a temperature-based

index as a water stress indicator. This was intended to normalize canopy temperature minus air

temperature, To-Ta, for environmental variability. It was introduced as the Crop Water Stress

Index (CWSI). CWSI produces a species-specific linear relationship between air and canopy

temperature, which is irrespective of other environmental factors except cloud cover (Slack et al.,

1981). CWSI can be described as the "fractional decrease of potential ET" which can be

presented in terms of the canopy temperature minus the air temperature (To-Ta) or

CWSI= (Toc-T~a)_S c-Ta)1


Where Te= canopy temperature, Ta= actual temperature, a= actual, u= upper limit, and 1= lower

limit (Jalali-Farahani et al., 1994)

The CWSI was found to have potential in turf irrigation scheduling (Throssel et al., 1987).

Further investigations of this model spawned numerous theories accounting for several









BIOGRAPHICAL SKETCH

Jason Hamilton Frank was born in 1982, in Deland, FL. He lived there until 2001when he

graduated from Deland high school and was accepted to the University of Central Florida. He

spent a year there as a business administration maj or before realizing he wanted to attend the

University of Florida. He transferred to Daytona Beach Community College in August of 2002

and received his A.A. degree in July of 2003. He then transferred to the University of Florida in

August of 2003 where he completed his B.S. in turfgrass science with a minor in business

administration in December of 2005. He then went on to graduate school, also at the University

of Florida, where he will graduate with his M. S. in horticulture sciences with a minor in

agricultural and biological engineering and interdisciplinary concentration in geographic

information sciences in May of 2008. Upon graduation Jason has accepted a 2nd assistant

superintendent position at Royal Poinciana Golf Club in Naples, FL.









Table 2-15. Partial Least Squares regression on hyperspectral data from ASD device with a range of 350-2500 nm and a spectral
resolution of 1 nm for prediction of soil volumetric water content (VWC), percent leaf nitrogen concentration (%N),
biomass, and visual quality collected from fairway height hybrid bermudagrass (Cynodon dactylon X C. transvaalensis,
Burtt-Davy), four irrigation rates (60, 80, 100, and 120% estimated ET values), four N rates (0, 25, 50, and 100 kg ha )~
averaged over all dates (22 June, 12 July, 25 July, 22 Aug., and 4 Sept. 2007).
Calibration No. of Factors* r2 SE n
VWC 8 0.72 0.030 240
% N 7 0.25 0.043 240
Biomass (g) 6 0.31 0.040 240
Visual Quality 7 0.55 0.031 240
*Number of factors required to achieve a minimal Predicted Residual Sum of Squares (PRESS) of prediction for the partial least
squares regression model.
z FieldSpec Pro; Analytical Spectral Devices, Inc., Boulder, CO










image based-application methods (Variable Rate and Site-Specific) to traditional blanket

applications for PGR' s in cotton, it was found that the broadcast applications were on average 19

% more costly than the site-specific method, and 27 % more costly than the variable rate method

(Bethel et al., 2003). These results demonstrate the potential economic benefit of this technology

for future use in other aspects of agriculture to reduce excess chemical runoff and leaching into

the surrounding environment (Bethel et al., 2003).

Statistical Methods

Older work involving remote sensing typically used multispectral data, which is collected

in individual or wide bandwidths and give a coarse representation of the electromagnetic

spectrum (Lillesand and Kiefer, 2003). With the introduction of hyperspectral data, which

typically acquires 200 or more bands in very narrow resolution throughout the visible, NIR, and

SWIR regions, researchers have more data to analyze (Lillesand and Keifer, 1987). These

enhanced techniques for obtaining the data have created the need to explore different statistical

methods to quantify and interpret the data. Traditionally linear regression is the simplest way to

understand variance. However, in interpreting reflectance data it has been found that other

regression techniques better assess plant status (Kruse et al., 2005).

Partial Least Squares (PLS) regression is a method used to predict variables when there are

a large number of factors; typically more than the number of observations (Tobias, 1997). The

use of multiple variables is generally tested through multiple linear regression (MLR) (Tobias,

1997). However, the problem with MLR is when the number of factors becomes too large the

model may become over-fitting (Tobias, 1997). This is observed when a large number of

variables are present but only a few account for most of the variation (Tobias, 1997). Almost

perfect models may be achieved; however they will not be highly predictive of new data.










Table B-1. Continued


Cumulative
Difference
(cm)
0.15
3.07
2.89
2.53
2.05
1.54
1.09
3.22
2.79
5.07
4.64
4.34
3.88
3.42
2.99
2.51
2.08
1.72
1.64
1.29
1.87
1.44
0.96
0.50
0.04
(0.39)
(0.49)
(0.92)
(1.40)
(1.89)
(2.39)
(2.78)
(3.16)
(3.56)
2.99
2.94
3.63
3.52
3.12
2.91
2.64


Rainfall
(cm>a
0.53
3.35
0.28
0.00
0.00
0.00
0.00
2.54
0.00
2.62
0.00
0.15
0.00
0.00
0.00
0.00
0.00
0.00
0.41
0.03
0.84
0.00
0.00
0.00
0.00
0.00
0.33
0.00
0.00
0.00
0.00
0.00
0.00
0.05
4.34
0.28
0.89
0.33
0.00
0.08
0.05


Irrigation
Applied (cm)
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
2.54
0.00
0.00
0.00
0.00
0.00
0.00
0.00


Day
23
24
25
26
27
28
29
30
31
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
1


ET (cm)a
0.41
0.43
0.46
0.36
0.48
0.51
0.46
0.41
0.43
0.33
0.43
0.46
0.46
0.46
0.43
0.48
0.43
0.36
0.48
0.38
0.25
0.43
0.48
0.46
0.46
0.43
0.43
0.43
0.48
0.48
0.51
0.38
0.38
0.46
0.33
0.33
0.20
0.43
0.41
0.28
0.33


August





































September










sensed data and not just on an incremental scale. In addition, investigate more into the use of

PLS regression on hyperspectral data to formulate prediction models for assessment of the

biomass of the turf.

Crop circle. Test in various light and ambient conditions to assess different turfgrass

parameters and investigate its usability in an assortment of environmental settings. Turfgrass

managers work in all conditions, rain or shine. An assessment of how these devices will work in

varying environmental conditions must be evaluated before they are employed into real world

applications of turfgrass management.

Beyond these experiments. Conduct experiments on larger scale while adding other

technologies such as GPS, GIS, and VRA. This will allow researchers to better understand the

usability of the technology and the logistics of operating it as a turfgrass manager. Remote

sensing in a large scale setting is not very effective without the combination of other

technologies. You need GPS and GIS to georeference remotely sensed data to understand where

the data is coming from. Then by applying VRA technology some sort of economic impact can

be assessed to understand the potential benefit of the use all of these technologies. Remote

sensing is just the beginning of a long road of investigation in helping turfgrass managers reduce

irrigation and chemical use and apply these inputs only where needed.

Currently the use of remote sensing technology in the turfgrass management industry is

still in its infancy. However, the experiments described in this thesis along with other research

that has been referenced show it has potential to assist turfgrass managers to make better

decisions about managing their turf. Presently, this technology is not in a very usable form and

much research still needs to be conducted to understand its place in the field of turfgrass

management.












TABLE OF CONTENTS

page

ACKNOWLEDGMENTS .............. ...............3.....


LIST OF TABLES ................ ...............6............ ....


LIST OF FIGURES .............. ...............10....


AB S TRAC T ............._. .......... ..............._ 1 1..


CHAPTER


1 LITERATURE REVIEW ................. ...............13...............


Introducti on ................. ...............13......__ ......

Irrigation Importance ........_................. ..........._..........1
Nutrient Importance .................. ...............17.................
Plant Growth Regulator Importance ................. ...............18................
Remote Sensing .............. ...............18....
Vegetative Indices .............. ...............19....
Human Visual Ratings............... ...............20
Irrigation ................. ...............20.................
Nutri ent ................. ...............23.................
Biom ass .............. ...............24....
PGRs ............... .. .... ._ ...............24....
Statistical M ethods .............. ...............25....

Summary ............. ...... ._ ...............26...


2 DETECTION OF LEAF NITROGEN CONCENTRATION AND SOIL
VOLUMETRIC WATER CONTENT USINTG GROUND-BASED REMOTE
SENSING TECHNOLOGY ................. ...............28.......... .....


Introducti on ................. ...............28.................
Materials and Methods .............. ...............34....
Reflectance Measurements ................. ...............3.. 5..............
Statistical Analysis .............. ...............36....
Results and Discussion ................. .. ...............37..
Influence of Irrigation Treatments ................. ...............37................
Nitrogen Effects............... ...............37
Nitrogen X Date Interaction .............. ...............38....
Tissue N Concentration vs. Reflectance............... ...............3
VWC vs. Reflectance .............. ...............39....

Visual Ratings .............. ...............40....
Biomass ................... ...............41..

Crop Circle vs. ASD ................. ...............41................
Conclusions............... ..............4









Conclusions

This experiment shows the biomass reduction expected from increasing rates of PGRs in

two of the three PGRs applied (Table 3-1) (Figure 3-1, 3-2). Flurprimidol and TE reduced

biomass in response to increasing application rates compared to the untreated control; however,

the results are not always significant (Table 3-1) (Figure 3-1, 3-2). Similar trends were observed

in NDVI and LAI indices computed from Crop Circle and ASD instruments, as well as Stress1

and Stress2 indices computed from the ASD device (Table 3-2). Increases in NDVI and LAI

which model plant health, as well as reductions in Stress1 and Stress2 indices, could be

attributed to suppression of seed head production in addition to the decrease of overall biomass.

This overall decrease in biomass prevented the scalping of the turf in the study area that was

observed in the untreated control and surrounding areas not treated with PGRs.

The inability to adequately model biomass reduction, visual ratings, and VWC through

increasing rates of PGR applications may be attributed to a number of factors already mentioned.

PGRs have been found to limit vertical growth but increase prostrate growth, thus increasing

density (Branham, 1997). A general increase in density was observed in this experiment with

increasing application rates of TE and flurprimidol, however, they were not significantly

different from the untreated control (Table 3-1). The lack of significant differences in density

and all visual ratings affected the ability of the remote sensing devices to detect differences

among theses various parameters. Additionally, this experiment was performed on Princess 77

bermudagrass turf which produces seed heads. These seed heads are white in color and could

potentially skew remotely sensed data compared to a turf that does not produce seed heads.

Currently, turf managers apply PGRs in the form of broadcast applications to reduce

biomass. If remote sensing technology could be used to adequately detect biomass, and was

coupled with GPS and GIS technologies, applications of PGRs could be applied on a site specific









technological advances in their industry. These environmental concerns have pressed turf

managers to reduce nutrient and pesticide inputs used for turf maintenance (Bell et al., 2004). In

other industries, such as agronomic food production crops, precision agriculture has been used as

a management tool to maximize yield and minimize cost (Bethel et al., 2003). The basis of this

technology relies heavily on the Global Positioning System (GPS), Geographic Information

Systems (GIS), Variable Rate Application (VRA) and remote sensing technology. Using this

technology to apply nutrients, water, and pesticides only where they are needed can optimize

yield while minimizing cost. (Lee et al., 1999)

Plant growth regulators (PGRs) are used by turf managers to suppress vertical turf

growth, which ultimately reduces maintenance costs due to reduced mowing requirements. A

PGR is a substance which adjusts the growth and development of a plant. This is generally

achieved through the inhibition of the gibberellic acid (GA), a hormone which is responsible for

cell elongation. PGRs were first introduced for use on fine turf in 1987 (VanBibber, 2006).

Some data suggests that PGR-treated turf can produce greater root mass, recover from injury

faster, reduce water use rate, reduce disease incidence, and reduce Poa annua weed populations

(Branham, 1997). The first of these chemicals were flurprimidol (Trade name Cutless) and

paclobutrazol (Trade name Trimmet) which inhibit GA synthesis at relatively the same point in

the GA biosynthesis pathway resulting in very similar plant responses (Branham, 1997).

However, paclobutrazol is a more active compound and requires lower rates to achieve the same

response as higher rates of flurprimidol (Branham, 1997). The newest of the PGR' s, trinexapac-

ethyl (TE) (Trade name Primo), became commercially available in 1995 and has become the

standard for use throughout the turf industry (Branham, 1997). TE facilitates GA inhibition later










experiment. It is simple to use and is potentially non-restrictive of environmental conditions,

like cloud cover and variation in solar radiation. Currently, devices like this could provide fast

and easy supplemental information to turfgrass managers for assistance in making decisions.

Coupled with GPS and GIS technology, ground based remote sensing devices, like the Crop

Circle, can give turfgrass managers greater insight into managing their turf by potentially

allowing them to more efficiently schedule their irrigation, thereby reducing water consumption,

leaching, and runoff.

Furthermore, if devices like the Crop Circle could be improved to produce and quantify

additional wavelengths these devices could potentially be employed by turfgrass managers due to

their ease of use. However, further investigation into data from devices like the ASD using PLS

regression is still needed to investigate the entire spectral constituency for optimal wavelengths

to predict various turf parameters. Only then can prediction equations be formulated and

incorporated into devices like the Crop Circle that could produce the correct wavelengths to

calculate these prediction equations.









concluded that relationships between spectral reflectance and leaf water content can be direct or

indirect and are wavelength dependent. It is believed that elevated correlations in the 630-690

nm range with spectral reflectance data was due to a reduction in chlorophyll as the leaves dried

and not a true relation to actual plant water status (Ripple, 1986).

Nutrient

To a large extent, much remote sensing work has been done on the assessment of N and

chlorophyll levels. Overall, reflectance is low in the visible range of the electromagnetic

spectrum due to the high absorption by chlorophyll for use in photosynthetic processes

(Salisbury and Ross, 1992 ). For this reason, if there is a particular reflectance in this region this

could correlate to a certain physiological disruption in the plant. It is generally assumed that

there is a close relationship between plant color and chlorophyll content (Bell et al., 2004).

Research has shown reflectance at 550 nm to be the most sensitive assessment of N in many

agronomic crops, which is what the human eyes see as green (Blackmer et al., 1994, 1996, Yoder

and Pettigree-Crosby, 1995). Bell et al. (2002a) found vehicle mounted optical sensors provided

a good assessment of variable N applications. Xiong et al. (2007) found significant seasonal

responses to N treatments.

For other nutrient deficiencies such as P and K, limited research has been conducted to

establish their function in the spectral constituency. Research that has been conducted suggests

that reflectance wavelengths for detecting P concentration in corn (Zea mays L.) are in the NIR

(730 nm and 930 nm) region of the spectrum as well as reflectance from the blue region (440 nm

and 445 nm) (Osborne et al., 2002). This can possibly be attributed to the production of

anthocyanin which is produced in P deficient plants that absorb more green light and reflect

more blue light (Osborne et al., 2002). Research conducted by Kruse et al. (2005) found that

spectral reflectance data can be used as a good estimation of P concentration in creeping










Lillesand, T.M., and R.W. Keifer. 2000. Remote Sensing and Image Interpretation. 4th ed. John
Wiley & Sons, New York.

Martin, D.L, C.S. Throssel, and D.J. Wehner. 1994. Models for Predicting Lower Limit of The
Canopy-Air Temperature Difference of Two Cool Season Grasses. Crop Sci. 34: 192-198.

Min, M. and W. S. Lee. 2003. Spectral Based Nitrogen Sensing for Citrus. 27-30 July. ASAE
Annual Intemnational Meeting. Las Vegas, NV

Morgan, W.C., J. Letey, S.J. Richards, and N. Valoras. 1966. Physical Soil Amendments, Soil
Compaction, Irrigation, and Wetting Agents in Turfgrass Management Effects on
Compactibility, Water Infiltration Rates, Evapotranspiration, and Number oflIrrigations.
Agron. J. 58:525-527.

Morris, K.N. A Guide to NTEP Turfgrass Ratings. http://www.ntep. org/reports/ratings.htm. 14
Nov. 2007.

Mylavarapu, R.S. and E.D. Kennelly. UF/IFAS Extentsion Soil Testing Laboratory (ESTL)
Analytical Procedures and Training Manual .
http://edis.ifas.ufl .edu/pdffiles/S S/SS3 1200.pdf. 14 Dec. 2007.

Osbome, S.L., J.S. Schepers, D.D. Francis, and M.R. Schlemmer. 2002. Detection of Phosphorus
and Nitrogen Deficiencies in Corn Using Spectral Radiance Measurements. Agron. J.
94:1215-1221.

Ripple, W.J. 1986. Spectral Reflectance Relationships to Leaf Water Stress. Photogramm. Eng.
Rem. Sens. 52:1669-1675

Rouse, J.W., R.H. Haas, J.A. Schell, and D.W. Deering. 1973. Monitoring the vernal
advancement and retroradation (green wave effect) of Natural Vegetation. Prog. Rep. SRC
1978-1, Remote Sensing Center, Texas A&M Univ., College Station, 93p. (NTIS) No.
E73-106393.

Salisbury, F.B. and C.W. Ross. 1992. Plant Philology. 4th ed. Wadsworth Publ. Co. Belmont,
CA.

SAS Institute. 2002-2003. The SAS System for Windows. Release 9.1. SAS Inst., Cary, NC.

Slack, D.C., K.M. Geiser, K.W. Stange, and E.R. Allred. 1981. Irrigation Scheduling in Sub-
Humid Aareas with Infrared Thermometry. p. 116-124. In Proc. Irrigation Scheduling
Conf., Chicago, and IL. 14-15 Dec. 1981. ASAE Annual Intemnational Meeting, St Joseph,
MI.

Starks, P. J., D. Zhao, W.A. Phillips, and S.W. Coleman. 2006. Development of Canopy
Reflectance Algorithms for Real-Time Prediction of Bermudagrass Pasture Biomass and
Nutritive Values. Crop Sci. 46:927-934.













0.9

0.8

0.7

0.6

0.5

0.4


ASD
y = 52.74x 15.50
R2 = 0.57


Crop Circle
y = 65.19x 17.46
R2 = 0.71


VWC


Figure 2-2. Coefficient estimates (r2) Of Crop Circle and ASD reflectance data computed into a Normalized Difference Vegetation
Index (NDVI), to soil volumetric water content (VWC) collected from fairway height hybrid bermudagrass (Cynodon
dactylon X C. transvaalensistr~r~r~r~r~r~ Burtt-Davy), four irrigation rates (60, 80, 100, and 120% estimated ET values), four N rates
(0, 25, 50, and 100 kg ha )~, and averaged over all dates (22 June, 12 July, 25 July, 22 Aug., and 4 Sept. 2007).










Stowell, L. and W. Gelernter. 2006. Sensing the Future. Golf Course Management. March p. 107-
110.

Throssel, C.S., R.N. Carrow, and G.A. Milliken. 1987. Canopy Temperature Based Irrigation
Scheduling Indices for Kentucky Bluegrass Turf. Crop Sci. 27: 126-131.

Tobias, R.D. 1997. An Introduction to Partial Least Squares Regression (Online). Available at
SAS institute http://ftp. sas.com/techsup/download/technote/ts509.pd 17 April 2007.

Trenholm, L. E., R.N. Carrow, and R.R. Duncan. 1999b. Relationship of Multispectral
Radiometry Data to Qualitative Data in Turfgrass Research. Crop Sci. 39:763-769.

Trenholm, L.E., R.R. Duncan, and R.N. Carrow. 1999a. Wear Tolerance, Shoot Performance,
and Spectral Reflectance of Seashore Paspalum and Bermudagrass. Crop Sci.
39:1147:1152.

Trenhholm, L.E., M.J. Schlossberg, G. Lee, and W. Parks. 2000. An Evaluation of Multispectral
Responses on Selected Turfgrass Species. Int. J. Remote Sensing. 21:709-721.

Turgeon. A.J. 2008. Turfgrass Management. 8th edition. Pearson Education Inc. Upper Saddle
River, NJ.

Unruh, J. and M.L. Elliot. 1999. Best Management Practices for Florida Golf Courses.
University of Florida., Gainesville.

VanBibber, L. 2006. Putting the Numbers to PGR's. http://grounds-
mag.com/chemicals/grounds~maintenance_puttignmespr/ 24 Oct. 2007.

Yoder, B. J. and R.E. Pettigree-Crosby. 1995. Predicting Nitrogen and Chlorophyll Content and
Concentrations From Reflectance Spectra (400-2500) at Leaf and Canopy Scales. Remote
Sens. Environ. 53:199-211.

Xiong, X, G.E. Bell, J.B. Solie, M.W. Smith, and B. Martin. 2007. Bermudagrass Seasonal
Responses to Nitrogen Fertilization and Irrigation Detected Using Optical Sensing. Crop
Sci. 47:1603-1610.









response to various plant parameters and not the actual plant parameter (Trenholm et al., 1999a).

Similarly, the ability of the remote sensing devices to model VWC through individual bands,

indices and ratios, could potentially be attributed to decline in turf health in response to VWC

and not a true predictor of actual VWC.

Furthermore, the high r2 value achieved by PLS regression to model soil VWC (Table 2-

15) (Figure 2-1) could potentially be highly predictive of new data and specific to actual VWC

and not a response to plant stress. This is due to the re-sampling method employed by the PLS

procedure mentioned in the materials and methods.

Visual Ratings

Visual ratings were reasonably predicted from reflectance data from both devices. Visual

assessment of color, quality, and density was best predicted by vegetative indices NDVI and LAI

from both instruments (Table 2-12, 2-14)). Furthermore, raw reflectance values 650 nm and 880

nm from the Crop Circle device achieved reasonable r2 ValUeS (Table 2-12), as well as, 510 nm,

535 nm, 545 nm, 550 nm, 635 nm, and 661 nm individual reflectance wavelengths (Table 2-13),

630-690 nm and 2080-23 50 nm range averages, and 605/5 15 nm, 915/975 nm, and 865/725 nm

ratios (Table 2-14) from the ASD device. Conversely, quality was not as well modeled by PLS

Regression (Table 2-15), compared to the vegetative indices, individual reflectance wavelengths,

and the ratios mentioned above.

The ability of vegetative indices, NDVI and LAI, to model visual assessments could

potentially be attributed to the fact that these indices were formulated to normalize the data.

Since regression functions were run across all dates the normalization of these indices could

potentially smooth out variation in reflectance due to differences in light intensity on various

evaluation dates.










2-12 Coefficient estimates (r2) Of Crop Circle reflectance data Normalized Difference
Vegetation Index (NDVI), Leaf Area Index (LAI), NIR, and Red, and crop water
stress index (CWSI) to soil volumetric water content (VWC) visual color, quality,
density ratings collected from fairway height hybrid bermudagrass (Cynodon
dactylon X C. transvaalensistr~r~r~r~r~r~ Burtt-Davy), four irrigation rates (60, 80, 100, and
120% estimated ET values), four N rates (0, 25, 50, and 100 kg ha- ), and averaged
over all dates (22 June, 12 July, 25 July, 22 Aug., and 4 Sept. 2007). ............. ................55

2-13 Coefficients estimates (r2) Of ASD reflectance readings of individual reflectance
wavelengths taken from an Analytical spectral device (ASD) with a range of 350-
2500 nm and a spectral resolution of 1 nm to %N concentration leaf tissue, soil
volumetric water content (VWC) visual color, quality, density ratings, and biomass
production of fairway height hybrid bermudagrass (Cynodon dactolon X C.
transvaalensistr~r~r~r~r~r~ Burtt-Davy), four irrigation rates (60, 80, 100, and 120% estimated
ET values), four N rates (0, 25, 50, and 100 kg ha- ), and averaged over all dates (22
June, 12 July, 25 July, 22 Aug., and 4 Sept. 2007) ................. ............... 56...........

2- 14 Coefficients estimates (r2) Of average range computations from 630-690 nm and
2080-23 50 nm, and computed indices Normalized difference vegetation index
(NDVI), Leaf area index (LAI) and Stress1 and Stress2 indices also taken from an
ASD with a range of 350-2500 nm and a spectral resolution of 1 nm to %N
concentration leaf tissue, soil volumetric water content (VWC) visual color, quality,
density ratings, and biomass production collected from fairway height hybrid
bermudagrass (Cynodon dactylon X C. transvaalensistr~r~r~r~r~r~ Burtt-Davy), four irrigation
rates (60, 80, 100, and 120% estimated ET values), four N rates (0, 25, 50, and 100
kg ha- ), and averaged over all dates (22 June, 12 July, 25 July, 22 Aug., and 4 Sept.
2007). ............. ...............57.....

2-15 Partial Least Squares regression on hyperspectral data from ASD device with a range
of 350-2500 nm and a spectral resolution of 1 nm for prediction of soil volumetric
water content (VWC), percent leaf nitrogen concentration (%N), biomass, and visual
quality collected from fairway height hybrid bermudagrass (Cynodon dactylon X C.
transvaalensistr~r~r~r~r~r~ Burtt-Davy), four irrigation rates (60, 80, 100, and 120% estimated
ET values), four N rates (0, 25, 50, and 100 kg ha- ), averaged over all dates (22
June, 12 July, 25 July, 22 Aug., and 4 Sept. 2007) ................. ............... 58...........

3-1 Evaluation of means of biomass (g), volumetric water content (VWC), visual ratings:
color, quality, density (rated on NTEP 1-9 scale 9 is best and 6 is acceptable), from
fairway height hybrid bermudagrass (Cynodon dactylon X C. transvaalensis,
'Princess 77) over five dates (22 June, 12 July, 25 July, 22 Aug., and 4 Sept. 2007)
combined. ........._ ...... .. ...............74....

3-2 Evaluation of means of Normalized Difference Vegetation Index (NDVI), Leaf Area
Index (LAI) computed from Crop Circle and ASD reflectance data, as well as Stress1
and Stress2 indices computed from ASD reflectance data from fairway height hybrid
bermudagrass (Cynodon dactylon X C. transvaalensis, 'Princess 77) over five dates
(22 June, 12 July, 25 July, 22 Aug., and 4 Sept. 2007) combined ................. ................75

















80


70-


60-


S 50 -



E 40
ooo

30-






20




Untreated 0.05 0.1 0.2 0.4

Rate (kg a.i. ha l)



Figure 3-1. Biomass measurements collected from fairway height hybrid bermudagrass (Cynodon dactylon X C. transvaalensis
'Princess 77) treated with trinexapac-ethyl at 0.05, 0.1, 0.2, and 0.4 kg a.i. ha '. Plots measured 1.5 X 3.0 m with three
replications. Means for biomass measurements was computed using Fishers protected LSD (p< 0.05) from data over 5
dates (22 June, 12 July, 25 July, 22 Aug., and 4 Sept. 2007) combined.









in the GA biosynthesis pathway so that GA' s are formed but are not active and serve as

precursors for plant biochemical processes.

Historically PGRs are relatively expensive compared to other chemical applications made

by turf managers. In an effort to reduce the cost of using PGRs, researchers have investigated

new methods to reduce the use rate. Researchers used remote sensing, the global positioning

system (GPS), and geographic information systems (GIS), and variable rate application (VRA)

technology to make site specific applications ofPGRs. These image-based PGR applications

were tested in cotton (Gossypium sp.) to reduce the chemical input of PGRs while also reducing

cost. Among comparisons of different image based-application (Variable Rate and Site-Specific)

methods to traditional broadcast applications for PGRs in cotton, it was found that the broadcast

applications were on average 19 % more costly than the site-specific applications, and 27 %

more costly than applications using variable rate methods. This demonstrates the potential

economic benefit of remote sensing technology for PGR use (Bethel, 2003).

In the turf industry, research has been conducted to assess various parameters of turfgrass

stress using remotely sensed data. Several different indices have been developed that have been

correlated with plant stress. One of the most common indices to assess turfgrass stress is the

Normalized Difference Vegetation Index (NDVI) which is computed as the reflectance from the

Near Infrared (NIR) region minus the reflectance from the Red (R) region divided by a sum of

both (NIR-R )/(NIR+R) (Rouse et al., 1973). Fenstemaker-Shaulis et al. (1997) found a negative

correlation with NDVI and canopy temperature, (r2=0.74) and a positive correlation with NDVI

and plant moisture content (r2=0.90). They also concluded that NDVI was not as a strong

predictor of biomass (r2=0.37). Trenholm et al. (1999b) tested various wavelengths 507 nm, 559

nm, 661 nm, 706 nm, 760 nm, 813 nm, and 935 nm as well as NDVI, IR/R (LAI) computed as









CHAPTER 4
CONCLUSIONS

Currently many turf managers and researchers still use time and labor intensive techniques

to manage and assess turf that originate from decades ago. Many of these management practices

have been proven repeatedly to work in a variety of situations to assess turfgrass stress, but may

be time consuming and inconsistent. However, there are new developments in ways of assessing

and mapping stress that increase the efficiency by which one manages and assesses turf. With

increasing water and other environmental restrictions turf managers and researchers need be

aware of technological advances in their industry.

The research discussed in this thesis gives greater insight into the use of some the

technological advancements available to turfgrass managers demonstrating the sensitivity of

remote sensing instrumentation to various parameters of turfgrass stress. In conducting these

experiments many realizations were made about the usability and future potential for remote

sensing technology in the turfgrass management field. Some recommendations for anyone who

wishes to continue the research referred to in this thesis are as follows:

Irrigation X Nitrogen study. Conduct the experiment in the earlier part of the growing

season, when rainfall is not so prevalent. This will potentially reduce uniformity in soil VWC.

Also, apply N every two weeks as described by Kruse et al. (2005); this will continually promote

differences in growth. In addition, investigate more into the use of PLS regression on

hyperspectral data to formulate prediction models for N and soil VWC levels.

PGR study. Conduct the experiment on turfgrass that does not produce seedheads. In this

experiment seed heads could have potentially skewed remotely sensed data; this will at least take

that factor out. Also, obtain remote sensing data before and after PGR applications to assess the

differences they create. From this data possibly try to make applications based on remotely










Brown, K.W., R.L. Duble, and J.C. Thomas. 1977. Influence of Management and Season on Fate
of N Applied to Golf Greens. Agron. J. 69:667-671.

Ehrler, W.L., S.B. Idso, R.D. Jackson, and R.J. Reginato. 1978. Wheat canopy temperature:
Relation to Plant Water Potential. Agron. J. 70: 251-256.

Fenstermaker- Shaulis, L.K., A. Leskys, and D.A. Devitt. 1997. Utilization of Remotely Sensed
Data to Map and Evaluate Turfgrass Stress Associated with Drought. Journal of Turfgrass
Management. 2:65-81.

Geeske, Joel, J.A. Gamon, and C.B. Field. 1997. Production Effieiency in Sunflower: The Role
of Water and Nitrogen Stress. Remote Sens. Environ. 62:176-188.

Hopkins, W. G. 1999. Introduction to Plant Physiology. 2nd ed. John Wiley and Sons Inc. New
York, Ch. 22 The physiology of plants under stress. 453-457.

Horst, G.L, J.C. O'Toole, and K.L. Faver. 1989. Seasonal and Species Variation in Baseline
Functions for Determining Crop Water Stress Indices in Turfgrass. Crop Sci. 29:1227-
1232.

Hutto, K.C., R.L King, J.D. Byrd, and D.R. Shaw. 2006. Implementation of Hyper spectral
Radiometry in Irrigation Management of Creeping Bentgrass Putting Greens. Crop. Sci.
46:1564-1569.

Idso, S.B., R.J. Reginato, D.C Reicosky, and J.L. Hatfield. 1981. Determining Soil Induced Plant
Water Potential Depressions in Alfalfa by Means of Infrared Thermometry. Agron. J. 73:
826-830.

Ikemura, Y., and B. Leinauer. 2006. Remote Sensing Technology Detects Turfgrass Stress.
Golfdom, May, p.63-66.

Jalali-Farahani, H.R., A.D. Matthias, and D.C. Slack. 1993. Crop Water Stress Index Models for
Bermudgrass Turf: A Comparison. Agron. J. 85:1210-1217.

Jalali-Frahani, H.R., D.C. Slack, D.M. Kopec, A.D. Matthias, and P.W. Brown. 1994. Evaluation
of Resistances for Bermudagrass Turf Crop Water Stress Index Models. Agron. J. 86: 574-
581.

Kruse, J.K., N.E. Christians, and M.H. Chaplin. 2005. Remote Sensing of Phosphorus
Deficiencies in Agrostis Stolonifera. Inter. Turfgrass Soc. Res. J. 10:923-928.

Kruse, J.K., N.E. Christians, and M.H. Chaplin. 2006. Remote Sensing of Nitrogen Stress in
Creeping Bentgrass. Agron. J. 98:1640-1645.

Lee, Wonsuk, S.W. Searcy, and T. Kataoka. 1999. Assessing Nitrogen Stress in Corn Varieties
of Varying Color. 18-21 July. ASAE Annual International Meeting. Toronto, Ontario,
Canada.









When prediction is the goal, then PLS regression is a highly useful tool (Tobias, 1997).

Kruse et al. (2006) found PLS regression yielded a stronger relationship to predict N

concentration compared to NDVI in plant leaf tissue indicating its potential use in future

development models. Lee et al. (1999) also found that PLS and Principal Component Analysis

(PCA) prediction models performed similar to each other and both were better than MLR

regression to predict the N status in corn.

Summary

The use of remote sensing technology to predict turfgrass stress could decrease the time

and labor required to evaluate turfgrass health (Osborne et al., 2002). This would not only reduce

the environmental impact of turfgrass management practices, but could reduce maintenance costs

of the various inputs required by turfgrass managers. Current dependence on agricultural

chemicals have heightened many environmental concerns in Florida due to the sandy soils and

heavy rainfall which increases potential for runoff and leaching of chemicals if applied in excess

(Min and Lee, 2003). As stated earlier, real world applications of remote sensing technology for

utilization in turfgrass are still in their infancy. At the present time, an experienced turf

manager' s eye is still better than current computerized systems (Ikemura and Leinauer, 2006).

However, studies have shown that various remote sensing devices have strong potential to detect

a variety of turfgrass stresses (Ikemura and Leinauer, 2006).

Applying remote sensing technology to the field of turfgrass management could possibly

help reduce irrigation and provide more efficient watering and fertilization routines (Ikemura and

Leinauer, 2006; Bell et al., 2004). Combined with GPS, GIS, and variable rate application

technology, there is a potential to apply nutrients as well as irrigation based on plant need rather

than broadcast applications of entire areas (Bell et al., 2004). Currently, most of the remote

sensing research is conducted on agronomic food crops and forested areas. Only limited









and oven dried at 52 oC for 7d before being weighed again for determination of biomass

production.

Reflectance Measurements

Spectral measurements were taken from two remote sensing devices. A ground based

vehicle mounted optical sensor was used to assess turf health. A Crop Circle (Model ACS-210)

(Holland Scientific) was fitted onto a Toro Greensmaster 1000 walking greens mower 0.81 m

above ground which provided a 0.45 m2 field of view. The Crop Circle sensor produces it own

light source at 650 nm and 880 nm and measures reflected values to produce an NDVI which is

calculated as (880-650 nm)/(880+650 nm). No calibration method was used for the sensor since

it has the ability to detect its own light source from incoming ambient sunlight.

A hand-held hyperspectral spectroradiometer (FieldSpec Pro; Analytical Spectral

Devices, Inc., Boulder, CO) was also used to collect reflectance data. This radiometer has a

spectral range of 300-2,500 nm with a 1 nm resolution and a 23 o foreoptic. Reflectance readings

were collected from shoulder height on each plot which provided a 0.65 m2 field of view.

Radiance values are expressed in percent reflectance compared to standardization with a white

reference value across the entire spectral range. A white reference was used for calibration

purposes at the beginning of data collection with the ASD unit, and again every 15 minutes

depending on the length of time needed to collect data. Usually only one white reference was

needed to take data on all plots. All reflectance readings were taken in full sunlight between the

hours of 1100 and 1400 Central Standard Time (CST) to minimize variance caused by incoming

solar radiation.

Statistical Analysis

Analysis of variance (ANOVA) was performed using the PROC GLM method (SAS

Inst., 2003) to compare the differences among PGR treatments to biomass, soil VWC, visual









Plant Growth Regulator Importance

Plant Growth Regulators (PGRs) have become a very important tool in turfgrass

management. Many turfgrass managers use PGRs to suppress vertical turf growth which

ultimately reduces maintenance costs and improves turf quality. Some data suggests that PGR-

treated turf can produce greater root mass, recover from injury faster, reduce water use rate,

reduce disease incidence, and reduce Poa annua weed populations (Branham, 1997).

A PGR is a substance that adjusts the growth and development of a plant. This is generally

achieved through the inhibition of gibberellic acid (GA), a hormone which is responsible for cell

elongation. Plant growth regulators were first introduced for use on Eine turf in 1987

(VanBibber, 2006). The first of these chemicals were flurprimidol (Trade name Cutless) and

paclobutrazol (Trade name Trimmit), which inhibit GA synthesis at relatively the same point in

the GA biosynthesis pathway, resulting in very similar plant responses. However, paclobutrazol

is the more active compound and uses lower rates to achieve the same response as higher rates of

flurprimidol (Branham, 1997). The newest of the PGRs, trinexapac-ethyl (Trade name Primo),

was released commercially in 1995 and has become the standard for use throughout the industry

(Branham, 1997). It facilitates GA inhibition later in the GA biosynthesis pathway so that GA's

are formed but are not active and serves as precursors for plant biochemical processes. Plant

growth regulators are relatively expensive compared to other chemicals used by turf managers.

More efficient applications of PGRs could potentially reduce the application cost of these

chemicals.

Remote Sensing

Remote sensing can be defined as obtaining information about an obj ect, area, or

phenomenon by analyzing data acquired by a device that is not in contact with that obj ect, area,

or phenomenon (Lillesand and Keifer, 2000). For many years researchers have entertained the









these parameters to better meet the needs on a locational basis (Stowell and Gelernter, 2006).

However highly variable agronomic conditions create the need for further research to

incorporate the use of remote sensing into turfgrass management to reduce chemical inputs.

(Trenholm et al., 1999b). The obj ective of this research was to assess the ability of remote

sensing instruments to detect differences in turf growth parameters as influenced by applications

of PGRs on hybrid bermudagrass (Cynodon dactylon X C. transvaalensis~t~t~t~t~t~t~ 'Princess 77) turf.

Materials and Methods

The field experiment was conducted at the University of Florida, West Florida Research

and Education Center, Jay, FL. Plots measured 1.5 m x 3.0 m and treatments were arranged in a

factorial design observing three different PGRs, each at four incremental levels- V/2, 1, 2, and 4

times the labeled rates, with three replications per treatment. Plant growth regulators were

applied on 14 June 2007 and 20 July 2007. Plant growth regulators treatments included

trinexapac-ethyl applied at 0.05, 0.1, 0.2, and 0.4 kg a.i. ha- ethephon applied at 3.8, 7.6, 15.2,

30.4 kg a.i. ha- and flurprimidol applied at 0.3, 0.6, 1.1, and 2.3 kg a.i. ha- Plots were mowed

2 to 3 times per week at a height of 1.2 cm with a Toro Reelmaster 3100-D. Supplemental

irrigation was applied as needed. Data was collected five times throughout the growing season

in 2007 on 22 June, 12 July, 25 July, 22 Aug., and 4 Sept. When data were collected all

measurements were taken within a one hour time period.

Plots were rated visually for color, quality, and density based on the standard 1-9

National Turfgrass Evaluation Program (NTEP) rating scale where 9 is the highest and 6 is the

lowest acceptable rating. Soil volumetric water content readings were taken via Time Domain

Reflectometry with a Fieldscout TDR 300 soil moisture meter (Spectrum Technologies, Inc.).

Clipping samples for biomass determination were collected from a 4.3 m2 area immediately after

reflectance data and visual evaluations were obtained. Clipping tissue samples were weighed










LIST OF FIGURES


Figure page

2-1 Partial Least Squares regression on hyperspectral data from ASD device with a range
Of 3 50-2500 nm and a spectral range of 1 nm for prediction of volumetric water
content (VWC) collected from fairway height hybrid bermudagrass (Cynodon
dactylon X C. transvaalensistr~r~r~r~r~r~ Burtt-Davy), four irrigation rates (60, 80, 100, and
120% estimated ET values), four N rates (0, 25, 50, and 100 kg ha- ), averaged over
all dates (22 June, 12 July, 25 July, 22 Aug., and 4 Sept. 2007). ............. ....................59

2-2 Coefficient estimates (r2) Of Crop Circle and ASD reflectance data computed into a
Normalized Difference Vegetation Index (NDVI), to soil volumetric water content
(VWC) collected from fairway height hybrid bermudagrass (Cynodon dactylon X C.
transvaalensistr~r~r~r~r~r~ Burtt-Davy), four irrigation rates (60, 80, 100, and 120% estimated
ET values), four N rates (0, 25, 50, and 100 kg ha- ), and averaged over all dates (22
June, 12 July, 25 July, 22 Aug., and 4 Sept. 2007). ................. ....._.._................6

2-3 Coefficient estimates (r2) of Crop Circle vs. ASD using reflectance data to compute
a Normalized Difference Vegetation Index (NDVI) collected from fairway height
hybrid bermudagrass (Cynodon dactylon X C. transvaalensis, Burtt-Davy), four
irrigation rates (60, 80, 100, and 120% estimated ET values), four N rates (0, 25, 50,
and 100 kg ha-1), and averaged over all dates (22 June, 12 July, 25 July, 22 Aug.,
and 4 Sept. 2007). ............. ...............61.....

3-1 Biomass measurements collected from fairway height hybrid bermudagrass
(Cynodon dactylon X C. transvaalensis 'Princess 77) treated with trinexapac-ethyl at
0.05, 0.1, 0.2, and 0.4 kg a.i. ha- Plots measured 1.5 X 3.0 m with three
replications. Means for biomass measurements was computed using Fishers
protected LSD (p< 0.05) from data over 5 dates (22 June, 12 July, 25 July, 22 Aug.,
and 4 Sept. 2007) combined............... ...............80

3-2 Biomass measurements collected from fairway height hybrid bermudagrass
(Cynodon dactylon X C. transvaalensis 'Princess 77) treated with flurprimidol at 0.3,
0.6, 1.1, and 2.3 kg a.i. ha- Plots measured 1.5 X 3.0 m with three replications.
Means for biomass measurements was computed using Fishers protected LSD (p<
0.05) from data over 5 dates (22 June, 12 July, 25 July, 22 Aug., and 4 Sept. 2007)
combined. ........._._.. ...._... ...............81....

A-1 Experimental layout for detection of leaf nitrogen concentration and soil volumetric
water content using ground-based remote sensing technology............... ...............8









LIST OF TABLES


Table page

2-1 Analysis of variance of soil volumetric water content (VWC), percent leaf nitrogen
tissue concentration (%/N), visual ratings: color, quality, density (rated on NTEP 1-9
scale, where 9 is the best rating and 6 is acceptable), and biomass. Normalized
difference vegetation index (NDVI), Leaf area index (LAI), 880 nm, and 650 nm
reflectance were obtained from Crop Circle device. Crop water stress index (CWSI)
was computed from IR thermometer readings. All data was collected from fairway
height hybrid bermudagrass (Cynodon dactylon X C. transvaalensis, Burtt-Davy),
four irrigation rates (60, 80, 100, and 120% estimated ET values), four N rates (0, 25,
50, and 100 kg ha- ), and five dates (22 June, 12 July, 25 July, 22 Aug., and 4 Sept.
2007) combined. ............. ...............44.....

2-2 Analysis of variance of individual reflectance bands from ASD reflectance
hyperspectral data with a range of 350-2500 nm and spectral resolution of 1 nm
collected from fairway height hybrid bermudagrass (Cynodon dactylon X C.
transvaalensistr~r~r~r~r~r~ Burtt-Davy), four irrigation rates (60, 80, 100, and 120% estimated
ET values), four N rates (0, 25, 50, and 100 kg ha- ), and five dates (22 June, 12 July,
25 July, 22 Aug., and 4 Sept. 2007) combined. .............. ...............45....

2-3 Analysis of variance of average range computations from 630-690 nm and 2080-
23 50 nm; NDVI, LAI, Stress1 and Stress2 indices; and 605 nm/5 15 nm, 915/975
nm, and 865/725 nm ratios computed from an Analytical Spectral Device (ASD)
with a range of 350-2500nm and spectral resolution of 1 nm collected from fairway
height hybrid bermudagrass (Cynodon dactylon X C. transvaalensis, Burtt-Davy),
four irrigation rates (60, 80, 100, and 120% estimated ET values), four N rates (0, 25,
50, and 100 kg ha- ), and five dates (22 June, 12 July, 25 July, 22 Aug., and 4 Sept.
2007) combined. ............. ...............46.....

2-4 Evaluation of means of volumetric water content (VWC), leaf nitrogen tissue
concentration (%N), visual ratings: color, quality, density (rated on NTEP 1-9 scale 9
is best and 6 is acceptable), and biomass (g) collected from fairway height hybrid
bermudagrass (Cynodon dactylon X C. transvaalensistr~r~r~r~r~r~ Burtt-Davy), four irrigation
rates (60, 80, 100, and 120% estimated ET values), four N rates (0, 25, 50, and 100
kg ha- ), and five dates (22 June, 12 July, 25 July, 22 Aug., and 4 Sept. 2007)
combined ................. ...............47.................

2-5 Evaluation of means of Normalized difference vegetation index (NDVI), Leaf area
index (LAI), 850 nm, and 650 nm reflectance from Crop Circle device and crop
water stress index (CWSI) computed from IR thermometer readings collected from
fairway height hybrid bermudagrass (Cynodon dactylon X C. transvaalensistr~r~r~r~r~r~ Burtt-
Davy), four irrigation rates (60, 80, 100, and 120% estimated ET values), four N
rates (0, 25, 50, and 100 kg ha- ) and five dates (22 June, 12 July, 25 July, 22 Aug.,
and 4 Sept. 2007) combined............... ...............48









Quality

Improvements in quality were observed with PGR applications. Although they were not

significant, all rates of TE and flurprimidol, and the 3.8 kg a.i. ha-l rate of ethephon slightly

improved turf quality (Table 3-1). Similar to turf color ratings, only slight correlations existed

between remotely sensed data and visual quality ratings. The highest correlations were achieved

by the ASD device at wavelengths 661 nm (r2=0.53) and 510 nm (r2=0.48), and ratio 915/975 nm

(r2=0.42) (Table 3-4 and 3-5).

Density

Visual density ratings showed a general increase in density from all rates of TE and

flurprimidol, and the low rate of ethephon, however these increases were not significant. Only

slight correlations existed between remotely sensed data and visual density ratings. The highest

correlations were achieved by the ASD device at wavelengths 661 nm (r2=0.53) and 510 nm

(r2=0.48) (table 3-4).

Reflectance Indices vs. PGR Application rates

Data derived from the Crop Circle device show a general increase in NDVI and LAI for

all rates of TE and flurprimidol compared to the untreated, however, only the 0.05 and 0. 1 kg a.i.

ha-l rate of TE and the 0.3, 0.6, and 1.1 kg a.i. ha-l rate of flurprimidol are significant (Table 3-

2). Furthermore, NDVI and LAI derived from ASD data show the same trend, yet only the 2.2

kg ha-l is statistically different from the untreated control (Table 3-2). The general trend for the

STRESS 1 and STRESS 2 indices show similar results, however, STRESS 1 reveals that TE and

flurprimidol treated turf was under less stress than the untreated control (Table 3-2).

Crop Circle Device

Reflectance data obtained from the Crop Circle device did not demonstrate it's ability to

detect biomass differences attributed to PGR applications (Table 3-3). The ability of the device










2-6 Evaluation of means of individual band reflectance from the ASD device with a
range of 350-2500 nm and a spectral resolution of 1 nm collected from fairway
height hybrid bermudagrass (Cynodon dactylon X C. transvaalensis, Burtt-Davy),
four irrigation rates (60, 80, 100, and 120% estimated ET values), four N rates (0, 25,
50, and 100 kg ha- ) and five dates (22 June, 12 July, 25 July, 22 Aug., and 4 Sept.
2007) combined. ............. ...............49.....

2-7 Evaluation of means of ASD reflectance readings of average range computations
from 630-690 nm and 2080-23 50 nm, computed indices Normalized Difference
Vegetation Index (NDVI ), Leaf Area Index (LAI) and Stress 1 and Stress2 indices,
and ratios 605/515 nm, 915/975 nm, 865/725 nm taken from an ASD device with a
range of 350-2500 nm and a spectral resolution of 1 nm collected from fairway
height hybrid bermudagrass (Cynodon dactylon X C. transvaalensis, Burtt-Davy),
four irrigation rates (60, 80, 100, and 120% estimated ET values), four N rates (0, 25,
50, and 100 kg ha- ) and five dates (22 June, 12 July, 25 July, 22 Aug., and 4 Sept.
2007) combined. ............. ...............50.....

2-8 Evaluation of means of NDVI and LAI computed from Crop Circle instrument, and
LAI computed from ASD collected from fairway height hybrid bermudagrass
(Cynodon dactylon X C. transvaalensistr~r~r~r~r~r~ Burtt-Davy), four irrigation rates (60, 80,
100, and 120% estimated ET values), four N rates (0, 25, 50, and 100 kg ha-1), and
five dates (22 June, 12 July, 25 July, 22 Aug., and 4 Sept. 2007) combined, which
had N treatment X date interactions............... ..............5

2-9 Evaluation of Means of leaf N concentration tissue (%N) collected from fairway
height hybrid bermudagrass (Cynodon dactylon X C. transvaalensis, Burtt-Davy),
four irrigation rates (60, 80, 100, and 120% estimated ET values), four N rates (0, 25,
50, and 100 kg ha- ), and five dates (22 June, 12 July, 25 July, 22 Aug., and 4 Sept.
2007) combined, which had N treatment X date interactions............... ..............5

2-10 Evaluation of means of individual reflectance wavelengths at 813 nm and 935 nm
collected from fairway height hybrid bermudagrass (Cynodon dactylon X C.
transvaalensistr~r~r~r~r~r~ Burtt-Davy), four irrigation rates (60, 80, 100, and 120% estimated
ET values), four N rates (0, 25, 50, and 100 kg ha- ), and five dates (22 June, 12 July,
25 July, 22 Aug., and 4 Sept. 2007) combined, which had N treatment X date
interactions. .............. ...............53....

2-11 Evaluation of means of ratios 915 nm/975 nm and 865 nm/925 nm computed from
analytical spectral device (ASD) with a range of 350-2500 nm and a spectral
resolution of 1 nm, collected from fairway height hybrid bermudagrass (Cynodon
dactylon X C. transvaalensistr~r~r~r~r~r~ Burtt-Davy), four irrigation rates (60, 80, 100, and
120% estimated ET values), four N rates (0, 25, 50, and 100 kg ha- ), and five dates
(22 June, 12 July, 25 July, 22 Aug., and 4 Sept. 2007) combined, which had N
treatment X date interactions. ............. ...............54.....









Table 2-4. Evaluation of means of volumetric water content (VWC), leaf nitrogen tissue concentration (%N), visual ratings: color,
quality, density (rated on NTEP 1-9 scale 9 is best and 6 is acceptable), and biomass (g) collected from fairway height
hybrid bermudagrass (Cynodon dactylon X C. transvaalensistr~r~r~r~r~r~ Burtt-Davy), four irrigation rates (60, 80, 100, and 120%
estimated ET values), four N rates (0, 25, 50, and 100 kg ha- ), and five dates (22 June, 12 July, 25 July, 22 Aug., and 4
Sept. 2007) combined.
Visual Ratings
N Rate VWC % N Color Quality Density biomass (g)
0 kg ha-l 22.04 3.38 7.44b 7.12b 7.01 48.47
25 kg hal 22.07 3.61 7.73a 7.29ab 6.99 55.85
50 kg hal 21.87 3.62 7.76a 7.22ab 7.03 53.23
100 kg hal 22.48 3.99 7.85a 7.40a 7.12 57.05
LSD N.S. -- 0.20 0.21 N.S. N.S.
"--" denotes LSD is not valid due to a significant date by N treatment interaction. Reported separately in Table 2-5 based on
interaction.
All values with a given number for the LSD denotes p < 0.05









ACKNOWLEDGMENTS

I wish to acknowledge the guidance and direction of Dr. J. Bryan Unruh for all his wisdom

and support during the last two years. It has been a learning experience like none other. Sincere

appreciation also goes to all my committee members, Dr. Barry Brecke, Dr. Laurie Trenholm,

Dr. Won Suk "Daniel" Lee, and Dr. Jasmeet Judge. I appreciate their helping me whenever

asked. Special thanks also go to all the current and past members of the West Florida Research

and Education Center: Dr. Ken Hutto, Raymond Edwards, Rex Lawson, Chris Adkison, Jason

Ging, Chase McKeithen, Phil Moon, and Dr. Darcy Partridge. I appreciate their help and

support. This undertaking could not have been accomplished without them. Bryan Schwartz and

Dr. Jason Dettman-Kruse thank you for all the assistance in SAS: This paper would have never

been written in time without your help.

I appreciate the support and wisdom received from my parents. Their guidance throughout

my life has inspired me to better myself not only through hard work in my career but as person.

I am grateful to the University of Florida for providing me with the most valuable resource

I will ever have, an education and the inspiration to never stop learning and always strive for the

best. I am a proud part of the "Gator Nation." The time I have spent with the University of

Florida will never be forgotten.





















Visual Ratings -


Table 2-1. Analysis of variance of soil volumetric water content (VWC), percent leaf nitrogen tissue concentration (%N), visual
ratings: color, quality, density (rated on NTEP 1-9 scale, where 9 is the best rating and 6 is acceptable), and biomass.
Normalized difference vegetation index (NDVI), Leaf area index (LAI), 880 nm, and 650 nm reflectance were obtained
from Crop Circle device. Crop water stress index (CWSI) was computed from IR thermometer readings. All data was
collected from fairway height hybrid bermudagrass (Cynodon dactylon X C. transvaalensis, Burtt-Davy), four irrigation
rates (60, 80, 100, and 120% estimated ET values), four N rates (0, 25, 50, and 100 kg ha )~, and five dates (22 June, 12
July, 25 July, 22 Aug., and 4 Sept. 2007) combined.


Crop Circlez
880
LAIW nm
.0001 0.0001
.0001 0.0001
.1709 0.0515
.9631 0.9970


1RY

CWSI"
0.2616
0.0001
0.7858
0.9896

0.5505
0.8610
0.7184


Source of
Variation df VWC % N Dry W Color Quality Density
Rep 2 0.0001 0.0044 0.0001 0.0102 0.0048 0.0015
Date 4 0.0001 0.0001 0.0001 0.0001 0.0001 0.0001
Irrigation 3 0.3106 0.8398 0.3815 0.9382 0.5418 0.4083
Dxl 12 0.9981 0.9934 0.9401 0.9968 0.6717 0.9902
Error a 12
Nitrogen 3 0.1447 0.0001 0.1100 0.0001 0.0153 0.1087
IxN 9 0.0004 0.0502 0.3567 0.6267 0.2050 0.0065
DxN 12 0.2988 0.0002 0.4345 0.0595 0.7875 0.5126
error b 36
Probability of greater F ratio (P>F) for Date, Irrigation, and N treatment components
z Model AC S-210, Holland Scientific
SIR thermometer Raytek Raynger@ STTw, TTI, Inc.
x NDVI= (880-650 nm)/(880+650 nm)
LAI= 880/650 nm

Where Te= Canopy Temperature, Ta= Actual Temperature, a= actual, u= upper limit


650
nm
0.0001
0.0001
0.2266
0.9860

0.0001
0.3908
0.0612


NDVIx
0.0001
0.0001
0.2885
0.9921

0.0001
0.8093
0.0029


0
0
0
0


I


0.0001
0.8038
0.0001


0.2223
0.4815
0.0309


:, and 1= lower limit









Biomass

Biomass was not modeled well by reflectance data compared to other factors tested.

However, some weak correlations exist (Tables 2-12, 2-13, 2-14, and 2-15). This is similar to

research conducted by Kruse et al., 2005, 2006; Fenstemaker-Shaulis et al., 1997; and Trenholm

et al., 1999b.

The inability to effectively model biomass in this experiment could be attributed to the

methodology used. Nitrogen treatments were only applied twice one month apart. In other

research where biomass was adequately modeled (Kruse et al., 2005, 2006) N treatments were

applied every 2 weeks thus continually promoting greater differences in turf growth.

Crop Circle vs. ASD

The Crop Circle and ASD instruments are two very different remote sensing devices that

differ dramatically in price and use-ability. The ASD device is a highly scientific instrument

with great ability to give a researcher a plethora of reflectance data from a broad region of the

electromagnetic spectrum. However, it is limited in its ability because it requires full sunlight

for use and only collects reflectance data when manually prompted by the user. Nonetheless,

when reflectance data in this magnitude is analyzed using the modern regression techniques, like

PLS regression, the correlations could potentially be very specific to the parameter tested and

highly predictive of new data. Therefore, prediction models could likely be constructed for

specific forms of turfgrass stress.

Conversely, optical sensing devices like the Crop Circle instrument are less complicated to

use but only give a limited amount of information typically in the visible and NIR regions.

However, this data can be used to compute NDVI, LAI, and a variety of other indices to make

the data more usable. Furthermore, they produce their own light source and, thus, can be

mounted on a vehicle or mower for autonomous reflectance measurement collection.









Table 2-7. Evaluation of means of ASD reflectance readings of average range computations from 630-690 nm and 2080-23 50 nm,
computed indices Normalized Difference Vegetation Index (NDVI ), Leaf Area Index (LAI) and Stress1 and Stress2
indices, and ratios 605/515 nm, 915/975 nm, 865/725 nm taken from an ASD device with a range of 350-2500 nm and a
spectral resolution of 1 nm collected from fairway height hybrid bermudagrass (Cynodon dactylon X C. transvaalensistr~r~r~r~r~r~
Burtt-Davy), four irrigation rates (60, 80, 100, and 120% estimated ET values), four N rates (0, 25, 50, and 100 kg ha-l ),
and five dates (22 June, 12 July, 25 July, 22 Aug., and 4 Sept. 2007) combined.
Ranne Averane (ASD z) Venetative Indices (ASD z) Ratios (ASD z
605/515 915/975 865/725
N Rate 630-690 nm 2080-23 50 nm NDVI LAI x Stress 1 Stress 2 nm nm nm

0 kg ha-l 0.071a 0.175 0.715ab 7.06 0.367ab 0.788ab 1.02a 0.99 1.88

25 kg hal 0.073a 0.182 0.707b 7.26 0.375a 0.794a 1.03a 0.99 1.89

50 kg hal 0.743a 0.186 0.700b 7.28 0.383a 0.791a 1.03a 0.99 1.90

100 kg hal 0.657b 0.170 0.729a 8.21 0.354b 0.779b 1.03b 1.00 1.97
LSD 0.0053 N.S. 0.018 -- 0.019 0.009 0.01---
"--" denotes LSD is not valid due to a significant date by N treatment interaction
z FieldSpec Pro; Analytical Spectral Devices, Inc., Boulder, CO
YNDVI= (880-650 nm)/(880+650 nm)
x LAI= 880/650 nm
SStressl= 706/760 nm
a Stress2= 706/813 nm









(Branham, 1997). In other research where biomass has been adequately modeled, PGRs were

not used (Kruse et al., 2005, 2006) and biomass was modeled from plots with increasing N rate

applications applied at frequent intervals.

Volumetric Water Content

PGRs applied to bermudagrass in this study had only a slight effect on soil VWC. Only

TE and flurprimidol applied at the lowest rate gave significant differences from the untreated

control (Table 3-1). Slight correlations were observed (Table 3-3) in modeling soil VWC with

NIR reflectance quantified from the Crop Circle device as well as reflectance in the SWIR region

at 2,132 nm (Table 3-4) and the 2,080-2,350 nm range (Table 3-5) quantified from the ASD

device. This is in agreement with research conducted by Ripple (1986) and Hutto et al. (2006).

Visual Ratings

Color

Color rating means (Table 3-1) show an increase in turf color by TE applications at 0.05 and 0. 1

kg a.i. ha- flurprimidol applications at 0.3, 0.6, and 1.1 kg a.i. ha- and ethephon applications at

the low rate compared to the untreated control, however, these increases were not significant

(Table 3-1). This general increase could be attributed to seed head reduction achieved by PGR

applications, since seed heads are generally less green causing a waning in color on the turf.

Additionally, increasing rates of PGRs resulted in a trend towards reduced turf color which could

be caused by a slight injury effect.

Only slight correlations in visual color ratings to remotely sensed data from either device

were achieved (Tables 3-3, 3-4, 3-5, 3-6). The highest correlations were observed in data taken

with the ASD device at wavelengths 661 nm (r2=0.42) and 5 10nm (r2=0.3 5), and ratio 605/5 15

nm (r2=0.41) (Table 3-4 and 3-5).









to correlate with visual human evaluations was weak. However, the NIR (880 nm) reflectance

values from the device slightly modeled changes in soil VWC (r2=0.43) throughout the

experiment (Table 3-3).

The inability of the Crop circle device to adequately detect differences in biomass, visual

ratings, and VWC can be attributed to a number of things previously mentioned. Among these

are the seedhead production of the turf which could have potentially skewed the reflectance data,

and the lack of significant differences among the parameters tested.

ASD Device

Individual wavelengths, computed indices, and various ratios obtained from ASD

reflectance data did not demonstrate the ability to detect differences in biomass attributed to PGR

applications. Some weak correlations exist between visual ratings and NDVI, LAI, Stress1, and

Stress2 indices computed from the raw data generated by the ASD instrument (Table 3-3).

Similar, and sometimes stronger, correlations exist with visual ratings with individual

wavelengths investigated throughout the visible range (Table 3-4). Again, this inability to

adequately model visual ratings, VWC, and biomass could be attributed to the lack of significant

differences in the various parameters of the treated verses the untreated control and the seedhead

production of the turf.

PLS Regression

The use of PLS regression on the hyperspectral data generated from the ASD instrument

did not perform well in predicting differences in biomass created by various applications of

PGRs. Furthermore PLS regression only slightly predicted differences in soil VWC (r2=0.36)

and visual quality (r2=0.26) (Table 3-6). This again is likely due to the lack of significant

differences in soil VWC and visual quality from the untreated control.









In this experiment VWC was best modeled by both remote sensing devices, Crop Circle

and ASD. NDVI from the Crop Circle (r2=0.71) (Table 2-12) (Figure 2-2) device and PLS

regression from the ASD device (r2=0.72) were the highest correlations achieved (Table 2-15)

(Figure 2-1). NDVI from the ASD device was close in comparison to the Crop Circle (r2=0.57)

to model VWC (Figure 2-2).

Even though both of these devices utilize completely different sources of light they both

calculate NDVI similarly and are close in comparison to model VWC (Figure 2-3). Granted the

prediction of VWC with PLS regression on ASD data could be highly predictive of new data and

specific to actual VWC and not just turf health, the Crop Circle device shows there is some

sensitivity to this parameter.

Conclusions

The results indicate reflectance data best modeled soil VWC, visual ratings, and some

treatment effects from N. Based on research by Ripple (1986), the elevated levels of prediction

for VWC could be due to a reduction in chlorophyll as the leaves dried in response to limiting

soil VWC and not a true relation to actual soil water status. Nonetheless, this research gives

greater insight into the use of remote sensing technology and its place for use by turf managers

and researchers demonstrating the sensitivity of remote sensing instrumentation to various

parameters of turfgrass stress. Previous research found optical sensing measurements using

NDVI could provide a fast and accurate alternative to traditional visual rating systems (Trenholm

et al., 1999b; Bell et al., 2000) as well as assessing turfgrass stress (Trenholm et. al., 1999a,

Fenstemaker-Shaulis et al., 1997). Our results from two different remote sensing instruments are

in agreement with this previous work.

The Crop Circle instrument performed equally well with the ASD device in modeling

stress attributed to reduced soil VWC (Figure 2-2) and was consistent with visual ratings in this









Normalized Difference Vegetation Index (NDVI) which is computed as the reflectance in the

NIR region minus the reflectance in the red region divided by a sum of both (NIR-R)/ (NIR+R).

Bahrum et al. (2003) found that NDVI values from remotely sensed data can be used to assess

turfgrass stress. Fenstemaker-Shaulis et al. (1997) found a negative correlation between NDVI

and canopy temperature (r2=0.74), and a positive correlation between NDVI and plant moisture

content (r=0.90). However, NDVI was not as strong a predictor of biomass (r2=0.37). In

addition, spatial mapping of NDVI values can provide valuable insight into the status of

irrigation system uniformity and management practices (Fenstemaker-Shaulis et al., 1997). Bell

et al. (2002b) found strong correlations between NDVI and turfgrass color ratings. Additionally,

the use of NDVI ratings obtained from vehicle mounted optical sensors were effective for

measuring herbicide damage (Bell et al., 2000), variable N applications (Bell et al., 2002a), and

significant seasonal responses to N treatments (Xiong et al., 2007). It has also been reported that

NDVI is a better estimator of chlorophyll content than visual color evaluation (Bell et al., 2004).

Trenholm et al. (1999) tested four indices to detect stress in turfgrass. These indices were:

* NDVI computed as (935-661 nm)/ (935+661 nm)
* Leaf Area Index (LAI) computed as 93 5/661 nm
* Stress 1 computed as 706/760 nm
* Stress 2 computed as 706/813 nm

They concluded NDVI, IR/R, and Stress 2 indices, as well as individual wavelength

measurements at 661 nm and 813 nm were well correlated to visual ratings.

Older work involving remote sensing typically used multispectral data, which is collected

in individual or wide bandwidths and gave a coarse representation of the electromagnetic

spectrum (Lillesand and Kiefer, 2003). With the introduction of hyperspectral data, which

typically acquires 200 or more wavelength bands in very narrow resolution throughout the

visible, NIR, and MIR regions, researchers have more data to analyze (Lillesand and Keifer,










bentgrass (Agrostris stolinifera L.). Reflectance wavelengths for predicting P concentration of

plant tissue were in the blue (480 nm), yellow (565 nm), orange (595 nm), and red (650 nm)

regions with an overall r2 Value Of 0.73. This reflectance response relates to the spectral

characteristic of anthocyanin which increased in concentration under P stress conditions (Kruse

et al., 2005). Hutto et al. (2006) found that potassium deficiency could be detected at

bandwidths from 750 nm to 785 nm.

Biomass

Green plant material has high reflectance in the NIR (700-1100 nm) region, while dead or

dying plant material is just the opposite (Geeske et al., 1997). Yoder and Pettigree-Crosby

(1995) related this correlation in the NIR region to biomass, or energy status of the plant.

Similar results were found by Kruse et al. (2005) on creeping bentgrass, where reflectance in the

green (5 10 nm and 53 5 nm), red (63 5 nm), and NIR (73 5 nm) were significant in predicting

biomass. Starks et al. (2006) found crude protein concentration, biomass production, and crude

protein availability of bermudagrass (Cynodon dactylon L. Pers.) grown in pastures closely

correlated with canopy reflectance ratios of 605/515 nm, 915/975 nm, and 865/725 nm.

Conversely, Fenstemaker-Shaulis et al. (1997) found that NDVI is not a strong predictor of

biomass (r2=0.37). Kruse et al. (2005) also observed no correlation between biomass and NDVI.

Kruse et al. (2006) found implementing Partial Least Squares regression on hyperspectral data to

be an adequate predictor of biomass in creeping bentgrass compared to indices NDVI, LAI,

Stress, and Stress2.

PGRs

Not a large body of literature exists in the area of remote sensing and the effects of PGR

applications. Some image-based applications have been tested in cotton (Gosspypium spp.) to

reduce PGR use for economic reasons (Bethel et al., 2003). Among comparisons of different









DETECTION OF TURF GRASS STRESS USING GROUND BASED REMOTE SENSING


By

JASON HAMILTON FRANK
















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

UNIVERSITY OF FLORIDA

2008









Tissue N Concentration vs. Reflectance

Reflectance values in this experiment did not prove to be a good predictor of leaf tissue N

concentration in bermudagrass using linear or PLS regression to model percent N verses

reflectance measurements However, some weak correlations exist (Tables 2-12, 2-13, 2-14, 2-

15). This could potentially be due to the experimental design where N treatments were only

applied twice, one month apart, causing the waning effect mentioned in the N X Date interaction

section. Comparatively, research conducted by Kruse et al. (2005) and Xiong et al. (2007) where

N treatments were applied every two weeks promoted a more consistent response to treatments

across the growing season.

VWC vs. Reflectance

The highest r2 ValUeS for estimation of VWC are derived from PLS regression (r2= 0.72)

calculated from reflectance data from the ASD device (Table 2-15) (Figure 2-1), and NDVI ( 2=

0.71) derived from reflectance data from the Crop Circle device (Table 2-12) (Figure 2-2). Also,

LAI (r2= 0.59), NIR reflectance (r2= 0.61), and RED reflectance (r2= 0.52) derived from Crop

Circle values provide adequate correlation as well (Table 2-12). This corresponds to vegetative

indice values NDVI (r2= 0.57), LAI (r2= 0.55), Stress1 (r2= 0.61), Stress2 (r2= 0.64), and ratios

605/515 nm (r2= 0.66), 915/975 nm (r2= 0.68), 865/725 nm (r2= 0.44) derived from the ASD unit

(Table 2-14). Some correlations at individual bands 755 nm (r2= 0.53), 813 (r2= 0.53), and 935

nm (r2= 0.49) were also notable (Table 2-13).

The high r2 ValUeS for estimation of VWC with remotely sensed data from Crop Circle and

various individual bands, indices, and ratios from the ASD device could be attributed to the

extremely low VWC values recorded on 22 Aug 2007. On this date, cumulative ET calculations

were in deficit 2.39 cm causing a significant decline in VWC. This resulted in a maj or decline in

turf health. In previous research many indices such as NDVI do well in predicting turf stress in









Table A-1. Description of treatments for Figure A-1


Whole Plot
Irrigation
60% ET
60% ET
60% ET
60% ET
80% ET
80% ET
80% ET
80% ET
100% ET
100% ET
100% ET
100% ET
120% ET
120% ET
120% ET
120% ET


Subplot
Nitrogen
0 kg ha
25 kg ha l
50 kg ha l
100 kg ha l
0 kg ha-
25 kg ha l
50 kg ha l
100 kg ha l
0 kg ha-
25 kg ha l
50 kg ha l
100 kg ha l
0 kg ha-
25 kg ha l
50 kg ha l
100 kg ha l


Rep
2
704
702
703
701
802
803
804
801
503
502
501
504
602
601
603
604


Treatment #
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16


1
101
102
103
104
201
202
203
204
301
302
303
304
401
402
403
404


3
1201
1203
1202
1204
901
903
904
902
1003
1002
1004
1001
1101
1104
1102
1103










LIST OF REFERENCES


Anonymous. 2007. Proxy (specimen label). Bayer Scientific, Triangle Park, NC. U.S. Patent
4,240,819. Date issued: 25 Apr. 2005.

Aston, A.R., and C.H.M. Van Bavel. 1972. Soil Surface Water Depletion and Leaf Temperature.
Agron. J. 64:368-373.

Augustin, B.J., and G.H. Snyder. 1984. Moisture Sensor Controlled Irrigation For Maintaining
Bermudagrass Turf. Agron. J. 76:848-850.

Bahrun, A., V.O. Mogensen, and C.R. Jensen. 2003. Water Stress Detection in Field Grown
Maize by Using Spectral Vegetation Index. Commun. Soil Sci. Plant Anal. 34:65-79.

Barret, J. 2003. Golf Course irrigation: Environmental Design and Management Practices. 413-
417. John Wiley & Sons, Inc., Hoboken, NJ. 413-417.

Bethel, M., T. Gress, S. White, J. Johnson, T. Sheely, B. Roberts, N. Gat, G. Scriven, G.
Hagglund, M. Paggi, and N. Groenenberg. 2003. Image-Based Variable Rate Plant Growth
Regulator Application in Cotton at Sheely Farms in California. 6-10 Jan. 2003. Beltwide
Cotton Conferences, Nashville, TN.

Bell, G.E., B.M. Howell, G.V. Johnson, W.R. Raun, M.L. Stone, and J.B. Solie. 2004. Optical
Sensing of Turfgrass Chlorophyll Content and Tissue Nitrogen. Hortic. Sci. 39(5): 1130-
1132.

Bell, G.E., D.L. Martin, R.M. Kuzmic, M.L. Stone, and J.B. Solie. 2000. Herbicide Tolerance of
Two Cold-Resistant Bermudagrass (Cynodon spp.) Cultivars Determined by Visual
Assessment and Vehicle Mounted Optical Sensing. Weed Tech. 14:635-641.

Bell, G.E., D.L. Martin, M.L. Stone, J.B. Solie, and G.V. Johnson. 2002a. Turf area Mapping
Using Vehicle Mounted Optical Sensors. Crop Sci. 42:648-651.

Bell, G.E., D.L. Martin, S.G. Wiese, D.D. Dobsom, M.W. Smith, M.L. Stone, and J.B. Solie.
2002b. Vehicle- Mounted Optical Sensing An Obj ective Means for Evaluating Turf
Quality. Crop Sci. 42:197-201.

Blackmer, T.M., J.S. Schepers, and G. E. Varvel. 1994. Light reflectance Compared with Other
Nitrogen Stress Measurements in Corn Leaves. Agron. J. 86: 934-938.

Blackmer, T.M., J.S. Schepers, G.E. and Varvel, E.A. Walter-Shea. 1996.Nitrogen Deficiency
Detection Using Reflected Shortwave Radiation From Irrigated Corn Canopies. Agron. J.
88: 1-5.

Branham, Bruce. "Plant Growth Regulators for Fine Turf Use."
http ://www.interactiveturf.com/tips/1 997_02.htm. 24 Oct. 2007.





































O 2008 Jason Hamilton Frank









This is why research focusing on improving water use management in agriculture is very

important. It has both environmental and economic consequences, especially as water

availability becomes more limited. According to a survey conducted in 2000 by the Golf Course

Superintendents Association of America, the median annual water usage of a golf course in the

United States is 32.49 cm applied to an average of 31.44 ha resulting in 108 million L used

annually (Barret, 2003).

Historically the maj ority of turfgrass managers scheduled irrigation usage based on their

experience and set schedule of time intervals, both of which can result in over watering

(Augustin and Snyder, 1984). Irrigation schedules are often based on calendar dates such as 3

or 7 times a week. Studies have shown that this style of irrigation may promote over watering

and provide too much moisture to the turf (Unruh and Elliott, 1999). This wastes water and

energy, produces poor playing and agronomic conditions, and leads to degradation of the

environment from excessive runoff (Barret, 2003). Tools are needed to monitor turf water status

to aid in decisions that result in more efficient irrigation (Bahrun et al., 2003). Irrigation systems

should be operated so that the addition of water never exceeds that lost by ET in any given area

(Brown et al., 1977). This will increase irrigation efficiency and possibly reduce plant water

stress in those areas.

Researchers and turf managers over the past two decades have investigated many ways to

determine when to irrigate. Many of the techniques developed to assess plant water status are

time consuming, spatially restrictive, and costly (Ripple, 1986). Currently, visual symptoms, ET

rates, and tensiometers are all methods for determining irrigation needs (Turgeon, 2008).

However, when visual symptoms of drought are apparent, it is already too late because the turf is

already stressed. Likewise, ET values are generally not site specific and actual ET can differ










APPENDIX B
DAILY RAINFALL AND ET DATA FOR JAY, FL


Table B-1. Daily Rainfall and ET data for Jay, FL


Cumulative
Difference
(cm)
(0.10)
(0.31)
(0.18)
1.57
1.24
3.30
4.09
3.50
2.97
2.51
2.21
1.85
1.60
1.24
1.04
0.61
0.45
(0.08)
(0.51)
(0.82)
(0.26)
0.35
0.33
(0.16)
0.17
(0.36)
(0.84)
(1.32)
(0.92)
(1.37)
(1.71)
0.78
0.56
0.76
0.28
(0.18)
(0.64)
0.43


Rainfall
(cm>a
0.38
0.00
0.51
2.29
0.00
2.36
1.27
0.00
0.00
0.00
0.00
0.00
0.10
0.00
0.25
0.03
0.38
0.00
0.00
0.00
0.89
0.94
0.46
0.00
0.76
0.00
0.00
0.00
0.89
0.00
0.00
2.79
0.10
0.71
0.00
0.00
0.00
1.40


Irrigation
Applied (cm)
0.20b
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.20b
0.00


ET (cm)a
0.48
0.41
0.38
0.53
0.33
0.30
0.48
0.58
0.53
0.46
0.30
0.36
0.36
0.36
0.46
0.46
0.53
0.53
0.43
0.30
0.33
0.33
0.48
0.48
0.43
0.53
0.48
0.48
0.48
0.46
0.33
0.30
0.33
0.51
0.48
0.46
0.46
0.53


Day
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21


June


July









which causes a decrease in the structural integrity of many important elements of the

photosynthetic process (Hopkins, 1999). This in turn will decrease electron transport and

photophosphorolation leading to damage in the thylakoid membrane and adenine triphosphate

(ATP) synthetase protein (Hopkins, 1999). The hormone abscisic acid (ABA) also plays an

important role in the plant' s ability to endure water stress. Under stress, ABA accumulates in the

leaves causing an efflux of potassium from the guard cells, resulting in stomatal closure and

reducing transpiration (Hopkins, 1999).

Plant water stress can occur if a plant receives too much water or if it does not receive

enough. Flooding leads to a lack of oxygen, this leads to decreased respiration, nutrient uptake,

and root functionality, thus reducing photosynthesis (Hopkins, 1999). Drought stress from lack

of water leads to increased solute concentration in the protoplasm, which in turn affects many

physiological processes (Hopkins, 1999). However, mild drought stress can cause the roots of a

plant to grow deeper. This is why irrigation events should be spaced out as long as possible

(Unruh and Elliott, 1999).

Water is an important resource; however, it is not a renewable resource (Barret, 2003).

Ninety-seven percent of all the earth's water is salt water, three percent is fresh water; however,

two-thirds of this is frozen in polar ice caps (Barret, 2003). Therefore, one percent of the Earth' s

fresh water must meet the water needs of manufacturing, mining, agriculture, and turf not met by

rainfall (Barret, 2003). In Florida the average rainfall does not meet the watering requirements

to maintain golf courses at the expectations of players and industry viewers (Unruh and Elliott,

1999). As a result, the irrigation system is the single most important tool available for turfgrass

management to a golf course superintendent or turf manager (Barret, 2003).










Table 3 -2. Evaluation of means of Normalized Difference Vegetation Index (NDVI), Leaf Area Index (LAI) computed from Crop
Circle-and ASD reflectance data, as well as Stress1 and Stress2 indices computed from ASD reflectance data from fairway
height hybrid bermudagrass (Cynodon dactylon X C. transvaalensistr~r~r~r~r~r~ 'Princess 77) over five dates (22 June, 12 July, 25
July, 22 Aug., and 4 Sept. 2007) combined.
Crop Circlez ASDY
NDVIx LAI" NDVIx LAI" Stress1v Stress2u
Trinexapac-ethyl
0.05 kg a.i. hal 0.66 ab 5.08 ab 0.76 ab 7.42 a-c 0.38 bc 0.79 cd
0.1 kg a.i. ha-l 0.67 ab 5.13 ab 0.77 ab 7.76 ab 0.36 c 0.77 cd
0.2 kg a.i. ha-l 0.65 a-d 4.90 a-c 0.76 ab 7.43 a-c 0.38 bc 0.78 cd
0.4 kg a.i. ha-l 0.66 a-c 5.00 a-c 0.76 ab 7.46 a-c 0.37 bc 0.78 cd
Flurprimidol
0.3 kg a.i. ha-l 0.67 ab 5.10 ab 0.77 ab 7.51 ab 0.37 bc 0.78 cd
0.6 kg a.i. ha-l 0.66 ab 5.01 ab 0.76 ab 7.34 a-d 0.37 bc 0.78 cd
1.1 kg a.i. ha-l 0.68 a 5.24 a 0.78 a 7.82 a 0.36 c 0.77 d
2.3 kg a.i. ha-l 0.66 a-c 5.03 ab 0.76 ab 7.49 a-c 0.37 bc 0.78 cd
Ethephon
3.8 kg a.i. ha-l 0.65 a-d 4.82 a-c 0.75 a-c 7.16 a-d 0.39 a-c 0.78 b-d
7.6 kg a.i. ha-l 0.64 b-d 4.64 b-d 0.75 a-d 6.96 a-d 0.39 a-c 0.78 a-d
15.2 kg a.i. hal 0.62 de 4.45 dc 0.72 cd 6.50 cd 0.41 a 0.79 ab
30.4 kg a.i. hal 0.60 e 4.23 d 0.72 d 6.39 d 0.42 a 0.80 a

13) Untreated 0.63 c-e 4.64 b-d 0.74 b-d 6.81 b-d 0.40 ab 0.79 a-c
Means in the same column followed by the same letter are not significantly different (LSD; P< 0.05)
z Model ACS-210, Holland Scientific
YFieldSpec Pro; Analytical Spectral Devices, Inc., Boulder, CO
xNDVI= (880-650nm)/(880+650nm)
" LAI= 880/650nm
SStressl= 706/760nm
a Stress2= 706/813nm









APPENDIX A
EXPERIMENTAL LAYOUT FOR DETECTION OF LEAF NITROGEN CONCENTRATION
AND SOIL VOLUMETRIC WATER CONTENT USING GROUND-BASED REMOTE
SENSING TECHNOLOGY










Table B-1. Continued


Cumulative
Difference
(cm)
4.64
5.56


Rainfall
Day (cm>a ET (
3 2.34 0.
4 1.32 0.
aBased on data from http://fawn.ifas.ufl .edu/.
SIrrigation applied to water in fertilizer application


Irrigation
Applied (cm)
0.00
0.00


(cm~a
28
41









1987). These enhanced techniques for obtaining the data have created the need to explore

different statistical methods to quantify and interpret the large amount of data produced.

Traditionally, linear regression is the simplest way to understand variance. However in

interpreting reflectance data it has been found that other regression techniques better assess plant

status (Kruse et al., 2005).

Partial Least Squares (PLS) regression is a method used to predict variables when there are

a large number of factors; typically more than the number of observations (Tobias, 1997). The

use of multiple variables is generally tested through Multiple Linear Regression (MLR) (Tobias,

1997). However, the limitation with MLR is when the number of factors becomes too large the

model may become over-fitting (Tobias, 1997). This is observed when a large number of

variables are present but only a few account for most of the variation (Tobias, 1997). Almost

perfect models may be achieved; however they will not be highly predictive of new data.

When prediction is the obj ective then PLS regression is a highly useful tool (Tobias,

1997). Kruse et al. (2006) found PLS regression yielded a stronger relationship to predict N

concentration compared to NDVI in creeping bentgrass indicating its potential use in future

development models. Lee et al. (1999) also found that PLS and Principal Component Analysis

(PCA) prediction models performed similar to each other and both were better than MLR

regression to predict the N status in corn (Zea mays).

Currently, the maj ority of remote sensing research is conducted on agronomic food crops

and forested areas. Only limited research is available in the area of remote sensing and turfgrass

management (Trenholm et al., 1999b). If conditions across the entire golf course were uniform

in microclimates, soil types, turf varieties, pest densities, nutrient status, and irrigation

performance there would be little need for advanced sensor systems to measure the variability of
















ET (cm)
4.92125
5.5118
8.46455
10.4775
12.5984
12.192
12.5984
12.2047
10.0965
7.28345
4.7625
3.34645


Rainfall-ET (cm)
2.36855
8.1026
3.4544
4.8768
-4.7879
11.89355
11.71575
8.0391
5.9436
-1.59385
7.23265
7.3533


Based on data from http://fawn.ifas.ufl .edu/


January
February
March
April
May
June
July
August
September
October
November
December


APPENDIX D
MOTHLYRAINFALL AND ET AVERAGES FROM 2003-2006 FOR JAY, FL


Table D-1.


MothlyRainfall and ET averages from 2003-2006 for Jay, FL


Rainfall
(cm)
7.2898
13.6144
11.91895
15.3543
7.8105
24.08555
24.31415
20.2438
16.0401
5.6896
11.99515
10.69975


Cumulative Total
Rainfall-ET (cm)
2.36855
10.47115
13.92555
18.80235
14.01445
25.908
37.62375
45.66285
51.60645
50.0126
57.24525
64.59855












Table 2- 14. Coefficients estimates (r2) Of average range computations from 630-690 nm and 2080-23 50 nm, and computed indices
Normalized difference vegetation index (NDVI), Leaf area index (LAI) and Stress1 and Stress2 indices also taken from an
ASD with a range of 3 50-2500 nm and a spectral resolution of 1 nm to %N concentration leaf tissue, soil volumetric water
content (VWC) visual color, quality, density ratings, and biomass production collected from fairway height hybrid
bermudagrass (Cynodon dactylon X C. transvaalensistr~r~r~r~r~r~ Burtt-Davy), four irrigation rates (60, 80, 100, and 120% estimated
ET values), four N rates (0, 25, 50, and 100 kg ha )~, and averaged over all dates (22 June, 12 July, 25 July, 22 Aug., and 4
Sept. 2007).
Range Average (ASDz) Vegetative Indices (ASDz) Ratios (ASDz)
630-690 2080-2350 605/515 915/975 865/725
Parameter nm nm NDVIY LAIx Stress 1" Stress 2" nm nm nm
%/N 0.02 0.02* 0.13*** 0.20** 0.15*** 0.16*** 0.21*** 0.22*** 0.14***
VWC 0.22*** 0.29*** 0.57*** 0.53*** 0.61*** 0.64*** 0.66*** 0.68*** 0.44***
Color 0.38*** 0.40*** 0.53*** 0.55*** 0.53*** 0.55*** 0.45*** 0.52*** 0.51***
Quality 0.62*** 0.69*** 0.71*** 0.66*** 0.67*** 0.69*** 0.46*** 0.61*** 0.67***
Density 0.65*** 0.71*** 0.69*** 0.63*** 0.65*** 0.67*** 0.44*** 0.58*** 0.67***
Biomass (g) 0.26*** 0.25*** 0.26*** 0.35*** 0.24*** 0.27*** 0.15*** 0.24*** 0.67***


*, **, *** Significant at the 0.05, 0.01, and 0.001 probability levels, respectively
z FieldSpec Pro; Analytical Spectral Devices, Inc., Boulder, CO
YNDVI= (880-650 nm)/(880+650 nm)
x LAI= 880/650 nm
" Stressl= 706/760 nm
SStress2= 706/813 nm


ul









CHAPTER 1
LITERATURE REVIEW

Introduction

The maj ority of turf managers and researchers still use outdated time and labor intensive

techniques to manage and assess turf. Many of these management practices have been

repeatedly proven to work in a variety of situations to assess turfgrass stress, but may be time

consuming and inconsistent. For example, in the turfgrass research field visual evaluations can

be labor intensive and variable among evaluators (Trenholm et al., 1999). However, there are

new developments in ways of assessing and mapping stress that will increase the efficiency by

which one manages and assesses turf. With increasing water restrictions and other

environmental constraints turf managers and researchers need be aware of technological

advances in their industry. Recent environmental concerns, such as nutrient and pesticide

leaching and runoff, have pressed turf managers to reduce inputs used for turf maintenance (Bell

et al., 2002).

There are a variety of new or improved methods available to detect water and nutrient

stresses before they are visible. The basis of this technology relies heavily on the global

positioning system (GPS), geographic information systems (GIS), and remote sensing

technology. Much of this technology is in the beginning stages, however, the theory for use of

this technology has been proven in other agronomic industries such as food production systems.

This technology has been used as a management tool to maximize crop yield and minimize cost

through site-specific application of inputs by applying nutrients, water and pesticides only where

needed (Lee et al., 1999).









idea of measuring plant stress using remote sensing technology. Plant light interception

influences growth and physiological responses (Salisbury and Ross, 1992). When light is

intercepted by the plant, it is absorbed, transmitted, or reflected (Turgeon, 2008). Therefore, by

quantifying the reflected light, remote sensing technology can help to detect the onset of

turfgrass stress (Ikemura and Leinauer, 2006).

Presently, real world applications of remote sensing technology in turfgrass management

are still in their infancy (Ikemura and Leinauer, 2006). Studies have shown that image analysis

and various remote sensing devices have strong potential to detect a variety of turfgrass stresses

(Ikemura and Leinauer, 2006). Spectral radiometry (Trenholm et al., 2000), and infrared

thermometry (Slack et al., 1981) have been shown to adequately assess light reflectance at

various wavelengths. By differentiating certain wavelength characteristics, insights into growth

and adaptive characteristics and how a plant responds to stress can be seen (Trenholm et al.,

2000). There have been several aspects of remote sensing technology explored to reveal the level

of stress that may be occurring in a plant. Researchers have created a variety of theories utilizing

remote sensing techniques and exploring many parts of the electromagnetic spectrum to describe

various plant stresses.

Vegetative Indices

Researchers have found certain relationships exist between various aspects of the

electromagnetic spectrum and a variety of plant parameters and stress (Bell et al., 2000, 2002a,

2002b, 2004; Trenholm et al., 1999a, 1999b, 2000; Kruse et al., 2005; Fenstemaker-Shaulis et

al., 1997; Bahrum et al., 2003; Xiong et al., 2007). These relationships are formulated into

indices to normalize data for varying illumination conditions (Lillesand and Kiefer, 2003). One

of the most common indices is the Normalized Difference Vegetation Index (NDVI) computed









Table 2-5. Evaluation of means of Normalized difference vegetation index (NDVI), Leaf area index (LAI), 850 nm, and 650 nm
reflectance from Crop Circle device and crop water stress index (CWSI) computed from IR thermometer readings
collected from fairway height hybrid bermudagrass (Cynodon dactylon X C. transvaalensis, Burtt-Davy), four irrigation
rates (60, 80, 100, and 120% estimated ET values), four N rates (0, 25, 50, and 100 kg ha-l ), and five dates (22 June, 12
July, 25 July, 22 Aug., and 4 Sept. 2007) combined.
Crop Circle Reflectance Readingsz IR ThermometerY
N Rate NDVIx LAI" 880 nm 650 nm CWSI"
0 kg ha-l 0.59 4.35 1.38 .34a 0.44
25 kg hal 0.61 4.58 1.38 .33a 0.44
50 kg hal 0.60 4.61 1.38 .33a 0.45
100 kg hal 0.62 4.99 1.39 .31b 0.43
LSD -- -- N.S. 0.01 N.S.
"--" denotes LSD is not valid due to a significant date by N treatment interaction. Reported separately in Table 5 based on interaction.
All values with a given number for the LSD denotes p < 0.05
z Model ACS-210, Holland Scientific
SIR thermometer Raytek Raynger@ STTw, TTI, Inc.
P xN\DVI= (880-650 nm)/(880+650 nm)
LAI= 880/650 nm
"CWSI= (To-Ta)a (To-Ta)1 / (To-Ta)u (To-Ta)1
Where Te= Canopy Temperature, Ta= Actual Temperature, a= actual, u= upper limit, and 1= lower limit









CHAPTER 2
DETECTION OF LEAF NITROGEN CONCENTRATION AND SOIL VOLUMETRIC
WATER CONTENT USING GROUND-BASED REMOTE SENSING TECHNOLOGY

Introduction

Remote sensing is defined as obtaining information about an obj ect, area, or phenomenon

by analyzing data acquired by a device that is not in contact with that obj ect, area, or

phenomenon (Lillesand and Keifer, 1987). For many years researchers have entertained the idea

of measuring plant stress using remote sensing technology. Plant light interception significantly

influences growth and physiological responses (Salisbury and Ross, 1992). When light is

intercepted by the plant it is absorbed, transmitted, or reflected (Salisbury and Ross, 1992).

Using remote sensing technology to quantify the light that is reflected could help to detect the

onset of turfgrass stress (Ikemura and Leinauer, 2006; Kruse et al., 2005; Hutto et al., 2006;

Trenholm et al., 1999a). Real world applications of remote sensing technology for use in

turfgrass management are still in their infancy; however, studies have shown that image analysis

and various remote sensing devices have strong potential to detect a variety of turfgrass stresses

(Ikemura and Leinauer, 2006).

Recent investigations into remote sensing instrumentation have produced promising results

for turf managers. However, the majority of turf managers and researchers still use outdated

time and labor intensive techniques to manage and assess turf. Nonetheless, there are new

developments in ways of assessing and mapping plant stress that may increase the efficiency by

which we manage and assess turf. The use of remote sensing technology to estimate turfgrass

stress could significantly decrease the time and labor required to assess these levels in traditional

ways, thus reducing cost as well (Osborne et al., 2002).

The many great achievements in U.S. agricultural productivity in the past decades can be

attributed to the use of agricultural chemicals, including fertilizer and pesticides (Lee et al.,

































0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1


ASD


Figure 2-3. Coefficient estimates (r2) of Crop Circle vs. ASD using reflectance data to compute a Normalized Difference Vegetation
Index (NDVI) collected from fairway height hybrid bermudagrass (Cynodon dactylon X C. transvaalensistr~r~r~r~r~r~ Burtt-Davy),
four irrigation rates (60, 80, 100, and 120% estimated ET values), four N rates (0, 25, 50, and 100 kg ha-1), and averaged
over all dates (22 June, 12 July, 25 July, 22 Aug., and 4 Sept. 2007).









were taken via Time Domain Reflectometry (TDR) with a Field Scout TDR 300 Soil Moisture

meter (Spectrum Technologies, Inc., Plainfield, IL). Clipping tissue samples were immediately

weighed after collection and oven dried at 52 oC for 7d before being weighed again for

determination of biomass production. The oven dried tissue was ground and then analyzed for

Total Kj eldahl N content using the RFA Method, number A303-SO75-01 (Alpkem Corporation,

1990) as described by Mylavarapu and Kennelly (2007).

Reflectance Measurements

Spectral measurements were taken from three remote sensing devices. An infrared

thermometer (Raytek Raynger@ STTw, TTI, Inc.) was used to determine canopy temperature of

the turf and to develop a CWSI calculated as: CWSI= (To-Ta) a- (To-Ta)1/ (To-Ta)u (To-Ta)1,

Where Te= Canopy Temperature, Ta= Actual Temperature, a= actual, u= upper limit, and 1=

lower limit (Jalali-Farahani et al., 1994).

A Crop Circle (Model ACS-210) (Holland Scientific) was fitted onto a Toro Greensmaster

1000 walking greens mower 0.81m above the turf canopy which provided a 0.45 m2 field of

view. The Crop Circle sensor produces its own light source at 650 nm and 880 nm and measures

reflected values to produce an NDVI which is calculated as (880-650 nm)/ (880+650 nm). No

calibration method was used for the sensor since it has the ability to detects it own light source

from incoming ambient sunlight.

A hand-held hyperspectral spectroradiometer (Field Spec Pro; Analytical Spectral Devices,

Inc., Boulder, CO) was also used to collect reflectance data. This radiometer has a spectral range

of 300-2,500 nm with a 1 nm resolution and a 23 o foreoptic. Reflectance readings were collected

from shoulder height on each plot which provided a 0.65 m2 field of view. Radiance values are

expressed in percent reflectance compared to standardization with a white reference value across

the entire spectral range. A white reference was used for calibration purposes at the beginning of












3 USE OF GROUND-BASED REMOTE SENSING TECHNOLOGY TO ASSESS
BIOMASS PRODUCTION IN RESPONSE TO PLANT GROWTH REGULATOR
APPLICATIONS ................. ...............62.......... ......


Introducti on ................. ...............62.................
Materials and Methods .............. ...............66....
Reflectance Measurements ................. ...............67.......... ......
S tati sti cal Analy si s .............. ...............67....
Results and Discussion .............. ...............68....
B iom ass .............. ............. ..............6
Volumetric Water Content .............. ...............69....
Visual Ratings .............. ...............69....
Color............... ...............69.

Quality ................ ...............70.......... ......
D ensity ...................... ... .. .................7
Reflectance Indices vs. PGR Application rates .......___......... .........___......70
Crop Circle Device ............ ..... .__ ...............70...
ASD Device............... ...............71.
PLS Regression .............. ...............71....
Conclusions............... ..............7


4 CONCLUSIONS .............. ...............82....


APPENDIX


A EXPERIMENTAL LAYOUT FOR DETECTION OF LEAF NITROGEN
CONCENTRATION AND SOIL VOLUMETRIC WATER CONTENT USING
GROUND-BASED REMOTE SENSING TECHNOLOGY ................. .......................84


B DAILY RAINFALL AND ET DATA FOR JAY, FL................ ...............87..


C CUMALITIVE WEEKLY RAINFALL AND ET DATA FOR JAY, FL .............................90


D MOTHLYRAINFALL AND ET AVERAGES FROM 2003-2006 FOR JAY, FL...............91

LIST OF REFERENCES ............ ...............92.....


BIOGRAPHICAL SKETCH .............. ...............96....










1999). However, the increased dependence on pesticides and fertilizers have heightened many

environmental concerns especially in Florida due to the sandy soils and heavy rainfall, which

increases potential for heavy runoff and leaching of chemicals (Min and Lee, 2003). With

increasing water and other environmental restrictions, turf managers and researchers must be

aware of technological advances in their industry.

Recent concerns have pressed turf managers to reduce nutrient and pesticide inputs used

for turf maintenance (Bell et al., 2002a). In other industries, such as agronomic food production

crops, precision agriculture has been used as a management tool to maximize yield and minimize

cost (Bethel et al., 2003). The basis of this technology relies heavily on the Global Positioning

System (GPS), Geographic Information Systems (GIS), and Variable-Rate Application (VRA)

and remote sensing technology. Using this technology for site-specific management of nutrients,

water, and pesticides could optimize yield while minimizing cost (Lee et al., 1999).

Water is an important resource in turfgrass management; however, it is not a renewable

resource (Barret, 2003). The Earth's water reserves are 97 % salt water and 3 % fresh water. Of

the fresh water reserve, two-thirds is frozen in polar ice caps (Barret, 2003). Therefore, only 1 %

of the Earth' s water is available to meet the water needs of manufacturing, mining, agricultural,

and turf not met by rainfall (Barret, 2003). In addition, the average rainfall in Florida does not

meet the watering requirements to maintain golf course quality at the level expected by players

and industry viewers (Unruh and Elliot, 1999). As a result, the irrigation system is the single

most important tool available to a golf course superintendent or turf manager (Barret, 2003). As

water restrictions increase and pressure rises from environmental groups to restrict water use on

golf courses, it is increasingly imperative that turfgrass managers improve water use efficiency

(Martin et al., 1994).









environmental factors as well. Horst et al. (1989) suggested that canopy temperature minus air

temperature plotted as a function of Vapor Pressure Deficit (VPD), which is the difference in the

amount of moisture in the air and the amount of moisture the air can hold. This may be species

and mowing height dependent and may require further adjustment of the function for each

species under individual environmental regimes (Horst et al., 1989). Jalali-Frahani et al. (1993)

proposed that CWSI of turfgrass was not only a function of VPD but also net radiation from

sunlight. Martin et al. (1994) found that using variables accounting for VPD, wind speed, and

net radiation increased coefficient of determination (r2) ValUeS dramatically across several

varieties of turfgrass. Ehrler et al. (1978) concluded that temperature difference (T,-Ta) in wheat

(Triticum spp., durham 'Produra') is a reliable method for monitoring plant stress from aircraft or

satellite. Frequent measurements were necessary to reduce the variability in the data. Throssel

et al. (1987) found that temperature difference (Te-Ta) appeared to be indicative of water stress

for well-watered Kentucky bluegrass (Poa pratensis L.).

More recent investigations into plant and soil water status rely on spectral reflectance data

measured from plant or turfgrass canopies. Fenstemaker-Shaulis et al. (1997) found a negative

correlation with NDVI and canopy temperature (r2=0.74) and a positive correlation to plant

moisture content (r2=0.90). In addition, they concluded, that mapping NDVI values can also

provide valuable insight into the status of irrigation system uniformity and management

practices. Hutto et al. (2006) found that drought stressed species had a higher amount of

reflectance in the short wave infrared region (SWIR) (1,300-2,500 nm) than that of the non-

stressed control. Ripple (1986) found that reflectance in the 630-690 nm and 2,080-2,350 nm

ranges is strongly correlated with leaf water content. As reflectance decreased in these ranges

there was an inverse linear relationship with leaf relative water content. However, Ripple (1986)









Aside from the importance of water to plants, available nutrients also drive plant growth

and development. Nutrient availability affects many of the physiological processes a plant must

perform in order to develop properly. Nitrogen (N) is the most limiting nutrient in production of

non-legumous crops (Osborne et al., 2002). It is essential to the building of several important

plant structures including, proteins, nucleic acids, hormones, and most importantly chlorophyll,

which harvests light energy used in photosynthesis (Hopkins, 1999).

Remote sensing technology has been shown to assess stress that may be occurring in a

plant. Researchers have created a variety of theories and indices exploring many parts of the light

spectrum to describe various stresses. These stresses include water, nutrient, soil compaction,

pests, and disease.

One of the early theories involving remote sensing to predict plant water status was the

Crop Water Stress Index (CWSI), which correlated plant water potential with canopy

temperature (Ehrler et al., 1978). More recent investigations rely on spectral reflectance data

measured from plant or turfgrass canopies. Several spectral reflectance bands have been

identified for use in modeling various turfgrass stresses, especially those in the visible (400-700

nm) and near infrared (NIR) regions (700-1,300 nm) of the spectrum, as well as short wave

infrared (SWIR) regions (1,300-2,500 nm) (Hutto et al, 2006). Reflectance from healthy plant

canopies is characteristically low in the visible range of the spectrum due to high absorption by

chlorophyll for use in photosynthetic processes (Trenholm et al., 2000), If there is a particularly

high reflectance in this region, this could correlate to a physiological disruption in the plant

(Geeske et al., 1997).

Green plant material has high reflectance in the NIR range, while dead or dying plant

material has low reflectance (Geeske et al., 1997). Yoder and Pettigree-Crosby (1995) relate this









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

DETECTION OF TURF GRASS STRESS USING GROUND BASED REMOTE SENSING

By

Jason Hamilton Frank

May 2008

Chair: J. Bryan Unruh
Major: Horticulture Sciences

Many turf managers and researchers still use time and labor intensive techniques to

manage and assess turf that originate from decades ago. Many of these management practices

have been proven repeatedly to work in a variety of situations to assess turfgrass stress, but may

be time consuming and inconsistent. However, there are new developments in ways of assessing

and mapping stress that increase the efficiency by which one manages and assesses turf. With

increasing water and other environmental restrictions turf managers and researchers need be

aware of technological advances in their industry.

Two Hield experiments were conducted at the University of Florida, West Florida Research

and Education Center near Jay, FL. The first was the detection of leaf nitrogen concentration

and soil volumetric water content using remote sensing technology. Plots were arranged in a

randomized complete block design and treatments were arranged in a split plot factorial design

with four N levels split across four irrigation regimes with three replications per treatment. The

second experiment used remote sensing technology to assess biomass production created by

applications of plant growth regulators. Plots were arranged in a factorial design observing three

different PGRs, each at four incremental levels with three replications per treatment. Spectral

measurements were taken from two remote sensing devices.











Table 2-8. Evaluation of means of NDVI and LAI computed from Crop Circle instrument, and LAI computed from ASD collected
from fairway height hybrid bermudagrass (Cynodon dactylon X C. transvaalensistr~r~r~r~r~r~ Burtt-Davy), four irrigation rates (60,
80, 100, and 120% estimated ET values), four N rates (0, 25, 50, and 100 kg ha-1), and five dates (22 June, 12 July, 25
July, 22 Aug., and 4 Sept. 2007) combined, which had N treatment X date interactions.


NDVIz (Crop Circle )
25 kg 50 kg 100 kg
ha-l ha-l ha-
0.73a23 0.73a2 0.75al
0.69b2 0.69b2 0.71b1
0.69b2 0.70abl2 0.71b1
0.42dl2 0.39d2 0.53cl
0.52cl 0.49c1 0.5201


LAIx (Crop Circle)
25 kg 50 kg 100 kg
ha-l ha-l ha-
6.34a23 6.58a2 7.34al
5.49b2 5.50b2 5.99b1
5.41b2 5.69bl2 5.95 67b 1
2.45dl2 2.96cl2 2.41d2
3 .20cl 2.34dl 3.28c1


LAIx (ASD")
25 kg 50 kg
ha-l ha-
11.11a23 11.65a2
8.59b2 8.73bl2
9.29b2 9.32b2
3.39c1 3.1901
3.93cl 3.54cl


0 kg
ha-
0.70a3
0.67b3
0.66b3
0.53cl
0.44dl


0 kg
ha-
5.79a3
5.10b3
4.87b3
2.58dl
3.39c1


0 kg
ha-
10.20a23
8.28b2
8.25b3
3.66c1
4.89c1


100 kg
ha-l
13.08al
9.60b1
10.31b1
3.41cl
4.64d1


Date
6/22
7/12
7/25
8/22
9/4


Means in the same column and category followed by the same letter are not significantly different (LSD; P, 0.05); means in the same
row and category followed by the same number are not significantly different (LSD; P, 0.05).
zNDVI= (880-650 nm)/(880+650 nm)
SModel AC S-210, Holland Scientific
x LAI= 880/650 nm
" Field Spec Pro; Analytical Spectral Devices, Inc., Boulder, CO










Table 2-3. Analysis of variance of average range computations from 630-690 nm and 2080-23 50 nm; NDVI, LAI, Stress1 and
Stress2 indices; and 605 nm/515 nm, 915/975 nm, and 865/725 nm ratios computed from an Analytical Spectral Device
(ASD) with a range of 350-2500nm and spectral resolution of 1 nm collected from fairway height hybrid bermudagrass
(Cynodon dactylon X C. transvaalensistr~r~r~r~r~r~ Burtt-Davy), four irrigation rates (60, 80, 100, and 120% estimated ET values),
four N rates (0, 25, 50, and 100 kg ha )~, and five dates (22 June, 12 July, 25 July, 22 Aug., and 4 Sept. 2007) combined.
ASDz
Source of 630-690 2080- 605/515 915/975 865/725
Variation DF nm 2350 nm NDVIY LAI x Stress1 Stress2 0 nm nm nm
Rep 2 0.1209 0.0627 0.0313 0.1656 0.0099 0.0545 0.0206 0.0025 0.0018
Date 4 0.0001 0.0001 0.0001 0.0001 0.0001 0.0001 0.0001 0.0001 0.0001
Irrigation 3 0.2010 0.1034 0.1562 0.6883 0.1654 0.4132 0.8121 0.3246 0.6495
Dxl 12 0.0712 0.0768 0.1737 0.9830 0.1296 0.4497 0.8643 0.8901 0.9094
error a 12
Nitrogen 3 0.0105 0.0869 0.0050 0.0001 0.0089 0.0019 0.0001 0.0001 0.0001
IxN 9 0.6521 0.7096 0.8275 0.9484 0.8028 0.9325 0.6612 0.9299 0.9001
DxN 12 0.5014 0.7276 0.3930 0.0008 0.2417 0.3474 0.1934 0.0299 0.0001
Error b 36
Probability of greater F ratio (P>F) for Date, Irrigation, and N treatment components
zFieldSpec Pro; Analytical Spectral Devices, Inc., Boulder, CO
YNDVI= (880-650 nm)/(880+650 nm)
x LAI= 880/650 nm
SStressl= 706/760 nm
a Stress2= 706/813 nm




Full Text

PAGE 1

1 DETECTION OF TURFGRASS STRESS US ING GROUND BASED REMOTE SENSING By JASON HAMILTON FRANK A THESIS PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLOR IDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF SCIENCE UNIVERSITY OF FLORIDA 2008

PAGE 2

2 2008 Jason Hamilton Frank

PAGE 3

3 ACKNOWLEDGMENTS I wish to acknowledge the guidance and directio n of Dr. J. Bryan Unruh for all his wisdom and support during the last two year s. It has been a learning expe rience like none other. Sincere appreciation also goes to all my committee members, Dr. Barry Brecke, Dr. Laurie Trenholm, Dr. Won Suk Daniel Lee, and Dr. Jasmeet Judge I appreciate their helping me whenever asked. Special thanks also go to all the current and past member s of the West Florida Research and Education Center: Dr. Ken Hutto, Raymon d Edwards, Rex Lawson, Chris Adkison, Jason Ging, Chase McKeithen, Phil Moon, and Dr. Darcy Partridge. I appreciate their help and support. This undertaking could not have been accomplished without them. Bryan Schwartz and Dr. Jason Dettman-Kruse thank you for all the assi stance in SAS: This paper would have never been written in time without your help. I appreciate the support and wis dom received from my parent s. Their guidance throughout my life has inspired me to better myself not only through hard work in my career but as person. I am grateful to the University of Florida for providing me with the most valuable resource I will ever have, an education and the inspiration to never stop learning and always strive for the best. I am a proud part of the Gator Nation. The time I have spent wi th the University of Florida will never be forgotten.

PAGE 4

4 TABLE OF CONTENTS page ACKNOWLEDGMENTS...............................................................................................................3 LIST OF TABLES................................................................................................................. ..........6 LIST OF FIGURES................................................................................................................ .......10 ABSTRACT....................................................................................................................... ............11 CHAPTER 1 LITERATURE REVIEW.......................................................................................................13 Introduction................................................................................................................... ..........13 Irrigation Importance.......................................................................................................... ....14 Nutrient Importance............................................................................................................ ....17 Plant Growth Regulator Importance.......................................................................................18 Remote Sensing................................................................................................................. .....18 Vegetative Indices...........................................................................................................19 Human Visual Ratings.....................................................................................................20 Irrigation..................................................................................................................... .....20 Nutrient....................................................................................................................... .....23 Biomass........................................................................................................................ ...24 PGRs........................................................................................................................... .....24 Statistical Methods..........................................................................................................25 Summary........................................................................................................................ .........26 2 DETECTION OF LEAF NITROG EN CONCENTRATION AND SOIL VOLUMETRIC WATER CONTENT USING GROUND-BASED REMOTE SENSING TECHNOLOGY...................................................................................................28 Introduction................................................................................................................... ..........28 Materials and Methods.......................................................................................................... .34 Reflectance Measurements..............................................................................................35 Statistical Analysis..........................................................................................................36 Results and Discussion......................................................................................................... ..37 Influence of Irrigation Treatments...................................................................................37 Nitrogen Effects...............................................................................................................37 Nitrogen X Date Interaction............................................................................................38 Tissue N Concentration vs. Reflectance..........................................................................39 VWC vs. Reflectance......................................................................................................39 Visual Ratings.................................................................................................................40 Biomass........................................................................................................................ ...41 Crop Circle vs. ASD........................................................................................................41 Conclusions.................................................................................................................... .........42

PAGE 5

5 3 USE OF GROUND-BASED REMOTE SENSING TECHNOLOGY TO ASSESS BIOMASS PRODUCTION IN RESPONSE TO PLANT GROWTH REGULATOR APPLICATIONS................................................................................................................... .62 Introduction................................................................................................................... ..........62 Materials and Methods.......................................................................................................... .66 Reflectance Measurements..............................................................................................67 Statistical Analysis..........................................................................................................67 Results and Discussion......................................................................................................... ..68 Biomass........................................................................................................................ ...68 Volumetric Water Content..............................................................................................69 Visual Ratings.................................................................................................................69 Color.........................................................................................................................69 Quality......................................................................................................................70 Density.....................................................................................................................70 Reflectance Indices vs. PGR Application rates...............................................................70 Crop Circle Device..........................................................................................................70 ASD Device.....................................................................................................................71 PLS Regression...............................................................................................................71 Conclusions.................................................................................................................... .........72 4 CONCLUSIONS....................................................................................................................82 APPENDIX A EXPERIMENTAL LAYOUT FOR D ETECTION OF LEAF NITROGEN CONCENTRATION AND SOIL VOLUMETRIC WATER CONTENT USING GROUND-BASED REMOTE SENSING TECHNOLOGY.................................................84 B DAILY RAINFALL AND ET DATA FOR JAY, FL............................................................87 C CUMALITIVE WEEKLY RAINFALL AND ET DATA FOR JAY, FL.............................90 D MOTHLYRAINFALL AND ET AVERAGES FROM 2003-2006 FOR JAY, FL...............91 LIST OF REFERENCES .......................................................................................................... ..92 BIOGRAPHICAL SKETCH.........................................................................................................96

PAGE 6

6 LIST OF TABLES Table page 2-1 Analysis of variance of soil volumetric wa ter content (VWC), percent leaf nitrogen tissue concentration (%N), visual ratings: co lor, quality, density (rated on NTEP 1-9 scale, where 9 is the best rating and 6 is acceptable), and biomass. Normalized difference vegetation index (NDVI), Leaf area index (LAI), 880 nm, and 650 nm reflectance were obtained from Crop Circle device. Crop water stress index (CWSI) was computed from IR thermometer readings. All data was collected from fairway height hybrid bermudagrass ( Cynodon dactylon X C. transvaalensis, Burtt-Davy), four irrigation rates (60, 80, 100, and 120% es timated ET values), four N rates (0, 25, 50, and 100 kg ha-1), and five dates (22 June, 12 July, 25 July, 22 Aug., and 4 Sept. 2007) combined.................................................................................................................44 2-2 Analysis of variance of individual reflectance bands from ASD reflectance hyperspectral data with a range of 350-2500 nm and spectral resolution of 1 nm collected from fairway he ight hybrid bermudagrass ( Cynodon dactylon X C. transvaalensis, Burtt-Davy), four irrigation ra tes (60, 80, 100, and 120% estimated ET values), four N rates (0, 25, 50, and 100 kg ha-1), and five dates (22 June, 12 July, 25 July, 22 Aug., and 4 Sept. 2007) combined..................................................................45 2-3 Analysis of variance of average ra nge computations from 630 nm and 2080 2350 nm; NDVI LAI Stress1 and Stress2 indices; and 605 nm/515 nm, 915/975 nm, and 865/725 nm ratios computed from an Analytical Spec tral Device (ASD) with a range of 350-2500nm and spectral reso lution of 1 nm collected from fairway height hybrid bermudagrass ( Cynodon dactylon X C. transvaalensis, Burtt-Davy), four irrigation rates (60, 80, 100, and 120% es timated ET values), four N rates (0, 25, 50, and 100 kg ha-1), and five dates (22 June, 12 July, 25 July, 22 Aug., and 4 Sept. 2007) combined.................................................................................................................46 2-4 Evaluation of means of volumetric wa ter content (VWC), leaf nitrogen tissue concentration (%N), visual ra tings: color, quality, density (rated on NTEP 1-9 scale 9 is best and 6 is acceptable), and biomass (g) collected from fairway height hybrid bermudagrass ( Cynodon dactylon X C. transvaalensis, Burtt-Davy), four irrigation rates (60, 80, 100, and 120% estimated ET va lues), four N rates (0, 25, 50, and 100 kg ha-1), and five dates (22 June, 12 Ju ly, 25 July, 22 Aug., and 4 Sept. 2007) combined....................................................................................................................... .....47 2-5 Evaluation of means of Normali zed difference vegetation index (NDVI ), Leaf area index (LAI ), 850 nm, and 650 nm reflectance from Crop Circle device and crop water stress index (CWSI) computed from IR thermometer readings collected from fairway height hybrid bermudagrass ( Cynodon dactylon X C. transvaalensis, BurttDavy), four irrigation rates (60, 80, 100, and 120% estimated ET values), four N rates (0, 25, 50, and 100 kg ha-1) and five dates (22 June 12 July, 25 July, 22 Aug., and 4 Sept. 2007) combined...............................................................................................48

PAGE 7

7 2-6 Evaluation of means of individual band reflectance from the ASD device with a range of 350-2500 nm and a spectral resolu tion of 1 nm collected from fairway height hybrid bermudagrass ( Cynodon dactylon X C. transvaalensis, Burtt-Davy), four irrigation rates (60, 80, 100, and 120% es timated ET values), four N rates (0, 25, 50, and 100 kg ha-1) and five dates (22 June, 12 July, 25 July, 22 Aug., and 4 Sept. 2007) combined.................................................................................................................49 2-7 Evaluation of means of ASD reflectance readings of average range computations from 630 nm and 2080 nm, computed indices Normalized Difference Vegetation Index (NDVI y), Leaf Area Index (LAI) and Stress1 and Stress2 indices, and ratios 605/515 nm, 915/975 nm, 865/725 nm taken from an ASD device with a range of 350-2500 nm and a spectral resolu tion of 1 nm collected from fairway height hybrid bermudagrass ( Cynodon dactylon X C. transvaalensis, Burtt-Davy), four irrigation rates (60, 80, 100, and 120% es timated ET values), four N rates (0, 25, 50, and 100 kg ha-1) and five dates (22 June, 12 July, 25 July, 22 Aug., and 4 Sept. 2007) combined.................................................................................................................50 2-8 Evaluation of means of NDVI and LAI com puted from Crop Circle instrument, and LAI computed from ASD collected from fairway height hybrid bermudagrass ( Cynodon dactylon X C. transvaalensis Burtt-Davy), four ir rigation rates (60, 80, 100, and 120% estimated ET values), four N rates (0, 25, 50, and 100 kg ha-1), and five dates (22 June, 12 July, 25 July, 22 Aug., and 4 Sept. 2007) combined, which had N treatment X date interactions..................................................................................51 2-9 Evaluation of Means of leaf N concentr ation tissue (%N) coll ected from fairway height hybrid bermudagrass ( Cynodon dactylon X C. transvaalensis, Burtt-Davy), four irrigation rates (60, 80, 100, and 120% es timated ET values), four N rates (0, 25, 50, and 100 kg ha-1), and five dates (22 June, 12 July, 25 July, 22 Aug., and 4 Sept. 2007) combined, which had N treatment X date interactions............................................52 2-10 Evaluation of means of individual re flectance wavelengths at 813 nm and 935 nm collected from fairway he ight hybrid bermudagrass ( Cynodon dactylon X C. transvaalensis, Burtt-Davy), four irrigation ra tes (60, 80, 100, and 120% estimated ET values), four N rates (0, 25, 50, and 100 kg ha-1), and five dates (22 June, 12 July, 25 July, 22 Aug., and 4 Sept. 2007) combin ed, which had N treatment X date interactions................................................................................................................... ......53 2-11 Evaluation of means of ratios 915 nm/ 975 nm and 865 nm/925 nm computed from analytical spectral device (ASD) with a range of 350-2500 nm and a spectral resolution of 1 nm, collected from fairway height hybrid bermudagrass ( Cynodon dactylon X C. transvaalensis, Burtt-Davy), four irrigation rates (60, 80, 100, and 120% estimated ET values), four N rates (0, 25, 50, and 100 kg ha-1), and five dates (22 June, 12 July, 25 July, 22 Aug., and 4 Sept. 2007) combined, which had N treatment X date interactions.............................................................................................54

PAGE 8

8 2-12 Coefficient estimates (r2) of Crop Circle reflectance data Normalized Difference Vegetation Index (NDVI), Leaf Area Index (LAI), NIR, and Red, and crop water stress index (CWSI ) to soil volumetric water conten t (VWC) visual color, quality, density ratings collected from fa irway height hybrid bermudagrass ( Cynodon dactylon X C. transvaalensis, Burtt-Davy), four irrigation rates (60, 80, 100, and 120% estimated ET values), four N rates (0, 25, 50, and 100 kg ha-1), and averaged over all dates (22 June, 12 July, 25 July, 22 Aug., and 4 Sept. 2007)...............................55 2-13 Coefficients estimates (r2) of ASD reflectance readings of individual reflectance wavelengths taken from an Analytical sp ectral device (ASD) w ith a range of 3502500 nm and a spectral resolution of 1 nm to %N concentration leaf tissue, soil volumetric water content (VWC) visual colo r, quality, density ratings, and biomass production of fairway height hybrid bermudagrass ( Cynodon dactolon X C. transvaalensis, Burtt-Davy), four irrigation ra tes (60, 80, 100, and 120% estimated ET values), four N rates (0, 25, 50, and 100 kg ha-1), and averaged over all dates (22 June, 12 July, 25 July, 22 Aug., and 4 Sept. 2007)............................................................56 214 Coefficients estimates (r2) of average range computations from 630 nm and 2080 nm, and computed indices Normalized difference vegetation index (NDVI), Leaf area index (LAI) and Stress1 and Stress2 indices also taken from an ASD with a range of 350-2500 nm and a spectral resolution of 1 nm to %N concentration leaf tissue, soil volumetric wa ter content (VWC) visu al color, quality, density ratings, and biomass production co llected from fairway height hybrid bermudagrass ( Cynodon dactylon X C. transvaalensis, Burtt-Davy), four irrigation rates (60, 80, 100, and 120% estimated ET va lues), four N rates (0, 25, 50, and 100 kg ha-1), and averaged over all dates (22 June, 12 July, 25 July, 22 Aug., and 4 Sept. 2007).......................................................................................................................... ........57 2-15 Partial Least Squares regression on hypers pectral data from ASD device with a range of 350-2500 nm and a spectral resolution of 1 nm for prediction of soil volumetric water content (VWC), percent leaf nitrogen concentration (%N), biomass, and visual quality collected from fairway height hybrid bermudagrass ( Cynodon dactylon X C. transvaalensis, Burtt-Davy), four irrigation ra tes (60, 80, 100, and 120% estimated ET values), four N rates (0, 25, 50, and 100 kg ha-1), averaged over all dates (22 June, 12 July, 25 July, 22 Aug., and 4 Sept. 2007)............................................................58 3-1 Evaluation of means of biomass (g), volum etric water content (VWC), visual ratings: color, quality, density (rated on NTEP 1-9 s cale 9 is best and 6 is acceptable), from fairway height hybrid bermudagrass ( Cynodon dactylon X C. transvaalensis Princess 77) over five dates (22 June, 12 July, 25 July, 22 Aug., and 4 Sept. 2007) combined....................................................................................................................... .....74 3-2 Evaluation of means of Normalized Di fference Vegetation Index (NDVI), Leaf Area Index (LAI) computed from Crop Circle and ASD reflectance data, as well as Stress1 and Stress2 indices computed from ASD reflectan ce data from fairway height hybrid bermudagrass ( Cynodon dactylon X C. transvaalensis Princess 77) over five dates (22 June, 12 July, 25 July, 22 Aug., and 4 Sept. 2007) combined.....................................75

PAGE 9

9 3-3 Coefficient estimates (r2) of Crop Circle reflectance data, Normalized Difference Vegetation Index (NDVI), Leaf Area I ndex (LAI), 880nm, and 650nm, and, ASD reflectance data computed into NDVI LAI Stress1, and Stress2 indices to volumetric water content (VWC) visual colo r quality, density ratings, and biomass production from fairway hei ght hybrid bermudagrass ( Cynodon dactylon X C. transvaalensis Princess 77) averaged over five dates (22 June, 12 July, 25 July, 22 Aug., and 4 Sept. 2007).....................................................................................................76 3-4 Coefficient estimates (r2) of ASD reflectance data at in dividual wavelengths, as well as range averages from 630-690nm a nd 2080nm to volumetric water content (VWC) visual color, quality, density rati ngs, and biomass production from fairway height hybrid bermudagrass ( Cynodon dactylon X C. transvaalensis Princess 77) averaged over five dates (22 June, 12 July, 25 July, 22 Aug., and 4 Sept. 2007). All numbers in column headings ar e expressed nanometers (nm)...........................................77 3-5 Coefficient estimates (r2) of range averages from 630-690 nm and 2080 nm and ratios 605/515 nm, 915/975 nm, and 865/ 725 nm to volumetric water content (VWC) visual color, quality, density rati ngs, and biomass production from fairway height hybrid bermudagrass ( Cynodon dactylon X C. transvaalensis Princess 77) averaged over five dates (22 June, 12 July, 25 July, 22 Aug., and 4 Sept. 2007). All numbers in column headings ar e expressed nanometers (nm)...........................................78 3-6 Partial Least Squares regression coeffi cients on hyperspectral data from ASD device to predict volumetric water content (VWC), biomass, and visual quality (rated on NTEP 1-9 scale 9 is best and 6 is accep table) in fairway height bermudagrass ( Cynodon dactylon X C. transvaalensis Princess 77) averaged over five dates (22 June, 12 July, 25 July, 22 Aug., and 4 Sept. 2007)............................................................79 A-1 Description of treat ments for Figure A-1...........................................................................86 B-1 Daily Rainfall and ET data for Jay, FL..............................................................................87 C-1 Cumalitive weekly rainfall and et data for Jay, Fl.............................................................90 D-1 MothlyRainfall and ET averages from 2003-2006 for Jay, FL.........................................91

PAGE 10

10 LIST OF FIGURES Figure page 2-1 Partial Least Squares regression on hypers pectral data from ASD device with a range 0f 350-2500 nm and a spectral range of 1 nm for prediction of volumetric water content (VWC) collected from fair way height hybrid bermudagrass ( Cynodon dactylon X C. transvaalensis, Burtt-Davy), four irrigation rates (60, 80, 100, and 120% estimated ET values), four N rates (0, 25, 50, and 100 kg ha-1), averaged over all dates (22 June, 12 July, 25 July, 22 Aug., and 4 Sept. 2007).......................................59 2-2 Coefficient estimates (r2) of Crop Circle and ASD reflect ance data computed into a Normalized Difference Vegetation Index ( NDVI), to soil volumetric water content (VWC) collected from fairway height hybrid bermudagrass ( Cynodon dactylon X C. transvaalensis, Burtt-Davy), four irrigation ra tes (60, 80, 100, and 120% estimated ET values), four N rates (0, 25, 50, and 100 kg ha-1), and averaged over all dates (22 June, 12 July, 25 July, 22 Aug., and 4 Sept. 2007)............................................................60 2-3 Coefficient estimates (r2) of Crop Circle vs. ASD using reflectance data to compute a Normalized Difference Vegetation Index (NDVI) collected from fairway height hybrid bermudagrass ( Cynodon dactylon X C transvaalensis Burtt-Davy), four irrigation rates (60, 80, 100, and 120% estima ted ET values), four N rates (0, 25, 50, and 100 kg ha-1), and averaged over all dates (22 June, 12 July, 25 July, 22 Aug., and 4 Sept. 2007).............................................................................................................. .61 3-1 Biomass measurements collected fr om fairway height hybrid bermudagrass ( Cynodon dactylon X C. transvaalensis Princess 77) treated with trinexapac-ethyl at 0.05, 0.1, 0.2, and 0.4 kg a.i. ha-1. Plots measured 1.5 X 3.0 m with three replications. Means for biomass meas urements was computed using Fishers protected LSD (p< 0.05) from data over 5 dates (22 June, 12 July, 25 July, 22 Aug., and 4 Sept. 2007) combined...............................................................................................80 3-2 Biomass measurements collected fr om fairway height hybrid bermudagrass ( Cynodon dactylon X C. transvaalensis Princess 77) treated w ith flurprimidol at 0.3, 0.6, 1.1, and 2.3 kg a.i. ha-1. Plots measured 1.5 X 3.0 m with three replications. Means for biomass measurements was com puted using Fishers protected LSD (p< 0.05) from data over 5 dates (22 June, 12 July, 25 July, 22 Aug., and 4 Sept. 2007) combined....................................................................................................................... .....81 A-1 Experimental layout for detection of l eaf nitrogen concentration and soil volumetric water content using ground-base d remote sensing technology..........................................85

PAGE 11

11 Abstract of Thesis Presen ted to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Master of Science DETECTION OF TURFGRASS STRESS US ING GROUND BASED REMOTE SENSING By Jason Hamilton Frank May 2008 Chair: J. Bryan Unruh Major: Horticulture Sciences Many turf managers and researchers still use time and labor inte nsive techniques to manage and assess turf that originate from decad es ago. Many of these management practices have been proven repeatedly to work in a variety of situations to assess turfgrass stress, but may be time consuming and inconsistent. However, th ere are new developments in ways of assessing and mapping stress that increase the efficiency by which one manages an d assesses turf. With increasing water and other environmental restrict ions turf managers and researchers need be aware of technological advan ces in their industry. Two field experiments were conducted at the Univ ersity of Florida, West Florida Research and Education Center near Jay, FL. The first wa s the detection of leaf nitrogen concentration and soil volumetric water content using remote sensing technology. Plots were arranged in a randomized complete block design and treatments we re arranged in a split plot factorial design with four N levels split across four irrigation regi mes with three replications per treatment. The second experiment used remote sensing tec hnology to assess biomass production created by applications of plant growth re gulators. Plots were arranged in a factorial design observing three different PGRs, each at four incremental levels w ith three replications pe r treatment. Spectral measurements were taken from two remote sensing devices.

PAGE 12

12 A ground based vehicle mounted optical sensor Crop Circ le (Model ACS-210) (Holland Scientific) which produces its own light to compute a normali zed difference vegetation index (NDVI) which is calculated as (880nm-650nm) / (880nm+650nm). A hand-held hyperspectral spectroradiometer (FieldSpec Pro; Analytical Spectral Devices, Inc., Boulder, CO) with a spectral range of 300-2500 nm was also used to colle ct reflectance data. A ll reflectance readings were taken in full sunlight between the hours of 1100 and 1400 Central Standard Time (CST) to minimize variance caused by solar radiation. Plots were ground-truthed using time domain reflectometry for soil moisture content (VWC) a nd clipping samples obtained from plots were dried, weighed for biomass, and tested for Total Kj eldahl N content. Visu al ratings were also taken. The results from the nitrogen and irrigation ex periment indicate re flectance data best modeled soil VWC, visual ratings, and some treatme nt effects from N. Highest correlations for VWC were achieved from PLS regression on ASD data (r2=0.71) and the Crop Circle device (r2=0.70). The results from the PGR experiment modeled the biomass reduction expected from increasing rates of flurprimidol and trinexapac-ethyl compared to the untreated control, but not ethephon. Similar trends were observed in NDVI and leaf area index (LAI) indices computed from Crop Circle and ASD instruments, as well as several stress indices computed from the ASD device. If remote sensing technology could be used to adequately assess turf parameters, and was coupled with the global positioning system (GPS ) and geographic information systems (GIS) technologies, irrigations and chemi cal applications could be applie d on a site specific basis. In this setting, turf managers could potentially reduce inputs thereby reducing cost and environmental impact.

PAGE 13

13 CHAPTER 1 LITERATURE REVIEW Introduction The majority of turf managers and researchers still use outdated time and labor intensive techniques to manage and assess turf. Many of these management practices have been repeatedly proven to work in a variety of situat ions to assess turfgrass stress, but may be time consuming and inconsistent. For example, in the turfgrass research field visual evaluations can be labor intensive and variable among evaluators (Trenholm et al., 1999). However, there are new developments in ways of assessing and mapp ing stress that will increase the efficiency by which one manages and assesses turf. With increasing water restrictions and other environmental constraints turf managers and researchers need be aw are of technological advances in their industry. R ecent environmental concerns, such as nutrient and pesticide leaching and runoff, have pressed turf managers to reduce inputs used for turf maintenance (Bell et al., 2002). There are a variety of new or improved met hods available to dete ct water and nutrient stresses before they are visible. The basi s of this technology relie s heavily on the global positioning system (GPS), geographic informa tion systems (GIS), and remote sensing technology. Much of this technol ogy is in the beginning stages, however, the theory for use of this technology has been proven in other agronomic industries such as food production systems. This technology has been used as a management tool to maximize crop yield and minimize cost through site-specific application of inputs by applying nutrients, wa ter and pesticides only where needed (Lee et al., 1999).

PAGE 14

14 Irrigation Importance Water is the primary requirement for devel opment and survival of turfgrass and many other agronomic crops (Turgeon, 200 8). Plant water potential is essential for plant growth as well as most physiological processes the plant must perform (Idso et al., 1981). Therefore, water availability is one of the most critical factors in plant health (Ripple, 1986). Plants are made up of cells, which are filled with water to mainta in their turgor. Howeve r, if turgor is not maintained the plant begins to wilt (Unruh and Elliott, 1999). Many plants contain 75-85 % water and begin to die if this percentage decreases below 65 % (Unruh and Elliott, 1999). Water is taken up by the roots in the soil and is m oved throughout the plant and eventually released through the stomata through a pr ocess known as transpiration (Salisbury and Ross, 1992). Stomata are responsible for gas ex change in the plant and they close when the plant is under a water deficit or stress (Salisbury and Ross, 1992) This ultimately re duces photosynthesis and stunts leaf growth (Unruh and Elliott, 1999). G eeske et al. (1997) found that this stunting in growth is the plants defense mechanism to reduce nutrient and water requirements. The reduction in leaf area serves as a beneficial fact or that ultimately reduces transpiration and the need for increased physiological pr ocesses (Geeske et al., 1997). In addition, transpiration serves many differe nt purposes such as nutrient movement through the plant and ev aporative cooling, also known as evapotranspiration (ET) (Turgeon, 2008). Evapotranspiration is the total water lost by all transpiratory movement of water from soil, through the plant and ultimately into the atmosphere, where it evaporates (Turgeon, 2008). Many factors affect this process such as light intensity, humidity, wind ve locity and temperature (Unruh and Elliott, 1999). Water stress has a direct affect on the rate of photosynthesis. It first causes stomatal closure, which causes a reduced supply of CO2. Secondly, water stress reduces water potential,

PAGE 15

15 which causes a decrease in the structural integrity of many important elements of the photosynthetic process (Hopkins, 1999). This in tu rn will decrease electron transport and photophosphorolation leading to damage in the thylakoid membrane and adenine triphosphate (ATP) synthetase protein (Hopkins, 1999). The hormone abscisic acid (ABA) also plays an important role in the plants ability to endure wa ter stress. Under stress, ABA accumulates in the leaves causing an efflux of potassium from the guard cells, resulting in stomatal closure and reducing transpiration (Hopkins, 1999). Plant water stress can occur if a plant receives too much wa ter or if it does not receive enough. Flooding leads to a lack of oxygen, this l eads to decreased respiration, nutrient uptake, and root functionality, thus reducing photosynt hesis (Hopkins, 1999). Drought stress from lack of water leads to increased solute concentrati on in the protoplasm, which in turn affects many physiological processes (Hopkins, 1999). However, mild drought stress can cause the roots of a plant to grow deeper. This is why irrigation events should be spaced out as long as possible (Unruh and Elliott, 1999). Water is an important resource; however, it is not a renewa ble resource (Barret, 2003). Ninety-seven percent of all the earths water is salt water, three percent is fresh water; however, two-thirds of this is frozen in polar ice caps (B arret, 2003). Therefore, on e percent of the Earths fresh water must meet the water needs of manuf acturing, mining, agriculture, and turf not met by rainfall (Barret, 2003). In Florida the average rainfall does not meet the watering requirements to maintain golf courses at the expectations of players and industry viewers (Unruh and Elliott, 1999). As a result, the irrigation system is the si ngle most important tool available for turfgrass management to a golf course superintende nt or turf manager (Barret, 2003).

PAGE 16

16 This is why research focusing on improving wate r use management in agriculture is very important. It has both environmental and ec onomic consequences, especially as water availability becomes more limited. According to a survey conducted in 2000 by the Golf Course Superintendents Association of Am erica, the median annual water usage of a golf course in the United States is 32.49 cm applied to an average of 31.44 ha resulting in 108 million L used annually (Barret, 2003). Historically the majority of turfgrass mana gers scheduled irrigation usage based on their experience and set schedule of time intervals, both of which can result in over watering (Augustin and Snyder, 1984). Irrigation schedules are often based on calendar dates such as 3 or 7 times a week. Studies have shown that th is style of irrigation may promote over watering and provide too much moisture to the turf (U nruh and Elliott, 1999). This wastes water and energy, produces poor playing and agronomic c onditions, and leads to degradation of the environment from excessive runoff (Barret, 2003). Tools are needed to monitor turf water status to aid in decisions that result in more efficient irrigation (Bahrun et al., 2003). Irrigation systems should be operated so that the addition of water never exceeds that lost by ET in any given area (Brown et al., 1977). This will increase irrigati on efficiency and possibly reduce plant water stress in those areas. Researchers and turf managers over the past two decades have investigated many ways to determine when to irrigate. Many of the techni ques developed to assess plant water status are time consuming, spatially restrictiv e, and costly (Ripple, 1986). Currently, visual symptoms, ET rates, and tensiometers are all methods fo r determining irrigation needs (Turgeon, 2008). However, when visual symptoms of drought are apparent, it is alrea dy too late because the turf is already stressed. Likewise, ET values are gene rally not site specific and actual ET can differ

PAGE 17

17 greatly from one location to anothe r. It is assumed that all the turf within a given area is precisely the same and responds exactly the sa me to environmental conditions. Conversely, tensiometers have proven to be an adequate asse ssment of plant water status, but require periodic maintenance, and may have to be removed duri ng periods of cold weather (Unruh and Elliott, 1999). In addition, the tensiometer reading is only appropriate in the area adjacent to the placement of the ceramic tips and do not indicate water status for a large area (Unruh and Elliott, 1999). Therefore, because conservation of this limited resource is critical, it is increasingly necessary to develop technology to help turf managers allocate ir rigation water more efficiently. Nutrient Importance Aside from the importance of water to plants available nutrients al so drive growth and development. Nutrient availability affects ma ny physiological processes the plant must perform in order to develop properly. Nitrogen (N) is the most limiting nutri ent in production of nonlegumous crops (Osborne et al., 200 2). It is essential to the bu ilding of several important plant components including proteins, nucleic acids, hor mones, and chlorophyll, which harvests light energy used in photosynthesis (Hopkins, 1999). P hosphorus (P) plays an important role in the transfer of energy within the pl ant in the form of compounds like adenosine triphosphate (ATP), adenosine diphosphate (ADP), and phosphate (Pi). P also is an importa nt factor in root growth (Hopkins, 1999). Potassium (K) is also a very impor tant nutrient in plant growth and is required in large amounts. Potassium serves to activ ate a number of enzymes and regulates stomatal conductance; a very important pr ocess controlling tr anspiration (Hopkins, 1999). All of these nutrients, if in deficit of pl ant requirement, can cause a numbe r of physiological disruptions which are detrimental to plant growth and develo pment. Once these stresses become visible it may be too late to correct the damage. Therefor e, methods are needed to predict plant nutrient status that can detect physio logical stress before it becomes visible to the naked eye.

PAGE 18

18 Plant Growth Regulator Importance Plant Growth Regulators (PGRs) have beco me a very important tool in turfgrass management. Many turfgrass managers use PGRs to suppress vertical turf growth which ultimately reduces maintenance costs and improves turf quality. Some data suggests that PGRtreated turf can produce greater root mass, recover from injury faster, reduce water use rate, reduce disease incidence, and reduce Poa annua weed populations (Branham, 1997). A PGR is a substance that adjusts the growth an d development of a plant. This is generally achieved through the inhibition of gibberellic acid (GA), a hormone which is responsible for cell elongation. Plant growth regul ators were first introduced for use on fine turf in 1987 (VanBibber, 2006). The first of these chemicals were flurprimidol (Trade name Cutless) and paclobutrazol (Trade name Trimmit), which inhib it GA synthesis at relatively the same point in the GA biosynthesis pathway, resulting in very similar plant responses. However, paclobutrazol is the more active compound and uses lower rates to achieve the same respons e as higher rates of flurprimidol (Branham, 1997). The newest of the PGRs, trinexapac-ethyl (Trade name Primo), was released commercially in 1995 and has become the standard for use throughout the industry (Branham, 1997). It facilitates GA inhibition la ter in the GA biosynthesi s pathway so that GAs are formed but are not active and serves as prec ursors for plant biochemical processes. Plant growth regulators are relatively expensive compared to other chemicals used by turf managers. More efficient applications of PGRs could potentially reduce the application cost of these chemicals. Remote Sensing Remote sensing can be defined as obtaini ng information about an object, area, or phenomenon by analyzing data acquired by a device that is not in contact with that object, area, or phenomenon (Lillesand and Keifer, 2000). For ma ny years researchers have entertained the

PAGE 19

19 idea of measuring plant stress using remote sensing technology. Plant light interception influences growth and physiological respons es (Salisbury and Ross, 1992). When light is intercepted by the plant, it is absorbed, transmitted, or reflec ted (Turgeon, 2008). Therefore, by quantifying the reflected light, remote sensing technology can help to detect the onset of turfgrass stress (Ikemura and Leinauer, 2006). Presently, real world applications of remote sensing technology in turfgrass management are still in their infancy (Ikemu ra and Leinauer, 2006). Studies have shown that image analysis and various remote sensing devices have strong pote ntial to detect a variet y of turfgrass stresses (Ikemura and Leinauer, 2006). Spectral radiom etry (Trenholm et al., 2000), and infrared thermometry (Slack et al., 1981) have been show n to adequately assess light reflectance at various wavelengths. By differen tiating certain wavelength characteri stics, insights into growth and adaptive characteristics and how a plant responds to stress can be seen (Trenholm et al., 2000). There have been several aspects of remote sensing technology explored to reveal the level of stress that may be occurring in a plant. Researchers have created a variety of theories utilizing remote sensing techniques and exploring many part s of the electromagnetic spectrum to describe various plant stresses. Vegetative Indices Researchers have found certain relationships exist between vari ous aspects of the electromagnetic spectrum and a va riety of plant parameters and stress (Bell et al., 2000, 2002a, 2002b, 2004; Trenholm et al., 1999a, 1999b, 2000; Krus e et al., 2005; Fenstemaker-Shaulis et al., 1997; Bahrum et al., 2003; Xiong et al., 2007). These relationships are formulated into indices to normalize data for varying illumina tion conditions (Lillesand and Kiefer, 2003). One of the most common indices is the Normalized Difference Vegetation Index (NDVI) computed

PAGE 20

20 as: reflectance in the Near Infrared (NIR) regi on minus reflectance in the Red (R) region divided by a sum of both (NIR-R / NI R+R) (Rouse et al., 1973) Human Visual Ratings Traditionally, visual observation has been the standard for assessing turf stress for turf managers. Researchers quantified this observati on by creating a scale. The National Turfgrass Evaluation Program (NTEP) scale is a 1-9 assessment of turf color, quality and density, where 6 is the least acceptable and has been the standard for turf research for many years (Morris, 2007). However, recent investigations have looked beyond this scale to compare remote sensing systems with traditional human ev aluation. Trenholm et al. (1999) tested four of the current indices and compared them to visual evaluations. These indices were: NDVI computed as (935-661 nm)/ (935+661 nm) Leaf area index (LAI) computed as 935/661 nm Stress 1 computed as 706/760 nm Stress 2 computed as 706/813 nm They concluded that NDVI, LAI, and Stress 2 indices, and individual wavelength reflectance measurements at 661 nm and 813 nm were compar able to visual evaluations (Trenholm et al., 1999). Bell et al. (2002b) found strong correlations between NDVI and turf color, and subsequent work by Bell et al. (2004) concluded that NDVI was a better estimator of chlorophyll content than visual color evaluation. Irrigation Early research involving remote sensing to pr edict plant water status related plant canopy temperature to plant water potentia l (Ehrler et al., 1978). This wa s derived from the fact that transpiration plays a major role in the plants ab ility to cool itself through ET. As soil water decreases, evaporative cooling al so decreases causing the canopy te mperature to rise. This is what many researchers have used to relate ca nopy temperature to relative soil water content

PAGE 21

21 (Aston and Van-Bavel, 1972). Stom atal closure of sunlit leaves results in increased leaf temperature if factors such as wind speed and va por pressure remain c onstant (Ehrler et al., 1978). This temperature difference between canopy and air can be determined in several ways. Originally, microthermocouples we re used, but this only worked on a small scale. However, advances in infrared thermometr y have increased the ability to detect crop water stress through the utilization of remote sensing de vices (Slack et al., 1981). Infrar ed thermometers registered in a narrow bandpass (8-14 m or 10.5 12.5 m) are cap able of measuring crop temperatures with 0.5 C accuracy (Ehrler et al., 1978). They can be used for measuring canopy temperature at ground level, from aircraft, or satellite. Th ese methods are done in a much timelier manner compared to old methods such as tensiometers and microthermocouples (Ehrler et al., 1978). From this relation of canopy temperature to plant water potential, many indices have been developed to best predict plant water status. Idso et al. ( 1981) proposed a temperature-based index as a water stress indicator. This was in tended to normalize canopy temperature minus air temperature, Tc-Ta, for environmental variability. It was introduced as the Crop Water Stress Index (CWSI). CWSI produces a species-spec ific linear relationshi p between air and canopy temperature, which is irrespective of other envi ronmental factors except cloud cover (Slack et al., 1981). CWSI can be described as the fractiona l decrease of potential ET which can be presented in terms of the canopy temp erature minus the air temperature (Tc-Ta) or CWSI= (T c -T a ) a (T c -T a ) l (Tc-Ta)u (Tc-Ta)l Where Tc= canopy temperature, Ta= actual temperature, a= actual, u= upper limit, and l= lower limit (Jalali-Farahani et al., 1994) The CWSI was found to have pot ential in turf irrigation scheduling (T hrossel et al., 1987). Further investigations of this model spaw ned numerous theories accounting for several

PAGE 22

22 environmental factors as well. Horst et al. (1989) suggested that ca nopy temperature minus air temperature plotted as a function of Vapor Pressure Deficit (VPD), which is the difference in the amount of moisture in the air a nd the amount of moisture the air can hold. This may be species and mowing height dependent and may require further adjustment of the function for each species under individual environmen tal regimes (Horst et al., 1989). Jalali-Frahani et al. (1993) proposed that CWSI of turfgrass was not only a function of VPD but also net radiation from sunlight. Martin et al. (1994) found that usi ng variables accounting for VPD, wind speed, and net radiation increased coefficient of determination (r2) values dramatically across several varieties of turfgrass. Eh rler et al. (1978) concluded that temperature difference (Tc-Ta) in wheat ( Triticum spp ., durham Produra) is a reliable method for monitoring plant stress from aircraft or satellite. Frequent measurements were necessary to reduce the va riability in the data. Throssel et al. (1987) found that temperature difference (Tc-Ta) appeared to be indicative of water stress for well-watered Kentucky bluegrass ( Poa pratensis L.). More recent investigations into plant and soil water status rely on spectral reflectance data measured from plant or turfgrass canopies. Fe nstemaker-Shaulis et al (1997) found a negative correlation with NDVI and canopy temperature (r2=0.74) and a positive correlation to plant moisture content (r2=0.90). In addition, they concluded, that mapping NDVI values can also provide valuable insight into the status of irrigation system uniformity and management practices. Hutto et al. (2006) found that dr ought stressed species had a higher amount of reflectance in the short wave infrared regi on (SWIR) (1,300-2,500 nm) than that of the nonstressed control. Ripple (1986) found that reflectance in the 6 30 nm and 2,080-2,350 nm ranges is strongly correlated with leaf water content. As refl ectance decreased in these ranges there was an inverse linea r relationship with leaf relative wate r content. However, Ripple (1986)

PAGE 23

23 concluded that relationships betw een spectral reflectance and leaf wa ter content can be direct or indirect and are wavelength depe ndent. It is believed that elevated correlations in the 630 nm range with spectral reflectance data was due to a reduction in chlorophyll as the leaves dried and not a true relation to actual plan t water status (Ripple, 1986). Nutrient To a large extent, much remote sensing work has been done on the assessment of N and chlorophyll levels. Overall, re flectance is low in the visible range of the electromagnetic spectrum due to the high absorption by chlo rophyll for use in photosynthetic processes (Salisbury and Ross, 1992 ). For th is reason, if there is a particular reflectance in this region this could correlate to a certain physio logical disruption in the plant. It is generally assumed that there is a close relationship between plant colo r and chlorophyll conten t (Bell et al., 2004). Research has shown reflectance at 550 nm to be the most sensitive assessment of N in many agronomic crops, which is what the human eyes see as green (Blackmer et al., 1994, 1996, Yoder and Pettigree-Crosby, 1995). Bell et al. (2002a) found vehicle mounted optical sensors provided a good assessment of variable N applications Xiong et al. (2007) found significant seasonal responses to N treatments. For other nutrient deficiencies such as P and K, limited research has been conducted to establish their function in the sp ectral constituency. Research that has been conducted suggests that reflectance wavelengths for detecting P concentration in corn ( Zea mays L.) are in the NIR (730 nm and 930 nm) region of the spectrum as well as reflectance from the blue region (440 nm and 445 nm) (Osborne et al., 2002). This can possibly be attributed to the production of anthocyanin which is produced in P deficient plants that absorb more green light and reflect more blue light (Osborne et al., 2002). Res earch conducted by Kruse et al. (2005) found that spectral reflectance data can be used as a good estimation of P concentration in creeping

PAGE 24

24 bentgrass ( Agrostris stolinifera L.). Reflectance wavelengths fo r predicting P concentration of plant tissue were in the blue (480 nm), yellow (565 nm), orange (595 nm), and red (650 nm) regions with an overall r2 value of 0.73. This reflectance response relates to the spectral characteristic of anthocyanin which increased in concentration under P stress conditions (Kruse et al., 2005). Hutto et al. (2006) found that potassium deficiency could be detected at bandwidths from 750 nm to 785 nm. Biomass Green plant material has high reflectance in the NIR (700 nm) region, while dead or dying plant material is just the opposite (Gees ke et al., 1997). Yoder and Pettigree-Crosby (1995) related this correlation in the NIR region to biomass, or energy status of the plant. Similar results were found by Kruse et al. (2005) on creeping bentgrass, wh ere reflectance in the green (510 nm and 535 nm), red (635 nm), and NIR (735 nm) were significant in predicting biomass. Starks et al. (2006) found crude protein concentrat ion, biomass production, and crude protein availability of bermudagrass ( Cynodon dactylon L. Pers.) grown in pastures closely correlated with canopy reflectance ratios of 605/515 nm, 915/975 nm, and 865/725 nm. Conversely, Fenstemaker-Shaulis et al. (1997) found that NDVI is not a strong predictor of biomass (r2=0.37). Kruse et al. (2005) also observe d no correlation between biomass and NDVI. Kruse et al. (2006) found implementing Partial Leas t Squares regression on hyperspectral data to be an adequate predictor of biomass in cree ping bentgrass compared to indices NDVI, LAI, Stress1, and Stress2. PGRs Not a large body of literature exists in the ar ea of remote sensing and the effects of PGR applications. Some image-based appli cations have been tested in cotton ( Gosspypium spp. ) to reduce PGR use for economic reasons (Bethel et al., 2003). Among compar isons of different

PAGE 25

25 image based-application methods (Variable Rate and Site-Specific) to traditional blanket applications for PGRs in cotton, it was found that the broadcast applications were on average 19 % more costly than the site-spe cific method, and 27 % more costly than the variable rate method (Bethel et al., 2003). These results demonstrate th e potential economic benefit of this technology for future use in other aspects of agriculture to reduce excess chemical runoff and leaching into the surrounding environment (Bethel et al., 2003). Statistical Methods Older work involving remote sensing typically used multispectral data, which is collected in individual or wide bandwidt hs and give a coarse representation of the electromagnetic spectrum (Lillesand and Kiefer, 2003). With the introduction of hyperspectral data, which typically acquires 200 or more bands in very narr ow resolution throughout the visible, NIR, and SWIR regions, researchers have more data to analyze (Lillesand and Keifer, 1987). These enhanced techniques for obtaining th e data have created the need to explore different statistical methods to quantify and interpret the data. Tradit ionally linear regression is the simplest way to understand variance. However, in interpreting re flectance data it has been found that other regression techniques better assess plan t status (Kruse et al., 2005). Partial Least Squares (PLS) regression is a met hod used to predict variables when there are a large number of factors; typi cally more than the number of observations (Tobias, 1997). The use of multiple variables is generally tested through multiple linear regression (MLR) (Tobias, 1997). However, the problem with MLR is when the number of factors becomes too large the model may become over-fitting (Tobias, 1997). This is observed when a large number of variables are present but only a few account for mo st of the variation (T obias, 1997). Almost perfect models may be achieved; however they will not be highly predictive of new data.

PAGE 26

26 When prediction is the goal, then PLS regressi on is a highly useful tool (Tobias, 1997). Kruse et al. (2006) found PLS regression yielded a stronger relationship to predict N concentration compared to NDVI in plant leaf tissue indicating its pot ential use in future development models. Lee et al. (1999) also f ound that PLS and Princi pal Component Analysis (PCA) prediction models performed similar to each other and both were better than MLR regression to predict the N status in corn. Summary The use of remote sensing technology to pred ict turfgrass stress could decrease the time and labor required to evaluate tu rfgrass health (Osborne et al ., 2002). This would not only reduce the environmental impact of turfgrass manageme nt practices, but could reduce maintenance costs of the various inputs required by turfgrass managers. Curre nt dependence on agricultural chemicals have heightened many environmental c oncerns in Florida due to the sandy soils and heavy rainfall which increases potential for runoff and leaching of chemical s if applied in excess (Min and Lee, 2003). As stated earlier, real wo rld applications of remote sensing technology for utilization in turfgrass are sti ll in their infancy. At the present time, an experienced turf managers eye is still better than current co mputerized systems (Ikemura and Leinauer, 2006). However, studies have shown that various remote sensing devices have strong potential to detect a variety of turfgrass stresses (Ikemura and Leinauer, 2006). Applying remote sensing technology to the fi eld of turfgrass management could possibly help reduce irrigation and provide more efficient watering and fertil ization routines (Ikemura and Leinauer, 2006; Bell et al., 2004). Combined with GPS, GIS, and variable rate application technology, there is a potential to apply nutrients as well as irriga tion based on plant need rather than broadcast applications of entire areas (Bell et al., 2004). Currently, most of the remote sensing research is conducte d on agronomic food crops and forested areas. Only limited

PAGE 27

27 research is available in the area of turfgr ass management (Trenholm et al., 1999). Further research is needed to incorporate remote sens ing into turfgrass management (Trenholm et al., 1999). If conditions across the entire golf course we re uniform in microclimates, soil types, turf varieties, pest densities, nutrien t status, and irrigation performance there would be little need for advanced sensor systems to measure the variabilit y of these parameters to better meet the needs on a locational basis (Stowell and Gelerntner, 2006). However, that is not the case. Optimistically, Bell et al. (2002a) concludes If an optical sensing (In reference to remote sensing) system and software can be economi cally produced, reasonabl y priced, and mounted effectively on normal maintenance equipment, a turf practitioner could save enough money in fertilizers and pesticides to pay for the equipmen t. This approach would be useful for reducing the amount of fertilizers and pesticides needed to adequately manage large turf areas. The use of optical sensing to determine fertilizer rate be fore or during applica tion could increase turf uniformity and possibly, turf health. Sensor maps of large turf areas could be used to signal turf decline and provide an early warn ing system for turf mangers.

PAGE 28

28 CHAPTER 2 DETECTION OF LEAF NITROGEN CO NCENTRATION AND SOIL VOLUMETRIC WATER CONTENT USING GROUND-BA SED REMOTE SENSING TECHNOLOGY Introduction Remote sensing is defined as obtaining info rmation about an object, area, or phenomenon by analyzing data acquired by a device that is not in contact with that object, area, or phenomenon (Lillesand and Keifer, 1987). For many years researchers have entertained the idea of measuring plant stress using remote sensi ng technology. Plant light interception significantly influences growth and physiological responses (Salisbury and Ross, 1992). When light is intercepted by the plant it is ab sorbed, transmitted, or reflecte d (Salisbury and Ross, 1992). Using remote sensing technology to quantify the light that is reflected coul d help to detect the onset of turfgrass stress (Ikemura and Leinau er, 2006; Kruse et al., 2005; Hutto et al., 2006; Trenholm et al., 1999a). Real world applicatio ns of remote sensing technology for use in turfgrass management are still in their infancy; however, studies have shown that image analysis and various remote sensing devices have strong pote ntial to detect a variet y of turfgrass stresses (Ikemura and Leinauer, 2006). Recent investigations into remo te sensing instrumentation have produced promising results for turf managers. However, the majority of turf managers and researchers still use outdated time and labor intensive techniques to manage and assess turf. Noneth eless, there are new developments in ways of assessing and mapping plant stress that may increase the efficiency by which we manage and assess turf. The use of remote sensing t echnology to estimate turfgrass stress could significantly decrease the time and labor required to assess these levels in traditional ways, thus reducing cost as well (Osborne et al., 2002). The many great achievements in U.S. agricultural productivity in the past decades can be attributed to the use of agricultural chemicals, including fertilizer and pesticides (Lee et al.,

PAGE 29

29 1999). However, the increased dependence on pest icides and fertilizers have heightened many environmental concerns especially in Florida due to the sandy soils and heavy rainfall, which increases potential for heavy runoff and l eaching of chemicals (Min and Lee, 2003). With increasing water and other environmental restrict ions, turf managers and researchers must be aware of technological advan ces in their industry. Recent concerns have pressed turf managers to reduce nutrient and pesticide inputs used for turf maintenance (Bell et al ., 2002a). In other industries, such as agronomic food production crops, precision agriculture has be en used as a management tool to maximize yield and minimize cost (Bethel et al., 2003). The basis of this technology relies heavily on the Global Positioning System (GPS), Geographic Information Systems (GIS), and Variable-R ate Application (VRA) and remote sensing technology. Using this techno logy for site-specific management of nutrients, water, and pesticides could optimize yiel d while minimizing cost (Lee et al., 1999). Water is an important resource in turfgrass management; however, it is not a renewable resource (Barret, 2003). The Ea rths water reserves are 97 % salt water and 3 % fresh water. Of the fresh water reserve, two-thirds is frozen in polar ice caps (Barret, 2003). Therefore, only 1 % of the Earths water is available to meet the water needs of manufacturing, mining, agricultural, and turf not met by rainfall (Barre t, 2003). In addition, the averag e rainfall in Florida does not meet the watering requirements to maintain golf c ourse quality at the le vel expected by players and industry viewers (Unruh and El liot, 1999). As a result, the ir rigation system is the single most important tool available to a golf course su perintendent or turf ma nager (Barret, 2003). As water restrictions increase and pressure rises fr om environmental groups to restrict water use on golf courses, it is increasingly im perative that turfgrass managers improve water use efficiency (Martin et al., 1994).

PAGE 30

30 Aside from the importance of water to plants, available nutrients also drive plant growth and development. Nutrient availability affect s many of the physiological processes a plant must perform in order to develop properly. Nitrogen (N) is the most limiting nutrient in production of non-legumous crops (Osborne et al ., 2002). It is essential to th e building of several important plant structures including, prot eins, nucleic acids, hormones, a nd most importantly chlorophyll, which harvests light energy used in photosynthesis (Hopkins, 1999). Remote sensing technology has been shown to assess stress that may be occurring in a plant. Researchers have created a variety of theo ries and indices explorin g many parts of the light spectrum to describe various stresses. These stresses include water, nutrient, soil compaction, pests, and disease. One of the early theories involving remote se nsing to predict plant water status was the Crop Water Stress Index (CWSI), which corr elated plant water potential with canopy temperature (Ehrler et al., 1978). More recent investigations rely on spectral reflectance data measured from plant or turfgrass canopies. Several spectral reflectance bands have been identified for use in modeling various turfgrass stresses, especially thos e in the visible (400-700 nm) and near infrared (NIR) regions (700-1,300 nm ) of the spectrum, as well as short wave infrared (SWIR) regions (1,300-2,5 00 nm) (Hutto et al, 2006). Reflectance from healthy plant canopies is characteristically lo w in the visible range of the spectrum due to high absorption by chlorophyll for use in phot osynthetic processes (Trenholm et al., 2000), If there is a particularly high reflectance in this region, this could correlate to a physio logical disruption in the plant (Geeske et al., 1997). Green plant material has high reflectance in the NIR range while dead or dying plant material has low reflectance (Geeske et al., 1997). Yoder and Pettigree-Crosby (1995) relate this

PAGE 31

31 correlation in the NIR region to biomass, or the en ergy status of the plant. Similar results were found by Kruse et al. (2005) on creeping bentgrass ( Agrostis stolinifera L. ), where reflectance in the green (510 nm and 535 nm), red (635 nm), a nd NIR (735 nm) were significant in predicting biomass. Trenholm et al. (1999a) found that wear stress in several varieties of hybrid bermudagrass ( Cynodon dactylon x C. transvalensis Burtt-Davy) and seashore paspalum ( Paspalum vaginatum Swartz) is associated w ith the specific wavelengt h at 813 nm. Starke et al., (2006) found that crude protein concentr ation, biomass production, and crude protein availability of bermudagrass ( C. dactylon L. Pers.) pastures is closely related with canopy reflectance ratios of 605/515 nm, 915/975 nm, and 865/725 nm. Researchers have been exploring the idea that plant water status can be determined by measuring reflectance in the short wave infr ared (SWIR) region (1,300-2,500 nm). Ripple (1986) found that SWIR reflectance is strongly correlated with leaf water content and that significant correlations exist between certain spectral bands (630 nm and 2,080-2,350 nm ranges), with SWIR reflectance having the highe st correlations. As reflectan ce decreased in the 630-690 nm range there was an inverse linear relationship with leaf relative water content. However, Ripple (1986) concluded that relationships between spectral reflectance and leaf water content can be direct or indirect and ar e wavelength dependent. It is believed that elevated correlations with spectral reflectance data was due to a reduction in chlorophyll as the leaves dried and not a true relation to actual plant water st atus (Ripple, 1986). Hutto et al. (2006) also documented that drought stressed creeping bentgras s had a higher amount of reflectan ce in the SWIR regions than that of the non-stressed control. Several different spectral respons es to light to assess indicatio ns of plant stress, nutrient and water status have been tested in the form of indices. One of the most common indices is the

PAGE 32

32 Normalized Difference Vegetation Index (NDVI) which is computed as the reflectance in the NIR region minus the reflectance in the red region divided by a su m of both (NIR -R)/ (NIR+R). Bahrum et al. (2003) found that NDVI values from remotely sensed data can be used to assess turfgrass stress. Fenstemaker-Shaulis et al (1997) found a negative correlation between NDVI and canopy temperature (r2=0.74), and a positive correlation be tween NDVI and plant moisture content (r2=0.90). However, NDVI was not as strong a predictor of biomass (r2=0.37). In addition, spatial mapping of NDVI values can pr ovide valuable insight into the status of irrigation system uniformity and management pract ices (Fenstemaker-Shau lis et al., 1997). Bell et al. (2002b) found strong correlations between NDVI and turfgr ass color ratings. Additionally, the use of NDVI ratings obtaine d from vehicle mounted optical sensors were effective for measuring herbicide damage (Bell et al., 2000), variable N applica tions (Bell et al., 2002a), and significant seasonal responses to N treatments (Xiong et al., 2007). It has also been reported that NDVI is a better estimator of chlo rophyll content than visual colo r evaluation (Bell et al., 2004). Trenholm et al. (1999) tested four indices to detect stress in tu rfgrass. These indices were: NDVI computed as (935-661 nm)/ (935+661 nm) Leaf Area Index (LAI) computed as 935/661 nm Stress 1 computed as 706/760 nm Stress 2 computed as 706/813 nm They concluded NDVI, IR/R, and Stress 2 i ndices, as well as i ndividual wavelength measurements at 661 nm and 813 nm were well correlated to visual ratings. Older work involving remote sensing typically used multispectral data, which is collected in individual or wide bandwidt hs and gave a coarse repres entation of the electromagnetic spectrum (Lillesand and Kiefer, 2003). With the introduction of hyperspectral data, which typically acquires 200 or more wavelength band s in very narrow resolution throughout the visible, NIR, and MIR regions, researchers have more data to analy ze (Lillesand and Keifer,

PAGE 33

33 1987). These enhanced techniques for obtaining th e data have created the need to explore different statistical methods to quantify and interpret the large amount of data produced. Traditionally, linear regression is the simplest way to understand variance. However in interpreting reflectance data it has been found that other regre ssion techniques better assess plant status (Kruse et al., 2005). Partial Least Squares (PLS) regression is a met hod used to predict variables when there are a large number of factors; typi cally more than the number of observations (Tobias, 1997). The use of multiple variables is generally tested through Multiple Linear Regression (MLR) (Tobias, 1997). However, the limitation with MLR is when the number of factors becomes too large the model may become over-fitting (Tobias, 1997). This is observed when a large number of variables are present but only a few account for mo st of the variation (T obias, 1997). Almost perfect models may be achieved; however they will not be highly predictive of new data. When prediction is the objective then PLS re gression is a highly us eful tool (Tobias, 1997). Kruse et al. (2006) found PLS regression yielded a str onger relationship to predict N concentration compared to NDVI in creeping bent grass indicating its pote ntial use in future development models. Lee et al. (1999) also fo und that PLS and Princi pal Component Analysis (PCA) prediction models performed similar to each other and both were better than MLR regression to predict the N status in corn ( Zea mays ). Currently, the majority of remote sensing research is conducted on agronomic food crops and forested areas. Only limited research is ava ilable in the area of remote sensing and turfgrass management (Trenholm et al., 1999b). If conditions across the entire golf course were uniform in microclimates, soil types, turf varieties, pest densities, nutrient status, and irrigation performance there would be little need for advanced sensor systems to measure the variability of

PAGE 34

34 these parameters to better meet the needs on a locational basis (Stowell and Gelernter, 2006). However highly variable agronomic conditions create the need for further research to incorporate the use of remote sensing into turfgrass management (Trenhol m et al., 1999b). The objective of this research was to assess the ability of three remote sensing instruments to detect soil moisture and leaf nitrogen levels of Tifsport bermudagrass turf. Materials and Methods The field experiment was conducted at the Univ ersity of Florida, West Florida Research and Education Center near Jay, FL. Plots were arranged in a randomized complete block design and treatments were arranged in a split plot factorial design with four N levels (sub plot) split across four irrigation regimes (w hole plot) with three replicati ons per treatment (Appendix A). Whole plots were 4.88m x 4.88m in size and subp lots within the whol e plots were 1.22m x 4.88m in size. Whole plots were spaced 4.88m apar t on each side to minimi ze the impact of drift from irrigation from plot to plot. Ammonium sulfate was applied on 14 June 2007 and 20 July 2007 at 0, 25, 50, and 100 kg ha-1 with a Gandy drop spreader and th en watered in with 0.5 cm of supplemental irrigation. Irrigation was appl ied at 60%, 80%, 100%, and 120% of weekly estimated ET values obtained from an onsite we ather station that is part of the Florida Agricultural Weather Network (http://fawn.ifas.ufl.edu/ ) when ET was in deficit of 1.27cm or more. This occurred only on 24 August 2007 (Appendix B-D). Plots were mowed 2 to 3 times per week at a height of 1.2 cm with a Toro R eelmaster 3100-D. Data were collected five times (22 June, 12 July, 25 July, 22 Aug., and 4 Sept .) throughout the growi ng season in 2007. All measurements were taken within a one hour time period and collected between 1100-1300 hrs. Plots were rated visually for color, quality, and density based on the standard 1-9 National Turfgrass Evaluation Program (NTE P) rating scale where 9 is the highest possible rating, 6 is the lowest acceptable rating, and 1 indicates dead turf. Volumetric water content (VWC) readings

PAGE 35

35 were taken via Time Domain Reflectometry (T DR) with a FieldScout TDR 300 Soil Moisture meter (Spectrum Technologies, Inc ., Plainfield, IL). Clipping tissue samples were immediately weighed after collection and oven dried at 52 C for 7d before being weighed again for determination of biomass production. The oven dried tissue was ground and then analyzed for Total Kjeldahl N content using the RFA Me thod, number A303-S075-01 (Alpkem Corporation, 1990) as described by Mylavarapu and Kennelly (2007). Reflectance Measurements Spectral measurements were taken from three remote sensing devices. An infrared thermometer (Raytek Raynger ST, TTI, Inc.) wa s used to determine canopy temperature of the turf and to develop a CWSI calculated as: CWSI= (Tc-Ta) a (Tc-Ta) l / (Tc-Ta) u (Tc-Ta) l, Where Tc= Canopy Temperature, Ta= Actual Temperature, a= act ual, u= upper limit, and l= lower limit (Jalali-Farahani et al., 1994). A Crop Circle (Model ACS-210) (Holland Scien tific) was fitted onto a Toro Greensmaster 1000 walking greens mower 0.81m above the turf canopy which provided a 0.45 m2 field of view. The Crop Circle sensor produces its own light source at 650 nm and 880 nm and measures reflected values to produce an NDVI which is calculated as (880-650 nm)/ (880+650 nm). No calibration method was used for the sensor since it has the ability to de tects it own light source from incoming ambient sunlight. A hand-held hyperspectral spectroradiometer (F ieldSpec Pro; Analytical Spectral Devices, Inc., Boulder, CO) was also used to collect reflec tance data. This radiometer has a spectral range of 300-2,500 nm with a 1 nm resolution and a 23 foreoptic. Reflectance re adings were collected from shoulder height on each plot which provided a 0.65 m2 field of view. Radiance values are expressed in percent reflectance compared to st andardization with a white reference value across the entire spectral range. A white reference was used for calibration purposes at the beginning of

PAGE 36

36 data collection with the ASD unit, and again every 15 minutes depending on the length of time needed to collect data. Usually only one white re ference was needed to take data on all plots. All reflectance readings were taken in full sunlight be tween the hours of 1100 and 1400 Central Standard Time (CST) to minimize vari ance caused by incoming solar radiation. Several bands were extracted from ASD refl ectance data for comparison based on previous research. These bands consisted of wavelengths 550 nm (Blackmer et al., 1994; Blackmer et al., 1996; Yoder and Pettigree-Crosby, 1995); 510 nm 535 nm, 635 nm, and 735 nm (Kruse et al., 2005); 545 nm, 755 nm and 935 nm (Starks et al., 2006); and 661 nm and 813 nm (Trenholm et al., 1999a). Spectral ranges 630-690 nm and 2080-2350 nm were averaged (Ripple 1986) and ratios 605/515 nm, 915/975 nm, and 865/725 nm were calculated (Starks et al., 2006). Also NDVI ((880-650 nm)/ (880+650 nm)) and LAI (8 80/650 nm) indices were calculated for comparison against Crop Circle NDVI and LAI ca lculations, as well as Stress1 (706/760 nm) and Stress2 (706/813 nm) indices (T renholm et al., 2000). Statistical Analysis Analysis of variance (ANOVA) was perf ormed using the PROC GLM method (SAS Inst., 2003) to compare the differences among N a nd irrigation treatments at various dates to visual ratings, Crop Circle read ings, IR temperature readings calculated into a CWSI, various individual reflectance wavelengt hs quantified from the ASD unit, soil VWC, and leaf N concentration. Regression analysis was perf ormed using the PROC REG method (SAS Inst., 2003) to correlate relationships between visual ra tings, Crop Circle readi ngs, and IR temperature readings, various individual reflectance wavele ngths quantified from the ASD unit, soil VWC, and leaf nitrogen concentration. Additionally, du e to the large amount of data generated by the ASD device, ASD data were rendered using PROC PLS (SAS Inst., 2003) for prediction of soil VWC, leaf N concentration, vi sual quality ratings, and biom ass production. Equations were

PAGE 37

37 validated through a single sample cross-validatio n. For cross-validation, ten percent of the sample was omitted for prediction purposes so th at the number of factors chosen creates the minimal predicted residual sum of squares (PRESS) This process was repeated so that every observation was used exactly once for cross-validation. Results and Discussion Influence of Irrigation Treatments ANOVA (Tables 2-1, 2-2, 2-3) indi cated no irrigation treatment effects were significant. This is likely due to rainfall received during the experiment which eliminated differential soil VWC. As stated previously in the material s and methods, irrigation treatments were only applied once during the experiment due to the fa ct that the weekly cumulative ET calculations (Appendix C) based on daily rainfall and ET values (Appendix B) we re only in deficit of more than 1.27cm once during the experiment. In review of the average rainfall and ET data from the last 3 years, the experiment s hould be run earlier in the year when cumulative ET minus rainfall differences are not as great (Appendix D). Nitrogen Effects ANOVA indicated a large source of variation in bermudagrass response to N treatments. Leaf tissue N concentration, color, quality, NDVI and LAI from both optical sensing devices, as well as many individual reflectan ce wavelengths and reflectance ratios indicated a significant response to N treatments (Tables 21, 2-2, 2-3). This can be attri buted to the ability of N to increase overall turf health si nce it is the single most impor tant macronutrient to the plant (Hopkins, 1999). NDVI and LAI from the Crop Circ le device indicated a sampling date effect (Table 2-1). Additionally, reflectance at wavelength 813 nm, LAI, and 915/975 nm and 865/725 nm ratios quantified from the ASD device also indicated a sampling date effect (Tables 2-1, 2-2, 2-3).

PAGE 38

38 Research conducted by Xiong et al (2007) showed seasonal NDVI and LAI effects in response to N treatments. There was a significant improvement in visual color between no N applied and N applied at all levels across all dates (Table 2-4). Visual Quality rating (Table 2-4) and 650 nm reflectance from the Crop Circle device were bo th greater at the highe st N rate (100 kg ha-1) compared to no N applied (Table 2-5). Simila rly, reflectance wavelengths (510 nm, 535 nm, 545 nm, 550 nm, and range average 630-690 nm) (Tab le 2-6), ratios (605/515 nm, 915/975 nm, and 865/725 nm) (Table 2-7), and indices (NDVI, St ress1, and Stress2) derived from ASD data detected the high rates of applied N (Table 2-7). Nitrogen X Date Interaction It is apparent from the N X Date interaction means that there is a significant sequential response to increasing N treatments on percent leaf N concentration the first three sampling dates (22 June, 12 July, and 25 July 2007) (Tables 2-8, 2-9, 2-10, and 2-11). NDVI and LAI indices calculated from the Crop Circle device (Table 2-8) also show this response as well as LAI (Table 2-8), 813 nm and 935 nm (Table 2-10), and 915 /975 nm and 865/925 nm ratios (Table 2-11) calculated from the ASD reflectance data. This response in reflectan ce data can be attributed to the ability of the devices to detect overall incr eases in turf health due to a sequential vigor response in reaction to varying N treatments. This agrees with the work of Bell et al. (2002a, 2004). The last two sampling dates (22 Aug. a nd 4 Sept. 2007) do not show a response to increasing N treatments for any parameter measur ed that was significant by date (Tables 2-8, 29, 2-10, and 2-11). This is likely due to a waning N response, since N had not been applied for 33 and 46 days respectively for the final two sampling dates.

PAGE 39

39 Tissue N Concentration vs. Reflectance Reflectance values in this experiment did not prove to be a good predictor of leaf tissue N concentration in bermudagrass using linear or PLS regression to model percent N verses reflectance measurements However, some weak correlations exist (T ables 2-12, 2-13, 2-14, 215). This could potentially be due to the expe rimental design where N treatments were only applied twice, one month apart, causing the waning effect mentioned in the N X Date interaction section. Comparatively, research conducted by Kr use et al. (2005) and Xi ong et al. (2007) where N treatments were applied every two weeks promot ed a more consistent response to treatments across the growing season. VWC vs. Reflectance The highest r2 values for estimation of VWC ar e derived from PLS regression (r2= 0.72) calculated from reflectance data from the ASD device (Table 2-15) (F igure 2-1), and NDVI (r2= 0.71) derived from reflectance data from the Crop Ci rcle device (Table 2-12) (Figure 2-2). Also, LAI (r2= 0.59), NIR reflectance (r2= 0.61), and RED reflectance (r2= 0.52) derived from Crop Circle values provide adequate correlation as well (Table 2-12). This corresponds to vegetative indice values NDVI (r2= 0.57), LAI (r2= 0.55), Stress1 (r2= 0.61), Stress2 (r2= 0.64), and ratios 605/515 nm (r2= 0.66), 915/975 nm (r2= 0.68), 865/725 nm (r2= 0.44) derived from the ASD unit (Table 2-14). Some correlations at individual bands 755 nm (r2= 0.53), 813 (r2= 0.53), and 935 nm (r2= 0.49) were also notable (Table 2-13). The high r2 values for estimation of VWC with remo tely sensed data from Crop Circle and various individual bands, indices, and ratios from the ASD devi ce could be attributed to the extremely low VWC values record ed on 22 Aug 2007. On this date, cumulative ET calculations were in deficit 2.39 cm causing a significant decline in VWC. This resulted in a major decline in turf health. In previous research many indices such as NDVI do well in predicting turf stress in

PAGE 40

40 response to various plant parameters and not th e actual plant parameter (Trenholm et al., 1999a). Similarly, the ability of the remote sensing devices to model VWC through individual bands, indices and ratios, could potentially be attributed to decline in turf health in response to VWC and not a true predictor of actual VWC. Furthermore, the high r2 value achieved by PLS regressi on to model soil VWC (Table 215) (Figure 2-1) could potentia lly be highly predictive of new data and specific to actual VWC and not a response to plant stress. This is due to the re-sampling method employed by the PLS procedure mentioned in the materials and methods. Visual Ratings Visual ratings were reasonably predicted from reflectance data from both devices. Visual assessment of color, quality, and density was best predicted by vegetative indices NDVI and LAI from both instruments (Table 2-12, 2-14)). Furt hermore, raw reflectance values 650 nm and 880 nm from the Crop Circle device achieved reasonable r2 values (Table 2-12), as well as, 510 nm, 535 nm, 545 nm, 550 nm, 635 nm, and 661 nm individual reflectance wavelengths (Table 2-13), 630-690 nm and 2080-2350 nm range averages, and 605/515 nm, 915/975 nm, and 865/725 nm ratios (Table 2-14) from the ASD device. C onversely, quality was not as well modeled by PLS Regression (Table 2-15), compared to the vege tative indices, individua l reflectance wavelengths, and the ratios mentioned above. The ability of vegetative indices, NDVI and LAI, to model visual assessments could potentially be attributed to the fact that these indices were formulated to normalize the data. Since regression functions were run across all dates the normaliz ation of these indices could potentially smooth out variation in reflectance due to differences in light intensity on various evaluation dates.

PAGE 41

41 Biomass Biomass was not modeled well by reflectance da ta compared to other factors tested. However, some weak correlations exist (Tables 2-12, 2-13, 2-14, and 2-15). This is similar to research conducted by Kruse et al., 2005, 2006; Fenstemaker-Shaulis et al., 1997; and Trenholm et al., 1999b. The inability to effectively model biomass in this experiment could be attributed to the methodology used. Nitrogen treatments were only applied twice one month apart. In other research where biomass was adequately modele d (Kruse et al., 2005, 2006) N treatments were applied every 2 weeks thus continually promo ting greater differences in turf growth. Crop Circle vs. ASD The Crop Circle and ASD instruments are two ve ry different remote sensing devices that differ dramatically in price and use-ability. Th e ASD device is a highly scientific instrument with great ability to give a rese archer a plethora of reflectance data from a broad region of the electromagnetic spectrum. However, it is limite d in its ability because it requires full sunlight for use and only collects reflectance data when manually prompted by the user. Nonetheless, when reflectance data in this magnitude is analyz ed using the modern regression techniques, like PLS regression, the correlations could potentially be very specific to the parameter tested and highly predictive of new data. Therefore, pred iction models could likely be constructed for specific forms of turfgrass stress. Conversely, optical sensing devices like the Crop Circle instrume nt are less complicated to use but only give a limited amount of informati on typically in the visi ble and NIR regions. However, this data can be used to compute NDVI, LAI, and a variety of other indices to make the data more usable. Furthermore, they produce their own light s ource and, thus, can be mounted on a vehicle or mower for autonomous reflectance measurement collection.

PAGE 42

42 In this experiment VWC was best modeled by both remote sensing devices, Crop Circle and ASD. NDVI from the Crop Circle (r2=0.71) (Table 2-12) (Figure 2-2) device and PLS regression from the ASD device (r2=0.72) were the highest correla tions achieved (Table 2-15) (Figure 2-1). NDVI from the ASD device was close in comp arison to the Crop Circle (r2=0.57) to model VWC (Figure 2-2). Even though both of these device s utilize completely different sources of light they both calculate NDVI similarly and are close in comp arison to model VWC (Fi gure 2-3). Granted the prediction of VWC with PLS regr ession on ASD data could be highl y predictive of new data and specific to actual VWC and not just turf health, the Crop Circ le device shows there is some sensitivity to this parameter. Conclusions The results indicate reflectance data best m odeled soil VWC, visual ratings, and some treatment effects from N. Based on research by Ripple (1986), the elevated levels of prediction for VWC could be due to a reduction in chlorop hyll as the leaves dried in response to limiting soil VWC and not a true relation to actual soil water status. None theless, this research gives greater insight into the use of remote sensi ng technology and its place for use by turf managers and researchers demonstrating the sensitivity of remote sensing instrumentation to various parameters of turfgrass stress. Previous re search found optical sens ing measurements using NDVI could provide a fast and accurate alternativ e to traditional visual rating systems (Trenholm et al., 1999b; Bell et al., 2000) as well as assessing turfgra ss stress (Trenholm et. al., 1999a, Fenstemaker-Shaulis et al., 1997). Our results from two different remote sensing instruments are in agreement with this previous work. The Crop Circle instrument performed equa lly well with the ASD device in modeling stress attributed to reduced soil VWC (Figure 2-2) and was consistent with visual ratings in this

PAGE 43

43 experiment. It is simple to use and is poten tially non-restrictive of environmental conditions, like cloud cover and variation in so lar radiation. Currently, device s like this could provide fast and easy supplemental information to turfgrass managers for assistance in making decisions. Coupled with GPS and GIS technology, ground ba sed remote sensing devices, like the Crop Circle, can give turfgrass mana gers greater insight into mana ging their turf by potentially allowing them to more efficien tly schedule their irrigation, ther eby reducing water consumption, leaching, and runoff. Furthermore, if devices like the Crop Circ le could be improved to produce and quantify additional wavelengths these devices could potentia lly be employed by turfgrass managers due to their ease of use. However, further investigatio n into data from devices like the ASD using PLS regression is still needed to investigate the entire spectral constituency for optimal wavelengths to predict various turf parame ters. Only then can predicti on equations be formulated and incorporated into devices like the Crop Circle that could produce the correct wavelengths to calculate these prediction equations.

PAGE 44

44Table 2-1. Analysis of variance of soil volumetric water content (VWC), percent leaf nitrogen tissue concentration (%N), visual ratings: color, quality, density (rated on NTEP 1-9 scale, where 9 is the best rating and 6 is acceptable), and biomass. Normalized difference vegetation index (NDVI), Leaf area index (L AI), 880 nm, and 650 nm reflectance were obtained from Crop Circle device. Crop water stress i ndex (CWSI) was computed from IR thermometer readings. All data was collected from fairway he ight hybrid bermudagrass ( Cynodon dactylon X C. transvaalensis, Burtt-Davy), four irrigation rates (60, 80, 100, and 120% estimated ET valu es), four N rates (0, 25, 50, and 100 kg ha-1), and five dates (22 June, 12 July, 25 July, 22 Aug., and 4 Sept. 2007) combined. _____Visual Ratings______ _________Crop Circlez_________ __I Ry__ Source of Variation df VWC % N Dry W Color Quality Density NDVIx LAIw 880 nm 650 nm CWSIv Rep 2 0.0001 0.0044 0.0001 0.0102 0.0048 0.0015 0.0001 0.0001 0.0001 0.0001 0.2616 Date 4 0.0001 0.0001 0.0001 0.0001 0.0001 0.0001 0.0001 0.0001 0.0001 0.0001 0.0001 Irrigation 3 0.3106 0.8398 0.3815 0.9382 0.5418 0.4083 0.2885 0.1709 0.0515 0.2266 0.7858 DxI 12 0.9981 0.9934 0.9401 0.9968 0.6717 0.9902 0.9921 0.9631 0.9970 0.9860 0.9896 error a 12 Nitrogen 3 0.1447 0.0001 0.1100 0.0001 0.0153 0.1087 0.0001 0.0001 0.2223 0.0001 0.5505 IxN 9 0.0004 0.0502 0.3567 0.6267 0.2050 0.0065 0.8093 0.8038 0.4815 0.3908 0.8610 DxN 12 0.2988 0.0002 0.4345 0.0595 0.7875 0.5126 0.0029 0.0001 0.0309 0.0612 0.7184 error b 36 Probability of greater F ratio (P>F) for Date, Irrigation, and N treatment components z Model ACS-210, Holland Scientific y IR thermometer Raytek Raynger ST, TTI, Inc. x NDVI= (880-650 nm)/(880+650 nm) w LAI= 880/650 nm v CWSI= (T c -T a ) a (T c -T a ) l / (T c -T a ) u (T c -T a ) l Where Tc= Canopy Temperature, Ta= Actual Temperature, a= actual, u= upper limit, and l= lower limit

PAGE 45

45 Table 2-2. Analysis of variance of indi vidual reflectance bands from ASD reflectan ce hyperspectral data with a range of 350-250 0 nm and spectral resolution of 1 nm collected from fairway height hybrid bermudagrass ( Cynodon dactylon X C. transvaalensis, Burtt-Davy), four irrigation rates (60, 80, 100, and 120% es timated ET values), four N rates (0, 25, 50, and 100 kg ha-1), and five dates (22 June, 12 July, 25 Ju ly, 22 Aug., and 4 Sept. 2007) combined. _____________________________________ASDz (nm)___________________________________________ Source DF 510 535 545 550 635 661 735 755 813 935 2132 Rep 2 0.0429 0.0002 0.0001 0.0001 0.0596 0.1303 0.0001 0.0001 0.0001 0.0001 0.0584 Date 4 0.0001 0.0001 0.0001 0.0001 0.0001 0.0001 0.0001 0.0001 0.0001 0.0001 0.0001 Irrigatio 3 0.2045 0.4189 0.4541 0.4650 0.2572 0.1908 0.0999 0.1148 0.1094 0.1134 0.1102 DxI 12 0.0826 0.2307 0.2667 0.2756 0.0948 0.0659 0.9956 0.9554 0.9698 0.9933 0.0774 error a 12 Nitroge 3 0.0029 0.0002 0.0001 0.0001 0.0039 0.0116 0.9039 0.0275 0.0053 0.0073 0.097 IxN 9 0.5871 0.3050 0.2683 0.2631 0.6254 0.6496 0.0874 0.4093 0.4387 0.4533 0.726 DxN 12 0.4677 0.2312 0.2048 0.2034 0.4701 0.5078 0.4558 0.7430 0.0222 0.1810 0.7521 error b 36 Probability of greater F ratio (P>F) for Date, Irrigation, and N treatment components z FieldSpec Pro; Analytical Spec tral Devices, Inc., Boulder, CO

PAGE 46

46 Table 2-3. Analysis of variance of average range computations from 630690 nm and 2080 nm; NDVI LAI Stress1 and Stress2 indices; and 605 nm/515 nm, 915/975 nm, and 865/725 nm ratio s computed from an Analytical Spectral Device (ASD) with a range of 350-2500nm and spectral resolution of 1 nm collected from fairway height hybrid bermudagrass ( Cynodon dactylon X C. transvaalensis, Burtt-Davy), four irriga tion rates (60, 80, 100, and 120% estimated ET values), four N rates (0, 25, 50, and 100 kg ha-1), and five dates (22 June, 12 July, 25 July, 22 Aug., and 4 Sept. 2007) combined. ________________________________________ASDz____________________________________________ Source of Variation DF 630-690 nm 20802350 nm NDVI y LAI x Stress1 v Stress2 u 605/515 nm 915/975 nm 865/725 nm Rep 2 0.1209 0.0627 0.0313 0.1656 0.0099 0.0545 0.0206 0.0025 0.0018 Date 4 0.0001 0.0001 0.0001 0.0001 0.0001 0.0001 0.0001 0.0001 0.0001 Irrigation 3 0.2010 0.1034 0.1562 0.6883 0.1654 0.4132 0.8121 0.3246 0.6495 DxI 12 0.0712 0.0768 0.1737 0.9830 0.1296 0.4497 0.8643 0.8901 0.9094 error a 12 Nitrogen 3 0.0105 0.0869 0.0050 0.0001 0.0089 0.0019 0.0001 0.0001 0.0001 IxN 9 0.6521 0.7096 0.8275 0.9484 0.8028 0.9325 0.6612 0.9299 0.9001 DxN 12 0.5014 0.7276 0.3930 0.0008 0.2417 0.3474 0.1934 0.0299 0.0001 error b 36 Probability of greater F ratio (P>F) for Date, Irrigation, and N treatment components z FieldSpec Pro; Analytical Spec tral Devices, Inc., Boulder, CO y NDVI= (880-650 nm)/(880+650 nm) x LAI= 880/650 nm v Stress1= 706/760 nm u Stress2= 706/813 nm

PAGE 47

47 Table 2-4. Evaluation of means of volumetric water content (VWC), leaf nitrogen tissue concentrati on (%N), visual ratings: col or, quality, density (rated on NTEP 1-9 scal e 9 is best and 6 is acceptable), and bi omass (g) collected from fairway height hybrid bermudagrass ( Cynodon dactylon X C. transvaalensis, Burtt-Davy), four irriga tion rates (60, 80, 100, and 120% estimated ET values), four N rates (0, 25, 50, and 100 kg ha-1), and five dates (22 June, 12 July, 25 July, 22 Aug., and 4 Sept. 2007) combined. _____________Visual Ratings______________ N Rate VWC % N Color Quality Density biomass (g) 0 kg ha-1 22.04 3.38 7.44b 7.12b 7.01 48.47 25 kg ha-1 22.07 3.61 7.73a 7.29ab 6.99 55.85 50 kg ha-1 21.87 3.62 7.76a 7.22ab 7.03 53.23 100 kg ha-1 22.48 3.99 7.85a 7.40a 7.12 57.05 LSD N.S. -0.20 0.21 N.S. N.S. "--" denotes LSD is not valid due to a significant date by N treatment interaction. Reporte d separately in Table 2-5 based on interaction. All values with a given number for the LSD denotes p < 0.05

PAGE 48

48 Table 2-5. Evaluation of means of Norm alized difference vegetation index (NDVI ), Leaf area index (LAI ), 850 nm, and 650 nm reflectance from Crop Circle device and crop water st ress index (CWSI) computed from IR thermometer readings collected from fairway he ight hybrid bermudagrass ( Cynodon dactylon X C. transvaalensis, Burtt-Davy), four irrigation rates (60, 80, 100, and 120% estimated ET valu es), four N rates (0, 25, 50, and 100 kg ha-1) and five dates (22 June, 12 July, 25 July, 22 Aug., and 4 Sept. 2007) combined. ____________Crop Circle Reflectance Readingsz______________ IR Thermometery N Rate NDVIx LAIw880 nm 650 nm CWSIv 0 kg ha-1 0.59 4.35 1.38 .34a 0.44 25 kg ha-1 0.61 4.58 1.38 .33a 0.44 50 kg ha-1 0.60 4.61 1.38 .33a 0.45 100 kg ha-1 0.62 4.99 1.39 .31b 0.43 LSD --N.S. 0.01 N.S. "--" denotes LSD is not valid due to a significant date by N treatment interaction. Reported separately in Table 5 based on in teraction. All values with a given number for the LSD denotes p < 0.05 z Model ACS-210, Holland Scientific y IR thermometer Raytek Raynger ST, TTI, Inc. x NDVI= (880-650 nm)/(880+650 nm) w LAI= 880/650 nm v CWSI= (Tc-Ta)a (Tc-Ta)l / (Tc-Ta)u (Tc-Ta)l Where Tc= Canopy Temperature, Ta= Actual Temperature, a= actual, u= upper limit, and l= lower limit

PAGE 49

49 Table 2-6. Evaluation of means of individual band reflectance from the ASD device with a range of 350-2500 nm and a spectral resolution of 1 nm collected from fairway height hybrid bermudagrass ( Cynodon dactylon X C. transvaalensis, BurttDavy), four irrigation rates (60, 80, 100, and 120% estimated ET values), four N rates (0, 25, 50, and 100 kg ha-1) and five dates (22 June, 12 July, 25 July 22 Aug., and 4 Sept. 2007) combined. _____________________________________________ASD z (nm)_________________________________________ N Rate 510 535 545 550 635 661 735 755 813 935 2132 0 kg ha-1 0.054a 0.081a 0.086a 0.088a 0.074a 0.069ab 0.29 0.366a 0.407 0.437 0.183 25 kg ha-1 0.055a 0.081a 0.086a 0.088a 0.075a 0.071a 0.29 0.363a 0.404 0.433 0.190 50 kg ha-1 0.055a 0.081a 0.086a 0.088a 0.076a 0.072a 0.29 0.361a 0.402 0.431 0.194 100 kg ha-1 0.051b 0.077b 0.082b 0.084b 0.068b 0.064b 0.29 0.367a 0.409 0.437 0.178 LSD 0.002 0.002 0.002 0.002 0.005 0.005 N.S. 0.007 N.S. -N.S. "--" denotes LSD is not valid due to a significant date by N treatment interaction z FieldSpec Pro; Analytical Spec tral Devices, Inc., Boulder, CO

PAGE 50

50 Table 2-7. Evaluation of means of ASD reflectance readings of average range computations from 630 nm and 2080 nm, computed indices Normalized Difference Vegetation Index (NDVI y), Leaf Area Index (LAI) and Stress1 and Stress2 indices, and ratios 605/515 nm, 915/975 nm, 865/725 nm take n from an ASD device with a range of 350-2500 nm and a spectral resolution of 1 nm collected fr om fairway height hybrid bermudagrass ( Cynodon dactylon X C. transvaalensis, Burtt-Davy), four irrigation rates (60, 80, 100, and 120% es timated ET values), four N rates (0, 25, 50, and 100 kg ha-1) and five dates (22 June, 12 July, 25 Ju ly, 22 Aug., and 4 Sept. 2007) combined. Range Average (ASD z) Vegetative Indices (ASD z) Ratios (ASD z) N Rate 630-690 nm 2080-2350 nmNDVI y LAI x Stress 1 v Stress 2 u 605/515 nm 915/975 nm 865/725 nm 0 kg ha-1 0.071a 0.175 0.715ab7.06 0.367ab 0.788ab 1.02a 0.99 1.88 25 kg ha-1 0.073a 0.182 0.707b 7.26 0.375a 0.794a 1.03a 0.99 1.89 50 kg ha-1 0.743a 0.186 0.700b 7.28 0.383a 0.791a 1.03a 0.99 1.90 100 kg ha-1 0.657b 0.170 0.729a 8.21 0.354b 0.779b 1.03b 1.00 1.97 LSD 0.0053 N.S. 0.018 -0.019 0.009 0.01 --"--" denotes LSD is not valid due to a significant date by N treatment interaction z FieldSpec Pro; Analytical Spec tral Devices, Inc., Boulder, CO y NDVI= (880-650 nm)/(880+650 nm) x LAI= 880/650 nm v Stress1= 706/760 nm u Stress2= 706/813 nm

PAGE 51

51 Table 2-8. Evaluation of means of NDVI and LAI computed from Crop Circle instrument, and LAI computed from ASD collected from fairway height hybrid bermudagrass ( Cynodon dactylon X C. transvaalensis Burtt-Davy), four irrigation rates (60, 80, 100, and 120% estimated ET values), four N rates (0, 25, 50, and 100 kg ha-1), an d five dates (22 June, 12 July, 25 July, 22 Aug., and 4 Sept. 2007) combined, which had N treatment X date interactions. NDVIz (Crop Circley) LAIx (Crop Circle) LAIx (ASDw) Date 0 kg ha-1 25 kg ha-1 50 kg ha-1 100 kg ha-1 0 kg ha-1 25 kg ha-1 50 kg ha-1 100 kg ha-1 0 kg ha-1 25 kg ha-1 50 kg ha-1 100 kg ha-1 6/22 0.70a3 0.73a23 0.73a2 0.75a1 5.79a3 6.34a23 6.58a2 7.34a1 10.20a23 11.11a23 11.65a2 13.08a1 7/12 0.67b3 0.69b2 0.69b2 0.71b1 5.10b3 5.49b2 5.50b2 5.99b1 8.28b2 8.59b2 8.73b12 9.60b1 7/25 0.66b3 0.69b2 0.70ab12 0.71b1 4.87b3 5.41b2 5.69b12 5.9567b1 8.25b3 9.29b2 9.32b2 10.31b1 8/22 0.53c1 0.42d12 0.39d2 0.53c1 2.58d1 2.45d12 2.96c12 2.41d2 3.66c1 3.39c1 3.19c1 3.41c1 9/4 0.44d1 0.52c1 0.49c1 0.52c1 3.39c1 3.20c1 2.34d1 3.28c1 4.89c1 3.93c1 3.54c1 4.64d1 Means in the same column and category fo llowed by the same letter are not significantly different (LSD; P, 0.05); means in the same row and category followed by the same number are not significantly different (LSD; P, 0.05). z NDVI= (880-650 nm)/(880+650 nm) y Model ACS-210, Holland Scientific x LAI= 880/650 nm w FieldSpec Pro; Analytical Spec tral Devices, Inc., Boulder, CO

PAGE 52

52 Table 2-9. Evaluation of Means of leaf N concentration tissue (%N) collected from fairway height hybrid bermudagrass ( Cynodon dactylon X C. transvaalensis, Burtt-Davy), four irrigation rates (60, 80, 100, and 120% estimated ET values), four N rates (0, 25, 50, and 100 kg ha-1), and five dates (22 June, 12 July, 25 July, 22 Aug., and 4 Sept. 2007) combined, which had N treatment X date interactions. __________________________________ %N _________________________________________ Date 0 kg ha-1 25 kg ha-1 50 kg ha-1 100 kg ha-1 6/22 3.51b3 4.02a2 4.19a2 4.66a1 7/12 3.24bc3 3.25b2 3.38b2 3.81b1 7/25 3.26bc3 3.91a2 3.91a2 4.59a1 8/22 2.90c1 2.91b1 2.75c1 2.83c1 9/4 4.01a1 3.98a1 3.88a1 4.10b1 Means in the same column and category fo llowed by the same letter are not significantly different (LSD; P, 0.05); means in the same row and category followed by the same number are not significantly different (LSD; P, 0.05).

PAGE 53

53 Table 2-10. Evaluation of means of individua l reflectance wavelengths at 813 nm and 935 nm collected from fairway height hybrid bermudagrass ( Cynodon dactylon X C. transvaalensis, Burtt-Davy), four irrigation rate s (60, 80, 100, and 120% estimated ET values), four N rates (0, 25, 50, and 100 kg ha-1), and five dates (22 June, 12 Ju ly, 25 July, 22 Aug., and 4 Sept. 2007) combined, which had N treatment X date interactions. ___________________813 nm (ASDz)________________ __________________935 nm (ASDz)__________________ Date 0 kg ha-1 25 kg ha-1 50 kg ha-1 100 kg ha-1 0 kg ha-1 25 kg ha-1 50 kg ha-1 100 kg ha-1 6/22 0.47a1 0.47a1 0.48a1 0.49a1 0.49a1 0.49a1 0.50a1 0.51a1 7/12 0.43bc1 0.43b1 0.43b1 0.44b1 0.45b1 0.45c1 0.45c1 0.46b1 7/25 0.42c1 0.43b1 0.43b1 0.43b1 0.45b1 0.46bc1 0.46bc1 0.46b1 8/22 0.26d1 0.25c12 0.24c2 0.24c2 0.30c1 0.29d12 0.28d2 0.28c2 9/4 0.45ab1 0.44b1 0.43b1 0.45b1 0.49a1 0.48ab1 0.47b1 0.49a1 Means in the same column and category fo llowed by the same letter are not significantly different (LSD; P, 0.05); means in the same row and category followed by the same number are not significantly different (LSD; P, 0.05). z FieldSpec Pro; Analytical Spectr al Devices, Inc., Boulder, CO

PAGE 54

54 Table 2-11. Evaluation of means of ratios 915 nm/975 nm and 865 nm/925 nm computed from analytical spectral device (ASD) with a range of 350-2500 nm and a spectral resolution of 1 nm, collected from fairway height hybrid bermudagrass ( Cynodon dactylon X C. transvaalensis, Burtt-Davy), four irrigation rates (60, 80, 100, and 120% estimated ET values), four N rates (0, 25, 50, and 100 kg ha-1), and five dates (22 June, 12 July, 25 July, 22 Aug., and 4 Sept. 2007) combined, which had N treatment X date interactions. 915/975 nm _(ASDz) 865/925 nm_(ASDz) Date 0 kg ha-1 25 kg ha-1 50 kg ha-1 100 kg ha-1 0 kg ha-1 25 kg ha-1 50 kg ha-1 100 kg ha-1 6/22 1.03a2 1.03a2 1.03a2 1.04a1 2.12a3 2.20a23 2.23a2 2.35a1 7/12 1.01b2 1.02b12 1.02b12 1.02b1 2.00b2 2.02b2 2.04b2 2.13b1 7/25 1.01b3 1.02b23 1.02b2 1.03b1 1.96b3 2.03b23 2.04b2 2.12b1 8/22 0.95d1 0.95d12 0.94d12 0.94d2 1.66c1 1.63c12 1.61c12 1.63c2 9/4 0.99c1 0.98c1 0.98c1 0.99c1 1.68c 1.60c1 1.57c1 1.63c1 Means in the same column and category fo llowed by the same letter are not significantly different (LSD; P, 0.05); means in the same row and category followed by the same number are not significantly different (LSD; P, 0.05). z FieldSpec Pro; Analytical Spectr al Devices, Inc., Boulder, CO

PAGE 55

55 Table 2-12. Coefficient estimates (r2) of Crop Circle reflectance data Normalized Difference Vegetation I ndex (NDVI), Leaf Area Index (LAI), NIR, and Red, and crop water stress index (CWSI ) to soil volumetric water content (VWC) visual color, quality, density ratings collected from fairway height hybrid bermudagrass ( Cynodon dactylon X C. transvaalensis, BurttDavy), four irrigation rates (60, 80, 100, and 120% esti mated ET values), four N rates (0, 25, 50, and 100 kg ha-1), and averaged over all dates (22 June, 12 July, 25 July, 22 Aug., and 4 Sept. 2007). _________________________Crop Circlez___________________________ IR Thermometery Parameter NDVIx LAIw 880 nm 650 nm CWSIv %N 0.19*** 0.21*** 0.15*** 0.15*** 0.02* VWC 0.71*** 0.59*** 0.61*** 0.52*** 0.02* Color 0.59*** 0.57*** 0.39*** 0.61*** 0.13*** Quality 0.70*** 0.67*** 0.50*** 0.71*** 0.33*** Density 0.67*** 0.65*** 0.48*** 0.69*** 0.38*** Biomass 0.26*** 0.33*** 0.14*** 0.34*** 0.05** *, **, *** Significant at the 0.05, 0.01, and 0.001 probability levels, respectively z Model ACS-210, Holland Scientific y IR thermometer Raytek Raynger ST, TTI, Inc. x NDVI= (880-650 nm)/(880+650 nm) w LAI= 880/650 nm v CWSI= (Tc-Ta)a (Tc-Ta)l / (Tc-Ta)u (Tc-Ta)l Where Tc= Canopy Temperature, Ta= Actual Temperature, a= actual, u= upper limit, and l= lower limit

PAGE 56

56 Table 2-13. Coefficients estimates (r2) of ASD reflectance readings of individual reflectance wavelengt hs taken from an Analytical spectral device (ASD) with a range of 3502500 nm and a spectral resolution of 1 nm to %N concentration leaf tissue, soil volumetric water content (VWC) visual co lor, quality, density ratings, and biom ass production of fair way height hybrid bermudagrass ( Cynodon dactolon X C. transvaalensis, Burtt-Davy), four irrigation rate s (60, 80, 100, and 120% estimated ET values), four N rates (0, 25, 50, and 100 kg ha-1), and averaged over all dates (22 June, 12 July, 25 July, 22 Aug., and 4 Sept. 2007). ASDz Reflectance (nm) Parameter 510 535 545 550 635 661 735 755 813 935 2132 %N 0.00 0.02* 0.02* 0.02* 0.01 0.02* 0.25*** 0.29*** 0.29*** 0.28*** 0.02* VWC 0.08*** 0.01 0.00 0.36 0.19*** 0.23*** 0.38*** 0.53*** 0.52*** 0.49*** 0.29*** Color 0.28*** 0.15*** 0.13*** 0.13*** 0.36*** 0.38*** 0.09*** 0.23*** 0.24*** 0.19*** 0.40*** Quality 0.52*** 0.32*** 0.31*** 0.31*** 0.60*** 0.62*** 0.06*** 0.22*** 0.23*** 0.17*** 0.69*** Density 0.55*** 0.37*** 0.34*** 0.35*** 0.62*** 0.65*** 0.17** 0.18*** 0.19*** 0.13*** 0.71*** Biomass 0.25*** 0.19*** 0.18*** 0.18*** 0.26*** 0.25*** 0.01 0.06*** 0.07*** 0.05** 0.25*** *, **, *** Significant at the 0.05, 0.01, and 0.001 probability levels, respectively z FieldSpec Pro; Analytical Spectr al Devices, Inc., Boulder, CO

PAGE 57

57 Table 214. Coefficients estimates (r2) of average range computations from 630 nm and 2080 nm, and computed indices Normalized difference vegetation index (NDV I), Leaf area index (LAI) and Stress1 a nd Stress2 indices also taken from an ASD with a range of 350-2500 nm and a spectra l resolution of 1 nm to %N concentra tion leaf tissue, so il volumetric water content (VWC) visual color, quality, density ratings, and biomass production co llected from fairway height hybrid bermudagrass ( Cynodon dactylon X C. transvaalensis, Burtt-Davy), four irrigation rate s (60, 80, 100, and 120% estimated ET values), four N rates (0, 25, 50, and 100 kg ha-1), and averaged over all dates (22 June, 12 July, 25 July, 22 Aug., and 4 Sept. 2007). Range Average (ASDz) Vegetative Indices (ASDz) Ratios (ASDz) Parameter 630-690 nm 2080-2350 nm NDVIy LAIx Stress 1w Stress 2v 605/515 nm 915/975 nm 865/725 nm %N 0.02 0.02* 0.13*** 0.20** 0.15*** 0.16*** 0.21*** 0.22*** 0.14*** VWC 0.22*** 0.29*** 0.57*** 0.53*** 0.61*** 0.64*** 0.66*** 0.68*** 0.44*** Color 0.38*** 0.40*** 0.53*** 0.55*** 0.53*** 0.55*** 0.45*** 0.52*** 0.51*** Quality 0.62*** 0.69*** 0.71*** 0.66*** 0.67*** 0.69*** 0.46*** 0.61*** 0.67*** Density 0.65*** 0.71*** 0.69*** 0.63*** 0.65*** 0.67*** 0.44*** 0.58*** 0.67*** Biomass (g) 0.26*** 0.25*** 0.26*** 0.35*** 0.24*** 0.27*** 0.15*** 0.24*** 0.67*** *, **, *** Significant at the 0.05, 0.01, and 0.001 probability levels, respectively z FieldSpec Pro; Analytical Spec tral Devices, Inc., Boulder, CO y NDVI= (880-650 nm)/(880+650 nm) x LAI= 880/650 nm w Stress1= 706/760 nm v Stress2= 706/813 nm

PAGE 58

58 Table 2-15. Partial Least Squares regressi on on hyperspectral data from ASD device with a range of 350-2500 nm and a spectral resolution of 1 nm for prediction of soil volumetric water co ntent (VWC), percent leaf n itrogen concentration (%N), biomass, and visual quality collected fr om fairway height hybrid bermudagrass ( Cynodon dactylon X C. transvaalensis, Burtt-Davy), four irrigation rates (60, 80, 100, and 120% es timated ET values), four N rates (0, 25, 50, and 100 kg ha-1), averaged over all dates (22 June, 12 July, 25 July, 22 Aug., and 4 Sept. 2007). Calibration No. of Factors* r2 SE n VWC 8 0.72 0.030 240 % N 7 0.25 0.043 240 Biomass (g) 6 0.31 0.040 240 Visual Quality 7 0.55 0.031 240 *Number of factors required to achieve a minimal Predicted Resi dual Sum of Squares (PRESS) of prediction for the partial least squares regression model. z FieldSpec Pro; Analytical Spec tral Devices, Inc., Boulder, CO

PAGE 59

59 ASD y = 0.7166x + 6.231 R2 = 0.7083 -5 0 5 10 15 20 25 30 35 40 45 0510152025303540 Actual VWCPredicted VWC Figure 2-1. Partial Least Squares regression on hyperspectral data from ASD device with a range 0f 350-2500 nm and a spectral r ange of 1 nm for prediction of volumetric water content (VWC ) collected from fairway he ight hybrid bermudagrass ( Cynodon dactylon X C. transvaalensis, Burtt-Davy), four irrigation rates (60, 80, 100, and 120% estimated ET values), four N rates (0, 25, 50, and 100 kg ha-1), averaged over all dates (22 June, 12 July, 25 July, 22 Aug., and 4 Sept. 2007).

PAGE 60

60 Crop Circle y = 65.19x 17.46 R2 = 0.71 ASD y = 52.74x 15.50 R2 = 0.57 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0510152025303540 VWCNDVI Figure 2-2. Coefficient estimates (r2) of Crop Circle and ASD reflectance data com puted into a Normalized Difference Vegetation Index (NDVI), to soil volumetric water content (VWC) collected from fairway height hybrid bermudagrass ( Cynodon dactylon X C. transvaalensis, Burtt-Davy), four irrigation rates (60, 80, 100, and 120% estimated ET values), four N rates (0, 25, 50, and 100 kg ha-1), and averaged over all da tes (22 June, 12 July, 25 Ju ly, 22 Aug., and 4 Sept. 2007).

PAGE 61

61 y = 0.8478x + 0.0025 R2 = 0.88 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 00.10.20.30.40.50.60.70.80.91 ASDCrop Circle Figure 2-3. Coefficient estimates (r2) of Crop Circle vs. ASD using reflectance data to compute a Normalized Difference Vegetat ion Index (NDVI) collected from fair way height hybrid bermudagrass ( Cynodon dactylon X C transvaalensis Burtt-Davy), four irrigation rates (60, 80, 100, and 120% estimated ET values), four N rates (0, 25, 50, and 100 kg ha-1), and averaged over all dates (22 June, 12 Jul y, 25 July, 22 Aug., and 4 Sept. 2007).

PAGE 62

62 CHAPTER 3 USE OF GROUND-BASED REMOTE SENSIN G TECHNOLOGY TO ASSESS BIOMASS PRODUCTION IN RESPONSE TO PLANT GROWTH REGULATOR APPLICATIONS Introduction Remote sensing is defined as obtaining info rmation about an object, area, or phenomenon by analyzing data acquired by a device that is not in contact with that object, area, or phenomenon (Lillesand and Keifer, 1987). For many years researchers have entertained the idea of measuring various plant parameters using remo te sensing technology. Plant light interception significantly influences growth and physiologi cal responses (Salisbury and Ross, 1992). When light is intercepted by the plant it is absorbe d, transmitted, or reflected (Salisbury and Ross, 1992). Using remote sensing technology to quantify the light that is refl ected could help to detect the onset of turfgrass st ress (Ikemura and Leinauer, 2006). Real world applications of remote sensing technology for use in turfgrass ma nagement are still in their infancy; however, studies have shown that image analysis and variou s remote sensing devices have strong potential to detect a variety of turfgrass stresses (Ikemur a and Leinauer, 2006). Recent investigations into this technology show that there is potential for turfgrass managers to manage and assess turf more efficiently. The use of remote sensi ng technology to estimate turfgrass stress could significantly decrease the time and labor required to assess these levels in traditional ways, thus reducing cost as well (Osborne et al., 2002). The many great achievements in U.S. agricultural productivity can be attributed to the use of agricultural chemicals in cluding fertilizer and pesticides (Lee et al., 1999). However, the increased dependence on pesticides and fer tilizers have heightened many environmental concerns especially in Florida due to the sandy soils and heavy ra infall, which increases potential for heavy runoff and leaching of chemicals, if applied in excess (Min and Lee, 2003). With increasing water and environmental restrictions, turf managers and researchers must be aware of

PAGE 63

63 technological advances in their industry. Thes e environmental concerns have pressed turf managers to reduce nutrient and pesticide inputs used for turf maintenance (Bell et al., 2004). In other industries, such as agr onomic food production crops, precision agriculture has been used as a management tool to maximize yield and minimize cost (Bethel et al., 2003) The basis of this technology relies heavily on the Global Positi oning System (GPS), Geographic Information Systems (GIS), Variable Rate Application (V RA) and remote sensing technology. Using this technology to apply nutrients, wate r, and pesticides only where they are needed can optimize yield while minimizing cost. (Lee et al., 1999) Plant growth regulators (PGRs) are used by turf managers to s uppress vertical turf growth, which ultimately reduces maintenance costs due to reduced mowing requirments. A PGR is a substance which adjusts the growth and development of a plant. This is generally achieved through the inhibition of the gibberellic acid (GA), a horm one which is responsible for cell elongation. PGRs were first introduced for use on fine turf in 1987 (VanBibber, 2006). Some data suggests that PGR-treat ed turf can produce greater r oot mass, recover from injury faster, reduce water use rate, reduce disease incidence, and reduce Poa annua weed populations (Branham, 1997). The first of these chemicals were flurprimidol (Trade name Cutless) and paclobutrazol (Trade name Trimmet) which inhib it GA synthesis at relativ ely the same point in the GA biosynthesis pathway resulting in very similar plant responses (Branham, 1997). However, paclobutrazol is a more active compound and requires lower rates to achieve the same response as higher rates of flur primidol (Branham, 1997). The newe st of the PGRs, trinexapacethyl (TE) (Trade name Primo), became commer cially available in 199 5 and has become the standard for use throughout the turf industry (B ranham, 1997). TE facilitates GA inhibition later

PAGE 64

64 in the GA biosynthesis pathway so that GAs are formed but are not active and serve as precursors for plant biochemical processes. Historically PGRs are relatively expensive comp ared to other chemical applications made by turf managers. In an effort to reduce the co st of using PGRs, resear chers have investigated new methods to reduce the use rate. Research ers used remote sensi ng, the global positioning system (GPS), and geographic information system s (GIS), and variable rate application (VRA) technology to make site specific applications of PGRs. These image-based PGR applications were tested in cotton ( Gossypium sp.) to reduce the chemical i nput of PGRs while also reducing cost. Among comparisons of different image base d-application (Variable Ra te and Site-Specific) methods to traditional broadcast applications for PGRs in cott on, it was found that the broadcast applications were on average 19 % more costly than the site-specific applications, and 27 % more costly than applications using variable rate methods. Th is demonstrates the potential economic benefit of remote sensing t echnology for PGR use (Bethel, 2003). In the turf industry, research has been conducted to assess vari ous parameters of turfgrass stress using remotely sensed data. Several differe nt indices have been developed that have been correlated with plant stress. One of the most common indices to assess turfgrass stress is the Normalized Difference Vegetation Index (NDVI) whic h is computed as the reflectance from the Near Infrared (NIR) region minus the reflectance from the Red (R) region divided by a sum of both (NIR-R )/(NIR+R) (Rouse et al., 1973). Fenstemaker-Shaulis et al. (1997) found a negative correlation with NDVI and canopy temperature, (r2=0.74) and a positive correlation with NDVI and plant moisture content (r2=0.90). They also concluded that NDVI was not as a strong predictor of biomass (r2=0.37). Trenholm et al. (1999b) test ed various wavelengths 507 nm, 559 nm, 661 nm, 706 nm, 760 nm, 813 nm, and 935 nm as well as NDVI, IR/R (LAI) computed as

PAGE 65

65 935/661 nm, Stress1 computed as 706/760 nm, an d Stress2 computed as 706/813 nm. The results indicated that reflectance data was not successful in expl aining variability in the models for clipping yield. However, Kruse et al (2006) found conducting Pa rtial Least Squares regression on hyperspectral data (400,100nm) wa s an adequate predictor of biomass in creeping bentgrass ( Agrostis stolinifera L.) With the introduction of hyperspectral data, which typically measures more than 200 wavelengths, there has also been investigation of different statistical methods to quantify and interpret remotely sensed data. Traditionally, linear regression is the simplest way to understand variance. However, in interpreting reflectance data, it has been found that other regression techniques better assess plant stat us. Partial Least Squares Regre ssion (PLS) is a method used to predict variables when there are a large number of factors, typically more than the number of observations (Tobias, 1997). The use of multiple variables is generally tested through Multiple Linear Regression (MLR) (Tobias, 1997). Howe ver, the limitation with MLR is when the number of factors becomes too large the model may become over-fitting (Tobias, 1997). This is when it uses a large number of va riables but only a few account fo r most of the variation, also called latent factors (Tobias, 1997) Almost perfect models may be achieved; however they will not be highly predictive of new data. When pr ediction is the objective then PLS regression is a highly useful tool (Tobias, 1997). Currently, the majority of remote sensing research is done on agronomic food crops and forested areas. There is onl y limited research available in the area of remote sensing and turfgrass management (Trenholm et al., 1999b). If conditions across the entire golf course were uniform in microclimates, soil type s, turf varieties, pest densiti es, nutrient status, and irrigation performance there would be little need for advanced sensor systems to measure the variability of

PAGE 66

66 these parameters to better meet the needs on a locational basis (Stowell and Gelernter, 2006). However highly variable agronomic conditions create the need for further research to incorporate the use of remote se nsing into turfgrass management to reduce chemical inputs. (Trenholm et al., 1999b). The objective of this re search was to assess th e ability of remote sensing instruments to detect diffe rences in turf growth parameters as influenced by applications of PGRs on hybrid bermudagrass ( Cynodon dactylon X C. transvaalensis Princess 77) turf. Materials and Methods The field experiment was conducted at the Univ ersity of Florida, West Florida Research and Education Center, Jay, FL. Plots measured 1.5 m x 3.0 m and treatments were arranged in a factorial design observing three different PGRs, each at four incremental levels, 1, 2, and 4 times the labeled rates, with three replications per treatment. Plant growth regulators were applied on 14 June 2007 and 20 July 2007. Plan t growth regulators treatments included trinexapac-ethyl applied at 0.05, 0.1, 0.2, and 0.4 kg a.i. ha-1, ethephon applied at 3.8, 7.6, 15.2, 30.4 kg a.i. ha-1, and flurprimidol applied at 0.3, 0.6, 1.1, and 2.3 kg a.i. ha-1. Plots were mowed 2 to 3 times per week at a height of 1.2 cm with a Toro Reelmast er 3100-D. Supplemental irrigation was applied as needed. Data was co llected five times throughout the growing season in 2007 on 22 June, 12 July, 25 July, 22 Aug., and 4 Sept. When data were collected all measurements were taken within a one hour time period. Plots were rated visually for color, qual ity, and density based on the standard 1-9 National Turfgrass Evaluation Prog ram (NTEP) rating scale where 9 is the highest and 6 is the lowest acceptable rating. Soil volumetric water content readings were taken via Time Domain Reflectometry with a Fieldscout TDR 300 soil mo isture meter (Spectrum Technologies, Inc.). Clipping samples for biomass determin ation were collected from a 4.3 m2 area immediately after reflectance data and visual evaluations were obtained. Clipping tissue samples were weighed

PAGE 67

67 and oven dried at 52 C for 7d before being weighed agai n for determination of biomass production. Reflectance Measurements Spectral measurements were taken from two remote sensing devices. A ground based vehicle mounted optical sensor was used to asse ss turf health. A Crop Circle (Model ACS-210) (Holland Scientific) wa s fitted onto a Toro Greensmaste r 1000 walking greens mower 0.81 m above ground which provided a 0.45 m2 field of view. The Crop Circ le sensor produces it own light source at 650 nm and 880 nm and measures re flected values to produce an NDVI which is calculated as (880-650 nm)/(880+650 nm). No calib ration method was used for the sensor since it has the ability to detect its own light source from incoming ambient sunlight. A hand-held hyperspectral spectroradiomete r (FieldSpec Pro; Analytical Spectral Devices, Inc., Boulder, CO) was also used to co llect reflectance data. This radiometer has a spectral range of 300-2,500 nm with a 1 nm resolution and a 23 foreoptic. Reflectance readings were collected from shoulder height on each plot which provided a 0.65 m2 field of view. Radiance values are expressed in percent reflect ance compared to standardization with a white reference value across the entire spectral range. A white reference was used for calibration purposes at the beginning of data collection with the ASD unit, and again every 15 minutes depending on the length of time needed to coll ect data. Usually only one white reference was needed to take data on all plots. All reflectance readings were taken in full sunlight between the hours of 1100 and 1400 Central Standard Time (C ST) to minimize variance caused by incoming solar radiation. Statistical Analysis Analysis of variance (ANOVA) was perf ormed using the PROC GLM method (SAS Inst., 2003) to compare the differences among PG R treatments to biomass, soil VWC, visual

PAGE 68

68 ratings, Crop Circle readings (NDVI and LAI) and ASD readi ngs (NDVI, LAI, Stress1, and Stress2). Regression analysis was performed using the PROC REG method (SAS Inst., 2003) to correlate relationships between visual ratings Crop Circle readings, various individual reflectance wavelengths quantified from the ASD unit, and soil VWC. Additionally, due to the large amount of data generated by the ASD device, ASD data were rendered using partial least squares regression PROC PLS (SAS Inst., 2003) to soil VWC, visual quality ratings, and biomass production. Equations were validated th rough a single sample cross-validation. For cross-validation, ten percent of the sample wa s omitted for prediction purposes so that the number of factors chosen create s the minimal predicted residual sum of squares (PRESS). This process was repeated so that every observat ion was used exactly on ce for cross-validation. Results and Discussion Biomass ANOVA reveals that there was a significant reduction in biomass in the 0.1 kg a.i. ha-1 and 0.4 kg a.i. ha-1 rates of TE, the highest rate of flurprimidol (2.3 kg a.i. ha-1 ) and the low rate of ethephon (30.4 kg a.i. L ha-1) (Table 3-1). Although not alwa ys significant, a dose response trend was observed with TE (Figur e 3-1) and flurprimidol (Figur e 3-2). This dose response was not observed with Ethephon due to the fact th at ethephon is primarily used for seedhead suppression as opposed to biomass suppr ession in turf (Anonymous 2007). Biomass did not correlate well with any remote ly sensed parameter quantified from either device (Tables 3-3, 3-4, 3-5, and 3-6) This data is in agreement with the research conducted by Fenstemaker-Shaulis et al. (1997) and Trenholm et al. (1999b) who found re motely sensed data was not an adequate predictor of biomass. Th e inability to adequately model biomass reduction through increasing rates of PGR app lications may be attributed to a number of factors. PGRs have been found to limit vertical growth but incr ease prostrate growth, thus increasing density

PAGE 69

69 (Branham, 1997). In other research where biom ass has been adequately modeled, PGRs were not used (Kruse et al., 2005, 2006) and biomass was modeled from plots with increasing N rate applications applied at frequent intervals. Volumetric Water Content PGRs applied to bermudagrass in this study ha d only a slight effect on soil VWC. Only TE and flurprimidol applied at th e lowest rate gave significant differences from the untreated control (Table 3-1). Slight co rrelations were obser ved (Table 3-3) in modeling soil VWC with NIR reflectance quantified from the Crop Circle de vice as well as reflectance in the SWIR region at 2,132 nm (Table 3-4) and the 2,080-2,350 nm ra nge (Table 3-5) quantified from the ASD device. This is in agreement with research cond ucted by Ripple (1986) and Hutto et al. (2006). Visual Ratings Color Color rating means (Table 3-1) s how an increase in turf color by TE applications at 0.05 and 0.1 kg a.i. ha-1, flurprimidol applications at 0.3, 0.6, and 1.1 kg a.i. ha-1, and ethephon applications at the low rate compared to the untreated control, however, these increases were not significant (Table 3-1). This general increase could be at tributed to seed head reduction achieved by PGR applications, since seed heads are generally le ss green causing a waning in color on the turf. Additionally, increasing rates of PG Rs resulted in a trend towards reduced turf color which could be caused by a slight injury effect. Only slight correlations in visu al color ratings to remotely se nsed data from either device were achieved (Tables 3-3, 3-4, 3-5, 3-6). The highe st correlations were observed in data taken with the ASD device at wavelengths 661 nm (r2=0.42) and 510nm (r2=0.35), and ratio 605/515 nm (r2=0.41) (Table 3-4 and 3-5).

PAGE 70

70 Quality Improvements in quality were observed with PGR applications. A lthough they were not significant, all rates of TE and fl urprimidol, and the 3.8 kg a.i. ha-1 rate of ethephon slightly improved turf quality (Table 3-1). Similar to turf color ratings, only slight correlations existed between remotely sensed data a nd visual quality ratings. The hi ghest correlations were achieved by the ASD device at wavelengths 661 nm (r2=0.53) and 510 nm (r2=0.48), and ratio 915/975 nm (r2=0.42) (Table 3-4 and 3-5). Density Visual density ratings showed a general increase in density from all rates of TE and flurprimidol, and the low rate of ethephon, however these increases were not significant. Only slight correlations existed between remotely sens ed data and visual dens ity ratings. The highest correlations were achieved by the AS D device at wavelengths 661 nm (r2=0.53) and 510 nm (r2=0.48) (table 3-4). Reflectance Indices vs. PGR Application rates Data derived from the Crop Circle device s how a general increase in NDVI and LAI for all rates of TE and flurprimidol compared to the untreated, ho wever, only the 0.05 and 0.1 kg a.i. ha-1 rate of TE and the 0.3, 0.6, and 1.1 kg a.i. ha-1 rate of flurprimidol are significant (Table 32). Furthermore, NDVI and LAI derived from ASD data show the same trend, yet only the 2.2 kg ha-1 is statistically different from the untreated control (Table 3-2). Th e general trend for the STRESS 1 and STRESS 2 indices show similar re sults, however, STRESS 1 reveals that TE and flurprimidol treated turf was under less stre ss than the untreated control (Table 3-2). Crop Circle Device Reflectance data obtained from the Crop Circle device did not demonstrate its ability to detect biomass differences attributed to PGR applic ations (Table 3-3). Th e ability of the device

PAGE 71

71 to correlate with visual huma n evaluations was weak. Howeve r, the NIR (880 nm) reflectance values from the device slightly modeled changes in soil VWC (r2=0.43) throughout the experiment (Table 3-3). The inability of the Crop circle device to adeq uately detect differences in biomass, visual ratings, and VWC can be attributed to a number of things previously mentioned. Among these are the seedhead production of the turf which could have potentially skewed the reflectance data, and the lack of significant differe nces among the parameters tested. ASD Device Individual wavelengths, computed indices and various ratios obtained from ASD reflectance data did not demonstrate the ability to detect differences in biomass attributed to PGR applications. Some weak correlations exist betw een visual ratings and NDVI, LAI, Stress1, and Stress2 indices computed from the raw data generated by the ASD instrument (Table 3-3). Similar, and sometimes stronger, correlations exist with visual ratings with individual wavelengths investigated throughout the visible range (Table 3-4) Again, this inability to adequately model visual ratings, VWC, and biomass could be attributed to the lack of significant differences in the various parameters of the trea ted verses the untreated control and the seedhead production of the turf. PLS Regression The use of PLS regression on the hyperspectral data generated from the ASD instrument did not perform well in predicting differences in biomass created by various applications of PGRs. Furthermore PLS regression only slig htly predicted differences in soil VWC (r2=0.36) and visual quality (r2=0.26) (Table 3-6). This again is likely due to the lack of significant differences in soil VWC and visual qu ality from the untreated control.

PAGE 72

72 Conclusions This experiment shows the biomass reduction ex pected from increasing rates of PGRs in two of the three PGRs applied (Table 3-1) (Figure 3-1, 3-2). Flurprimidol and TE reduced biomass in response to increasi ng application rates compared to the untreated control; however, the results are not always signifi cant (Table 3-1) (Figure 3-1, 3-2) Similar trends were observed in NDVI and LAI indices computed from Crop Circ le and ASD instruments, as well as Stress1 and Stress2 indices computed from the ASD de vice (Table 3-2). Increases in NDVI and LAI which model plant health, as well as reductions in Stress1 and Stress2 indices, could be attributed to suppression of seed head production in addition to th e decrease of overall biomass. This overall decrease in biomass prevented the scalping of the tu rf in the study area that was observed in the untreated c ontrol and surrounding areas no t treated with PGRs. The inability to adequately model biom ass reduction, visual ra tings, and VWC through increasing rates of PGR applications may be attributed to a number of factors already mentioned. PGRs have been found to limit vertical growth but increase prostrate growth, thus increasing density (Branham, 1997). A general increase in de nsity was observed in this experiment with increasing application rates of TE and flurpr imidol, however, they were not significantly different from the untreated cont rol (Table 3-1). The lack of significant differences in density and all visual ratings affected the ability of th e remote sensing devices to detect differences among theses various parameters. Additionally, this experiment was performed on Princess 77 bermudagrass turf which produces seed heads. These seed heads are white in color and could potentially skew remotely sensed data compared to a turf that does not produce seed heads. Currently, turf managers apply PGRs in the form of broadcast applications to reduce biomass. If remote sensing technology could be used to adequately detect biomass, and was coupled with GPS and GIS technologi es, applications of PGRs coul d be applied on a site specific

PAGE 73

73 basis. In this setting, turf ma nagers could potentially reduce chem ical input thereby reducing the high cost of broadcast PGR applications.

PAGE 74

74 Table 3-1. Evaluation of means of biomass (g), volumetric water content (VWC), visual ratings: color, qual ity, density (rated o n NTEP 1-9 scale 9 is best a nd 6 is acceptable), from fairwa y height hybrid bermudagrass ( Cynodon dactylon X C. transvaalensis Princess 77) over five date s (22 June, 12 July, 25 July, 22 Aug., and 4 Sept. 2007) combined. Biomass (g) VWC Color Quality Density Trinexapac-ethyl 0.05 kg a.i. ha-1 52.47 a-c 35.73 a 8.00 ab 8.00 ab 7.87 a-c 0.1 kg a.i. ha-1 45.20 bc 34.45 a-c 8.10 ab 8.05 ab 8.00 ab 0.2 kg a.i. ha-1 47.20 a-c 32.87d 7.73 a-c 7.80 a-c 7.93 ab 0.4 kg a.i. ha-1 43.80 bc 34.20 b-d 7.83 a-c 7.98 ab 8.00 ab Flurprimidol 0.3 kg a.i. ha-1 53.93 a-c 35.85 a 8.10 ab 8.17 a 8.10 ab 0.6 kg a.i. ha-1 49.93 a-c 34.85 a-c 8.19 a 8.11 a 8.17 a 1.1 kg a.i. ha-1 48.13 a-c 35.23 ab 8.07 ab 8.08 a 8.10 ab 2.3 kg a.i. ha-1 36.27 c 34.72 a-c 7.83 ab 7.93 ab 7.83 a-c Ethephon 3.8 kg a.i. ha-1 45.73 bc 35.21 ab 7.96 ab 7.91 ab 7.80 a-c 7.6 kg a.i. ha-1 56.07 a-c 35.24 ab 7.63 bc 7.58 b-d 7.67 b-d 15.2 kg a.i. ha-1 66.13 ab 33.88 b-d 7.47 c 7.33 cd 7.40 cd 30.4 kg a.i. ha-1 66.27 ab 33.37 cd 7.40 c 7.25 d 7.30 d Untreated 71.60 a 34.01 b-d 7.87 a-c 7.77 a-c 7.73 a-d Means in the same column followed by the same lett er are not significantly different (LSD; P< 0.05)

PAGE 75

75 Table 3-2. Evaluation of means of Normalized Difference Vegeta tion Index (NDVI), Leaf Area In dex (LAI) computed from Crop Circle and ASD reflectance data, as well as Stress1 and Stress2 indices computed from ASD re flectance data from fairway height hybrid bermudagrass ( Cynodon dactylon X C. transvaalensis Princess 77) over five dates (22 June, 12 July, 25 July, 22 Aug., and 4 Sept. 2007) combined. ______Crop Circlez_____ _______________________ASDy__________________________ NDVIx LAIw NDVIx LAIw Stress1v Stress2u Trinexapac-ethyl 0.05 kg a.i. ha-1 0.66 ab 5.08 ab 0.76 ab 7.42 a-c 0.38 bc 0.79 cd 0.1 kg a.i. ha-1 0.67 ab 5.13 ab 0.77 ab 7.76 ab 0.36 c 0.77 cd 0.2 kg a.i. ha-1 0.65 a-d 4.90 a-c 0.76 ab 7.43 a-c 0.38 bc 0.78 cd 0.4 kg a.i. ha-1 0.66 a-c 5.00 a-c 0.76 ab 7.46 a-c 0.37 bc 0.78 cd Flurprimidol 0.3 kg a.i. ha-1 0.67 ab 5.10 ab 0.77 ab 7.51 ab 0.37 bc 0.78 cd 0.6 kg a.i. ha-1 0.66 ab 5.01 ab 0.76 ab 7.34 a-d 0.37 bc 0.78 cd 1.1 kg a.i. ha-1 0.68 a 5.24 a 0.78 a 7.82 a 0.36 c 0.77 d 2.3 kg a.i. ha-1 0.66 a-c 5.03 ab 0.76 ab 7.49 a-c 0.37 bc 0.78 cd Ethephon 3.8 kg a.i. ha-1 0.65 a-d 4.82 a-c 0.75 a-c 7.16 a-d 0.39 a-c 0.78 b-d 7.6 kg a.i. ha-1 0.64 b-d 4.64 b-d 0.75 a-d 6.96 a-d 0.39 a-c 0.78 a-d 15.2 kg a.i. ha-1 0.62 de 4.45 dc 0.72 cd 6.50 cd 0.41 a 0.79 ab 30.4 kg a.i. ha-1 0.60 e 4.23 d 0.72 d 6.39 d 0.42 a 0.80 a 13) Untreated 0.63 c-e 4.64 b-d 0.74 b-d 6.81 b-d 0.40 ab 0.79 a-c Means in the same column followed by the same le tter are not significantly different (LSD; P< 0.05) z Model ACS-210, Holland Scientific y FieldSpec Pro; Analytical Spectr al Devices, Inc., Boulder, CO x NDVI= (880-650nm)/(880+650nm) w LAI= 880/650nm v Stress1= 706/760nm u Stress2= 706/813nm

PAGE 76

76 Table 3-3. Coefficient estimates (r2) of Crop Circle reflectance data, Normalized Differen ce Vegetation Index (NDVI), Leaf Area Index (LAI), 880nm, and 650nm, and, ASD reflectance data computed into NDVI LAI Stress1, and Stress2 indices to volumetric water content (VWC) visual co lor quality, density ratings, and bioma ss production from fair way height hybrid bermudagrass ( Cynodon dactylon X C. transvaalensis Princess 77) averaged over five dates (22 June, 12 July, 25 July, 22 Aug., and 4 Sept. 2007). Crop Circlez ASDy r2 NDVIx LAIw 880nm 650nm NDVIx LAIw Stress1v Stress2u VWC 0.10*** 0.12*** 0.43*** 0.00 0.11*** 0.10*** 0.09*** 0.10*** Color 0.18*** 0.16*** 0.00 0.28*** 0.23*** 0.22*** 0.19*** 0.21*** Density 0.22*** 0.19*** 0.03* 0.21*** 0.28*** 0.23*** 0.24*** 0.25*** Quality 0.22*** 0.18*** 0.02 0.24*** 0.29*** 0.24*** 0.23*** 0.26*** Biomass 0.06** 0.10*** 0.00 0.09*** 0.03* 0.08*** 0.05* 0.05*** *, **, *** Significant at the 0.05, 0.01, and 0.001 probability levels, respectively z Model ACS-210, Holland Scientific y FieldSpec Pro; Analytical Spectr al Devices, Inc., Boulder, CO x NDVI= (880-650nm)/(880+650nm) w LAI= 880/650nm v Stress1= 706/760nm u Stress2= 706/813nm

PAGE 77

77 Table 3-4. Coefficient estimates (r2) of ASD reflectance data at in dividual wavelengths, as well as range averages from 630-690nm and 2080nm to volumetric water content (VWC) visual colo r, quality, density ratings, and biomass production from fairway height hybrid bermudagrass ( Cynodon dactylon X C. transvaalensis Princess 77) averaged over five dates (22 June, 12 July, 25 July, 22 Aug., and 4 Sept. 2007). All numbe rs in column headings are expressed nanometers (nm). ___________________________________________ASD z (nm)____________________________________________ r2 510 535 545 550 661 735 755 813 935 2132 VWC 0.12*** 0.06** 0.05* 0.04* 0.07*** 0.02* 0.04* 0.06* 0.11*** 0.43*** Color 0.35*** 0.25*** 0.25*** 0.25*** 0.42*** 0.05* 0.01 0.01 0.02* 0.11*** Density 0.51*** 0.39*** 0.38*** 0.39*** 0.53*** 0.07* 0.01 0.01 0.02 0.06** Quality 0.48*** 0.35*** 0.35*** 0.35*** 0.53*** 0.06* 0.01 0.01 0.02 0.08*** Biomass 0.00 0.00 0.00 0.01 0.01 0.02 0.03* 0.04* 0.05* 0.01 *, **, *** Significant at the 0.05, 0.01, and 0.001 probability levels, respectively z FieldSpec Pro; Analytical Spectr al Devices, Inc., Boulder, CO

PAGE 78

78 Table 3-5. Coefficient estimates (r2) of range averages from 630-690 nm and 2080 nm and ratios 605/515 nm, 915/975 nm, and 865/725 nm to volumetric water content (VWC ) visual color, quality, density rati ngs, and biomass production from fairway height hybrid bermudagrass ( Cynodon dactylon X C. transvaalensis Princess 77) averaged ove r five dates (22 June, 12 July, 25 July, 22 Aug., and 4 Sept. 2007). All numbers in column headings are expressed nanometers (nm). ___________________________________ASD z (nm)______________________________________ r2 630-690 2080-2350 605/515 nm 915/975 nm 865/725 nm VWC 0.19*** 0.40*** 0.12** 0.14** 0.03 Color 0.19*** 0.12*** 0.41*** 0.32*** 0.24*** Density 0.30*** 0.08*** 0.32*** 0.39*** 0.22*** Quality 0.29*** 0.10 0.26*** 0.42*** 0.23*** Biomass 0.00 0.01 0.08* 0.06* 0.13** *, **, *** Significant at the 0.05, 0.01, and 0.001 probability levels, respectively z FieldSpec Pro; Analytical Spectr al Devices, Inc., Boulder, CO

PAGE 79

79 Table 3-6. Partial Least Squares regression co efficients on hyperspectral data from ASD device to predict volumetric water content (VWC), biomass, and visual quality (rated on NTEP 1-9 scale 9 is best and 6 is acceptable) in fairway height bermudagrass ( Cynodon dactylon X C. transvaalensis Princess 77) averaged over five date s (22 June, 12 July, 25 July, 22 Aug., and 4 Sept. 2007). Calibration No. of Factorsy r2 SE n VWC 5 0.36 0.05 195 Biomass 2 0.03 0.05 195 Visual Quality 8 0.26 0.03 195 y Number of factors required to achieve a mi nimal Predicted Residual Sum of Squares (P RESS) of prediction for the partial least squares regression model. z FieldSpec Pro; Analytical Spectr al Devices, Inc., Boulder, CO

PAGE 80

80 Figure 3-1. Biomass measurements collected from fairway height hybrid bermudagrass ( Cynodon dactylon X C. transvaalensis Princess 77) treated with trinexapac -ethyl at 0.05, 0.1, 0.2, and 0.4 kg a.i. ha-1. Plots measured 1.5 X 3.0 m with three replications. Means for biomass measurements was comput ed using Fishers protected LSD (p< 0.05) from data over 5 dates (22 June, 12 July, 25 July, 22 Aug., and 4 Sept. 2007) combined. 0 10 20 30 40 50 60 70 80 Untreated 0.05 0.1 0.2 0.4Rate (kg a.i. ha1 ) Biomass (g)

PAGE 81

81 Figure 3-2. Biomass measurements collected from fairway height hybrid bermudagrass ( Cynodon dactylon X C. transvaalensis Princess 77) treated with flurprimidol at 0.3, 0.6, 1.1, and 2.3 kg a.i. ha-1. Plots measured 1.5 X 3.0 m with three replications. Means for biomass measurements was comput ed using Fishers protected LSD (p< 0.05) from data over 5 dates (22 June, 12 July, 25 July, 22 Aug., and 4 Sept. 2007) combined. 0 10 20 30 40 50 60 70 80 Untreated 0.30.61.12.3 Rate (kg a.i. ha-1) Biomass (g)

PAGE 82

82 CHAPTER 4 CONCLUSIONS Currently many turf managers and researchers still use time and labor intensive techniques to manage and assess turf that originate from d ecades ago. Many of these management practices have been proven repeatedly to work in a variety of situations to assess turfgrass stress, but may be time consuming and inconsistent. However, th ere are new developments in ways of assessing and mapping stress that increase the efficiency by which one manages an d assesses turf. With increasing water and other environmental restrict ions turf managers and researchers need be aware of technological advan ces in their industry. The research discussed in this thesis give s greater insight into the use of some the technological advancements available to turfgr ass managers demonstrat ing the sensitivity of remote sensing instrumentation to various parame ters of turfgrass stress. In conducting these experiments many realizations we re made about the usability a nd future potential for remote sensing technology in the turfgr ass management field. Some recommendations for anyone who wishes to continue the research referr ed to in this thesis are as follows: Irrigation X Nitrogen study. Conduct the experiment in th e earlier part of the growing season, when rainfall is not so prevalent. This will potentially reduce un iformity in soil VWC. Also, apply N every two weeks as described by Kr use et al. (2005); this will continually promote differences in growth. In addition, investig ate more into the use of PLS regression on hyperspectral data to formulate predic tion models for N and soil VWC levels. PGR study. Conduct the experiment on turfgrass that does not produce seedheads. In this experiment seed heads could have potentially skewed remotely sensed data; this will at least take that factor out. Also, obtain remote sensing da ta before and after PGR a pplications to assess the differences they create. From this data possi bly try to make applications based on remotely

PAGE 83

83 sensed data and not just on an in cremental scale. In addition, i nvestigate more into the use of PLS regression on hyperspectral data to formul ate prediction models for assessment of the biomass of the turf. Crop circle. Test in various light and ambient conditions to assess different turfgrass parameters and investigate its usability in an assortment of environmental settings. Turfgrass managers work in all conditions, rain or shine. An assessment of how these devices will work in varying environmental conditions must be evalua ted before they are employed into real world applications of turfgrass management. Beyond these experiments. Conduct experiments on larger scale while adding other technologies such as GPS, GIS, and VRA. Th is will allow researcher s to better understand the usability of the technology and th e logistics of operating it as a turfgrass manager. Remote sensing in a large scale setti ng is not very eff ective without the co mbination of other technologies. You need GPS and GIS to georefer ence remotely sensed data to understand where the data is coming from. Then by applying VR A technology some sort of economic impact can be assessed to understand the potential benefit of the use all of these technologies. Remote sensing is just the beginning of a long road of investigation in helping turfgrass managers reduce irrigation and chemical use and appl y these inputs only where needed. Currently the use of remote sensing technol ogy in the turfgrass ma nagement industry is still in its infancy. However, the experiments desc ribed in this thesis al ong with other research that has been referenced show it has potential to assist turfgrass managers to make better decisions about managing their turf Presently, this technology is not in a very usable form and much research still needs to be conducted to understand its place in the field of turfgrass management.

PAGE 84

84 APPENDIX A EXPERIMENTAL LAYOUT FOR DETECTION OF LEAF NITROGEN CONCENTRATION AND SOIL VOLUMETRIC WATER CON TENT USING GROUND-BASED REMOTE SENSING TECHNOLOGY

PAGE 85

85 101 201 301 401 102 202 302 402 103 203 303 403 104 204 304 404 501 601 701 801 502 602 702 802 503 603 703 803 504 604 704 804 901 1001 1101 1201 902 1002 1102 1202 903 1003 1103 1203 904 1004 1104 1204 Figure A-1. Experimental layout for detecti on of leaf nitrogen concentration and soil volumetric water content using ground-bas ed remote sensing technology

PAGE 86

86 Table A-1. Description of treatments for Figure A-1 Whole Plot Subplot Rep Treatment # Irrigation Nitrogen 1 2 3 1 60% ET 0 kg ha-1 101 704 1201 2 60% ET 25 kg ha-1 102 702 1203 3 60% ET 50 kg ha-1 103 703 1202 4 60% ET 100 kg ha-1 104 701 1204 5 80% ET 0 kg ha-1 201 802 901 6 80% ET 25 kg ha-1 202 803 903 7 80% ET 50 kg ha-1 203 804 904 8 80% ET 100 kg ha-1 204 801 902 9 100% ET 0 kg ha-1 301 503 1003 10 100% ET 25 kg ha-1 302 502 1002 11 100% ET 50 kg ha-1 303 501 1004 12 100% ET 100 kg ha-1 304 504 1001 13 120% ET 0 kg ha-1 401 602 1101 14 120% ET 25 kg ha-1 402 601 1104 15 120% ET 50 kg ha-1 403 603 1102 16 120% ET 100 kg ha-1 404 604 1103

PAGE 87

87 APPENDIX B DAILY RAINFALL AND ET DATA FOR JAY, FL Table B-1. Daily Rainfall and ET data for Jay, FL Day Rainfall (cm)a ET (cm)a Cumulative Difference (cm) Irrigation Applied (cm) June 14 0.38 0.48 (0.10) 0.20b 15 0.00 0.41 (0.31) 0.00 16 0.51 0.38 (0.18) 0.00 17 2.29 0.53 1.57 0.00 18 0.00 0.33 1.24 0.00 19 2.36 0.30 3.30 0.00 20 1.27 0.48 4.09 0.00 21 0.00 0.58 3.50 0.00 22 0.00 0.53 2.97 0.00 23 0.00 0.46 2.51 0.00 24 0.00 0.30 2.21 0.00 25 0.00 0.36 1.85 0.00 26 0.10 0.36 1.60 0.00 27 0.00 0.36 1.24 0.00 28 0.25 0.46 1.04 0.00 29 0.03 0.46 0.61 0.00 30 0.38 0.53 0.45 0.00 July 1 0.00 0.53 (0.08) 0.00 2 0.00 0.43 (0.51) 0.00 3 0.00 0.30 (0.82) 0.00 4 0.89 0.33 (0.26) 0.00 5 0.94 0.33 0.35 0.00 6 0.46 0.48 0.33 0.00 7 0.00 0.48 (0.16) 0.00 8 0.76 0.43 0.17 0.00 9 0.00 0.53 (0.36) 0.00 10 0.00 0.48 (0.84) 0.00 11 0.00 0.48 (1.32) 0.00 12 0.89 0.48 (0.92) 0.00 13 0.00 0.46 (1.37) 0.00 14 0.00 0.33 (1.71) 0.00 15 2.79 0.30 0.78 0.00 16 0.10 0.33 0.56 0.00 17 0.71 0.51 0.76 0.00 18 0.00 0.48 0.28 0.00 19 0.00 0.46 (0.18) 0.00 20 0.00 0.46 (0.64) 0.20b 21 1.40 0.53 0.43 0.00

PAGE 88

88 Table B-1. Continued Day Rainfall (cm)a ET (cm)a Cumulative Difference (cm) Irrigation Applied (cm) 23 0.53 0.41 0.15 0.00 24 3.35 0.43 3.07 0.00 25 0.28 0.46 2.89 0.00 26 0.00 0.36 2.53 0.00 27 0.00 0.48 2.05 0.00 28 0.00 0.51 1.54 0.00 29 0.00 0.46 1.09 0.00 30 2.54 0.41 3.22 0.00 31 0.00 0.43 2.79 0.00 August 1 2.62 0.33 5.07 0.00 2 0.00 0.43 4.64 0.00 3 0.15 0.46 4.34 0.00 4 0.00 0.46 3.88 0.00 5 0.00 0.46 3.42 0.00 6 0.00 0.43 2.99 0.00 7 0.00 0.48 2.51 0.00 8 0.00 0.43 2.08 0.00 9 0.00 0.36 1.72 0.00 10 0.41 0.48 1.64 0.00 11 0.03 0.38 1.29 0.00 12 0.84 0.25 1.87 0.00 13 0.00 0.43 1.44 0.00 14 0.00 0.48 0.96 0.00 15 0.00 0.46 0.50 0.00 16 0.00 0.46 0.04 0.00 17 0.00 0.43 (0.39) 0.00 18 0.33 0.43 (0.49) 0.00 19 0.00 0.43 (0.92) 0.00 20 0.00 0.48 (1.40) 0.00 21 0.00 0.48 (1.89) 0.00 22 0.00 0.51 (2.39) 0.00 23 0.00 0.38 (2.78) 0.00 24 0.00 0.38 (3.16) 0.00 25 0.05 0.46 (3.56) 2.54 26 4.34 0.33 2.99 0.00 27 0.28 0.33 2.94 0.00 28 0.89 0.20 3.63 0.00 29 0.33 0.43 3.52 0.00 30 0.00 0.41 3.12 0.00 31 0.08 0.28 2.91 0.00 September 1 0.05 0.33 2.64 0.00

PAGE 89

89 Table B-1. Continued Day Rainfall (cm)a ET (cm)a Cumulative Difference (cm) Irrigation Applied (cm) 3 2.34 0.28 4.64 0.00 4 1.32 0.41 5.56 0.00 a Based on data from http://fawn.ifas.ufl.edu/ b Irrigation applied to water in fertilizer application

PAGE 90

90 APPENDIX C CUMALITIVE WEEKLY RAINFALL AND ET DATA FOR JAY, FL Table C-1. Cumalitive weekly rain fall and et data for Jay, Fl ______________Cumulative________________ Week Ending On Rainfall (cm)a ET (cm)a Difference (cm) Irrigation Applied (cm) June 15 0.38 0.89 (0.31) 0.00 22 6.43 3.15 2.97 0.00 29 0.38 2.74 0.61 0.00 July 6 2.67 2.95 0.33 0.00 20 3.61 2.87 (0.64) 0.20b 27 5.56 3.07 2.05 0.00 August 3 5.31 3.02 4.34 0.00 10 0.41 3.10 1.64 0.00 17 0.86 2.90 (0.39) 0.00 24 0.33 3.10 (3.16) 2.54 31 5.97 2.44 2.91 0.00 a Based on data from http://fawn.ifas.ufl.edu/ b Irrigation applied to water in fertilizer application

PAGE 91

91 APPENDIX D MOTHLYRAINFALL AND ET AVERAGES FROM 2003-2006 FOR JAY, FL Table D-1. MothlyRainfall and ET averages from 2003-2006 for Jay, FL Rainfall (cm) ET (cm) Rainfall-ET (cm) Cumulative Total Rainfall-ET (cm) January 7.2898 4.92125 2.36855 2.36855 February 13.6144 5.5118 8.1026 10.47115 March 11.91895 8.46455 3.4544 13.92555 April 15.3543 10.4775 4.8768 18.80235 May 7.8105 12.5984 -4.7879 14.01445 June 24.08555 12.192 11.89355 25.908 July 24.31415 12.5984 11.71575 37.62375 August 20.2438 12.2047 8.0391 45.66285 September 16.0401 10.0965 5.9436 51.60645 October 5.6896 7.28345 -1.59385 50.0126 November 11.99515 4.7625 7.23265 57.24525 December 10.69975 3.34645 7.3533 64.59855 Based on data from http://fawn.ifas.ufl.edu/

PAGE 92

92 LIST OF REFERENCES Anonymous. 2007. Proxy (specimen label). Bayer Scie ntific, Triangle Park, NC. U.S. Patent 4,240,819. Date issued: 25 Apr. 2005. Aston, A.R., and C.H.M. Van Bavel. 1972. Soil Su rface Water Depletion and Leaf Temperature. Agron. J. 64:368-373. Augustin, B.J., and G.H. Snyder. 1984. Moisture Se nsor Controlled Irri gation For Maintaining Bermudagrass Turf. Agron. J. 76:848-850. Bahrun, A., V.O. Mogensen, and C.R. Jensen. 2003. Water Stress Detec tion in Field Grown Maize by Using Spectral Vegetation Index. Co mmun. Soil Sci. Plant Anal. 34:65-79. Barret, J. 2003. Golf Course irrigation: Envi ronmental Design and Management Practices. 413417. John Wiley & Sons, Inc., Hoboken, NJ. 413-417. Bethel, M., T. Gress, S. White, J. Johnson, T. Sheely, B. Roberts, N. Gat, G. Scriven, G. Hagglund, M. Paggi, and N. Groenenberg. 2003. Im age-Based Variable Rate Plant Growth Regulator Application in Cotton at Sheely Farms in California. 6-10 Jan. 2003. Beltwide Cotton Conferences, Nashville, TN. Bell, G.E., B.M. Howell, G.V. Johnson, W.R. Raun, M.L. Stone, and J.B. Solie. 2004. Optical Sensing of Turfgrass Chlorophyll Content and Tissue Nitrogen. Hortic. Sci. 39(5):11301132. Bell, G.E., D.L. Martin, R.M. Kuzmic, M.L. Stone, and J.B. Solie. 2000. Herbicide Tolerance of Two Cold-Resistant Bermudagrass ( Cynodon spp.) Cultivars Determined by Visual Assessment and Vehicle Mounted Optic al Sensing. Weed Tech. 14:635-641. Bell, G.E., D.L. Martin, M.L. Stone, J.B. Solie, and G.V. Johnson. 2002a. Turf area Mapping Using Vehicle Mounted Optical Sensors. Crop Sci. 42:648-651. Bell, G.E., D.L. Martin, S.G. Wiese, D.D. D obsom, M.W. Smith, M.L. Stone, and J.B. Solie. 2002b. VehicleMounted Optical Sensing An Objective Means for Evaluating Turf Quality. Crop Sci. 42:197-201. Blackmer, T.M., J.S. Schepers, and G. E. Varv el. 1994. Light reflectan ce Compared with Other Nitrogen Stress Measurements in Co rn Leaves. Agron. J. 86: 934-938. Blackmer, T.M., J.S. Schepers, G.E. and Varv el, E.A. Walter-Shea. 1996.Nitrogen Deficiency Detection Using Reflected Shortwave Radiat ion From Irrigated Corn Canopies. Agron. J. 88: 1-5. Branham, Bruce. Plant Growth Regulators for Fine Turf Use. http://www.interactiveturf.com/tips/1997_02.htm 24 Oct. 2007.

PAGE 93

93 Brown, K.W., R.L. Duble, and J.C. Thomas. 1977. Influence of Management and Season on Fate of N Applied to Golf Greens. Agron. J. 69:667-671. Ehrler, W.L., S.B. Idso, R.D. Jackson, and R.J. Reginato. 1978. Wheat canopy temperature: Relation to Plant Water Pote ntial. Agron. J. 70: 251-256. Fenstermaker-Shaulis, L.K., A. Leskys, and D.A. Devitt. 1997. Utilization of Remotely Sensed Data to Map and Evaluate Turfgrass Stress A ssociated with Drought. Journal of Turfgrass Management. 2:65-81. Geeske, Joel, J.A. Gamon, and C.B. Field. 1997. Pr oduction Efficiency in Sunflower: The Role of Water and Nitrogen Stress. Remote Sens. Environ. 62:176-188. Hopkins, W. G. 1999. Introduction to Plant Physiology. 2nd ed. John Wiley and Sons Inc. New York, Ch. 22 The physiology of plants under stress. 453-457. Horst, G.L, J.C. OToole, and K.L. Faver. 1989. Seasonal and Species Variation in Baseline Functions for Determining Crop Water Stress Indices in Turfgrass. Crop Sci. 29:12271232. Hutto, K.C., R.L King, J.D. Byrd, and D.R. Shaw. 2006. Implementation of Hyper spectral Radiometry in Irrigation Management of Creeping Bentgrass Putting Greens. Crop. Sci. 46:1564-1569. Idso, S.B., R.J. Reginato, D.C Reicosky, and J.L. Hatfield. 1981. Determin ing Soil Induced Plant Water Potential Depressions in Alfalfa by M eans of Infrared Thermometry. Agron. J. 73: 826-830. Ikemura, Y., and B. Leinauer. 2006. Remote Se nsing Technology Detects Turfgrass Stress. Golfdom, May, p.63-66. Jalali-Farahani, H.R., A.D. Matthias, and D.C. Slack. 1993. Crop Water Stress Index Models for Bermudgrass Turf: A Comparison. Agron. J. 85:1210-1217. Jalali-Frahani, H.R., D.C. Slack, D.M. Kopec, A.D. Matthias, and P.W. Brown. 1994. Evaluation of Resistances for Bermudagrass Turf Crop Wa ter Stress Index Models. Agron. J. 86: 574581. Kruse, J.K., N.E. Christians, and M.H. Chaplin. 2005. Remote Sensing of Phosphorus Deficiencies in Agrostis Stolonifera. Inter. Turfgra ss Soc. Res. J. 10:923-928. Kruse, J.K., N.E. Christians, and M.H. Chap lin. 2006. Remote Sensing of Nitrogen Stress in Creeping Bentgrass. Agron. J. 98:1640-1645. Lee, Wonsuk, S.W. Searcy, and T. Kataoka. 1999 Assessing Nitrogen Stress in Corn Varieties of Varying Color. 18-21 July. ASAE A nnual International Meeting. Toronto, Ontario, Canada.

PAGE 94

94 Lillesand, T.M., and R.W. Keifer. 2000. Remote Sensing and Image Interpretation. 4th ed. John Wiley & Sons, New York. Martin, D.L, C.S. Throssel, and D.J. Wehner. 1994. Models for Predicting Lower Limit of The Canopy-Air Temperature Difference of Two C ool Season Grasses. Crop Sci. 34: 192-198. Min, M. and W.S. Lee. 2003. Spectral Based Nitrogen Sensing for Citrus. 27-30 July. ASAE Annual International Meeting. Las Vegas, NV Morgan, W.C., J. Letey, S.J. Richards, and N. Valoras. 1966. Physical Soil Amendments, Soil Compaction, Irrigation, and Wetting Agents in Turfgrass Management Effects on Compactibility, Water Infiltration Rates, Eva potranspiration, and Number of Irrigations. Agron. J. 58:525-527. Morris, K.N. A Guide to NTEP Turfgrass Ratings. http://www.ntep.org/reports/ratings.htm. 14 Nov. 2007 Mylavarapu, R.S. and E.D. Kennelly. UF/IFA S Extentsion Soil Testi ng Laboratory (ESTL) Analytical Procedures and Training Manual1. http://edis.ifas.ufl.edu/pdffiles/SS/SS31200.pdf 14 Dec. 2007. Osborne, S.L., J.S. Schepers, D.D. Francis, a nd M.R. Schlemmer. 2002. De tection of Phosphorus and Nitrogen Deficiencies in Corn Using Spectral Radiance Measurements. Agron. J. 94:1215-1221. Ripple, W.J. 1986. Spectral Reflectance Relationsh ips to Leaf Water Stress. Photogramm. Eng. Rem. Sens. 52:1669-1675 Rouse, J.W., R.H. Haas, J.A. Schell, and D.W. Deering. 1973. Monitoring the vernal advancement and retroradation (green wave effect) of Natural Vegetation. Prog. Rep. SRC 1978-1, Remote Sensing Center, Texas A&M Univ., College Station, 93p. (NTIS) No. E73-106393. Salisbury, F.B. and C.W. Ross. 1992. Plant Philology. 4th ed. Wadsworth Publ. Co. Belmont, CA. SAS Institute. 2002-2003. The SAS System for Windows. Release 9.1. SAS Inst., Cary, NC. Slack, D.C., K.M. Geiser, K.W. Stange, and E.R. Allred. 1981. Irrigation Scheduling in SubHumid Aareas with Infrared Thermometr y. p. 116-124. In Proc. Irrigation Scheduling Conf., Chicago, and IL. 14-15 Dec. 1981. ASAE Annual International Meeting, St Joseph, MI. Starks, P. J., D. Zhao, W.A. Phillips, a nd S.W. Coleman. 2006. Development of Canopy Reflectance Algorithms for Real-Time Predicti on of Bermudagrass Pasture Biomass and Nutritive Values. Crop Sci. 46:927-934.

PAGE 95

95 Stowell, L. and W. Gelernter. 2006. Sensing th e Future. Golf Course Management. March p.107110. Throssel, C.S., R.N. Carrow, and G.A. Milliken. 1987. Canopy Temperature Based Irrigation Scheduling Indices for Kentucky Bl uegrass Turf. Crop Sci. 27:126-131. Tobias, R.D. 1997. An Introduction to Partial Leas t Squares Regression (Online). Available at SAS institute http://ftp.sas.com/techsup/download/technote/ts509.pdf 17 April 2007. Trenholm, L. E., R.N. Carrow, and R.R. Duncan. 1999b. Relationship of Multispectral Radiometry Data to Qualitative Data in Turfgrass Research. Crop Sci. 39:763-769. Trenholm, L.E., R.R. Duncan, and R.N. Carrow. 1999a. Wear Tolerance, Shoot Performance, and Spectral Reflectance of Seashore Pa spalum and Bermudagrass. Crop Sci. 39:1147:1152. Trenhholm, L.E., M.J. Schlossberg, G. Lee, a nd W. Parks. 2000. An Evaluation of Multispectral Responses on Selected Turfgrass Specie s. Int. J. Remote Sensing. 21:709-721. Turgeon. A.J. 2008. Turfgrass Management. 8th edition. Pearson Education Inc. Upper Saddle River, NJ. Unruh, J. and M.L. Elliot. 1999. Best Manageme nt Practices for Florida Golf Courses. University of Florida., Gainesville. VanBibber, L. 2006. Putting the Numbers to PGR's. http://groundsmag.com/chemicals/grounds_maintenance_putting_numbers_pgrs/. 24 Oct. 2007 Yoder, B. J. and R.E. Pettigree-Crosby. 1995. Predicting Nitrogen and Chlorophyll Content and Concentrations From Reflectance Spectra ( 400-2500) at Leaf and Canopy Scales. Remote Sens. Environ. 53:199-211. Xiong, X, G.E. Bell, J.B. Solie, M.W. Smith, and B. Martin. 2007. Bermudagrass Seasonal Responses to Nitrogen Fertilization and Irrigat ion Detected Using Op tical Sensing. Crop Sci. 47:1603-1610.

PAGE 96

96 BIOGRAPHICAL SKETCH Jason Hamilton Frank was born in 1982, in Deland, FL. He lived there until 2001when he graduated from Deland high school and was accepted to the University of Central Florida. He spent a year there as a business administration ma jor before realizing he wanted to attend the University of Florida. He transferred to Daytona Beach Community College in August of 2002 and received his A.A. degree in Ju ly of 2003. He then transferred to the University of Florida in August of 2003 where he completed his B.S. in turfgrass science with a minor in business administration in December of 2005. He then went on to graduate school, also at the University of Florida, where he will gradua te with his M.S. in horticultu re sciences with a minor in agricultural and biologic al engineering and in terdisciplinary concen tration in geographic information sciences in May of 2008. Upon graduation Jason has accepted a 2nd assistant superintendent position at Royal Po inciana Golf Club in Naples, FL.