(1)
(2)
where TM3 and TM4 are reflectance (in decimals) in red and NIR portions of the EM spectrum,
respectively, and L is a coefficient that decreases with the increase in vegetation density. As per
suggestion of Huete (1988), a constant L value of 0.5 was used in this study. Although NDVI
has been extensively used in turfgrass studies, its value is sensitive to the changes in surface
wetness. To account for this issue, SAVI was developed in a fashion to be more resilient to
surface wetness variation (Huete 1988), a characteristic that was confirmed by numerous studies
(Glenn et al., 2011).
2.3 Crop Water Stress Index
Crop Water Stress Index (CWSI) approach was developed in 1981, when Idso et al. (1981) and
Jackson et al. (1981) proposed the empirical and theoretical methods of estimating CWSI,
respectively. According to Idso et al. (1981), CWSI can be estimated using equation 3.
(3)
where dT is the temperature difference between canopy and air (T
canopy
T
air
) and subscripts m,
LL, and UL represent measured, lower limit, and upper limit of dT, respectively. Since all
variable have the same units, CWSI is a dimensionless ratio. The lower limit of dT occurs under
non-water-stressed conditions when ET is only limited by atmospheric demand. On the other
hand, the upper limit of dT is reached under non-transpiring conditions when ET is stopped due
to the lack of water. Idso et al. (1981) proposed that under non-water-stressed conditions the
lower dT limit is a linear function of the air vapor pressure deficit (VPD, kPa):
dT
LL
= a + b VPD (4)
where a is the intercept and b is the slope of the linear relationship. Similarly, the upper
limit can be expressed as a linear function of vapor pressure gradient (VPG, kPa).
dT
UL
= a + b VPG (5)
where a and b are the same coefficients and VPG is the difference between saturated vapor
pressure at air temperature and at a higher temperature equal to air temperature plus the
coefficient a. dT
UL
can also be determined by measuring dT over a severely stress plant area
(turfgrass in this case). To verify the dT
UL
concept a grass plot was sprayed to bring ET to zero.
Page 7
In this study, dT
m
was calculated using turfgrass canopy temperature detected by the mobile or
handheld IRT (IRt/c.2) sensor and air temperature measured at about the same time at the on-site
weather station (Site C3). The lower limit of dT was determined by plotting VPD data vs. dT
m
values that were collected over healthy (non-stressed) turfgrass within two days after an
irrigation or precipitation event. The upper limit of dT was estimated using the proposed method
of Idso et al. (1981). In addition, estimated dT
UL
values were compared with the actual dT
measurements over a severely stressed (non-transpiring) patch of turfgrass to validate the
accuracy of this approach.
2.4 Evapotranspiration
Turfgrass ET was estimated based on two independent approaches. The first approach was
based on the CWSI concept. Jackson et al. (1981) showed that there is a unique mathematical
relationship between CWSI and the ET of the studied vegetative surface. The equation derived
by Jackson et al. (1981) can be rearranged into the following format:
ET
a
= (1 CWSI) ET
p
(6)
where ET
a
is actual ET and ET
p
is potential ET (grass-based reference ET, ET
o
in this case).
Equation 6 was used as the first approach to estimating grass ET using ground-based and
airborne remote sensing derived CWSI. The second implemented approach was a remote
sensing energy balance (RSEB) model known as SEBAL (Surface Energy Balance Algorithm for
Lands, Bastiaanssen et al. 1998). Similar to other RSEB models, SEBAL is based on the simple
form of energy balance equation at studied surfaces:
R
n
= G + H + LE (7)
where R
n
is net radiation, G is soil heat flux, H is sensible heat flux, and LE is the latent heat flux
in units of energy (W m
-2
) or depth of water (mm day
-1
). The R
n
, G, and H components are
modeled by integrating remotely sensed and in-situ data, and LE is calculated as the residual of
the above equation. SEBAL offers an innovative method for calculating H, in which spatially
distributed values are approximated iteratively by identifying two extreme pixels. One of the
extreme pixels is a hot/dry pixel (e.g., bare soil), where all of the available energy is used for
heating the soil and the air above the surface. As a result, the value of H over this pixel is equal
to the available energy (H = R
n
G). The other extreme pixel is a cold/wet pixel (e.g., water),
where the available energy is used in transforming the physical state of water from liquid to
vapor, resulting in a negligible H (H = 0). The value of H over all other pixels could be
interpolated between these two extreme conditions. In this study, the SEBAL model was applied
to the ground-based remote sensing data.
Page 8
3. Results and discussion
3.1 Remote sensing data
Remotely sensed data collected by the MSR5-IRt/c.2 sensor had values similar to those reported
in the literature for turfgrass surfaces (e.g., Trenholm et al., 2000; and Fitz-Rodriguez and Choi,
2002). In the visible part of the EM spectrum, average spectral reflectance of all experimental
plots was less than 6.7, 9.0 and 12.0% of the incoming shortwave radiation in TM1, TM2, and
TM3 wavebands, respectively.
As expected, turfgrass plots with lower quality (higher visible and SWIR and lower NIR
reflectance) had higher surface temperatures. Minimum and maximum average temperatures
were both detected over Tall Fescue plots of the D5 site, with values of 28.6 and 42.5 (C) over
D5-TF-4 and D5-TF-8 plots, respectively. Table 3 summarizes the average value of remotely
sensed spectral characteristics of the experimental plots for the seven dates of data collection. In
addition, Figures 2 and 3 demonstrate box plots of some of these parameters to facilitate a better
comparison among treatments.
3.2 Crop Water Stress Index
3.2.1. Upper and Lower dT limits
Several researchers (e.g. Martin et al., 1994) have reported that each turfgrass species has a
unique dT
LL
-VPD relationship, regardless of the variability in the cultural and climatological
conditions. Despite such species-dependence, Carrow (1993) stated that these relationships are
similar enough to be combined for a practical purpose such as irrigation scheduling. In this
study, the dT
LL
-VPD relationship was developed using the data collected within two days of an
irrigation and/or precipitation event over two experimental plots (D3-If Tall Fescue and C4
Kentucky Bluegrass) that had the highest values of NDVI and did not show sign of stress.
Tall Fescue: dT
LL
= 10.29 3.31 VPD, (r
2
= 0.97) (8)
Kentucky Bluegrass: dT
LL
= 10.36 3.27 VPD, (r
2
= 0.98) (9)
Interestingly, the developed relationships were almost identical. However, the slope and
intercept of the above equations are larger than the values reported by Al-Faraj et al. (2001) for
Tall Fescue and by Throssell et al., (1987) for Kentucky Bluegrass. As illustrated in Figure 4, the
cumulative amount of applied water at both plots was always larger than ET
o
over the study
period. It seems that obtaining larger coefficients in this study is most probably due to the
difference in climatological conditions. Previous studies have also showed that the slope of
dT
LL
-VPD relationship is a larger negative number under arid/semi-arid compared to humid
climates (Carrow 1993).
Page 9
Table 3. Turfgrass reflectance (%) in different wavebands, along with NDVI/SAVI (-) and surface temperature (C)
over experimental plots, averaged for the seven dates of data collection.
Plot TM1 TM2 TM3 TM4 TM5 NDVI SAVI T
surf
, C
D3-If 2.17 3.99 2.77 48.10 18.90 0.89 0.67 28.97
D3-Id 2.21 3.82 2.74 46.81 17.76 0.89 0.66 29.17
D5-WSM-2 4.46 7.41 6.60 34.86 26.84 0.68 0.46 31.06
D5-WSM-4 4.59 7.43 6.80 34.24 26.65 0.67 0.45 30.67
D5-WSM-6 4.76 7.75 7.34 33.30 27.32 0.64 0.43 31.83
D5-WSM-8 5.10 7.85 7.99 29.42 28.52 0.57 0.37 34.54
D5-AKB-2 3.06 5.18 4.81 36.92 26.87 0.77 0.53 33.64
D5-AKB -4 3.09 5.12 4.86 35.10 26.53 0.76 0.50 32.89
D5-AKB -6 3.19 5.25 4.99 36.15 26.95 0.76 0.51 32.82
D5-AKB -8 4.83 7.29 8.56 30.04 33.06 0.56 0.36 36.70
D5-THB-2 2.63 4.71 3.68 43.17 24.04 0.84 0.61 30.24
D5-THB -4 2.83 4.86 4.12 38.96 24.84 0.81 0.56 31.08
D5-THB -6 3.89 6.24 6.10 37.36 28.60 0.72 0.50 33.27
D5-THB -8 5.04 7.53 8.60 31.59 33.39 0.57 0.38 36.50
D5-FF-2 2.91 4.75 5.00 33.57 27.45 0.74 0.48 34.67
D5-FF-4 3.49 5.40 6.00 32.33 29.63 0.69 0.45 36.04
D5-FF-6 3.97 5.96 6.76 32.02 31.35 0.65 0.43 36.15
D5-FF-8 6.57 8.93 11.77 25.89 41.14 0.38 0.24 43.50
D5-TF-2 2.31 4.05 3.36 39.79 18.78 0.84 0.59 28.86
D5-TF-4 2.51 4.27 3.83 37.38 19.47 0.81 0.55 28.64
D5-TF-6 3.79 5.85 6.41 32.40 24.93 0.67 0.44 32.30
D5-TF-8 6.66 8.99 11.81 22.65 37.28 0.32 0.19 42.47
C4-Td/Ch 2.50 4.35 3.41 45.46 23.33 0.86 0.64 31.80
C4-Td/Cl 2.25 3.94 2.98 45.46 22.49 0.88 0.65 31.93
C4-Td/Cn 2.05 3.67 2.59 47.63 21.50 0.90 0.67 29.82
C4-Ts/Ch 2.23 3.85 2.99 44.99 22.17 0.87 0.64 32.66
C4-Ts/Cl 2.22 3.84 3.03 44.08 22.12 0.87 0.63 32.96
C4-Ts/Cn 2.00 3.51 2.49 46.76 20.85 0.90 0.67 31.08
C3-WS 2.45 4.51 3.34 44.53 22.88 0.86 0.63 32.37
Page 10
Figure 2. Turfgrass reflectance detected by the MSR5 at green (TM2, top), red (TM3, middle), and NIR (TM4,
bottom) wavebands.
Page 11
Figure 3. Turfgrass NDVI (top), SAVI (middle), and surface temperature (bottom).
Page 12
Figure 4. Cumulative amounts of reference ET
o
and irrigation/precipitation (I + P) over D3-If and C4 plots, all in
units of water depth (mm).
Given the similarity between the two dT
LL
-VPD equations developed in this study, they were
combined (averaged) and a single equation (dT
LL
= 10.3 3.3 VPD) was used to estimate the
lower limit of dT over the research plots. The dT
LL
-VPD coefficients were used to predict the
upper dT limit. Resulted dT
UL
values had a rather constant pattern, ranging from 17.0 to 21.9 C
over the seven dates of data collection. To evaluate the accuracy of Idso (1981) method,
estimated dT
UL
values were compared with dT
UL
values that were measured over a severely
stressed (non-transpiring) turfgrass patch. Based on this comparison, the mean bias error (MBE)
of modeled dT
UL
was 4.0 C. A closer investigation revealed that the error was influenced by
three measured dT
UL
values that were significantly lower than expected. These three
measurements were taken over the non-transpiring turfgrass patch on dates when the adjacent
D3-Id plot had been recently irrigated. Since the plot surface slope was toward the dry patch, the
lower measured dT
UL
seems to have been produced by irrigation water runoff from the D3-Id
plot which might have lowered the surface temperature at the dry plot/path area. Excluding these
three points from the analysis, the MBE of the modeled dT
UL
was reduced to only 0.7 C. Figure
5 shows measured as well as modeled limits of dT over the four plots of warm season grass mix.
As expected, measured dT falls within the range identified by the upper and lower limits. In
addition, measured dT values were closer to the dT
UL
as the distance from sprinklers increased.
3.2.2. CWSI computation using ground-based remote sensing inputs
Using modeled upper and lower dT limits and measured dT, CWSI was estimated for each of the
turfgrass plots on the 7 dates of ground-based remote sensing data collection. Except for the D5-
FF-8 and D5-TF-8 plots, estimated CWSI had a range of values between -0.1 and 0.8 (Figure 6).
150
250
350
450
550
7/20 8/10 8/31 9/21
D
e
p
t
h
o
f
w
a
t
e
r
(
m
m
)
D3-If (I+P)
C4 (I+P)
ETo
Page 13
Figure 5. Values of dT
m
(gray lines) and dT
LL
/dT
UL
(black lines) over D5-WSM plots.
As explained before, these two plots are located at larger distances from the sprinkler and thus
received the least amount of water. In addition, turfgrass was not at full cover at these locations
and part of the background soil was viewed by the IRT sensors. CWSI values for the farthest
distance of the other species (D5-WSM-8, D5-AKB-8, and D5-THB-8) were also larger than
their corresponding plots at a closer proximity to the sprinklers.
To compare the water status of each turfgrass species in site D5, CWSI values were averaged
over the two plots that were closer to the sprinkler and over the two plots that were farther from
sprinklers. Among studied species, the CWSI averaged for the first two plots was larger (0.44)
over Fine Fescue and smaller (0.02) over Tall Fescue. For the second two plots (farthest from
sprinklers) Fine Fescue still had the largest average CWSI (0.74), while the lower value (0.32)
belonged to the warm season mix. These results suggest that for the same amount of limited
irrigation (assuming similar sprinkler uniformities), Fine Fescue and warm season species have
the lowest and highest stress tolerance, respectively. This is not surprising as warm season
species are expected to perform better under hot and dry conditions. In addition, the range of
CWSI variation with distance from sprinklers was smaller for warm season mix (0.26) and larger
for Fine and Tall Fescue (0.63 and 0.91, respectively), confirming that the latter two species are
highly sensitive to the amount of applied water.
Within the soil preparation site (C4), plots with deeper tillage had a smaller CWSI compared to
plots with shallower tillage (0.17 vs. 0.23). This is probably due to the fact that a deeper tillage
destroys compacted soil layers and improves water movement and aeration in the root zone, thus
providing a more favorable growth environment. Among each tillage depth treatment, however,
the plots with no organic amendment had surprisingly smaller CWSI compared to the plots with
added compost. There was no significant difference in CWSI between the two rates of compost
application. Within site D3, the two treatments of irrigation amount showed almost no sign of
stress, with average CWSI values of 0.05 and 0.07 over D3-If and D3-Id plots, respectively.
-4
0
4
8
12
16
20
24
7/20 8/10 8/31 9/21
d
T
(
C
)
D5-WSM-2 D5-WSM-4 D5-WSM-6 D5-WSM-8
Page 14
Figure 6. Box plots of estimated CWSI (in decimals) over studied turfgrass plots.
Such a negligible difference is due to the fact that both plots received almost the same amount of
irrigation water. In addition, the sum of irrigation and precipitation was larger than ET
o
over the
study period, which explains why estimated CWSI was close to zero. In other words, the D3-Id
plot seems to not have been under a significant deficit irrigation practice.
3.2.3. CWSI computation using airborne remote sensing data
Models for identifying upper and lower dT limits that were developed using ground-based
remote sensing data were also applied to the airborne imagery to estimate CWSI on a distributed
basis (Figure 7). The results showed a significant variation in CWSI within each plot. However,
the general pattern was similar to the results of ground-based remote sensing. Over the D5 site,
CWSI values varied from almost zero at the south end to unity at the north end on July 19
th
and
August 12
th
flights (note that sprinklers were installed at the south end). Smaller CWSI values
were calculated on August 31
st
due to the cooler surface temperature. Lower temperatures were
caused by a combination of: wind gusts (> 4.0 m s
-1
) at the overpass time and irrigation events
that were either occurring at the time of the flight or occurred the day before.
3.3 Evapotranspiration
Turfgrass actual ET (ET
a
) was estimated using two different approaches, i.e. the CWSI and the
SEBAL model. For the seven dates of ground-based remote sensing data collection, average
daily ET
a
was very similar between the two methods. This result suggests that the CWSI can be
used as an effective alternative to the SEBAL model in the estimation of grass ET
a
. The average
daily ET
a
had a range of values from 0.0 to 4.0 mm d
-1
.
Page 15
Figure 7. False color airborne image of the study area (a) and maps of CWSI, ranging from zero (blue) to unity
(Red), for each overpass time: 07/19/2011 before (b) and after (c) solar noon, 08/12/2011 before (d) and after (e)
solar noon, and 08/31/2011 before solar noon (f).
The average grass reference ET
o
was 4.5 mm d
-1
during the same period. Based on both
methods, D5-FF-8 and D5-TF-8 plot had the smallest ET
a
, while D5-TF-2 and D5-TF-4 plots
had the largest average daily ET
a
. Within the D5 site, the average SEBAL-ET estimates over all
distances from sprinklers had the following highest to lowest order: D5-WSM, D5-AKB, D5-
THB, D5-TF, and D5-FF. SEBAL estimates also confirmed the results of CWSI for the C4 site.
For this site, the plots with no organic amendment had larger ET
a
rates compared to those that
received different rates of compost. The average daily ET
a
over the C3-WS (weather station)
plot was 3.2 and 3.4 mm d
-1
based on the CWSI and SEBAL methods, respectively. These
estimates were similar to the lowest ET
a
estimates over C4 site, which had the same turf species
(Kentucky Bluegrass) but received about 70 mm less irrigation water during the two months of
the study period. Thus, it seems that factors other than turf species and irrigation amount may
(a) (b) (c)
(d) (e) (f)
Page 16
have caused the observed ET
a
rates to be lower than expected over C3-WS plot. Average daily
SEBAL-ET
a
was the same over the two plots of D3 sites, which is consistent with their similar
surface spectral characteristics.
Daily ET
a
was calculated for each day during the study period by assuming that the ratio of
actual to reference ET (actual crop coefficient, K
ca
) varied linearly between two consecutive data
collection dates. Therefore, this ratio or K
ca
was interpolated between remote sensing days and
were multiplied by the alfalfa reference ET (ET
o
) to obtain an ET
a
value for each day during the
study period. Resulted daily ET
a
values were then summed up to provide an estimate of the total
ET
a
over the two months (seasonal) of the study. The maximum total (seasonal) ET
a
resulted
from the D5-TF-4 plot with 274 and 291 mm based on the CWSI and SEBAL methods,
respectively. Total ET
o
, from the weather station data, was 302 mm over the same period. The
minimum total ET
a
was also estimated over the D5-FF-8 and D5-TF-8 plots. On average, CWSI
results were 15% smaller than the SEBAL estimates. However, the difference was larger over
plots with lower ET
a
rates, which is most probably due to the lack of full canopy cover at these
plots. Since the majority of urban turfgrass systems (e.g., municipal parks, golf courses, athletic
fields, home gardens, etc.) maintain the turf at a full cover condition, all of the plots that had an
average NDVI of less than 0.6 were excluded from further comparison. This new analysis
reduced the difference between the two methods to 11%. This difference is acceptable by most
turf practitioners and may become even smaller as the study period expands to seasonal and
annual time frames. Table 4 presents average daily and total ET
a
rates for the two implemented
methods (CWSI and SEBAL) in SI and English units. This table is followed by Figure 8, which
shows box plots of daily ET
a
values for all of the days during the study period.
Turfgrass ET
a
was mapped by applying the CWSI approach to the airborne imagery. Daily ET
maps had a pattern similar to the pattern of ground-based CWSI-ET (Figure 9). For example, the
imagery acquired on August 12
th
and 31
st
clearly show that ET
a
was larger in the middle part of
C4 site, where the treatments with no added compost were located. Although the plots with
organic amendments, located toward the north and south end of this site, had smaller ET
a
rates,
the difference may not have been caused by the treatments as the pattern does not match the
rectangular shapes of plots. In addition, a gradient of ET
a
values was observed over the D5 site,
where water use decreases rather rapidly from the south to the north end of the site. Figure 9
also illustrates a significant variation within each treatment, affirming one of the most important
drawbacks of point-based measurements, which is selecting a representative location.
Page 17
Table 4. Turfgrass daily average and total ET
a
(during the study period), based on CWSI and SEBAL methods.
Plot
Avg. CWSI-ET
a
Total CWSI-ET
a
Avg. SEBAL-ET
a
Total SEBAL-ET
a
mm (in) d
-1
mm (in) mm (in) d
-1
mm (in)
D3-If 4.0 (0.16) 263 (10.4) 3.9 (0.15) 275 (10.8)
D3-Id 4.1 (0.16) 261 (10.3) 3.9 (0.15) 270 (10.6)
D5-WSM-2 3.5 (0.14) 237 (9.3) 3.4 (0.14) 245 (9.7)
D5-WSM-4 3.7 (0.15) 240 (9.4) 3.6 (0.14) 248 (9.8)
D5-WSM-6 3.4 (0.13) 218 (8.6) 3.3 (0.13) 232 (9.1)
D5-WSM-8 2.6 (0.10) 167 (6.6) 2.8 (0.11) 200 (7.9)
D5-AKB-2 3.2 (0.12) 208 (8.2) 3.3 (0.13) 231 (9.1)
D5-AKB -4 3.4 (0.13) 210 (8.3) 3.5 (0.14) 234 (9.2)
D5-AKB -6 3.2 (0.13) 188 (7.4) 3.3 (0.13) 220 (8.7)
D5-AKB -8 1.9 (0.07) 95 (3.7) 2.1 (0.08) 137 (5.4)
D5-THB-2 3.8 (0.15) 258 (10.2) 3.9 (0.15) 266 (10.5)
D5-THB -4 3.8 (0.15) 258 (10.2) 3.9 (0.15) 271 (10.7)
D5-THB -6 3.3 (0.13) 221 (8.7) 3.2 (0.13) 233 (9.2)
D5-THB -8 2.1 (0.08) 146 (5.7) 2.2 (0.08) 176 (6.9)
D5-FF-2 2.7 (0.11) 182 (7.1) 3.0 (0.12) 216 (8.5)
D5-FF-4 2.2 (0.09) 129 (5.1) 2.5 (0.10) 173 (6.8)
D5-FF-6 2.3 (0.09) 145 (5.7) 2.5 (0.10) 175 (6.9)
D5-FF-8 0.4 (0.02) 10 (0.4) 0.2 (0.01) 15 (0.6)
D5-TF-2 4.0 (0.16) 268 (10.5) 4.2 (0.16) 286 (11.2)
D5-TF-4 4.1 (0.16) 274 (10.8) 4.3 (0.17) 291 (11.5)
D5-TF-6 3.1 (0.12) 185 (7.3) 3.3 (0.13) 218 (8.6)
D5-TF-8 0.2 (0.01) 6 (0.2) 0.4 (0.02) 38 (1.5)
C4-Td/Ch 3.3 (0.13) 212 (8.3) 3.4 (0.13) 246 (9.7)
C4-Td/Cl 3.2 (0.13) 211 (8.3) 3.4 (0.13) 248 (9.7)
C4-Td/Cn 3.8 (0.15) 257 (10.1) 3.9 (0.15) 281 (11.1)
C4-Ts/Ch 3.1 (0.12) 205 (8.1) 3.3 (0.13) 248 (9.8)
C4-Ts/Cl 3.1 (0.12) 204 (8.0) 3.3 (0.13) 247 (9.7)
C4-Ts/Cn 3.6 (0.14) 238 (9.4) 3.6 (0.14) 268 (10.5)
C3-WS 3.2 (0.13) 209 (8.2) 3.4 (0.13) 249 (9.8)
Page 18
Figure 8. Box plots of estimated ET (mm.day
-1
) based on CWSI (top) and SEBAL (bottom).
Furthermore, maps of CWSI-ET
a
also showed that the value of daily ET
a
resulting from this
method depended on the time of remote sensing data acquisition. In Figure 9, for example, the
after solar noon airborne remote sensing system overpass generated somewhat different values of
daily ET
a
compared to the before solar noon flight on both July 19
th
and August 31
st
.
Vegetation water stress has been indicated to be better capture afternoon between 1 and 3 p.m.
Therefore, it is necessary to identify the time of the day at which remote sensing data acquisition
for estimating CWSI and daily ET
a
would be more appropriate and develop a method for
extrapolating/adjusting CWSI values when the remote sensing data have been acquired at non-
ideal times of the day.
Page 19
Figure 9. False color airborne image of the study area (a) and maps of daily ET (mm d
-1
), for each overpass time:
07/19/2011 before (b) and after (c) solar noon, 08/12/2011 before (d) and after (e) solar noon, and 08/31/2011 before
solar noon (f).
4. Conclusion
The results of this study indicate that variations in the amount of irrigation water applied caused
a difference in turfgrass multispectral response that was detected by the remotely sensed data.
Turfgrass canopy temperature as well as reflectance in the visible and SWIR wavebands had an
inverse, while the reflectance in the NIR band and VIs had a direct relationship with the amount
of applied water. Among studied grass species, the fine and tall Fescue showed the largest
sensitivity to the different amounts of water applied. This result suggests that irrigation
scheduling criteria for Fescue cultivars should be defined more conservatively compared to other
species. On the other hand, a mix of two warm season grasses showed the largest tolerance to
limited irrigation (less stress was detected).
(a) (b) (c)
(d) (e) (f)
Page 20
In this study, similar dT
LL
-VPD relationships were developed for Tall Fescue and Kentucky
Bluegrass. The coefficients of this relationship were used to estimate upper dT limit, using the
approach presented by Idso (1981). Modeled dT
UL
values were within 10% of the actual values,
measured over a severely stressed turfgrass patch. The variations in estimated CWSI over
experimental plots were consistent with changes in canopy reflectance, confirming that under
reduced amounts of applied water, warm season and Bluegrass turf experienced less water stress
than the Fescue grass. Although a deeper tillage resulted in smaller CWSI values, treatments
with added levels of compost surprisingly showed a larger stress compared to plots with no
organic amendment.
In this study, CWSI results were used to calculate turfgrass water use (ET). To evaluate the
performance of the CWSI-ET method, the results were compared with the ET estimate of the
SEBAL method; which is a complex surface energy balance model. Over turfgrass plots that
were at or close to full cover (where CWSI method can be applied), CWSI-ET results were only
11% smaller than SEBAL estimates on average. Such a low difference is promising, since the
CWSI method has several advantages over the SEBAL model. The main advantage is that it
requires far less input data. Once the relationships for upper and lower dT limits are identified
for a specific region and vegetation type, the CWSI method can be applied using only one
remote sensing (surface temperature) and two weather station measured weather (air temperature
and vapor pressure) variables. However, the CWSI to be applied successfully with airborne
remote sensing images an appropriate time of the day needs to be identify to image acquisition in
conjunction with an algorithm to adjust the CWSI when the images had been acquired outside of
the ideal opportunity time.
5. References
Al-Faraj, A., Meyer, G. E., & Horst, G. L. (2001). A crop water stress index for tall fescue (Festuca arundinacea
Schreb.) irrigation decision-making - a traditional method. Computers and Elec. in Agriculture, 31, 107-124.
Bastiaanssen, W. G. M., Menenti, M., Feddes, R. A., & Holtslang, A. A. (1998). A remote sensing surface energy
balance algorithm for land (SEBAL): 1. Formulation. Journal of Hydrology, 212213, 198-212.
Carrow, R. N. (1989). Turfgrass irrigation scheduling by infrared thermometry. In Proceedings of the 1989 Georgia
Water Resources Conference, May 16-17, 1989, Athens, Georgia, 63-65.
Fitz-Rodriguez, E., & Choi, C. Y. (2002). Monitoring turfgrass quality using multispectral radiometry. Transactions
of the ASAE, 45(3), 865-871.
Glenn, E. P., Neale, C. M. U., Hunsaker, D. J., & Nagler, P. L. (2011). Vegetation index-based crop coefficients to
estimate evapotranspiration by remote sensing in agricultural and natural ecosystems. Hydrological Processes,
25, 4050-4062.
Huete, A. R. (1988). A soil-adjusted vegetation index (SAVI). Remote Sensing of Environment, 25, 295-309.
Idso, S. B., Jackson, R. D., Pinter Jr., P. J., Reginato, R. J., & Hatfield, J. L. (1981). Normalizing the stress-degree-
day parameter for environmental variability. Agricultural Meteorology, 24(1), 45-55.
Page 21
Jackson, R. D., Idso, S. B., & Reginato, R. J. (1981). Canopy temperature as a crop water stress indicator. Water
Resources Research, 17(4), 1133-1138.
Martin, D. L., Wehner, D. J., & Throssell, C. S. (1994). Models for predicting the lower limit of the canopy-air
temperature difference of two cool season grasses. Crop Science, 34, 192-198.
Martin, D. L., Wehner, D. J., Throssell, C. S., & Fermanian, T. W. (2005). Evaluation of four crop water stress
index models for irrigation scheduling decisions on Penncross Creeping Bentgrass. International Turfgrass
Society Research Journal, 10, 373-386.
Payero, J. O., Neale, C. M. U., & Wright, J. L. (2005). Non-water-stressed baselines for calculating crop water stress
index (CWSI) for alfalfa and tall fescue grass. Transactions of the ASAE, 48(2), 653-661.
Throssell, C. S., Carrow, R. N., & Milliken, G. A. (1987). Canopy temperature-based irrigation scheduling indices
for Kentucky bluegrass turf. Crop Science, 27(1), 126-131.
Trenholm, L. E., Schlossberg, M. J., Lee, G., Parks, W., & Geer, S. A. (2000). An evaluation of multi-spectral
responses on selected turfgrass species. International Journal of Remote Sensing, 21(21), 709-721.