dyh formulations, 
ations may be used as a 
SAVI). This assumption 
1990) and Jackson et al. 
the vertices of the VH 
z for computation of r,. 
and d, in computation 
respectively; 
ely; and 
e dealing only with the 
n order to describe the 
>nd the boundaries [by 
t is the possible change 
:h would alter limits of 
' the composite surface, 
re based on radiometric 
is known, the relation 
.993) could be used to 
Ued with variations in 
(13) 
évapotranspiration rate 
refer to the minimum, 
îe CWSI for full-cover 
inces AC/AB in Figure 
«dirions. 
in et al. (1994a). WDI 
sasurements of T, and 
MAC), located 40 km 
:h field. The red, near 
I to the surface. Data 
processed for analysis. 
re t the harvest period 
rowth (Figures 2a and 
or right limits of the 
(DOY) 213, the SAVI 
d site A had not. This 
a slight increase in 
h the same amount of 
ite A remained higher 
at site A could be an 
o sites within the field 
had nearly identical SAVI and T f -T. values. This could indicate that the crop at site A was able to recover later 
in the harvest cycle, though no yield data were collected to verify this. 
Computations of WDI for sites A and B of the alfalfa field reflects the results of differing irrigation practices 
(Figure 3). The WDI of site A was nearly equal to 1.0 on DOY 229, just prior to irrigation, while the recent 
irrigation of site B on the same day resulted in a WDI value close to zero. The resultant lag in the vitality of 
the crop at site A was also apparent in the values of WDI. However, the two sites had nearly identical WDI 
at the end of the growing season. The SAVI becomes less sensitive to increases in Leaf Area Index for LAIs 
greater than about 3.0, while transpiration continues to increase, resulting in lower T.-T. values. The gradual 
decline in WDI after DOY 240 for both sites may have beat a result of this phenomenon. 
Ts-To (C) Ts-To (C) 
Figure 2a, 2b. Values of SAVI and (T.-TJ for sites A and B in the alfalfa field. The numbers within the 
graph represent the day of year (some dates are not listed for graphic clarity). Theoretically, the location of 
trapezoid vertices change each day. For these figures seasonal vertex locations were used. 
Unfortunately, we can’t quantify the differences in crop stress for these two sites because leaf turgor and 
soil moisture measurements were not made during this experiment. However, results show that the WDI has 
promise for use in scheduling irrigations of both sparsely- and densely-vegetated fields. 
4 - USING TREND VECTORS IN VIT SPACE TO ESTIMATE CANOPY TEMPERATURE 
In many instances a single agricultural field will contain a considerable variation in percent cover, resulting in 
a range of vegetation indices and surface temperatures. Assuming once more that T,-T. is a linear function 
of vegetation index, these SAVI-fT.-TJ pairs should form a linear cluster. A linear regression through these 
points, when extrapolated, would intersect the top and bottom lines of the Trapezoid at T„-T. and T 0 -T„ 
respectively. A uniform soil surface moisture is necessary for this to be true, and preliminary tests indicate that 
this is not always a safe assumption for surface irrigated crops, as areas with sparse vegetation tend to dry first 
as a result of exposure to higher levels of solar radiation. However, a crop supplied with subsurface drip 
irrigation would normally present a surface of uniform moisture content, as would a surface irrigated field well 
after the last irrigation. As this is the most critical period for monitoring crop water stress, the method would 
he useful for both drip and flood irrigated crops. Figure 4 illustrates two hypothetical cases; one set of points 
representing a mildly stressed crop with a moist soil background and the other set a non-stressed crop with a 
dry soil background. In this special case the non-stressed crop would have higher composite temperatures than 
the stressed crop, but the trend vectors clearly show the relative stress levels. 
Preliminary tests of the hypothesis have produced encouraging results. Two digital cameras sensitive to red 
(600nm - 670nm) and near infrared (790nm - 890nm) spectral bands were attached to a thermal infrared (8/un -