measurements and modeled SWC-values obtained. To directly relate WI-values at
the field level, we determined point-location estimates of WI by (i) taking into
account all WI-values, within a 500 Â 500 m area (consisting of a 20 Â 20 pixel
window) centered on the measurement sites, and (ii) averaging their values
(accounting for 400 pixel values). In the second comparison, as the SWAT model
provides spatially-averaged estimates of SWC, we averaged WI-values to provide a
mean value at spatial resolutions equivalent to those used in SWAT (i.e., 62–275 ha
resolution). To compare WI-values with SWC-values directly, we used least
squares regression (and the coefficient of determination, r
2 ) on WI-vs.-SWC data
pairs to determine the degree of agreement between the two independent variables
(i.e., WI and SWC).
15.4 Results and Discussion
Figure 15.3a provides a spatial distribution of WI-values over the study area based
on the 08 August, 2005, RADARSAT-1 image. It revealed that the WI-values fell
mostly in the wetness range of 20–45 % (~95 % of all values), with an average
wetness of 31 % (Fig. 15.3b).
Figure 15.4 shows the variations of WI and SWC at the measurement sites
(forests and bare field) for the radar-image acquisition dates (i.e., 15 July,
08 August, and 01 September, 2005). It revealed that both WI and SWC were the
lowest on 08 August, 2005, except for the measurements at the CL site. Temporal
variability in SWC is strongly coupled to episodes of rainfall within a 1–2 day
period prior to image acquisition (Table 15.2). The unchanging SWC conditions on
08 August at the CL site despite a lack of rainfall during a significant period prior
image acquisition, is most likely related to the site’s position within a prominent
landscape depression, ultimately leading to the site’s elevated soil water conditions
through lateral drainage from the surrounding landscape (Hassan et al. 2006).
Figure 15.5 shows a comparison between RADARSAT-1 ScanSAR-derived
WI-values and field measurements of volumetric SWC at a 10-cm depth as a
function of the three representative landcovers (i.e., dense forests, sparse forests,
and bare field). We preferred to use the field measurements of SWC at 10-cm depth
as it provided a better representation of available water to the vegetation. The
comparisons revealed reasonable agreement between the two variables for the three
areas, yielding r
2 -values of 74 % for the dense forests, 77 % for the sparselyvegetated forests, and 99 % for the bare field.
The C-band SAR backscattering has been previously demonstrated to be highly
influenced by vegetation water content in dense forests, opposed to SWC
(Pulliainen et al. 2004). So it is quite possible that our WI-values produced for
dense forests were affected similarly. However, since SWC is one of the most
critical variables for forest-site productivity (Bourque et al. 2000; Wang and Klinka
1996) and since it has an important role in plant physiology (by means of photosynthesis and evapotranspiration), we could envision canopy foliage water content
308
Q.K. Hassan and C.P.-A. Bourque
the field level, we determined point-location estimates of WI by (i) taking into
account all WI-values, within a 500 Â 500 m area (consisting of a 20 Â 20 pixel
window) centered on the measurement sites, and (ii) averaging their values
(accounting for 400 pixel values). In the second comparison, as the SWAT model
provides spatially-averaged estimates of SWC, we averaged WI-values to provide a
mean value at spatial resolutions equivalent to those used in SWAT (i.e., 62–275 ha
resolution). To compare WI-values with SWC-values directly, we used least
squares regression (and the coefficient of determination, r
2 ) on WI-vs.-SWC data
pairs to determine the degree of agreement between the two independent variables
(i.e., WI and SWC).
15.4 Results and Discussion
Figure 15.3a provides a spatial distribution of WI-values over the study area based
on the 08 August, 2005, RADARSAT-1 image. It revealed that the WI-values fell
mostly in the wetness range of 20–45 % (~95 % of all values), with an average
wetness of 31 % (Fig. 15.3b).
Figure 15.4 shows the variations of WI and SWC at the measurement sites
(forests and bare field) for the radar-image acquisition dates (i.e., 15 July,
08 August, and 01 September, 2005). It revealed that both WI and SWC were the
lowest on 08 August, 2005, except for the measurements at the CL site. Temporal
variability in SWC is strongly coupled to episodes of rainfall within a 1–2 day
period prior to image acquisition (Table 15.2). The unchanging SWC conditions on
08 August at the CL site despite a lack of rainfall during a significant period prior
image acquisition, is most likely related to the site’s position within a prominent
landscape depression, ultimately leading to the site’s elevated soil water conditions
through lateral drainage from the surrounding landscape (Hassan et al. 2006).
Figure 15.5 shows a comparison between RADARSAT-1 ScanSAR-derived
WI-values and field measurements of volumetric SWC at a 10-cm depth as a
function of the three representative landcovers (i.e., dense forests, sparse forests,
and bare field). We preferred to use the field measurements of SWC at 10-cm depth
as it provided a better representation of available water to the vegetation. The
comparisons revealed reasonable agreement between the two variables for the three
areas, yielding r
2 -values of 74 % for the dense forests, 77 % for the sparselyvegetated forests, and 99 % for the bare field.
The C-band SAR backscattering has been previously demonstrated to be highly
influenced by vegetation water content in dense forests, opposed to SWC
(Pulliainen et al. 2004). So it is quite possible that our WI-values produced for
dense forests were affected similarly. However, since SWC is one of the most
critical variables for forest-site productivity (Bourque et al. 2000; Wang and Klinka
1996) and since it has an important role in plant physiology (by means of photosynthesis and evapotranspiration), we could envision canopy foliage water content
308
Q.K. Hassan and C.P.-A. Bourque