The site was 1 km from the active river channel, depth to groundwater was 4 m,
and the aquifer was saline, producing stress on the plant stand (Nagler et al. 2008,
2009a,b). Two methods used moisture flux towers to measure moisture losses from
the canopy, but they differed in their principles of operation and underlying
assumptions. Bowen ratio towers recorded air temperature and humidity at two
heights over the canopy, and ET was calculated indirectly as a residual in the energy
balance equation. Eddy covariance towers measured moisture fluxes directly with a
precision hygrometer placed over the canopy. The other two methods used sap flux
sensors, also with different sources of error and uncertainty. Granier-type sap flow
sensors were placed in the trunk or main branch of saltcedar shrubs. ET was estimated
by introducing a source of heat into the xylem tissue and measuring the rate at which it
was dissipated away from the heat source in the transpiration stream. This required an
estimate of the conducting area of sapwood in the stem and its specific conductance,
which are problematic to measure. Heat balance sap flux sensors were placed around
small side branches of shrubs, and ET was calculated by solving a heat balance
equation based on temperature measurements upstream and downstream of the heat
coil. This is a more direct method compared to Granier sensors but requires more
scaling steps to estimate ET at the stand level. The four methods cover the range of
commonly used field methods for ET and illustrate that all methods have sources of
error and uncertainty in their application.
Eddy covariance ET results were about 50% higher than Bowen ratio results
(Figure 5.6), probably because the two methods measure ET over a different footprint
area and the Tamarix stands are not homogeneous over the site. Sap flow results fell
within the range of tower results, but Granier ET results were 30% higher than heat
balance results, attributed to different sets of assumptions in the calibration procedures, and possibly also because different sets of plants were sampled. Despite these
differences, all four methods showed that ET was only about a third of ET 0 despite the
fact that LAI was 2.66, indicating full plant cover. Hence, stress effects were
confirmed by all methods. Furthermore, ET measured by all methods decreased
over the growing season, attributed to an increase in depth to groundwater as water
was withdrawn from the aquifer faster than it could be replenished from the river. All
methods also showed a marked midday depression of ET which increased as depth to
groundwater increased over the growing season (data not shown). Hence, despite
variances in results, the combined methods produced a robust picture of Tamarix
water relations under stress conditions, with more confidence than any single method
could have produced.
Ground results were then compared to remote sensing results for the Diablo site
and for an unstressed Tamarix site (Slytherin) nearer the river and with lower salinity
in the aquifer (Figures 5.7a,b).
Remote sensing methods included a VI method (MODIS) and two thermal-band
methods, the surface energy balance algorithm (SEBAL) and the two-source energy
balance method (TSEB). The remote sensing methods were unable to detect stress
effects revealed by ground methods at the Diablo site, whereas both ground and
remote sensing methods produced similar results at the unstressed Slytherin site.
Reasons for the failure of the remote sensing methods at the Diablo site were negative
100
CHANGE DETECTION USING VEGETATION INDICES AND MULTIPLATFORM
and the aquifer was saline, producing stress on the plant stand (Nagler et al. 2008,
2009a,b). Two methods used moisture flux towers to measure moisture losses from
the canopy, but they differed in their principles of operation and underlying
assumptions. Bowen ratio towers recorded air temperature and humidity at two
heights over the canopy, and ET was calculated indirectly as a residual in the energy
balance equation. Eddy covariance towers measured moisture fluxes directly with a
precision hygrometer placed over the canopy. The other two methods used sap flux
sensors, also with different sources of error and uncertainty. Granier-type sap flow
sensors were placed in the trunk or main branch of saltcedar shrubs. ET was estimated
by introducing a source of heat into the xylem tissue and measuring the rate at which it
was dissipated away from the heat source in the transpiration stream. This required an
estimate of the conducting area of sapwood in the stem and its specific conductance,
which are problematic to measure. Heat balance sap flux sensors were placed around
small side branches of shrubs, and ET was calculated by solving a heat balance
equation based on temperature measurements upstream and downstream of the heat
coil. This is a more direct method compared to Granier sensors but requires more
scaling steps to estimate ET at the stand level. The four methods cover the range of
commonly used field methods for ET and illustrate that all methods have sources of
error and uncertainty in their application.
Eddy covariance ET results were about 50% higher than Bowen ratio results
(Figure 5.6), probably because the two methods measure ET over a different footprint
area and the Tamarix stands are not homogeneous over the site. Sap flow results fell
within the range of tower results, but Granier ET results were 30% higher than heat
balance results, attributed to different sets of assumptions in the calibration procedures, and possibly also because different sets of plants were sampled. Despite these
differences, all four methods showed that ET was only about a third of ET 0 despite the
fact that LAI was 2.66, indicating full plant cover. Hence, stress effects were
confirmed by all methods. Furthermore, ET measured by all methods decreased
over the growing season, attributed to an increase in depth to groundwater as water
was withdrawn from the aquifer faster than it could be replenished from the river. All
methods also showed a marked midday depression of ET which increased as depth to
groundwater increased over the growing season (data not shown). Hence, despite
variances in results, the combined methods produced a robust picture of Tamarix
water relations under stress conditions, with more confidence than any single method
could have produced.
Ground results were then compared to remote sensing results for the Diablo site
and for an unstressed Tamarix site (Slytherin) nearer the river and with lower salinity
in the aquifer (Figures 5.7a,b).
Remote sensing methods included a VI method (MODIS) and two thermal-band
methods, the surface energy balance algorithm (SEBAL) and the two-source energy
balance method (TSEB). The remote sensing methods were unable to detect stress
effects revealed by ground methods at the Diablo site, whereas both ground and
remote sensing methods produced similar results at the unstressed Slytherin site.
Reasons for the failure of the remote sensing methods at the Diablo site were negative
100
CHANGE DETECTION USING VEGETATION INDICES AND MULTIPLATFORM