Intercomparison of Remote Sensing-Based Models for Evapotranspiration.

Hatim M. E. Geli, Christopher M. U. Neale, James Verdin

Within the context of the U.S. Geological Survey (USGS) effort to improve estimates of consumptive water use, an inter-comparison of remote sensing based models was implemented. Consumptive water use in agricultural areas can be estimated by modeling evapotranspiration (ET) rates. Traditionally ET can be estimated using data from local weather stations applying methods such as crop coefficient, Blaney-Criddle, Penman-Monteith and other similar methods. These estimates of ET at local scales are usually assumed to represent regional scale estimates. This approach, however, could result in considerable uncertainties ET estimates at regional scales. Remote sensing approaches have already proven their ability in providing such estimates with a reasonable accuracy. As there are many different remote sensing method available in literature there is a need to conduct an inter-comparison analysis to identify their performance under different conditions. Such analysis will partly help in preparing guidelines and specifications for the use of remote sensing methods to estimate consumptive water that would be accepted by the different entities including modelers, end users, and decision makers. 

Different models were tested including the Operational Simplified Surface Energy Balance (SSEBop) (Senay et al. 2011), Mapping Evapotranspiration with Internalized Calibration (METRIC) (Allen et al., 2007), Surface Energy Balance System (Su 2002), and the Two Source Energy Balance (TSEB) (Norman et al., 1995). This analysis was conducted over irrigated agricultural areas at the Palo Verde Irrigation District (PVID), in California. The total area extent ~ 500 km² is covered mostly with irrigated alfalfa, cotton, forage and small vegetable crops. The used Landsat 5 imagery along with ground-based weather forcing data. Estimates of daily ETa during Landsat overpass dates were compared with ground-based Bowen ratio measurements. Preliminary results will include comparison of each model performance and an estimate of associated uncertainties in providing estimate of regional ET.