Description
Hydrologic shocks like droughts can change the spatial distribution of water use in water scarce basins, but assessment of this reallocation at the scales of river basins is complicated by data availability. Post-monsoon precipitation in the Krishna River Basin declined by an average of 24% from the normal year, October 2001-May 2002, to the drought year, October 2002-May 2003, though rainfall was higher in some parts of the basin in the drought year compared to the normal year. Surface water data were unavailable so the Surface Energy Balance Algorithm for Land (SEBAL), with MODIS Land Surface Temperature (LST) imagery and NASA Surface Radiation Budget (SRB) data, was used to assess the impact of the drought on post-monsoon seasonal evapotranspiration (ET). A SEBAL ET time series of 52 images per year was made possible by using an automated pixel selection approach and a time series processing algorithm that significantly reduced processing time and operator effort. Changes in ET were assessed over the whole Krishna Basin, by major sub-basin, and in the major canal irrigation projects. Validation approaches included comparison of SEBAL with a water balance over a large irrigated area, evaporation pans, ET modeled from lysimeters, and comparing ET time series' with NDVI and LST time series'. SEBAL overestimated drought impact to ET when compared with a water balance of the irrigated project, however the water balance did not account for precipitation, inflow from small tributaries, or changes in groundwater and reservoir storage. Pan evaporation compared favorably with SEBAL evaporation over water bodies near a meteorological station, though Net Radiation (Rn) may have been underestimated in the latter half of the normal year. Rainfall data from the Tropical Rainfall Measuring Mission (TRMM) was used as input to an empirical model of ET from rainfed agriculture, calibrated to lysimeter experiments in the basin, and compared favorably with SEBAL ET over rainfed areas. In areas where SEBAL ET was significantly increased or decreased from the normal to the drought year, the SEBAL ET time series tracked with Rn and not NDVI patterns, though areas of increased LST and decreased seasonal NDVI showed complementary decreases in seasonal ET. Most sub-basins showed increased SEBAL ET in the drought year, likely due to increased precipitation in some parts of the basin during the post-monsoon season, and to higher Rn in the drought year. Strong drought impact was evident in some upstream and far downstream irrigation projects, from the normal year to the drought year, and less drought impact in projects and sub-basins in the middle reaches of the basin where post-monsoon precipitation was increased. A sensitivity analysis was conducted to test the main assumptions of the SEBAL model at the regional scale, comparing SEBAL ET calculated over the whole basin (WB) with SEBAL ET calculated over each sub-basin individually (Sub). Disaggregating the basin for SEBAL analysis did not affect the results significantly, however, in sub-basins where the dry and wet pixel assumptions are likely not met, the two approaches showed some disagreement in volumetric changes and drought impacts. A possible systematic underestimation of Rn in the normal year (2001-02) warrants further investigation and possibly depressed normal year ET, causing underestimation of drought effects. If Rn and ET values were correct, ET from rainfed systems in the normal year was limited by Rn rather than soil moisture. Overall, ET was significantly reduced in a few upstream and downstream projects and increased at some projects in the upper and middle reaches, resulting in a large redistribution of consumptive water use during the drought.
