干旱区生态环境知识资源中心

Regional wheat yield was predicted using water use efficiency (WUE) and radiation use efficiency (RUE), District wise wheat WUE were generated from historical consumptive water use and reported yield. A single source energy balance algorithm was used to estimate actual evapotranspiration (AET) from MODIS (MODerate resolution Imaging Spectroradiometer) AQUA seven band reflectances and land surface temperature (LST). Time series AET from historical (2002-2003, 2003-2004) MODIS eight day composites and district wise crop calendar were used to generate wheat consumptive water use. In second approach, wheat yield was determined from the product of total net primary productivity (NPIP) accumulated during wheat growth period and harvest index (HI). NPP was estimated on spatial (1km) and temporal (8 day) domain by using LAI derived fAPAR and maximum radiation use efficiency (RUE..) of crop following CASA (Carinege Ames Stanford Approach). Temperature scalars estimated from MODIS AQUA day-night LST and water scalar computed from LST-albedo based evaporative fraction were used to constrain RUE(max).

Energy balance estimates derived from MODIS AQUA were validated with insitu attented and unattended observations for two wheat seasons, 2005 - 06 and 2006 - 07, within a 5 km x 5 km wheat growing region of Kheda district in Gujarat. Daily AQUA evapotranspiration estimates in terms of daytime latent heat fluxes were found to have root mean square error (RMSE) 29 Wm(-2) (r = 0.72) and 26 Wm(-2) (r = 0.8), respectively as compared to attended measurements during both the seasons. Correlation was substantially more and RIVISE were less when estimates were compared with in situ measurements using Large Aperture Scintillometer (LAS) sensible heat fluxes. The predicted district wheat yield for 2004-05 from WUE approach was found to produce RMSE 479 kg ha(-1)(18.5%) (r = 0.89) respectively. The error was almost similar in productivity based approach (RMSE = 471 kg ha(-1), r = 0.75). But latter does not require any pre-calibration with historical datasets.