The triangular method for evapotranspiration estimation from landsat data in arid and semi-arid regions

Abstract

The commonly applied triangular method was used to estimate regional evapotranspiration (ET). Based on the parameters of land surface, the applicability of NDVI-land surface temperature (NDVI-LST) and NDVI-albedo triangular methods were validated using the landsat7 data and the Heihe Watershed Allied Telemetry Experimental Research (HiWATER) observed data. Considering the affecting of soil moisture and vegetation water content on ET, a modified triangular method was developed using the Perpendicular Drought Index (PDI). Compared the estimated ET to the observed value, the result showed that LSTs retrieved by single channel method were closed to the measured data, and the correlation coefficient(R) was 0.8, the root mean square error (RMSE) was 5.72K and the mean absolute percent error (MAPE) was about 14.48%. The correlation coefficient between the PDI, soil moisture and vegetation water content was 0.77, so the PDI could respond the moisture condition of the surface well and could be used to estimate the ET. The ET estimated by NDVI-LST and NDVI-albedo triangular methods were in good agreement with those measured values and the correlation coefficients were 0.85 and 0.86, respectively. The NDVI and PDI showed a triangle relationship obviously, and the accuracy of estimated ET from NDVI-PDI triangular method had improved greatly, the RMSE reduced to 91.72 W·m-2 and the MAPE was 24.84%.