Assessing Accuracy of Vegetation Index Method to Estimate Actual Evapotranspiration

Abstract

The estimation of actual crop evapotranspiration (ET a ) maps using complex equations and remotely sensed shortwave and thermal infrared imagery can be challenging and may require input data that are not available. There is an opportunity, therefore to create a simpler and faster method to generate ET a maps using fewer input parameters for situations where limited input data is available or greater uncertainty in the resulting ET estimates are acceptable. We compared the estimates of ET a produced by a crop coefficient and NDVI-based (K c-NDVI ) method to a full energy balance (EB) method. Clear sky images from Landsat 7 and Landsat 8 were processed and used for the ET a estimations from maize during two growing seasons in eastern South Dakota, USA. The results showed that the ET a values from the K c-NDVI method were lower than the ET a values from the EB method by 18% for 2015 and 11% for 2016 growing season. During study period the accuracy of ET a estimation decreased 17% with the K c-NDVI method. For both years the mean bias error was 0.81 mm day -1 and the root mean square error (RMSE) was 0.37 mm day -1 . The average daily ET a of 5.3 mm day -1 . The K c-NDVI method performed acceptable for ET a estimations, indicating that this method can be used to estimate ET a with minimum input parameters at focused regional and field scales for short time periods.