Mapping evapotranspiration across Lebanon throughout 2006-2016 using landsat imagery and local weather data -

Abstract

Reference and actual evapotranspiration are two important variables in hydrologic analyses, agricultural crop production, determining irrigation water requirements, and irrigation management. Weather parameters that enable the quantification of evapotranspiration vary in time and space and the weather networks that measure them are too sparse for practical applications by water resource planners. In the present work, we combined remote sensing and in-situ measurements to estimate reference and actual evapotranspiration at the watershed and country level of Lebanon. The FAO-Penman Monteith method was used to calculate grass reference evapotranspiration from 42 local weather stations distributed across the country. Results were then spatially interpolated across the country using the Spline with barriers tool in ArcMap with the barrier shapefile dividing Lebanon into Western and Eastern ranges. Freely available Landsat surface reflectance product, the normalized difference vegetation index (NDVI), was used to calculate the fraction of vegetation cover (FoV) across the country. Maps of actual evapotranspiration were established as the product of reference evapotranspiration and fraction of vegetation cover maps. Zonal statistics was applied at the watershed level for the major 15 watersheds of Lebanon. The results were compared with NASA’s MODIS ET products. Landsat summer ETc (1829 mcm) across the study period was 22percent higher than that of MODIS (1421 mcm) while winter ETc from Landsat (1160 mcm) was 21percent lower than that of MODIS (1400 mcm). Mean annual Landsat ETc (286 mcm) was 6percent higher than that of MODIS (270 mcm). Our results showed that actual evapotranspiration values computed using Landsat 30 meter resolution images are generally higher than ETc values computed using the 1 kilometer resolution images of MODIS. The relatively finer resolution of the Landsat satellite allows evapotranspiration estimates to account more to the heterogeneity of the landscapes as each pixel of the Landsat image that captures