Remote Sensing (Oct 2022)

Medium-Resolution Mapping of Evapotranspiration at the Catchment Scale Based on Thermal Infrared MODIS Data and ERA-Interim Reanalysis over North Africa

  • Alhousseine Diarra,
  • Lionel Jarlan,
  • Saïd Khabba,
  • Michel Le Page,
  • Salah Er-Raki,
  • Riad Balaghi,
  • Soufyane Charafi,
  • Abdelghani Chehbouni,
  • Rafiq El Alami

DOI
https://doi.org/10.3390/rs14205071
Journal volume & issue
Vol. 14, no. 20
p. 5071

Abstract

Read online

Accurate quantification of evapotranspiration (ET) at the watershed scale remains an important research challenge for managing water resources in arid and semiarid areas. In this study, daily latent heat flux (LE) maps at the kilometer scale were derived from the two-source energy budget (TSEB) model fed by the MODIS leaf area index (LAI), land surface temperature (LST) products, and meteorological data from ERA-Interim reanalysis from 2001 to 2015 on the Tensift catchment (center of Morocco). As a preliminary step, both ERA-Interim and predicted LE at the time of the satellite overpass are evaluated in comparison to a large database of in situ meteorological measurements and eddy covariance (EC) observations, respectively. ERA-Interim compared reasonably well to in situ measurements, but a positive bias on air temperature was highlighted because meteorological stations used for the evaluation were mainly installed on irrigated fields while the grid point of ERA-Interim is representative of larger areas including bare (and hot) soil. Likewise, the predicted LE was in good agreement with the EC measurements gathered on the main crops of the region during 15 agricultural seasons with a correlation coefficient r = 0.70 and a reasonable bias of 30 W/m2. After extrapolating the instantaneous LE estimates to ET daily values, monthly ET was then assessed in comparison to monthly irrigation water amounts provided by the local agricultural office added to CRU precipitation dataset with a reasonable agreement; the relative error was more than 89% but the correlation coefficient r reached 0.80. Seasonal and interannual evapotranspiration was analyzed in relation to local climate and land use. Lastly, the potential use for improving the early prediction of grain yield, as well as detecting newly irrigated areas for arboriculture, is also discussed. The proposed method provides a relatively simple way for obtaining spatially distributed daily estimates of ET at the watershed scale, especially for not ungauged catchments.

Keywords