Amazirh A., Er-Raki S., Ojha N., Bouras E., Rivalland V., Merlin O., Chehbouni Abdelghani. (2022). Assimilation of SMAP disaggregated soil moisture and Landsat land surface temperature to improve FAO-56 estimates of ET in semi-arid regions. Agricultural Water Management, 260, p. 107290 [14 p.]. ISSN 0378-3774.
Titre du document
Assimilation of SMAP disaggregated soil moisture and Landsat land surface temperature to improve FAO-56 estimates of ET in semi-arid regions
Amazirh A., Er-Raki S., Ojha N., Bouras E., Rivalland V., Merlin O., Chehbouni Abdelghani
Source
Agricultural Water Management, 2022,
260, p. 107290 [14 p.] ISSN 0378-3774
Accurate estimation of evapotranspiration (ET) is of crucial importance in water science and hydrological process understanding especially in semi-arid/arid areas since ET represents more than 85% of the total water budget. FAO-56 is one of the widely used formulations to estimate the actual crop evapotranspiration (ETc (act)) due to its operational nature and since it represents a reasonable compromise between simplicity and accuracy. In this vein, the objective of this paper was to examine the possibility of improving ETc act estimates through remote sensing data assimilation. For this purpose, remotely sensed soil moisture (SM) and Land surface temperature (LST) data were simultaneously assimilated into FAO-dualK(c) . Surface SM observations were assimilated into the soil evaporation (E-s) component through the soil evaporation coefficient, and LST data were assimilated into the actual crop transpiration (T-c (act)) component through the crop stress coefficient. The LST data were used to estimate the water stress coefficient (K-s) as a proxy of LST (LST proxy ). The FAO-Ks was corrected by assimilating LST proxy derived from Landsat data based on the variances of predicted errors on K-s estimates from FAO-56 model and thermal-derived K-s. The proposed approach was tested over a semi-arid area in Morocco using first, in situ data collected during 2002-2003 and 2015-2016 wheat growth seasons over two different fields and then, remotely sensed data derived from disaggregated Soil Moisture Active Passive (SMAP) SM and Landsat-LST sensors were used. Assimilating SM data leads to an improvement of the ETc act model prediction: the mot mean square error (RMSE) decreased from 0.98 to 0.65 mm/day compared to the classical FAO-dualK(c) using in situ SM. Moreover, assimilating both in situ SM and LST data provided more accurate results with a RMSE error of 0.55 mm/day. By using SMAP-based SM and Landsat-LST, results also improved in comparison with standard FAO and reached a RMSE of 0.73 mm/day against eddy-covariance ETc act measurements.
