Algoritmo S-SEBI modificado para estimar a evapotranspiração usando dados de sensoriamento remoto no Semiárido Brasileiro.

Abstract

Evapotranspiration (ET) is one of the most underestimated components of the hydrological cycle, often being attributed as a residue of the water and energy balance components. However, ET measurements performed at the land surface are expensive and, therefore, difficult to replicate. On larger scales, its measurement is complex due to the need to represent hydrometeorological processes and the heterogeneity of the land surface. Remote Sensing (RS) is the most efficient way of monitoring the land surface and obtaining regional estimates of actual ET (ETa). Surface energy balance modeling is the most common procedure for estimating ETa by RS. The Simplified Surface Energy Balance Index (SSEBI) can provide a largescale ETa, at the cost of uncertainties associated with its simplifications of meteorological parameters. This study aims to improve the ETa through changes in the SSEBI algorithm, which include automatic selection of anchor pixels and incorporating soil moisture in calculating the evaporative fraction (EF). In addition, it was considered that the EF could be constant within a weekly period in order to remove outliers and missing data. SSEBI was implemented on the Google Earth Engine (GEE) platform. In this platform are available the data MODIS, ERA5Land and GLDAS for model application. The ETa estimates of the modifications imposed on the SSEBI were evaluated using data from the Eddy covariance system at the Estação Ecológica do Seridó (ESEC) for the year 2014. The spatial behavior of the ETa by the SSEBI was compared with the MOD16 product in three locations in the Brazilian semiarid region: Petrolina PE, Barreiras BA and Bom Jesus PI. The results revealed that the modifications in the SSEBI produced statistical metrics of performance with RMSE of 0.74mm/day, R² of 0.75 and NSE of 0.74, with the data observed in the ESEC. In the spatial analysis of the SSEBI, the annual accumulated ETa data ranged from 102 to 1448 mm/year, while in the MOD16 product, this variation was from 86 to 1873 mm/year. The results revealed that the MOD16 showed a spatial pattern more consistent with the land cover in the three regions when compared to the SSEBI. The modifications imposed to the SSEBI algorithm are useful since reliable ETa estimates are necessary for managing water resources, especially in semiarid regions.