Synthetic images to map daily evapotranspiration in field scale using SEBAL model and ESTARFM algorithm

Abstract

Many applications in water resource management require evapotranspiration (ET) information at the daily and field-level scales; however, no satellite system currently operating is able to capture all the resources of ET dynamics in an agricultural field. Thus, the objective of this study was to apply the SEBAL (Surface Energy Balance Algorithm for Land) and ESTARFM (Enhanced Adaptive Reflectance Fusion Model) methodology to estimate the daily ET in an agricultural area in the municipality of Cascavel, Paraná. We applied the ESTARFM algorithm to MODIS and Landsat 8 images to produce 8 synthetic images. The performance of the algorithm was evaluated by comparing predicted surface reflectance values obtained to real Landsat 8 images. SEBAL was applied to obtain daily evapotranspiration values (ET24) for 3 different targets (maize, soybean and stubble crop). The observed results showed that the predictions of ET using ESTARFM had a general determination coefficient of 0.80 in all analyzed images and ranged from 0.4 < R² < 0.81 when compared with real data from the Landsat 8 images, with soybean crop yielding the worst results. Low error values were found between the synthetic time series data of ET and the real data, with mean less than 1 mm day-1, meaning high reliability of synthetic data. ESTARFM tended to overestimate the ET values when compared with the real data, with the performance strongly affected by a change in the soil cover between the analyzed dates. Input data with the same soil cover is recommended for more accurate results.