Estimativa de evapotranspiração utilizando sebal automatizado em clima mediterrânico.

Abstract

For more than two decades, Surface Energy Balance Algorithms for Land (SEBAL) has been used successfully in different climatic regions to estimate actual evapotranspiration (ETr) from satellite imagery. The traditional version of SEBAL requires the operator (user) to visualize two anchor pixels (hot and cold) and the use of meteorological stations in the region of study. However, these two factors limit their application to jobs that require long time series and large areas. The objective of this work was to improve the algorithm of selection of reference pixels and reduce the use of meteorological data from surface stations by orbital sensor data. The work was developed for an intensive cultivation of olive trees located in the Mediterranean region of Portugal, bringing together 16 observations of the Landsat 5 and 7 satellites captured between 2010 and 2012. The SEBAL model and the algorithm of automatic selection of reference pixels used in this study was previously programmed in R by Cunha et al., 2018 and had adaptations performed in this work to allow the use of different configurations of selection of anchors and input of meteorological data from orbital data. The anchor pixel selection algorithm was evaluated for five different configurations (tests), resulting in five different ETr estimates. In order to reduce the demand for data from surface meteorological stations, this work used the available data of shortwave radiation and Surface Temperature by the Satellite Application Facility on Land Surface Analysis (LSA SAF), which are generated from the sensor data MSG. The evaluation of the different configurations (tests) was performed from the field instrumentation (Eddy covariance), always analyzing the daily ETr. The results show a high sensitivity of SEBAL in relation to the surface temperature quantiles of the candidate set of anchor pixels. When compared to the field data (Eddy Covariance), the best configuration of the anchor pixel selection algorithm allowed estimation of ETr and R²=0.734 and absolute mean error equal to 0.4mmd-1. The application of the automated SEBAL to regions of high seasonality with that used in this study allowed the evaluation in a wide range of values (1.64 - 4.40mmd-1). Also, the results found in this study show similar performance to those found by works that use SEBAL in the traditional way for similar regions of climate and soil cover. To prove the efficiency of the configuration that presented better performance (test 5), new evaluations should be performed for other climatic regions and soil cover. However, from this work it is observed that it is possible to obtain good performance for SEBAL with automatic selections of anchor pixels, this contribution will allow the reduction of the limitations and greater application of SEBAL to the areas of hydrology, agronomy, meteorology and environmental evaluation.