Open AccessPERFORMANCE OF THE SAFER MODEL IN ESTIMATING ACTUAL EVAPOTRANSPIRATION OF WHEAT AND BEAN IRRIGATED CROPS IN THE CENTRAL BRAZILIAN SAVANNAH
Author(s) -
JULIANA DIAS LOPES,
LINEU NEIVA RODRIGUES,
MORRIS SCHERER-WARREN,
ADRIANO CÉSAR PEREIRA DE PAULA,
THOMÁS ROCHA FERREIRA
Publication year - 2021
Publication title -
irriga
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.228
H-Index - 15
eISSN - 1808-8546
pISSN - 1413-7895
DOI - 10.15809/irriga.2021v26n3p476-489
Subject(s) - evapotranspiration , crop coefficient , mean squared error , calibration , irrigation , multispectral image , environmental science , crop , remote sensing , mathematics , hydrology (agriculture) , agronomy , statistics , forestry , geography , ecology , geotechnical engineering , engineering , biology
Meeting the ever-growing demand for food depends tightly on irrigation. In this scenario, it is essential to manage the scarce water resources available and so to understand the spatial and temporal variation of actual evapotranspiration (ETa). In the past decades, models for monitoring and mapping ETa applying remote sensing data were developed and assessed. In this work, the performance of the Simple Algorithm for Evapotranspiration Retrieving (SAFER) for estimating the evapotranspiration of two annual winter irrigated crops in the Cerrado (Brazilians Savannah) was evaluated. Multispectral and thermal images from the sensors ETM+/Landsat 7 and OLI-TIRS/Landsat 8 were used for this purpose. Initially, the model was calibrated for the wheat crop, then this calibration was tested for the bean crop, using Bowen ration ETa data as reference. After calibration, a coefficient of determination (r2) equal to 0.91 and the root mean square error (RMSE) equal to 0.58 mm d-1 were found for the wheat. For the bean, the r2 was equal to 0.65 and RMSE, equal to 0.77 mm d-1. The results are satisfactory, although they suggest that the model may present an improved performance with specific calibration for each crop and/or region, or adopting calibration in two stages, for lower and higher NDVI values.