ABSTRACT Monitoring of large agricultural lands is often hampered by data collection logistics at field level. To solve such a problem, remote sensing techniques have been used to estimate vegetation indices, which can subsidize crop management decision-making. Therefore, this study aimed to select vegetation indices to detect variability in irrigated corn crops. Data were collected in São Desidério, Bahia State (Brazil), using an OLI sensor (Operational Land Imager) embedded to a Landsat-8 satellite platform. Five corn growing plots under central pivot irrigation were assessed. The following vegetation indices were tested: NDVI (Normalized Difference Vegetation Index), EVI (Enhanced Vegetation Index), SAVI (Soil Adjusted Vegetation Index), GNDVI (Green Normalized Difference Vegetation Index), SR (Simple Ratio), NDWI (Normalized Difference Water Index), and MSI (Moisture Stress Index). Among the tested indices, SR was more sensitive to high corn biomass, while GNDVI, NDVI, EVI, and SAVI were more sensitive to low values. Overall, all indices were found to be concordant with each other, with high correlations among them. Despite this, the use of a set of these indices is advisable since some respond better to certain peculiarities than others.
ABSTRACT Surface temperature (Ts) is a determining factor to obtain energy balance parameters, being relevant to understand the influence of this variable on the estimation of evapotranspiration. Thus, the objective of this study was to simulate errors in Ts estimation to verify the consequences of actual evapotranspiration (ETa) estimated by the SAFER (Simple Algorithm for Evapotranspiration Retrieving) model. For this, an image of the Landsat-8 satellite was used to induce errors from 0.2K to 10K in the variable Ts, allowing verifying the consequences in the ETa data. After the estimations of Ts and ETa, the quantitative consequences and dynamics of Ts impact on the ETa data were verified along the different land uses in the study area. The results showed that the precise estimation of Ts is essential to obtain ETa accurately. The image of ETa errors presented the highest relative errors on the surface with exposed soils and with high Ts values. However, the highest residuals of ETa images occurred on the surfaces with milder Ts and higher evapotranspiration rates (irrigated surfaces).
ABSTRACT The importance of the precise estimation of evapotranspiration is directly related to sustainable water usage. Since agriculture represents 70% of Brazil’s water consumption, adequate and efficient application of water may reduce the conflicts over the use of water among the multiple users. Considering the importance of accurate estimation of evapotranspiration, the objective of the present study was to model and compare the reference evapotranspiration from different heuristic methodologies. The standard Penman-Monteith method was used as reference for evapotranspiration, however, to evaluate the heuristic methodologies with scarce data, two widely known methods had their performances assessed in relation to Penman-Monteith. The methods used to estimate evapotranspiration from scarce data were Priestley-Taylor and Thornthwaite. The computational techniques Stepwise Regression (SWR), Random Forest (RF), Cubist (CB), Bayesian Regularized Neural Network (BRNN) and Support Vector Machines (SVM) were used to estimate evapotranspiration with scarce and full meteorological data. The results show the robustness of the heuristic methods in the prediction of the evapotranspiration. The performance criteria of machine learning methods for full weather data varied from 0.14 to 0.22 mm d-1 for mean absolute error (MAE), from 0.21 to 0.29 mm d-1 for root mean squared error (RMSE) and from 0.95 to 0.99 coefficient of determination (r²). The computational techniques proved superior performance to established methods in literature, even in scenarios of scarce variables. The BRNN presented the best performance overall.
RESUMO A importância da estimativa precisa da evapotranspiração está diretamente relacionada ao uso sustentável da água. Uma vez que a agricultura representa 70% do consumo de água no Brasil, a aplicação adequada e eficiente de água reduz os conflitos sobre o uso da água entre os múltiplos usuários. Considerando a importância de uma estimativa precisa da evapotranspiração, o objetivo do presente estudo foi modelar e comparar a evapotranspiração de referência a partir de diferentes metodologias heurísticas. O método padrão Penman-Monteith foi utilizado como referência para evapotranspiração, porém, para avaliar as metodologias heurísticas com dados escassos, avaliou-se o desempenho de métodos difundidos na literatura em relação à Penman-Monteith. Os métodos utilizados para estimar a evapotranspiração a partir de dados escassos foram Priestley-Taylor, Thornthwaite. As técnicas computacionais Regressão Stepwise (SWR), Random Forest (RF), Cubist (CB), Rede Neural com Regularização Bayesiana (BRNN) e Máquinas de Vetor de Suporte (SVM) foram utilizados para estimar a evapotranspiração, tanto com dados meteorológicos escassos, quanto com dados completos. Os resultados mostram a robustez dos métodos de aprendizagem de máquina na predição da evapotranspiração. Os critérios de desempenho desses métodos para dados meteorológicos completos variaram de 0,14 a 0,22 mm d-1 para erro absoluto médio (MAE), de 0,21 a 0,29 mm d-1 para raiz do erro quadrático médio (RMSE) e de 0,95 a 0,99 para o coeficiente de determinação (r²). As técnicas computacionais mostraram desempenho superior em todos os cenários em relação aos métodos estabelecidos na literatura. A BRNN apresentou o melhor desempenho geral.