Modelagem matemática e computacional para estimativa da evapotranspiração de referência para o município de Botucatu/SP
Carregando...
Data
2022-11-30
Autores
Boso, Ana Cláudia Marassá Roza
Título da Revista
ISSN da Revista
Título de Volume
Editor
Universidade Estadual Paulista (Unesp)
Resumo
A quantificação da evapotranspiração de referência (ETo) é de grande importância na definição do balanço hídrico, principalmente na agricultura irrigada. Sua medida pode ser obtida por meio de metodologia direta (lisímetro) ou indireta (equações empíricas). No entanto, para pequenos produtores, a aquisição e manutenção do lisímetro em campo, se torna inviável. Nestes casos, metodologias alternativas, como o modelo recomendado pela FAO - Food and Agriculture Organization of the United Nations (Penman-Monteith), modelos empíricos e dados de sensoriamento remoto, podem ser aplicadas na estimativa da ETo. Dependendo das condições climáticas e da metodologia utilizada, como os modelos empíricos, é necessário realizar a calibração local dos coeficientes. Neste sentido o vigente trabalho, divido em três capítulos, apresenta a eficiência dos modelos empíricos (originais e calibrados) e do sensoriamento remoto, em estimar a ETo para o município de Botucatu – SP. O primeiro capítulo descreve o desempenho dos modelos empíricos em relação os valores reais, mensurados pelo lisímetro de pesagem. Os resultados demostram que os modelos originais, desenvolvidos para climas árido e semiárido, apresentaram uma superestimativa média da ETo, em relação à medida real, em 121,07% para período mensal e 49,14% para diário. Os modelos originais de Thornthwaite e Priestley e Taylor apresentaram médias próximas ao do lisímetro, uma diferença de 15,52% e 5,15% respectivamente. Entre os modelos calibrados, Hargreaves-Samani e Jensen-Haise exibiram o índice de Willmott (d) em 0,85 e 0,88 nessa ordem, indicando bom desempenho na estimativa. Para os modelos calibrados que ainda apresentaram diferença significativa entre as médias da ETo, foi realizado um ajuste para períodos acumulados de três, cinco e sete dias, onde verificou-se que os modelo de Turc, Camargo e Priestley e Taylor apresentaram desempenho satisfatório para os três períodos avaliado. Os três modelos estimaram valores da ETo equivalentes ao apresentado no lisímetro. O segundo capítulo, análogo ao primeiro, indica a eficiência dos modelos empíricos em relação ao modelo de Penman-Monteith (FAO 56). O modelo original de Tomar e O'toole apresentou valor médio inferior da ETo em 57,36%, sobre o valor estimado por Penman-Monteith (FAO 56). Novamente o modelo de Thornthwaite, original e ajustado, apresentou estimativas próximas ao modelo de referência. Uma diferença apenas de 4,25% e 2,82% respectivamente. A calibração local dos modelos proporcionou maior eficiências na estimativa da ETo, com redução de 82,75% do erro absoluto médio (MAE) e de 80,75% para raiz quadrada do erro médio (RMSE). Entre os modelos ajustados, Blaney-Criddle apresentou o maior valor do coeficiente de Pearson (r = 0,97), indicando a possibilidade de sua aplicação para o local de estudo. O terceiro capítulo descreve a eficiência do SR (sensoriamento remoto MODIS16) em estimar a ETo em relação ao valor real (lisímetro) e o estimado por PM (Penman-Monteith/FAO). O modelo SR superestimou a ETo em 5,42% e 12,19% em relação ao modelo PM e o lisímetro. Os dados também apresentaram boa correlação entre a relação SR/lisímetro (r = 0,5 e d = 0,66). Os dados estatísticos indicam que o modelo SR pode ser aplicado para estimativa da evapotranspiração, nas condições climáticas do município em estudo.
The quantification of reference evapotranspiration (𝐸𝑇𝑜) is of great importance in defining the water balance, mainly in irrigated agriculture. Its measurement can be controlled by means of direct methodology (lysimeter) or indirect methodology (empirical equations). However, for small producers, the acquisition and maintenance of the lysimeter in the field becomes unfeasible. In these cases, alternative methodologies, such as the model recommended by the FAO - Food and Agriculture Organization of the United Nations (Penman-Monteith), empirical models and remote sensing data, can be applied to estimate the 𝐸𝑇𝑜 . Depending on the weather conditions and the methodology used, such as empirical models, it is necessary to perform local calibration of the coefficients. In this sense, the current work, divided into three chapters, presents the efficiency of empirical models (original and calibrated) and remote sensing, in estimating the 𝐸𝑇𝑜 for the municipality of Botucatu - SP. The first chapter describes the performance of the empirical models in relation to the real values, measured by the weighing lysimeter. The results showed that the original models, carried out for arid and semi-arid climates, presented an overestimated average of 𝐸𝑇𝑜, in relation to the actual measurement, by 121.07% for the monthly period and 49.14% for the daily period. The original models by Thornthwaite and Priestley and Taylor presented means close to the lysimeter, a difference of 15.52% and 5.15% respectively. Among the calibrated models, Hargreaves-Samani and Jensen-Haise exhibit the Willmott index (𝑑) at 0.85 and 0.88 in that order, indicating good performance in the estimation. For the calibrated models that still showed a significant difference between the 𝐸𝑇𝑜 means, an adjustment was made for accumulated periods of three, five and seven days, where it was assumed that the model by Turc, Camargo and Priestley and Taylor performed for the three evaluation periods. The three models estimated 𝐸𝑇𝑜 values equivalent to those presented in the lysimeter. The second chapter, described in the first, indicates the efficiency of the empirical models in relation to the Penman-Monteith model (FAO 56). The original model by Tomar and O'toole showed a lower mean 𝐸𝑇𝑜 value of 57.36% compared to the value estimated by Penman-Monteith (FAO 56). Once again, the original and adjusted Thornthwaite model presented preterm infants close to the reference model. A difference of only 4.25% and 2.82% respectively. The local capacity of the administered models showed greater efficiencies in estimating the 𝐸𝑇𝑜, with a reduction of 82.75% in the mean absolute error (MAE) and 80.75% in the square root mean error (RMSE). Among the adjusted models, Blaney-Criddle had the highest Pearson coefficient value ( 𝑟 = 0.97), indicating the possibility of its application to the study location. The third chapter describes the efficiency of SR (MODIS16 remote sensing) in estimating the 𝐸𝑇𝑜 in relation to the real value (lysimeter) and the value estimated by PM (PenmanMonteith/FAO). The SR model overestimated the 𝐸𝑇𝑜 by 5.42% and 12.19% compared to the PM model and the lysimeter. The data also showed good dynamics between the SR/lysimeter ratio (𝑟 = 0.5 and 𝑑 = 0.66). Statistical data indicate that the SR model can be applied to estimate evapotranspiration, under the climatic conditions of the municipality under study
The quantification of reference evapotranspiration (𝐸𝑇𝑜) is of great importance in defining the water balance, mainly in irrigated agriculture. Its measurement can be controlled by means of direct methodology (lysimeter) or indirect methodology (empirical equations). However, for small producers, the acquisition and maintenance of the lysimeter in the field becomes unfeasible. In these cases, alternative methodologies, such as the model recommended by the FAO - Food and Agriculture Organization of the United Nations (Penman-Monteith), empirical models and remote sensing data, can be applied to estimate the 𝐸𝑇𝑜 . Depending on the weather conditions and the methodology used, such as empirical models, it is necessary to perform local calibration of the coefficients. In this sense, the current work, divided into three chapters, presents the efficiency of empirical models (original and calibrated) and remote sensing, in estimating the 𝐸𝑇𝑜 for the municipality of Botucatu - SP. The first chapter describes the performance of the empirical models in relation to the real values, measured by the weighing lysimeter. The results showed that the original models, carried out for arid and semi-arid climates, presented an overestimated average of 𝐸𝑇𝑜, in relation to the actual measurement, by 121.07% for the monthly period and 49.14% for the daily period. The original models by Thornthwaite and Priestley and Taylor presented means close to the lysimeter, a difference of 15.52% and 5.15% respectively. Among the calibrated models, Hargreaves-Samani and Jensen-Haise exhibit the Willmott index (𝑑) at 0.85 and 0.88 in that order, indicating good performance in the estimation. For the calibrated models that still showed a significant difference between the 𝐸𝑇𝑜 means, an adjustment was made for accumulated periods of three, five and seven days, where it was assumed that the model by Turc, Camargo and Priestley and Taylor performed for the three evaluation periods. The three models estimated 𝐸𝑇𝑜 values equivalent to those presented in the lysimeter. The second chapter, described in the first, indicates the efficiency of the empirical models in relation to the Penman-Monteith model (FAO 56). The original model by Tomar and O'toole showed a lower mean 𝐸𝑇𝑜 value of 57.36% compared to the value estimated by Penman-Monteith (FAO 56). Once again, the original and adjusted Thornthwaite model presented preterm infants close to the reference model. A difference of only 4.25% and 2.82% respectively. The local capacity of the administered models showed greater efficiencies in estimating the 𝐸𝑇𝑜, with a reduction of 82.75% in the mean absolute error (MAE) and 80.75% in the square root mean error (RMSE). Among the adjusted models, Blaney-Criddle had the highest Pearson coefficient value ( 𝑟 = 0.97), indicating the possibility of its application to the study location. The third chapter describes the efficiency of SR (MODIS16 remote sensing) in estimating the 𝐸𝑇𝑜 in relation to the real value (lysimeter) and the value estimated by PM (PenmanMonteith/FAO). The SR model overestimated the 𝐸𝑇𝑜 by 5.42% and 12.19% compared to the PM model and the lysimeter. The data also showed good dynamics between the SR/lysimeter ratio (𝑟 = 0.5 and 𝑑 = 0.66). Statistical data indicate that the SR model can be applied to estimate evapotranspiration, under the climatic conditions of the municipality under study
Descrição
Palavras-chave
Irrigação, Sensoriamento remoto, Evapotranspiração, Computação matemática, Meteorologia, Irrigation, Remote sensing, Evapotranspiration, Meteorology, Mathematical computing