Associação espaço-temporal de variáveis meteorológicas com técnicas de sensoriamento remoto para estimativa da evapotransporação
Resumo
Resumo: Teve-se por objetivo no presente estudo: i) Propor, testar e avaliar um método simplificado para estimar a evapotranspiração de referência (ETo) espacializada, baseando-se na associação de imagens de satélite e variáveis climáticas medidas em estações meteorológicas; e ii) Avaliar a variabilidade espacial da ETo diária na Bacia Hidrográfica do Rio Tibagi (BHRT), estimada com dados remotos, considerando eventos ENOS (El Nino, La Nina e Normalidade climática), bem como estimar a ETo na mesma bacia no intervalo entre imagens do satélite Landsat, utilizando algoritmo interpolador temporal. O trabalho foi estruturado em três capítulos: Capitulo I - Referencial teórico indicando a importância da estimativa espacial da evapotranspiração, com descrição dos aspectos técnicos e teóricos para subsidiar as discussões e metodologias do trabalho; Capítulo II - Avaliação das estimativas das temperaturas da superfície (Ts) e coluna total de vapor d’água precipitável (Wp), obtidas com imagem dos satélites Landsat 8 e Aqua MODIS, associadas em análises de regressão linear simples e múltipla com a temperatura e umidade relativa média do ar, respectivamente, medidas em estações meteorológicas. As equações de ajuste foram calibradas e validadas com dados diferentes. Estimativas da ETo com modelos alternativos utilizando dados meteorológicos espacializados de temperatura e umidade relativa foram comparados com a ETo estimada com o modelo Penman Monteith ASCE. As variáveis espaciais Ts e Wp estimaram a temperatura (TeSR(Wp;Ts) r = 0,87) e umidade relativa (UReSR(Wp;Ts); r = 0,41) do ar baseadas em sensoriamento remoto com desempenho satisfatório. O modelo Moretti-Jerszurki-Silva (MJS; EToMJS(Ra;psi ar)) apresentou desempenho satisfatório estatisticamente para estimar a ETo; Capitulo III - Avaliação da ETo estimada com modelo MJS, com entrada de variáveis espacializadas (temperatura e umidade relativa do ar) obtidas com modelo de regressão linear múltiplo na BHRT. A variabilidade espacial foi avaliada nas quatro estações do ano, com o uso de imagens de satélite, considerando os anos: 2011 (La Nina), 2013 (normal) e 2015 (El Nino). A variabilidade temporal da EToMJS(psi ar;Ra) foi testada com interpolação linear no período entre imagens do satélite Landsat 8, com algoritmo interpolador temporal "r.series.interp" disponível no software Qgis 3.10. A associação entre as ETo estimadas com metodologias padrão (Penman-Monteith ASCE - ETopm) e alternativa (MJS; EToMJS(psi ar;Ra)) foram avaliados estatisticamente com: NSE, d, RMSE, ERPAM% e r. Observou-se diferença da EToMJS(psi ar;Ra) média espacializada na estações do ano entre os cenários Normal e El Nino (0,36 mm dia-1) e entre os cenários Normal e La Nina (0,34 mm dia-1). O algoritmo interpolador apresentou precisão aceitável, mas distante do valor de referência EToPM. O modelo MJS utilizando variáveis baseadas em sensoriamento remoto obteve desempenho satisfatório, e o interpolador temporal linear apresentou estimativa aceitável da EToMJS(psi ar;Ra) diária espacializada, considerando a sua simplicidade. Abstract: The objective of this study was: i) To propose, test and evaluate a simplified methodology to estimate the spatialized reference evapotranspiration (ETo), based on the association of satellite images and climate variables measured in meteorological stations; and ii) Evaluate the spatial variability of the daily ETo in the Tibagi River Basin (TRB), estimated with remote data, considering ENSO events (El Nino, La Nina and Climatic Normality), as well as estimate the ETo in the same basin in the interval between images of the Landsat satellite, using the temporal interpolation algorithm. The work was structured in three chapters: Chapter I - Referential theoretical indicating the importance of the spatial estimation of evapotranspiration, with description of the technical and theoretical aspects to support the discussions and methodologies of work; Chapter II - Evaluation of the surface temperature estimates (Ts) and total column of precipitable water vapor (Wp), obtained with images from the Landsat 8 and Aqua MODIS satellites, associated with simple and multiple linear regression analysis with temperature and relative humidity air average, respectively, measured at meteorological stations. Single and multiple adjustment equations were calibrated and validated with different data. ETo estimates with alternative models using spatialized climate data of temperature and relative humidity, were compared with the ETo estimated with the Penman Monteith ASCE model. The spatial variables Ts and Wp estimated the temperature (TeSR(Wp;Ts); r = 0.87) and relative humidity (UReSR(Wp;Ts); r = 0.41) of the air based on remote sensing with performance satisfactory. The Moretti- Jerszurki-Silva model (MJS, EToMJS(psi ar;Ra)) presents a statistically satisfactory performance to estimate the ETo; Chapter III - ETo assessment estimated with MJS model, with input of specialized variables (temperature and relative humidity of the air) obtained with multiple linear regression model in TRB. Spatial variability was evaluated in the four seasons of the year, using satellite images, in climate scenarios considering the years: 2013 (Normal), 2015 (El Nino) and 2011 (La Nina). The temporal variability of EToMJS(psi ar;Ra) was tested with linear interpolation in the period between images from the Landsat 8 satellite, with the temporal interpolation algorithm "r.series.interp" available in the Qgis 3.10 software. The association between ETo estimated with standard (Penman-Monteith ASCE - EToPM) and alternative (MJS; EToMJS(psi ar;Ra)) methodologies were statistically evaluated with: NSE, d, RMSE, ERPAM% and r. It was observed difference EToMJS(psi ar;Ra) on the average spatialized seasons settings between Normal and El Nino (0.36 mm d_1) and between Normal and La Nina scenarios (0.34 mm day-1). The interpolator algorithm has acceptable precision, but far from the reference value EToPM. The MJS model using variables based on remote sensing achieved satisfactory performance, and the resulting linear temporal interpolator acceptable estimate of the spatialized daily EToMJS(psi ar.Ra), considering its simplicity.
Collections
- Dissertações [281]