Performance and validation of water surface temperature estimates from Landsat 8 of the Itaipu Reservoir, State of Paraná, Brazil

• Published in: Environmental monitoring and assessment Vol. 195; no. 1; p. 137
• Main Authors: Kramer, Gisieli, Filho, Waterloo Pereira, de Carvalho, Lino Augusto Sander, Trindade, Patricia Michele Pereira, da Rosa, Cristiano Niederauer, Dezordi, Rafael
• Format: Journal Article
• Language: English
• Published: Cham Springer International Publishing 01.01.2023; Springer Nature B.V
• Subjects: Algorithms; Atmospheric Protection/Air Quality Control/Air Pollution; Brazil; computer software; Earth and Environmental Science; Ecology; Ecotoxicology; Environment; Environmental Management; Environmental monitoring; Environmental science; Estimates; Infrared sensing; Inland waters; Land surface temperature; Landsat; Landsat satellites; Literature reviews; Mathematical models; Monitoring/Environmental Analysis; Python; regression analysis; Regression models; Remote sensing; Reservoirs; Satellite imagery; Statistical analysis; Surface temperature; Temperature; Temperature data; Water resources; Water temperature; Brazil; Thermal images; QGIS; Water temperature; Remote sensing
• ISSN: 0167-6369; 1573-2959; 1573-2959
• DOI: 10.1007/s10661-022-10677-6

Studies on water surface temperature (WST) from thermal infrared remote sensing are still incipient in Brazil, and for many water resources, they do not exist. Many algorithms have been developed to estimate surface temperature in satellite images. There are also many difficulties in implementing these algorithms due to their complexity, especially in free software, which restricts the satisfactory processing of these data by users of the technique. Thus, this work aimed to validate an algorithm used to estimate land surface temperature (LST) when applied to the surface of inland water bodies. Water surface temperature estimates (WSTe) were generated from Itaipu State of Paraná (PR) reservoir, Brazil, calculated from Landsat 8 – TIRS satellite images (WSTs) and water surface temperature data from 37 in situ stations (WSTi). A linear regression model of the WSTe was generated in 60% of the samples and its validation with the remaining 40%, subject to prior evaluation of some statistical indicators. The model was considered significant since the coefficient of determination ( r 2 ) was 0.90 (95% of confidence), root mean square deviation (RMSD) 0.8 °C, Willmott Index ( d ) = 0.97, and Nash–Sutcliffe efficiency coefficient (NSE) = 0.89. The methodology used to extract WSTs from the Python QGIS plugin was relatively quick to apply, easy to understand, and had a better performance of the estimates than those presented in the literature review.