Spatiotemporal variability of reference evapotranspiration estimated using satellite images in the Tibagi River Basin, Paraná State, Brazil

The reference evapotranspiration (ETo) is an essential tool in planning and management of water resources, but large-scale monitoring using traditional methods is impractical due to its cost and logistics. An alternative is to rely on remote data to model ETo. This study aimed to evaluate the spatial variability of daily ETo in the Tibagi River Basin (TRB), estimated with remote sensing data during years with ENSO events, and to estimate ETo between Landsat satellite images using a temporal interpolation algorithm. ETo was calculated using the Moretti-Jerszurki-Silva model (MJS; EToMJS(ψair;Ra)) and spatial data of temperature and relative humidity were estimated with a multiple linear regression model. Spatial variability was assessed using images that represented the seasons in 2013 (Normal), 2015 (El Niño), and 2011 (La Niña). The temporal variability of EToMJS(ψair;Ra) was tested with linear interpolation between Landsat 8 images in 2013, using the "r.series.interp" algorithm. The interpolated EToMJS(ψair;Ra)int was compared with EToPM calculated with the Penman-Monteith method using daily climatic data coming from local meterological stations. The spatialized ETo identified differences in the seasons under the analyzed climate scenarios, which was not possible with EToPM. The methodology for estimating spatialized EToMJS(ψair;Ra) over large areas showed acceptable accuracy, despite being laborious for extensive coverage. Temporal ETo showed satisfactory statistical accuracy (RMSE = 0,65 mm dia⁻¹; r = 0,73; MAPE = 5,94%; NSE = −1,2; d = 0,04), although the limitations of the images and the linear interpolation algorithm limited the monitoring of daily EToPM variations.