Calibration and validation of the safer evapotranspiration model using Landsat 8 and eddy covariance data in vineyards

Abstract

Accurate estimates of actual evapotranspiration (ET a ) are essential for irrigation management in vineyards. This study locally calibrated and validated the Simple Algorithm for Evapotranspiration Retrieving (SAFER) model for California vineyard systems using Landsat 8 imagery and multi-year eddy covariance (EC) measurements collected between 2013 and 2016 in the Central Valley of California. Two calibration approaches were implemented: k-fold cross-validation and a split-sample (two-year) strategy. SAFER performance was evaluated using both closed and unclosed energy balance EC fluxes to assess the influence of energy balance closure on model calibration and validation. Energy balance closure ratios ranged from 0.67 to 0.95, with moderate interannual variability and lower closure observed during some periods. Local calibration of SAFER resulted in regression coefficients that differed substantially from the original parametrization. Compared to the original coefficients, root mean square error (RMSE) was reduced by approximately 71–77%, with values below 0.70 mm d − 1 for both calibration approaches. Model validation showed strong agreement between SAFER ET a and EC observations after calibration. RMSE ranged from 0.57 to 0.71 mm d − 1 and mean absolute error (MAE) ranged from 0.46 to 0.55 mm d − 1 . Overall, the results indicate that SAFER provides reliable estimates of daily vineyard ET a when locally calibrated, supporting its application for field-scale irrigation management in perennial cropping systems.

Bibliographic Information