Performance of filling gaps in reference evapotranspiration data for the western region of Paraná

Abstract

The occurrence of failures in the readings of meteorological variables at surface stations can compromise the consistency of the historical series, making their use unfeasible or impairing. In this sense, the objective of this work was to evaluate the performance of methods for filling gaps in historical series of reference evapotranspiration data (ETo), considering the western region of Paraná. Two stations were used as reference, test station 1 (Foz do Iguaçu) and test station 2 (Marechal Cândido Rondon). Five alternatives for filling in gaps were evaluated: Regional Weighting with four neighboring stations (PR-4), Regional Weighting with one neighboring station (PR-1), Inverse Square Distance (IQD), Least Square (MQ) and data from NASA/POWER (NP). The methods were statistically compared using Pearson's correlation (r), Willmott's agreement (d), Nash-Sutcliffe efficiency coefficient (CE), Root Mean Square Error (RMSE) and Mean Absolute Error (EMA). Under the conditions in which the research was carried out, all the methods evaluated showed adequate performance, indicating the potential to be used as a method of filling in gaps. The IQD method stood out among the others, since their estimates were classified as “strong correlation” (r = 0.868 - test station 1 and r = 0.887 - test station 2) and “adequate” (CE = 0.999 - in both test stations). In addition to “d” values very close to 1 and low estimation errors (RMSE = 1.22 mm d⁻¹ and EMA = 0.82 mm d⁻¹, test station 1; RMSE = 0.87 mm d⁻¹ and EMA = 0.68 mm d⁻¹), being the recommended method for filling series failures in historical ETo data series.

Keywords
spatial interpolation; NASA/POWER; regional weighting; water resources