Hydrology and Earth System Sciences (Nov 2010)

Spatial uncertainty assessment in modelling reference evapotranspiration at regional scale

  • G. Buttafuoco,
  • T. Caloiero,
  • R. Coscarelli

DOI
https://doi.org/10.5194/hess-14-2319-2010
Journal volume & issue
Vol. 14, no. 11
pp. 2319 – 2327

Abstract

Read online

Evapotranspiration is one of the major components of the water balance and has been identified as a key factor in hydrological modelling. For this reason, several methods have been developed to calculate the reference evapotranspiration (ET<sub>0</sub>). In modelling reference evapotranspiration it is inevitable that both model and data input will present some uncertainty. Whatever model is used, the errors in the input will propagate towards the output of the calculated ET<sub>0</sub>. Neglecting the information about estimation uncertainty, however, may lead to improper decision-making and water resources management. One geostatistical approach to spatial analysis is stochastic simulation, which draws alternative and equally probable, realizations of a regionalized variable. Differences between the realizations provide a measure of spatial uncertainty and allows to carry out an error propagation analysis. <br><br> The aim of this paper is to assess spatial uncertainty of a monthly reference evapotranspiration model resulting from the uncertainties in the input attributes (mainly temperature) at a regional scale. A case study was presented for the Calabria region (southern Italy). Temperature data were jointly simulated by a conditional turning bands simulation with elevation as external drift and 500 realizations were generated. Among the evapotranspiration models, the Hargreaves-Samani model was used. <br><br> The ET<sub>0</sub> was then estimated for each set of the 500 realizations of the input variables, and the ensemble of the model outputs was used to infer the reference evapotranspiration probability distribution function. This approach allowed for the delineation of the areas characterised by greater uncertainty, to improve supplementary sampling strategies and ET<sub>0</sub> value predictions.