Comparison of three data-driven techniques in modelling the evapotranspiration process

摘要

Evapotranspiration is one of the main components of the hydrological cycle as it accounts fornmore than two-thirds of the precipitation losses at the global scale. Reliable estimates of actualnevapotranspiration are crucial for effective watershed modelling and water resourcenmanagement, yet direct measurements of the evapotranspiration losses are difficult andnexpensive. This research explores the utility and effectiveness of data-driven techniques innmodelling actual evapotranspiration measured by an eddy covariance system. The authorsncompare the Evolutionary Polynomial Regression (EPR) performance to Artificial Neural Networksn(ANNs) and Genetic Programming (GP). Furthermore, this research investigates the effect ofnprevious states (time lags) of the meteorological input variables on characterizing actualnevapotranspiration. The models developed using the EPR, based on the two case studies at thenMildred Lake mine, AB, Canada provided comparable performance to the models of GP andnANNs. Moreover, the EPR provided simpler models than those developed by the other data-drivenntechniques, particularly in one of the case studies. The inclusion of the previous states of theninput variables slightly enhanced the performance of the developed model, which in turnnindicates the dynamic nature of the evapotranspiration process.