Support material for article "Evapotranspiration Response to Climate Change in Semi-Arid Areas: Using Random Forest as Multi-Model Ensemble Method" published in the journal Water

Support material for article "Evapotranspiration Response to Climate Change in Semi-Arid Areas: Using Random Forest as Multi-Model Ensemble Method" published in the journal Water (https://doi.org/10.3390/w13020222).In the document “GomarizCastillo_etal_README.pdf” the work process is described, including its use, the description of the files and the reference data.This support material presents the R code used to carry out the ET0 ensembles using as example the weather station 7002A of the AEMET network. We have tried to implement a highly reproducible code to be used with one or more weather stations in different study areas and with different data sources. Although the main objective is to perform and evaluate individual regionalized ET0 models and multi-model ensembles, the process is applicable to any other problem related to ensembles, variables or data sources related to Climate Change. Two functions perform the process:<br>“<i>fun.ETPHargreaves</i>”: estimates ET0 series using the Hargreaves equation calibrated in the study area according to Gomariz-Castillo, Alonso-Sarria, and Cabezas-Calvo-Rubio (2018) (final in Section (A) Processing data in Fig. 2 of the paper). Among other options, the function allows a) to estimate ET0 either if Rs is known or if it has to be estimated from Ra, b) to modify the default parameters and c) to include calibration parameters based on the methodology proposed by Allen et al. (1994).“<i>fun.ensembles</i>” validates the individual regionalized models, estimates and validates the multi-model ensemble methods, and generates forecasts for the RCP scenarios (Section (B) Individual models and ensembles in Fig. 2 of the paper).The final subsection of this supplementary material describes in detail both functions and the required and optional parameters.The following files are included as additional material:<i>main.R</i>: Main code to launch the process.<i>data/*</i>: Three data files for HISTORICAL, RCP4.5 and RCP8.5 scenarios (weather station 7002A).<i>src/functions.R</i>: fun.ETPHargreaves and fun.ensembles functions.The example version includes the possibility of parallelize the <i>caret </i>(machine learning based ensembles) functions, but the rest of the process is not parallelized to ensure compatibility with any platform. The presented process has large computational requirements when applied to daily time series.