Estimating Reference Evapotranspiration using Artificial Neural Networks in Cold Semi-arid Regions: Study Performed in the Peruvian Altiplano

Evapotranspiration (ETo) is one of the most important variables of the water cycle when water requirements for irrigation, water resource planning or hydrological applications are analyzed. In this context, models based on artificial neural networks (ANNs) of the retro-propagation type can be an alternative method to estimate ETo over on Andean region. The objective of this study is to develop ANN models to estimate ETo for the Peruvian Altiplano using input variables such as maximum air temperature (Tmax), minimum air temperature (Tmin), hours of sunshine (Sh), relative humidity (Rh) and wind speed (Wv). Daily climatic datasets recorded at 12 meteorological stations between 1963 and 2015 were select in this study. For evaluation reason, the ETo calculated using the Penman - Monteith method (FAO - PM56) was also considered. The ET obtained from modeling was appropriately estimated by the ANN models (r=0.91). This appropriate estimation was determined using ANN models where the main input variable is Tmax, followed by Sh and Wv or combinations between them. These results also indicate that highest performances of the ANN models are associated with its location inthe semi-arid or semi-humid regions of the Peruvian Altiplano. Likewise the results also suggest that both the proximity to Lake Titicaca and the altitude may condition the efficiency of the ANN models to estimate adequately the ETo. Therefore, ANN models represent a great option to replace the FAO-PM56 method, when ETo data series are scarce.