Long-term and short-term estimations of snow water availability for the Barasona reservoir

The Barasona reservoir is one of the three main water supplies for the Aragon and Catalonia Irrigation District, the principal provider for 98.000 ha, including 37 municipalities and thousands of agricultural holdings and livestock farms. Thus, it is of great interest to have a correct understanding of the reservoir’s hydrological processes. In this study we combine remote sensing and in-situ data to estimate the amount of water in the form of snow available for the Barasona reservoir and predict its input flow, in addition to estimate the dates of the beginning and ending of thaw. First, we estimate the snow cover area (SCA) from MODIS and Sentinel 2. For Landsat 8, an equivalent algorithm is implemented to derive the snow cover. The result is a time series of SCA from 2009 to 2023. With this SCA together with snow height measured at the catchment we estimate the snow water equivalent (SWE). All this data along with in-situ information from weather stations is fed to an AI model that predicts the Barasona reservoir input flow in the short term. The base model achieved an accuracy of R2=0.691. If the flow from previous days was inputted as a variable, the performance increased to R2=0.781. Regarding long-term predictions, the estimated dates of the beginning and end of snowmelt are March 23 and May 26, with two weeks of standard deviation. The SWE available shows large inter-annual variability, ranging from 48.82 hm3 during years of drought to 228.77 hm3 when snow is abundant, with an average of 120.27 hm3, and a standard deviation of 58.15 hm3.