Comparison between Spatial Interpolation Methods of Daily Precipitation Stochastic Models Parameters — Application to Basins in Southern Spain.

Europe is immersed in an age in which the need to improve water management and planning models has become increasingly evident. This phenomenon is even more patent following the repeated announcements of climatic changes that will significantly affect areas like the Iberian Peninsula. In already deficit watersheds such as the Guadalquivir Basin of southern Spain, water supplies are expected to diminish, while water demands will increase. Given this scenario, it is essential to develop tools that will aid in determining the behavior of a fundamental climatic variable at the regional level, namely precipitation. Numerous methods which describe rainfall at the regional scale are used to generate synthetic daily precipitation records in locations where observed rainfall records are not available. These methods take into account the effect of geographical variables such as the proximity between stations, or are purely statistical such as the Kriging method. In order to improve the results of classic spatial interpolation methods, new methods are currently being developed to account for variables such as the "proximity between points" not only in the geographical sense (such as distance to the sea and the height above sea level), but also from a climatic standpoint (annual rainfall, average number of wet or dry days and mean precipitation per wet day). In order to estimate the coefficients of the Markov Chain-Mixed Exponential stochastic model, we compare simple interpolation methods such as the nearest station method, the arithmetic mean of the nearest stations or the inverse distance squared method to more complex methods including the Kriging method (with a statistical base) or methods which take into account the geographical and climatic variables described above, namely the multiple regression method, the weighting method developed by Moreno and Roldán and its subsequently improved version and the method based on the application of artificial neural networks to estimate climatic variables.