Spatio-temporal interpolation of rainfall data in western Mexico

One of the most common problems related to meteorological information is the missing registers. This lack of data generates uncertainties in the analysis of climate, hydrology, and natural disasters. In Mexico, very often, this problem is present in all the meteorological stations of the country. In this study, we apply two well-established spatial interpolation methods that have report competitive performance in the specialized literature: the Inverse Distance Weighting (IDW) and Modified Inverse Distance Weighting (MIDW); and they are compared with a proposal of spatio-temporal regression using an artificial neural network of the kind of multilayer perceptron (MLP). The results show that using a combination of spatial and temporal data with a low number of predictors is competitive with the comparing methods using a high number of predictors. We compare the methods through statistical measures of the error for 31 meteorological stations of the Jalisco state in the period of 2002-2006.