Infilling methods for monthly precipitation records with poor station network density in Subtropical Argentina

Abstract

Precipitation plays a crucial role from a social and economic perspective in Subtropical Argentina (STAr). Therefore, it renders the need for continuous and reliable precipitation records to develop serious climatological researches. However, precipitation records in this region are frequently inhomogeneous and scarce, which makes it necessary to deal with data filling methods. Choosing the best method to complete precipitation data series relies on rain gauge network density and on the complexity of orography, among other factors. Most comparative-method studies in the literature are focused on dense station networks while, contrastingly, the STAr's station network density is remarkably poor (between 10 and 1000 times lower). The research aims at assessing the performance of several interpolation methods in STAr. In this sense, the performance of a large number of interpolation methods was evaluated for dry and wet seasons, interpolating raw monthly data and their anomalies applied to different time-series subsets. In general, most methods performances improve when applied to anomalies in the seasonal time-series subset. Multiple Linear Regression (MLR) stands out as the method with the best performance for infilling precipitation records for most of the regions regardless of orography or season. Despite the bibliography invokes that kriging interpolation methods are the best ones, in this work the performance of kriging methods was similar to the one of the Inverse Distance Weighted method (IDW) and the Angular Distance Weighted method (ADW, the method used to generate CRU precipitation dataset).