Copula-based downscaling of spatial rainfall: a proof of concept

Abstract

Abstract. Fine-scale rainfall data is important for many hydrological applications. However, often the only data available is at a coarse scale. To bridge this gap in resolution, stochastic disaggregation methods can be used. Such methods generally assume that the distribution of the field is stationary, i.e. the distribution for the entire (fine-scale) field is the same as the distribution of a smaller region within the field. This assumption is generally incorrect and we provide a proof of concept of a method to estimate the distribution of a smaller region. In this method, a copula is used to construct a bivariate distribution describing the relation between the scales. This distribution is then used to estimate the distribution of the fine-scale rainfall within a single coarse-scale pixel, by conditioning on the coarse-scale rainfall depth.