GeoDS (v.1.0) : a simple Geographical DownScaling model for long-term precipitation data over complex terrains

Abstract. Global climate models offer the most comprehensive description of the climate system and its internal processes to date but current computational capabilities typically restrict their spatial resolution to the order of tens of kilometers when multi-decennial or longer simulations are required. For climate applications, it is notoriously difficult to generate high spatial resolution data over long timescales (typically millennial). Over the years, various downscaling techniques have been developed to generate fine scale data from climate models outputs but they often exhibit important limitations when applied over long periods of time. Building on previous efforts, we present a simple topography-based model (GeoDS) to downscale precipitation fields in complex areas, adapted to paleoclimate studies involving multi-millenia simulations. With a limited amount of inputs from a climate model and high resolved geographical information, the model computes, for each time step and every grid point, a topographic exposure index used to distribute precipitation into a high-resolution spatial grid. This dimensionless quantity represents the exposure of surfaces to dominant windward incoming airflows, assumed to bring most of the humidity, and only depends on large scale winds and terrain configuration. The model is first tested under current climate conditions over part of the European Alpine region due to the availability of field data for comparison ; the complexity of both regional topography and climate conditions making it a good test of the proposed methodology. The relative effects of the model’s parameters are assessed as well as the capacity of GeoDS to reproduce the spatial precipitation distribution of a well-resolved gridded target dataset. Despite uncertainties regarding the correct wind fields to choose as input, and the dependency of the model to the temporal resolution of the large-scale data to downscale, we show that the procedure is able to capture most of the patterns occurring at fine spatial scale while being computationally inexpensive. We also demonstrate that the physical base underlying our work grants the model valuable robustness when used outside the calibration framework. This notably opens promising prospects for the application of GeoDS in paleoclimate contexts while providing a flexible, open source and well documented downscaling tool for the climate community.