Evaluating state-of-art statistical downscaling and analogs approaches on historical climate statistics over European regions

Climate change impacts assessment crucially relies on climate information at high temporal and spatial resolutions, not available from global climate models (GCMs) involved in the coupled model intercomparison project (CMIP). At the same time, dynamically downscaled regional climate model simulations do not provide global-scale coverage and in several cases are computationally too expensive.

For this reason, downscaling techniques are commonly applied to bridge the resolution gap between GCM simulations and impact studies. The most common methodology is the statistical downscaling approach. However, statistical downscaling fast computation comes at a price, it does not account for physical and dynamic processes potentially inflates temporal variability of the original simulations’ resolution. Given this limitation, the analogs technique may represent a valuable alternative since it considers both large and local scales dynamics balanced by a reasonable increase in computational costs.

The present study explores differences, added value, and limitations characterizing state-of-the-art bias adjustment/statistical downscaling based on a stochastic quantile mapping approach and the analogs technique. In particular, the comparison applies to the data computed in the inter-sectoral impact model intercomparison project (ISIMIP) and data obtained by applying the analogs method based on the same ISIMIP reference dataset. The two approaches are compared and evaluated in terms of the historical period observed statistics reproduction for a few climate variables over European regions.

This study is performed in the framework of GoNEXUS and NEXOGENESIS European projects.