An ARF-based method to downscale sub-daily extreme precipitation return values

Estimating sub-daily precipitation return values at the kilometer-scale is critical for climate-risk assessments. However, this remains a challenge due to the strong intermittency of extremes, heterogeneous observational networks, and the breakdown of stationarity assumptions under climate change. Furthermore, coarse-resolution global climate models (e.g., CMIP6) systematically underestimate local extremes by smoothing convective processes and orographic gradients, thereby blurring hotspots that are critical for impact modelling.

We present an innovative statistical downscaling method that repurposes Areal Reduction Factors (ARFs), traditionally used to relate point rainfall to areal averages, as a resolution-bridging tool for extreme precipitation. By comparing return values derived from high-resolution (COMEPHORE, ~1 km2) and coarse-resolution (CERRA, ~25 km2) reanalyses, we compute spatially varying ARF maps. These maps quantify the attenuation of extremes induced by coarse spatial resolution and serve as multiplicative scaling factors to translate coarse-resolution outputs into 1-km products while preserving the large-scale climate signal.

This framework is validated against independent rain-gauge observations across multiple return periods and seasons. Finally, we apply the method to CMIP6 simulations to generate 3-hourly, 1-km precipitation return values for both historical and future periods.

This approach provides a computationally efficient and climate-change-consistent pathway to generate high-resolution hazard metrics, without the prohibitive cost of convection-permitting regional climate simulations.