Kriging-based methods for spatial downscaling of climate models

As the impacts of climate change persist, it becomes increasingly crucial to accurately assess future climate projections. High-resolution climate models play a pivotal role in conducting site-specific impact assessment studies, facilitating informed decision-making, and effective policy implementation. One of the main challenges regarding climate model data is to achieve the necessary spatial resolution while ensuring an accurate representation of climate variables. This study aims to enhance the resolution of climate model outputs provided within the CMIP6 initiative. These models have been run using General Circulation Models (GCMs) and often prove inadequate for regional or subregional studies. Meanwhile, Regional Climate Models (RCMs), which employ a computationally expensive dynamical downscaling, are still under development. To address the coarse resolution of GCMs, alternative commonly employed methods include spatial interpolation. Given that GCM data represent averaged characteristics across grid cells and to ensure physically consistent datasets, the use of conservative extrapolation procedures is recommended for downscaling. This study presents a geostatistical method for spatial downscaling based on block-cokriging. Elevation is used as an auxiliary variable to capture topographical influences on climate variables. The methodology is tested to downscale daily precipitation and maximum and minimum temperatures in the Emilia-Romagna region in northern Italy. Nevertheless, the designed methodology can be applied in other regions. A comparative analysis against alternative methods, such as nearest-neighbor and second-order conservative interpolation techniques, is also performed.