Evaluation of ESD methods for climate change projections over Spain

The study presents an evaluation and selection of global climate models for downscaled projections over Spain. The Spanish Meteorological Service (AEMET) is responsible for the elaboration of regionally downscaled climate projections over Spain according to the second National Adaptation Plan to Climate Change (PNACC-2 2021-2030). Regionally downscaled climate projections allow for increased spatial resolution compared to those provided by global climate models. AEMET develops regionally downscaled climate projections over Spain by applying empirical-statistical downscaling (ESD) methods to projections from a set of global climate models. Regionally downscaled climate projections are necessary for conducting impact and vulnerability studies that require data at high spatial resolution.

The present work has two objectives. The first one is to evaluate and select a reduced ensemble of global climate models from CMIP6 in order to downscale their projections. The second objective is to compare different ESD methods in order to choose the best performing method for downscaling daily data of the following climatic variables of interest: maximum temperature, minimum temperature, and accumulated precipitation. The evaluation of global climate models was carried out using the GCMeval tool [1]. The comparison of ESD methods is based on the analysis of indicators such as bias, coefficients of temporal and spatial correlation, root mean square error, and future trends, and it was carried out using pyClim-SDM [2], the statistical downscaling software developed at AEMET. The results indicate the selection of eleven global climate models for downscaling and the choice of specific methods for each climatic variable: regression-analogues (MLR-ANA) for temperatures and eXtreme Gradient Boost (XGB) for precipitation. Additionally, bias correction using Quantile Delta Mapping is recommended to improve projections. This research lays a robust groundwork for future studies on climate change in Spain, highlighting its implications for the population, socio-economic sectors, and ecosystems.

Ultimately, the key findings from the downscaling of projections over Spain using the representative ensemble of global climate models from CMIP6 and employing the resulting downscaling techniques from the prior assessment are presented.

 

[1] Parding, K. M. et al., (2020), GCMeval – An interactive tool for evaluation and selection of climate model ensembles. Climate Services, Volume 18, 100167. https://doi.org/10.1016/j.cliser.2020.100167

[2] Hernanz, A., et al. (2023), pyClim-SDM: Service for generation of statistically downscaled climate change projections supporting national adaptation strategies. Climate Services, Volume 32, December 2023, 100408. https://doi.org/10.1016/j.cliser.2023.100408