Spatial correlation structures of wind speed and irradiance in Europe as modelled in regional climate models and the ERA5 reanalysis

Abstract

Objective & Background For comprehensive energy systems analysis considering high shares of weather-dependent renewables, the spatiotemporal characteristics of meteorological input data (in particular, wind speed and irradiance) are of considerable interest. Studies aiming to evaluate the potential effects of future climate change rely on model-based projections of these variables. In order to reach as high a spatiotemporal resolution as possible, global climate models (GCM) can be used as drivers for regional climate models (RCM). In doing so, global climate effects are considered on large scales, while the higher resolution computations are limited to the domain of interest. Here, we analyse such regional climate model outputs as well as reanalysis data with regards to their suitability in energy systems analysis, using spatial correlation structures of the wind and solar resource anomalies. Method Based on 10 years of historical experiments of EURO-CORDEX regional climate model output and ERA5 reanalyses, we compute climate anomalies of both wind speed and irradiance for each point in space and time by subtracting the multi-year mean of each point from the instantaneous values at that point. In doing so, we largely remove effects of seasonal and diurnal cycles, especially in the case of irradiance. From each of these wind speed and irradiance anomaly fields, we sample distinct two-dimensional correlation structures (i.e., maps of the correlation coefficient between each pixel’s time series and a reference pixel’s series). Increasing the number of reference pixels hence increases the number of spatial correlation structures available in the analysis. We finally estimate the degree of similarity between the spatial correlation structures of each EURO-CORDEX member and its corresponding ERA5 counterpart using the coefficient of pattern correlation (i.e., the correlation coefficient of the mapped spatial correlation structures in vector form). Principal Findings Kernel density estimates and box-plot statistics of all available pattern correlations show that the wind speed anomaly spatial correlation structures tend to be better represented than their irradiance counterparts. Between different driving global models, the performance is relatively similar for wind speed anomaly correlation structures, while solar irradiance anomaly structures can be associated with a few unusually low pattern correlations for some global models. In terms of regional climate models, these cases of low pattern correlation values are largely associated with the SMHI-RCA4 regional model during the summer. Discussion This initial comparison of CORDEX model output and ERA5 reanalysis based on correlation structures in wind speed and solar irradiance anomalies allows to help in the selection of global and regional models when compiling an ensemble for use in energy systems analysis. In this context, the spatial correlation structures of wind speed need less attention when selecting model outputs, while the irradiance anomaly projections of the regional climate model SMHI-RCA4 should be handled with caution. Increasing the number of reference points used in the spatial correlation structure calculations may lead to further insights regarding potential regional differences in CORDEX model performances.