Characterizing solar radiation zones in the Canary Islands using cluster analysis

Abstract

This study categorizes solar radiation patterns across the Canary Islands using the clearness index (Kt) and K-means clustering. We have used a comprehensive dataset compiled from multiple meteorological sources to identify distinct radiation zones. The K-means algorithm, applied to the clearness index data, revealed four unique radiation zones within the archipelago. To validate these clusters, we compared them against observed radiation data using several performance metrics: Mean Absolute Bias Error (MABE), Mean Absolute Percentage Error (MAPE), Mean Bias Error (MBE), Root Mean Square Error (RMSE), and relative RMSE (rRMSE), resulting four zones. Each of the four identified zones is characterized by varying frequencies of cloudy, partially clear, and very clear days. Zone 1 exhibits the lowest radiation, defined by a high prevalence of cloudy days, while Zone 4 shows the highest radiation due to a significant proportion of clear days. Our findings emphasize the need for tailored calibration to improve predictive accuracy within each specific zone. Although high prediction errors were observed, this clustering approach effectively categorizes solar radiation patterns in the Canary Islands, suggesting that further model refinement could significantly enhance the accuracy of solar radiation forecasts. © 2025 The Author(s).