A hybrid transposition and sky radiance model for urban solar mapping at various latitudes

This study explores the impact of latitude, temporal resolution, and surface orientation on performance variability of solar irradiance model chains for urban solar mapping tools, highlighting the need for accurate model selection — through validation — when significant inter-model differences occur. However, as collecting high-quality observational data for validation is challenging, a hybrid transposition and sky radiance probabilistic model — designed to minimize model-specific biases and enhance the simulation of solar irradiance dynamics — was conceptualized, developed, and experimentally validated. To assess the variability of solar irradiance estimations, 450 model chains were initialized for 21 locations (latitude varies between 35° S to 65° N), 4 temporal resolutions (from one-minute to one-month), and 24 surface orientations. The spread of the global tilted irradiance (Egt) estimations from various models was quantified using the normalized interquartile range and the coefficient of variation. Afterwards, three probabilistic models were compared: the full transposition, the full sky radiance, and the hybrid transposition and sky radiance probabilistic models. Estimated Egt was validated using observations from Trondheim, Stuttgart, and Madrid. Results highlighted that model selection is important for latitudes higher than 50° N, time resolutions higher than one-day, and highly tilted surfaces. Further correlation analyses demonstrated that high variability among model outputs is associated to low solar accessibility (Pearson coefficient up to -0.85). Finally, using sky radiance models either as alternatives to or in combination with transposition models in the model chain improved the accuracy in Madrid and Trondheim case studies, while in Stuttgart, model performance resulted slightly reduced.