Abstract
Recent advancements in urban solar cadaster platforms have extended the investigated domain beyond rooftops to encompass vertical building façades. However, experimental validation remains limited despite the potential impact of these tools, which are openly accessible to the general public. This study evaluates the performance of the Geneva solar cadaster’s simulation engine (CadSol) in an urban canyon in Geneva, Switzerland, and a standalone building in Trondheim, Norway. Modeled façade irradiance is validated against pyranometer measurements for both the operational configuration of CadSol (hourly simulations for one representative day per month) and its possible application at 1-minute resolution. Under the operational configuration, CadSol adequately captures the seasonal variability of façade irradiance in both case studies, albeit with better agreement in Trondheim than in Geneva, where systematic overestimation affects the most irradiated façade and lower floors. At 1-minute resolution, model performance analysis confirms this trend, with Trondheim achieving higher accuracy (𝑅2=0.93, MAE = 19.23 𝑊∕m2) than Geneva (𝑅2=0.76, MAE = 26.77 𝑊∕m2). In Geneva, CadSol generally overestimates irradiance, with larger errors under overcast skies (MPE=−30%). In Trondheim, overestimation is less pronounced and is mainly observed during overcast periods (MPE up to −25%). Overall, the largest errors are associated with shadow mask inaccuracies (i.e., misclassifications of shaded/sunlit states), alongside biases in decomposition and transposition modeling. Improving the urban geometry model, the shadow masking algorithm, and the irradiance modeling is therefore important to increase the accuracy of façade irradiance outputs.