Analysis and comparison of weather models for solar irradiance forecasts in Sweden

Abstract: Solar radiation is affected in a variety of different ways through the atmosphere and on its way down to the Earth’s surface. Clouds are the main factor in this attenuation of solar radiation, but aerosols, ozone and other different gases have a significant impact too. Understanding these processes is fundamental to solar forecasting. However, the atmosphere is constantly changing, which makes forecasting weather a challenging task even with today’s complex numerical weather prediction models. In this study of solar irradiance, two different numerical weather models are analyzed: the global IFS-model and the regional ensemble prediction system MEPS. Both models include normal forecasts and ensemble forecasts. This analytical comparison was made in forecasts of up to 24 hours of three locations in Sweden over the time period of 2017-04-01 to 2017-06-30. The purpose was to see differences in the accuracy of the models, and also to see the influence of ensemble forecasts. Ensemble forecasts are used to better handle uncertainties in the weather and are more frequently used in today’s weather prediction. Observation data collected from SMHI’s radiation network was used as reference values in order to get forecasting errors. A generally better result for the ensemble forecasts was identified. Some variations between the stations were also detected where especially the result from one of the three stations was different. That was a location far away from the other two indicating that topography and climatology can affect the precision of the weather models. The regional high-resolution MEPS-model had overall better solar irradiance forecasts during the analyzed time, but the model had a consistent negative bias. A consistent problem with the MEPS-model is most likely that it predicts too much clouds which leads to less amount of solar irradiance in the forecasts.