Solar Forecasts During Broken Cloud Conditions: Improvements in WRF-Solar v2

Shallow cumuli present a significant challenge for solar forecasting applications due to the dynamic nature of the clouds. In addition, fields of broken clouds can lead to significant solar power ramps. Due to their small size, this particular type of cloud is poorly represented in mesoscale meteorological models, such as the Weather Research and Forecasting (WRF) model that is the basis of many numerical prediction systems. For this reason, the treatment of shallow cumuli is one area of focus for WRF-Solar version 2. Standard cumulus parameterizations utilized in WRF were developed to represent the impact of deep convective clouds on the thermodynamic structure of the atmosphere rather than their impact on surface irradiance. Different parameterizations have been designed to address this shortcoming and to improve forecasts of shallow cumuli and their impact on surface solar irradiance. In this study results from simulations using the Kain-Fritsch (KF), Kain-Fritsch Cumulus Potential (KF-CuP), and Deng parameterizations are compared. Three different geographic locations in the United States were selected based on the availability of high-quality data and a relatively large frequency of shallow cumuli, including the Department of Energy’s Atmospheric Radiation Measurement (ARM) User Facility’s Southern Great Plains (SGP) site, and NOAA’s Goodwin Creek (GWN) and Penn State (PSU) SURFRAD sites.

Overall, the KF-CuP parameterization outperformed the KF and Deng parameterizations in regard to the Global Horizontal Irradiance (GHI), with a reduction of the root-mean-squared error (RMSE) and mean absolute error (MAE) of greater than 10%. The overall improvement in Direct Normal Irradiance (DNI) was smaller, approximately 5% across the three sites. The results varied significantly with location. An improvement of 30% in RMSE of GHI was found at SGP. More moderate changes were found in RMSE of GHI GWN and PSU, which ranged between 7% improvement with the KF-CuP parameterization compared to the Deng parameterization, to a difference of -1% in the MAE of GHI at PSU using an interval of 60 minutes. The results for DNI were generally consistent to those for GHI, with the exception that the Deng parameterization outperformed the KF-CuP scheme at PSU for both RMSE and MAE regardless of the time interval that was used.