Sensitivity of Solar Generation Forecast Skill to WRF Parameterization Choices under Central Valley and Coastal Fog Conditions

The state of California has experienced a rapid rise in solar generation capacity to over 20 GW of installed capacity, with nearly all of this capacity installed in the past ten years. Much of this capacity lies within either within California’s Central Valley or along the Pacific Coast. Both locations are subject to occasional persistent fog that can sharply reduce generation from photovoltaic panels. Prediction of the occurrence and break-up of these fog events is thus crucial to day-to-day prediction of aggregate solar generation in California. The California Energy Commission has funded a project to improve these forecasts by deployment of additional observation stations and by modeling work using the WRF and WRF-Solar models aimed to define the best set of parameterization choices for fog and coastal cloud prediction in our study area and the best procedure to derive value from the additional observations.

Special observations deployed for this project include several radar and sodar wind profilers, three ceilometers and a microwave radiometer, located at six stations around the central valley and five locations in the Los Angeles Basin.

In this presentation we will present results from a series of experiments designed to identify the best choice of parameterization methods and model resolution and domain to skillfully predict central valley and coastal fog. Using an up-to-date WRF configuration that includes developments from the WRF-Solar project we compare results for multiple boundary layer schemes (Yonsei University with and without top-down mixing, Mellor-Yamada-Janjic, Grenier-Bretherton-McCaa), radiative schemes (FARMS, RRTMG), and microphysics schemes (Thompson and Thompson Aerosol Aware). Initial results show relatively strong sensitivity on boundary layer scheme choice, with weaker sensitivity to radiative scheme in Central Valley fog cases.