Multi-initial-conditions and Multi-physics Ensembles in the Weather Research and Forecasting Model to Improve Coastal Stratocumulus Forecasts for Solar Power Integration

Abstract:

In coastal Southern California, variation in solar energy production is predominantly due to the presence

of stratocumulus clouds (Sc), as they greatly attenuate surface solar irradiance and cover most

distributed photovoltaic systems on summer mornings. Correct prediction of the spatial coverage and

lifetime of coastal Sc is therefore vital to the accuracy of solar energy forecasts in California.

In Weather Research and Forecasting (WRF) model simulations, underprediction of Sc inherent in the

initial conditions directly leads to an underprediction of Sc in the resulting forecasts. Hence,

preprocessing methods were developed to create initial conditions more consistent with observational

data and reduce spin-up time requirements.

Mathiesen et al. (2014) previously developed a cloud data assimilation system to force WRF initial

conditions to contain cloud liquid water based on CIMSS GOES Sounder cloud cover. The Well-mixed

Preprocessor and Cloud Data Assimilation (WEMPPDA) package merges an initial guess of cloud liquid

water content obtained from mixed-layer theory with assimilated CIMSS GOES Sounder cloud cover to

more accurately represent the spatial coverage of Sc at initialization.

The extent of Sc inland penetration is often constrained topographically; therefore, the low inversion

base height (IBH) bias in NAM initial conditions decreases Sc inland penetration. The Inversion Base

Height (IBH) package perturbs the initial IBH by the difference between model IBH and the 12Z

radiosonde measurement.

The performance of these multi-initial-condition configurations was evaluated over June, 2013 against

SolarAnywhere satellite-derived surface irradiance data. Four configurations were run: 1) NAM initial

conditions, 2) RAP initial conditions, 3) WEMPPDA applied to NAM, and 4) IBH applied to NAM. Both

preprocessing methods showed significant improvement in the prediction of both spatial coverage and

lifetime of coastal Sc. The best performing configuration was then used to create a multi-parameter and

multi-physics ensemble. The ensemble forecast system is implemented operationally for San Diego Gas

& Electric Company to improve system operations.