However, marine layer cloud cover is a meteorologically complex weather phenomenon that is difficult to predict. Unique to southern California, marine layer clouds form when moist air near the ocean surface is mixed vertically. As the air rises, it expands, cools, and clouds condense. Contributing to marine layer formation is the presence of a strong temperature inversion. The oceanic temperature inversion occurs when dry, warm air is driven to sea via the large-scale circulation patterns. Surface air is cooled by the ocean while air aloft remains warm, resulting in the temperature inversion. This creates a stable layer that determines the thickness of marine layer clouds and suppresses further vertical mixing that would dissipate cloud cover. Generally, marine layer cloud cover is thickest shortly before sunrise. As temperature increases throughout the day, cloud cover dissipation occurs. Beginning at the cloud edge, dissipation uniformly proceeds towards the coast.
To accurately simulate this, physics-based models such as numerical weather prediction (NWP) are generally the most accurate methods. Day-ahead statistical models, though accurate when marine layer cloud cover is observed for consecutive days, cannot predict changes in the general circulation patterns that influence temperature inversion characteristics and the cloud cover extent. Thus, when synoptic conditions change, statistical models fail. Imagery-based methods, such as satellite cloud motion vector models, provide excellent characterizations of current cloud conditions. However, since marine layer dissipation is primarily dependent on local temperature changes and not cloud motion, these models have difficulty in predicting marine layer evaporation and are inaccurate. Only can physics-based models accurately predict the conditions for which marine layer clouds form and dissipate.
However, the operational NWP (e.g. the North American Mesoscale (NAM) model) have been shown to under-predict cloud cover near the coast for southern California. To address this, a high-resolution, cloud assimilating NWP using the Weather and Research Forecasting (WRF) model has been developed and implemented at the University of California, San Diego for solar irradiance forecasting. Using satellite imagery, clouds are populated into the initial conditions by directly modifying modeled water vapor mixing ratios. Overall, this model has been shown to reduce mean absolute error by over 10% when compared to the NAM for irradiance forecasts. Furthermore, simulating cloud cover at high-resolution allows for localized formation and dissipation effects to be explicitly resolved. Specifically, the Catalina Eddy and oceanic-inland temperature gradient affect local wind currents and the positioning of marine layer clouds. The accuracy of these phenomena and their affect on solar irradiance forecasting is quantified. Overall, it is determined that, under some conditions, modeled surface heating is too great, disrupting the large scale circulation pattern and reducing the accuracy of the positioning of marine layer cloud cover.