Abstract
We have applied the Model Output Statistics (MOS) approach to the prediction of cloudiness. Final guidance warm and cool season forecasting equations were developed by screening forecast fields from the primitive equation and trajectory models. We derived separate equations for each of 233 stations to estimate the probability of clear, scattered, broken and overcast conditions 12 to 48 h in advance. The same predictors were used in all four equations for any given station and projection. In like manner, we also derived a set of early guidance equations for the warm season by screening forecasts from the limited-area fine mesh model. Here, separate equations were developed for 230 stations and projections of 6 to 24 h. Weather parameters from surface reports were also included as potential predictors for the first two forecast projections to provide the latest observed conditions for the early and final guidance systems.
We verified both experimental and operational cloud forecasts made from the final guidance equations for approximately 90 widely distributed test stations. These objective cloud forecasts were compared with subjective National Weather Service local forecasts after transforming the objective probability estimates into categorical form. However, using the category with the highest probability produced too many forecasts of clear and overcast. So we transformed the objective estimates in such a way that the percentage of correct forecasts was still high, but with the restriction that the categorical forecasts were relatively unbiased (i.e., each category of cloud amount was forecast about as often as it occurred). The verification scores showed that both the experimental and operational objective forecasts compared favorably with the subjective forecasts.
