Forecasting Regional Chance of Occurrence through Aggregation of MOS PoPs (Invited Presentation)

Regional sets of point probabilities, e.g. MOS PoP forecasts, can be used to forecast both the expected areal coverage and the regional chance of occurrence. The former is just a data quality check given the statistical equivalence of the spatial average of point probabilities and the expected value of areal coverage. The latter, however, is a new product with utility to those users whose actions depend upon the occurrence of the event anywhere within their region of operation. While the method is demonstrated here using MOS PoP forecasts, potential applications include any weather variable for which point probability forecasts are available operationally. The forecasting of thunderstorm chance of occurrence for fire weather planning is one such application.

For this demonstration, MOS PoP forecasts are matched with the corresponding NCEP Stage IV precipitation analyses. Comparison of regional average PoP with fractional area coverage verifies their equivalence, but reveals a MOS station location bias to the drier lowlands in the intermountain West. Regional chance of occurrence is forecast via logistic regression with mean and standard deviation of the region's MOS PoP forecasts as predictors. Hindcast results show significant skill, but the regression equations vary by both season and location.