GFS-Based MOS Precipitation Forecast Guidance for Island Sites
in the Tropical Western Pacific Ocean
James C. Su
Meteorological Development Laboratory
Office of Science and Technology
National Weather Service, NOAA
Silver Spring, MD
New Model Output Statistics (MOS) precipitation forecast guidance for 15 island sites in the tropical western Pacific Ocean has been developed at the Meteorological Development Laboratory (MDL). The new guidance based on the Global Forecast System (GFS) has been scheduled to be implemented in April 2007, and will be added to the existing MOS forecast guidance products which include previously implemented wind forecast guidance. These islands are located in the area from 15° S to 30° N and from 130° E to 170° W. The guidance for two sites located to the east of the International Date Line (IDL) will be added to the Hawaii MOS forecast guidance products. The guidance for the other 13 sites located to the west of the IDL, will be added to an existing MOS forecast guidance products for the area. The new guidance includes probability of precipitation for a 12-hour period (PoP12), probability of precipitation for a 6-hour period (PoP6), and probability of precipitation on the hour (PoPO). The GFS model output and hourly observed data for the period from April 2000 to November 2006 were used in the development of the MOS forecast equations. Among the 15 island sites in the subject area, four sites had little or no observed precipitation data, and were not used in the MOS forecast equations development. As usually was done for the precipitation forecast guidance for the CONUS, the data used in the development were stratified into seasons, and meteorological stations were grouped into geographical regions, in order to increase the sample size. The data of monthly precipitation normals (for the period of 1971-2000) were provided by the NWS Pacific Region for 11 sites, and were investigated for the purpose of grouping them into geographical regions and stratifying their developmental data into seasons. Bogus climatic data were estimated for those four sites with little or no observed precipitation data, so that they could be included in appropriate geographical regions. The 15 sites were grouped into five regions, and their data were stratified into two seasons: November to April (season 1) and May to October (season 2). The forecast equations were developed for projections up to 84 hours, and for 0000 UTC and 1200 UTC cycles. Selected GFS model output and climatic variables were used as predictors in the development. No observations were used for predictors. In the forecast equations, the dominating predictors were different from one season to the other. The relevance between this fact and precipitation characteristics will be discussed in the presentation. Independent tests were conducted for the 0000 UTC cycle of both seasons, and on precipitation forecasts against verification standards derived from climatic data. The forecasts of PoP12 and PoP6 were tested against mean relative frequencies of precipitation in 12-hour and 6-hour periods, respectively. The forecasts of PoPO were tested against corresponding forecasts from the climatic PoPO forecast equations, which were developed by using climatic variables for predictors. The precipitation forecasts by the MOS test equations generally had 10% to 15% improvement over the climate for short projections, and gradually approached the climate for longer projections. The forecasts of PoP12 and PoP6 had overall improvements better than that of PoPO. All three elements had overall improvements better for season 1 than season 2. In this presentation, the performance of the MOS precipitation forecast guidance will be discussed.