The LAMP QPF Products. Part I: Model Development

Abstract

The Local AWIPS (Advanced Weather Interactive Processing System) MOS (model output statistics) Program (LAMP) quantitative precipitation forecast (QPF) system produces 1?22-h forecasts of precipitation over the conterminous United States. The forecasts are generated for points on a rectangular grid with a 20-km mesh and are in the form of probabilities for ≥0.10 (2.5), ≥0.25 (6.4), ≥0.50 (12.7), ≥1.00 (25.4), and ≥2.00 (50.8) in. (mm). The forecasts are also available in the form of the best category and expected precipitation, both of which are derived from the probabilities. The forecasts are valid for four 1-h periods in the 1?5-h range, two 3-h periods in the 4?10-h range, and two 6-h periods in the 7?22-h range. Upon implementation on the AWIPS system, the model could produce the forecasts as often as hourly. The LAMP QPFs are based on multiple regression equations, which are derived for 32 geographical regions, two seasons (warm and cool), and eight cycle times per day. The predictand is defined from 2500 to 3000 stations that compose the U.S. Climatic Hourly Precipitation Network. Predictors in the equations are derived from multiple data inputs, which include 1) centralized MOS QPF probability forecasts, 2) numerical model forecasts of 850-mb wind, 3) objectively analyzed and advected fields of variables based primarily on conventional hourly surface observations, 4) objectively analyzed and advected fields of 1- and 3-h antecedent precipitation, and 5) high-resolution topography and climatic monthly mean relative frequencies of precipitation categories. An innovative feature of the model is application of predictors that describe the interaction between finescale geoclimatic parameters and ambient low-level wind, moisture, and MOS QPF variables relative to precipitation occurrence. Predictors included in the QPF regression equations vary significantly as a function of forecast projection and geographical area. For the four 1-h valid periods that extend to 5 h, predictors are based largely on the observational data inputs. For the two 6-h periods that extend to 22 h, MOS QPFs are the dominant predictor input. The intervening 3-h valid periods exhibit a more uniform blend of the various predictor inputs. Geographically, the MOS input has more influence in the eastern United States than elsewhere, whereas the topographical, climatological, and antecedent precipitation inputs exert a strong influence in the west. Distinctive properties of the LAMP probability and derived forecast products are illustrated for a heavy rain event. Predictors that underlie the properties are identified.