9714291 Paegle The U.S. Weather Research Program (USWRP) is an interagency activity designed to perform and implement the research necessary to improve the delivery of weather services to the nation. Under this Program, the National Science Foundation, the National Oceanic and Atmospheric Administration, the National Aeronautics and Space Administration and the Office of Naval Research are jointly evaluating and supporting research of high priority to the USWRP. In this research, the Principal Investigator will investigate the relative contributions of regional and global-scale initial uncertainties on model forecasting sensitivity. Several recent studies emphasize benefits of locally targeted observations, based upon estimates of uncertainty growth produced by adjoint, "bred mode", and singular vector analyses. Such methods identify dynamically unstable regions within which details of the local initial state may have special relevance for subsequent forecast evolution. It is unclear whether those approaches distinguish forecast sensitivities due to slightly uncertain specification of the global-scale, background flow. Sample forecasts of the "superstorm" of 12-14 March 1996 suggest that details of the upstream domain that includes the initial potential vorticity maxima leading to the storm are important for a 96 h forecast. Such findings motivate the proposed work which will estimate uncertainty in specification of the atmospheric state, and will quantify impact of the uncertainty upon numerical weather prediction as a function of uncertainty location and scale. The working hypothesis of the research is that local weather prediction by numerical methods often depends more sensitively on relatively small uncertainties of the initial global-scale state than upon uncertainties of the initial regional state. Three separate global reanalysis products will be compared. The differences of different, equally credible atmospheric analyses prov ide a measure of current uncertainty in the initial state, and will be spectrally quantified. The uncertainties thus estimated will be inserted into the initial states of a global variable resolution model to determine the scale dependence of subsequent forecast sensitivity. Sensitivity studies will focus on a number of initial states selected from Spring and Winter conditions. Initially targeted domains will also include the Pacific data void. Successful completion of this research potentially will have important implications for data collection strategies for operational forecasting. ***