We seek to compare best practices in initializing models against those scenarios applicable to incident meteorology. This study examines the impact of decreased data ingested into models so less information has to be transferred across the NWP system. A hierarchical series of simulations has been designed to test the sensitivity of forecasts as the amount of information used to initialize the model is gradually scaled back. The best practices simulations are designed to emulate what is typically done in research and operational settings. The simulations using abridged meteorological data are more representative of the type of information that would be available on site during incident scenarios.
Two 24-hr periods during the April 1999 Upper Missouri Basin Pilot Project (UMRBPP) are simulated to test the sensitivity of model forecasts as first guess (FG) fields and observations are scaled back. We confirm that forecasts are highly sensitive to the type of FG field used to initialize the model. The largest differences occur when changing from the best practices research setting to the operational setting. The differences between the best practices operational setting and the abridged scenarios show less disparity. Furthermore, altering the initial field by including or excluding observational soundings can create perturbations that linger within the modeling domain well into the simulation. These perturbations propagate through the model domain in the direction of the prevailing winds.