P1.43 Using Abridged Atmospheric Data in Mesoscale Modeling: Applications to Incident Meteorology Scenarios Christopher L. Franks*, William J. Capehart, Mark R. Hjelmfelt South Dakota School of Mines and Technology, Institute of Atmospheric Science, Rapid City, SD 1. INTRODUCTION The arrival of more powerful computing per unit cost and user availability has made Numerical Weather Prediction (NWP) more economical in recent years. These advances, along with developments in paralleli- zation and distributed computing, has also led to corre- spondingly finer resolution numerical models more adept at studying complex mesoscale processes. These advancements allow mesoscale models to be used op- erationally even without access to supercomputing. However, finer resolution alone does not provide better model skill. A synergy of appropriate model resolution, physics, and data assimilation is necessary to produce the best possible NWP forecasts. With this in mind, nu- merical modeling done both operationally and in a re- search setting should involve careful considerations of model type, configuration, and the method and amount of data assimilated into the model. Research regarding NWP usually focuses on one (or more) of the aforementioned considerations. Typi- cally, the practice of modeling studies is to improve nu- merical forecasts by ingesting as much data as possi- ble. This study, however, takes the approach of testing the sensitivity of model forecasts when abridged atmos- pheric data is assimilated into the model. In particular, the role of assimilating truncated atmospheric data in incident/event meteorology scenarios will be addressed. We first provide the premise for this type of research and then describe our experiment design. Some pre- liminary results with case study simulations are also presented. 2. NWP AND INCIDENT METEOROLOGY Incident meteorology generally describes specific real-time weather and weather forecasts used by inci- dent management, resource management, and emer- gency response teams during episodic hazardous events. For example, since the 1930s, incident meteor- ology has often been used to aid wildfire management teams in charge of monitoring both controlled burns and wildfires. However, in recent years, incident meteorol- ogy has branched out to include more “all-risk” or non- wildland fire incidents (Querciagrossa-Sand, 2003). With its broader interpretation, incident meteorology can be expanded to include all hazardous situations where timely, site-specific forecasts are needed. Furthermore, these real-time, site-specific (mostly short-term) fore- casts are also an extremely important component of aviation meteorology as well as influencing decision making in various “weather-sensitive” industries and the planning of sporting events (May et al., 2004). For the remainder of this paper, the term incident meteorology will be used to describe event forecasting in all scenar- ios with both hazardous and non-hazardous implica- tions. Most real-time models run operationally at national centers, such as the National Centers for Environmental Prediction (NCEP), typically produce forecasts every 6- 12 hr. Even with modest horizontal grid spacing, large forecast areas require significant computational power, and the time it takes to run these models and transmit the data to a public domain is usually 2-3 hours past the initialization time. In addition, both the spatial and tem- poral resolution is too coarse to resolve mesoscale fea- tures with rapidly changing conditions (Mass and Kuo, 1998). As the result, most current operational forecast models are inadequate in providing the information needed in incident meteorology. NWP forecasting in incident meteorology is a particularly difficult area in numerical modeling because of the short time available to give timely forecasts to the public or incident manag- ers. The challenge is to produce forecasts that can be generated quickly without compromising the accuracy of predicted variables. Ultimately, incident meteorological forecasts will depend on the compromises that are made between computation time, physics options, and ingested data. Furthermore, operations at incident sites may have lim- ited bandwidth to acquire data and limited computing resources. One of the ways to try and compensate for this problem is to scale back the large-scale forecasts and observational data ingested into the model. This, in combination with the inclusion of less computationally intensive physics options and fewer vertical layers, will reduce the computational expense required to perform a numerical simulation. This is applicable to incident situa- tions where there may not be access to a large amount of data and high speed transmission. In this study, we are looking to go against “best practices” in initializing models. The term “best practices” is typically used to describe a model configured with adequate physics, that ingests enough data to provide an accurate forecast, and can be run at a reasonable cost relative to the time available to complete a simulation. We wish to further degrade the information being ingested into models, where decreasing the data ingested into models means less information has to be transferred across the NWP system. * Corresponding author address: Christopher L. Franks, South Dakota School of Mines and Tech- nology, Institute of Atmospheric Science, Rapid City, SD 57701; e-mail: christopher.franks@hardrockers.sdsmt.edu