Optimizing Numerical Weather Prediction Utility of the Maryland Mesonet with Observing System Simulation Experiments

The Maryland Mesonet Project (MMP) aims to construct a network of about 75 observing stations to improve regional forecasts and nowcasts in order to mitigate the impact of severe weather events across the state. It follows on similarly motivated projects launched by state governments across the country and abroad in recent years. Mesonets are defined by the generation of observational data for surface wind velocity, temperature, humidity, etc. with a spatial and temporal frequency sufficient to capture mesoscale processes and the development of severe convective storms. Such observing systems typically provide information at a granularity beyond the capabilities of pre-existing surface observing stations. The spatial configuration of stations within a mesonet is an important factor in the utility newly provided observations will have via data assimilation in terms of impact on forecast performance, making it desirable to optimize station placement in an objective manner. For this reason, we perform observing system simulation experiments (OSSEs) using a regional WRF modeling system based on the NSSL Warn on Forecast System (WoFS) to determine an optimal configuration for stations placed by the MMP. OSSEs are used to study the impact of new observing systems by comparing measures of forecast verification against a simulated nature run that is taken as the true system state. We evaluate multiple candidate network configurations, which include: a network placed to minimize the variance of distance between MMP and pre-existing stations, a network placed to minimize only the variance of distance between MMP stations, a network placed according to population density, and a network with greater density in the Appalachian plateau region of western Maryland, upstream of the prevailing motion of storm activity. We also perform experiments using only the pre-existing observation network without new construction, as a baseline for forecast improvements attributable to MMP configurations. Each of these configurations is tested for seven 18-hour case-study events featuring severe weather. We choose these events based on severe weather reports indicating significant community impact in the time period between 2020 and 2022. For each case-study, we generate a set of 3-hour forecasts from which we produce aggregated verification statistics using the fractions skill score (FSS) neighborhood-verification method. The tuning process for our OSSE experiments reveals unique challenges associated with regional OSSEs, and OSSEs designed for convective scale modeling systems. Solutions proposed for our study may prove useful in the optimization of station placement for future mesonet systems.