A set of observing system experiments (OSSEs) are currently underway to assess the impact of simulated GEMS observations on NWP model forecasts. The OSSEs are run over a regional-scale domain during a high-impact weather event utilizing simulated observations from a nature run. The model used for the nature run is the Advanced Regional Prediction System (ARPS) coupled with a Lagrangian particle model to simulate dispersion and collection of observations from an ensemble of GEMS probes. The model used for the OSSEs is the Pennsylvania State University/National Center for Atmospheric Research Fifth-generation Mesoscale Model (MM5). For the selected weather event, the MM5 is initialized 1-day prior to the time period of interest to generate a simulation with sufficient differences from the nature run. The simulated observations extracted from the ARPS nature simulation are assimilated into the MM5 forecasts during a 12-h integration window. An intermittent data assimilation technique is used to incorporate the simulated data and objective verification serves as the metric for measuring forecast improvement in subsequent OSSEs. A detailed evaluation of the forecast impact will be presented along with results from several proposed GEMS deployment scenarios.