The purpose of this paper is two-fold; first, to document and present a limited-area, moist primitive equation model and, second, to give some preliminary results of experiments testing the sensitivity of the model's quantitative precipitation forecasts to the initial horizontal and vertical relative humidity distribution. Three experiments were performed. The first case used a 1.5 km vertical grid and a Gandin humidity analysis based on standard rawinsonde observations. The results indicate that the model has some skill at forecasting precipitation amounts and location in regions of predominately stable rain and in regions of convective rain. However, some glaring defects in both the initialization of the mass-flow fields and the initialization of the moisture field were evident. The second experiment attempted to enhance the initial moisture field to reflect a narrow band of moisture which was suggested by satellite cloud observations. The inclusion of this moisture band increased the precipitation amounts in the squall-line region which was being fed by the enhanced moisture. The third experiment in which the low-level vertical grid increment was reduced indicates that in some areas the precipitation amounts are increased by as much as 25% due to increased resolution of the low-level moisture field. This precipitation forecast most nearly agreed with observations.