Find in Library

This paper examines the ability of the Weather Research and Forecasting model forecast to simulate moisture and precipitation during the North American Monsoon GPS Hydrometeorological Network field campaign that took place in 2017. A convective-permitting model configuration performs daily weather forecast simulations for northwestern Mexico and southwestern United States. Model precipitable water vapor (<i>PWV</i>) exhibits wet biases greater than 0.5 mm at the initial forecast hour, and its diurnal cycle is out of phase with time, compared to observations. As a result, the model initiates and terminates precipitation earlier than the satellite and rain gauge measurements, underestimates the westward propagation of the convective systems, and exhibits relatively low forecast skills on the days where strong synoptic-scale forcing features are absent. Sensitivity analysis shows that model <i>PWV</i> in the domain is sensitive to changes in initial <i>PWV</i> at coastal sites, whereas the model precipitation and moisture flux convergence (QCONV) are sensitive to changes in initial <i>PWV</i> at the mountainous sites. Improving the initial physical states, such as <i>PWV</i>, potentially increases the forecast skills.