In this note, we present a summary of results from a total of some 980 forecast experiments made with two simple high-resolution regional models. The models are based on the principles of conservation of absolute vorticity (a streamfunction barotropic model) and of potential vorticity (the shallow water equations with bottom topography). These regional forecast models are applied to a number of different domains (winter monsoon, central Pacific Ocean during Northern winter, Central America, Eastern Atlantic and South America, West African Monsoon, Indian summer monsoon and summer monsoon over Southeast Asia). The data sets for these experiments are those derived from the First GARP Global experiment (FGGE), the Asian Monsoon experiment (MONEX) and the West African monsoon experiment (WAMEX). Statistical evaluation includes a comparison of the root mean square vector wind error with the corresponding persistence errors. The results of this study show that these models have a varied skill over different domains. The single-level primitive equation model over West Africa, South America and the Atlantic Ocean during northern summer shows a skill beyond the persistance for a range of 4 days of integration. The performance of the single-level primitive model appears quite promising over these regions. These long series of experiments with the single-level primitive equation model were carried out with the choice of the mean free surface height of 2 km with a maximum height of the orography of 1 km. A further effort at the optimization of the single level model was explored. Here, a parameter space consisting of the free surface height and the maximum height of the orography was explored in order to assess the skill of the vector wind error. Marked improvement in skill was noted over a domain centred over China from an optimization of these parameters. DOI: 10.1111/j.1600-0870.1987.tb00296.x