NCEP short-range operational forecast and Limited Area HIBU (Federal Hydrometeorological Institute and Belgrade University) Model (LAHM) regional model performance during a 2-month period over the southern part of South America are evaluated through the analysis of bias and rmse's. While spatial structure of errors could be only examined using gridded operational analyses as the ''ground truth,'' observed data have been used at two radiosonde stations to have an independent control of forecast and analysis quality. LAHM precipitation forecast error has been also determined using observed 24-h accumulated precipitation over a subregion of interest.Bias and rmse are, in general, lower for Medium-Range Forecast Model (MRF) 24-h forecasts than for the regional model, though MRF errors appear to be larger than those reported by other studies carried out over the whole Southern Hemisphere, suggesting the necessity to perform regional verification analysis whenever gridded analyses and/or forecasts are being used. This recommendation particularly holds over data-void regions like South America.While geopotential and wind biases do not exhibit a particular pattern in either forecast, there is a clear tendency to cold biases over the whole troposphere, and for the MRF in particular, growing with height.The results obtained from LAHM evaluation suggest that continuous development is needed to keep this regional forecast system as a plausible counterpart of available global model products for fulfillment of local requirements.