The purpose of this paper is to discuss the major systematic errors of the U.S. Navy Operational Global Atmospheric Prediction System (NOGAPS), version 3.2, and to describe several tuning experiments of NOGAPS parameterizations. It is found that despite its overall good performance, major systematic errors exist in the forecast model. These errors lead to a warmer atmosphere with less precipitation and eddy kinetic energy than is observed. Some of the errors may be attributed to the lack of horizontal and vertical resolution, but most of the errors are due to inadequacies and incorrect assumptions in the physical parameterizations. We present a list of the systematic errors of the operational 5-day forecasts and results of a 1-yr integration with climatological sea surface temperatures. One of the prominent features of NOGAPS integrations is a large diurnal oscillation in the global mean averages. This oscillation is traced to large differences in total albedo over the land and sea areas. We present results of 31-day integrations for December 1989, December 1990, and December 1991, where we varied the value of the single-scattering albedo, limited the vertical region of the gravity-wave drag, and varied the magnitude of the vertical mixing coefficient.
In January 1992 a new version of the forecast model was implemented, which was designated as NOGAPS 3.3 and which incorporated changes to the cloud single-scattering albedo, the gravity-wave drag, and the vertical mixing parameterizations. We compare a 1-yr simulation of different versions of NOGAPS, and we show statistical results of forecasts using the two model versions. We believe that synoptically 3.3 is superior to 3.2, but the statistical skill, as measured by the anomaly correlation of the Northern Hemisphere's 500-mb height field, of the different versions is the same.