The National Meteorological Center's (NMC's) twice-daily, global 2.5° pressure analyses of temperature, relative humidity, and wind speed are compared, over the coterminous United States, to the National Climatic Data Center's twice-daily, upper-air rawinsonde observations and hourly, first-order, surface observations for the period 1 January 1988 through 31 December 1992. NMCs analyses have clearly improved during this time period. Still, there are some noticeable differences especially near the surface and at 1200 UTC. During the early morning there is a warm bias, relative humidity is too low, and the surface wind speed is too strong. Weaker systematic errors occur during the late afternoon: there is a cold bias, relative humidity is too high, and the surface wind speed is still too strong. Aloft, the bias is noticeably reduced except for the wind speed, which is somewhat too weak. The analysis wind speed also has too strong temporal variations near the surface and too weak temporal variations aloft. The analysis climatology can be corrected at each station by removing the bias. Transient variations can be corrected simply by multiplying the analysis anomalies by the ratio of the station standard deviation to the analysis standard deviation. Correcting for the biases and spatially interpolating the analysis and station collections to a 0.5° grid provides a useful guess for local conditions, especially if there is not a surface or rawinsonde station within about 200 km.