Autocorrelation and variance statistics were calculated for seven types of wind data in the western hemispheric tropics. Most of these data came from the Global Weather Experiment (GWE) in January 1979. They were: 1) cloud motion measurements from four different sources, 2) rawinsonde wind reports, 3) synoptic land station reports, 4) marine ship reports, 5) aircraft pilot reports, 6) automatic aircraft reports for the GWE, and 7) Seasat scatterometer winds from September 1978. Winds were analyzed within a target area from 30°N to 30°S latitude and 0° to 180°W longitude. The Seasat scatterometer data had the highest autocorrelations and lowest standard deviations over short distances (<500 km). Cloud motions and rawinsondes had lower autocorrelations than Seasat, while synoptic land stations, ship reports, and aircraft pilot reports had the poorest autocorrelations. These correlations imply that synoptic land stations, ship reports, and aircraft reports were either more sensitive to small‐scale fluctuations than other sensors, or had higher intrinsic noise levels. Structure function plots of autocovariances against separation distance between observations indicated that Seasat was most sensitive to wind field structure by having low autovariance at short distances (100 km) that also grew with distance. The other structure function plots for low‐level wind observations indicated a lack of structure sensitivity to scalar wind speeds because of very small growth rates of the autocovariances with distance. However, all observations were sensitive to structure in the wind direction patterns.