FIGURE 3.7 Winds obtained by tracking clouds in successive infrared images. The height of the cloud is determined by the cloud’s temperature. Note that where there are no trackable clouds, no winds can be retrieved. SOURCE: NOAA.

Figure 3.8 shows a time series of one measure of the skill of a representative numerical weather prediction model for 3-, 5-, 7-, and 10-day forecasts. The top line for each set of curves is for the northern hemisphere, where nonsatellite observations are plentiful; the bottom line is for the southern hemisphere, where nonsatellite observations are woefully few. Due largely to our increasing ability to use satellite observations effectively—that is, to assimilate the observations into numerical weather prediction models (e.g., Kalnay 2003)—the difference between northern hemisphere forecasts and southern hemisphere forecasts has steadily decreased, and the overall forecast skill has increased to the point that global 7-day forecasts are now as good as northern hemisphere 5-day forecasts were 25 years ago.

In addition, tests at NCEP show that data from just one satellite instrument, the Advanced Microwave Sounding Unit (AMSU), extend forecast usefulness by 1 day in the southern hemisphere and by about a half day in the data-rich northern hemisphere (Lord 2006). AIRS is also improving forecasting skill (Chahine et al. 2006), and there is evidence that satellite data, particularly QuikScat winds, are improving hurricane track forecasts (Zapotocny et al. 2007). Without question, improvement in numerical weather prediction—on which all forecasts more than a few hours ahead are based—is a major scientific accomplishment of Earth observations from space.