Info

To remedy the deficiency associated with the Goodberlet algorithm mentioned above, Krasnopolsky (1995a, 1995b) developed nonlinear algorithms based on neural network architectures, which were shown to be capable of modeling the nonlinear dependence of wind speed on brightness temperatures. The performance of neural network algorithms in the retrieval of SSM/I wind speed data discussed in Krasnopolsky (1995a, 1995b) has been tested mainly by using a well-prepared matchup database, i.e. their wind data were tested only in an experimental retrieval procedure, but not in any operational forecast system. For this reason, Yu et al. (1997) further conducted a global data assimilation experiment to investigate the impact of SSM/I wind speeds derived from the neural network algorithm. The assimilation period for this experiment was about three weeks, from May 16 to June 4, 1996. Detailed results of this investigation can be found in Yu et al. (1997). Figure 6 shows anomaly correlations for 500 hPa and 1000 hPa geopotential heights calculated from the control run (using SSM/I wind speeds derived from the Goodberlet algorithm) and the parallel run (using SSM/I

Figure 6. Anomaly correlations for 500 hPa (top panel) and 1000 hPa (bottom panel) geopotential heights calculated from the control run (with use of SSM/I winds from the Goodberlet algorithm — dashed line) and the parallel run (with use of SSM/I winds from the neural network algorithm — solid line) over the southern hemisphere.

Figure 6. Anomaly correlations for 500 hPa (top panel) and 1000 hPa (bottom panel) geopotential heights calculated from the control run (with use of SSM/I winds from the Goodberlet algorithm — dashed line) and the parallel run (with use of SSM/I winds from the neural network algorithm — solid line) over the southern hemisphere.

wind speeds derived from a neural network algorithm) over the southern hemisphere. The results clearly show that use of SSM/I wind speeds derived from a neural network algorithm improves the height forecasts at the 1000 hPa and 500 hPa levels over those derived by the

Goodberlet algorithm. Based on these results, the neural-network-derived SSM/I wind speeds were implemented in the NCEP global data assimilation system in 1997. It should be noted that the current operational SSM/I wind speeds at NCEP are still derived by the same neural network algorithm tested by Yu et al. (1997).

4. Applications of Ocean Surface

Wind Data in the Southwestern

Pacific Region

The results from the previous section on data assimilation experiments suggest that routine assimilation and forecast experiments may not show a very significant impact of satellite ocean surface winds from the gross statistics of anomaly correlations and root-mean-squared errors over many cases of analysis and forecasts. Nonetheless, they are important statistics, from which new data sets such as satellite ocean surface winds of ERS-1/2, SSM/I, and QuikSCAT were implemented operationally at NCEP, as has been discussed in the previous section. However, the impact of any satellite ocean surface winds may be most significant in some selected synoptic situations where a satellite has provided data over the region that is not covered by the conventional observations. Atlas et al. (1999) have shown an example of a very significant improvement to the cyclonic circulation due to the influence of NSCAT wind data over the extratropical Pacific oceans of the southern hemisphere. In the same paper, they also show an example of a substantial improvement in the forecast of the Christmas Day storm crossing northern Europe by using QuikSCAT wind data in the analysis. Similar case studies abound where use of satellite ocean surface winds has led to significant improvements in the analysis and forecasts of wind and sea level pressure fields over the global oceans.

The synoptic case chosen for the following discussions happens to occur on 0000 UTC,

May 2, 1994, over the southwestern Pacific region, and clearly demonstrates the important application of satellite ocean surface wind data (in this case, of ERS-1 scatterometer wind data) in identifying a closed cyclonic circulation over this region (Yu and Derber, 1995). Two analysis cases will be shown; one analysis case uses the ERS-1 scatterometer wind data (SCAT case), whereas the other does not use the wind data (control case) in the analyses. During the six-hour window centered at this analysis time, there were two swaths of ERS-1 scatterometer wind data passing through a well-developed cyclonic pressure circulation centered at a location between 150 and 155 east longitudes and between 55 and 60 south latitudes southeast of Tasmania in the southern hemisphere (see Fig. 7). The low pressure center is also well identified in the NOAA-12 visible imagery (Fig. 8), which serves as a ground truth for the assessment of analysis results. For this synoptic case, the NCEP surface wind analysis failed to depict a closed circulation center when compared to the satellite imagery. It is therefore of particular interest to see if additional ERS-1 observations of ocean surface wind data will improve the low level wind analysis in better defining the center of the storm circulation.

The vector winds at the lowest model level (40 m above the ocean surface) from the analysis which includes ERS-1 scatterometer wind data are shown in Fig. 9 (SCAT case). They should be compared with the analysis which were generated without the use of the ERS-1

scatterometer wind data shown in Fig. 10 (control case). One can see from comparing Figs. 9 and 10 that the analysis with the inclusion of the ERS-1 wind data shows a better defined circulation for the storm center than the control case. The increase in the cyclonic circulation contributed by the addition of the ERS-1 wind data is clearly shown in vector wind differences between the two analysis (see Fig. 11). Close inspection of the two analysis and their differences reveals that there are areas of large vector wind (about 20m/s) differences between the two analysis, and these differences occurred near the center of the storm over the passes of the two satellite swaths.

5. Summary and Conclusions

The forecasting skills of short range numerical weather predictions have been steadily improved since the beginning of NWP operations in the late 1950's. During the last two decades the forecasting skills have seen more drastic improvement, primarily owing to advances in the fields of satellite remote sensing and atmospheric analysis and data assimilation systems. This article first reviewed the evolution of atmospheric analysis schemes and global models at major NWP operational centers at NCEP and ECMWF during the last two decades. The procedures for the operational use of ocean surface

Renewable Energy Eco Friendly

Renewable Energy Eco Friendly

Renewable energy is energy that is generated from sunlight, rain, tides, geothermal heat and wind. These sources are naturally and constantly replenished, which is why they are deemed as renewable.

Get My Free Ebook


Post a comment