|
Panel: Wind Speed and Direction Moderator: Peter Taylor T. Liu:
Timothy Liu gave a summary of existing and future space-based ocean surface wind measurements.
Scatterometer experiments are NSCAT, QSCAT, ADEOS-2 and QCOM. Passive microwave experiments are WINDSAT and CMIS. Of all the above experiments, only ASCAT is an operational instrument. There will be very good scatterometer coverage during 2001-2006; global every 6 hours M. Freilich: Algorithms for determining winds from scatterometers are highly empirical; need realistic error model or you will tune the algorithm incorrectly Sampling related problems include the sampling pattern, field generation approach, time variability. Need multiple co-orbiting satellite missions to eliminate these errors Mike Freilich presented viewpoints showing the sampling errors in wind fields constructed from scatterometer data sets. Sampling errors in temporally and spatially averaged wind fields were constructed from single and tandem scatterometer data sets. Linear deterministic errors and random component errors both contribute to wind speed biases. The analysis of sampling errors in the ERS sampling pattern with its 2° x 2°, 4 day resolution gave large errors. The NSCAT data hass less error and the SeaWinds data was even better. The tandem QuikSCAT/SeaWinds data showed the least sampling error. The QuickSCAT data gives 12 hour global ocean coverage (high resolution global backscatter and ocean vector winds). This radiometer has a complete swath with no nadir gap. D. Chelton: Dudley Chelton showed comparisons between NSCAT and ECMWF vector winds and derivative wind fields. Also comparisons of NSCAT and ECMWF winds with TAO and NDBC bouys were made. The NSCAT-ECMWF vector squared correlations were worst in the eastern tropical region. The zonal component show a large rms difference and a large mean bias difference. C. Fairall mentions that problem areas (like the eastern tropical Pacific) maybe due to the presence of stratocumulus clouds. Viewgraphs of derivative fields, the divergence field and relative vorticity were also presented. P. Webster noted that the vorticity is tightly coupled to the divergence. X. Zeng asked to what extent precipitation interferes with retrievals Answer: this is a concern M. Kubota: Masahlsa Kubota presented a slide of GEOSAT windspeed distribution showing a banded structure due to aliasing errors. He mentioned various ways to remove aliasing errors.
D. Weissman: David Weissman demonstrated the use of scatterometer data to determine surface drag coefficients. His data sets included the NSCAT radar cross section and NDBC and TAO bouy measurements. The technique is to infer U* (friction velocity) from the radar cross section using a unique model function, then use this U* and buoy wind speeds to derive the drag coefficients. This assumes that the scatterometer is more directly related to stress than to wind speed. The results indicate that wave height does not influence the value of the drag coefficient significantly. During one 4 month period, data from the Pacific NDBC buoys give a higher drag coefficient than the data from Atlantic buoys. The Pacific data might be swell-dominated. Studies of the TAO buoys also show regional differences. The increase in the value of the drag coefficient in the upper latitude regions might be due to a strong equatorial counter-current in the region. Large scatter in data might be due to matching U* data from scatterometer data using a 25 km footprint to spot measurements of buoy wind speeds in a situation where the variables cannot go negative. A. Beljaars: Anton Beljaars gave a comparison of PACS bouy data and ECMWF wind speed caculations. He noted that there is a large scatter in the wind speed comparisons, and that the scatter becomes worse at lower wind speeds. M. Freilich noted that the scatter in data from scatterometer is smaller compared to bouy data. Summary by Anton: There is a large scatter in low-level wind speeds compared to the measurements. Scatterometer data used in the ECMWF calculations tend to improve the analysis. SSMI derived winds are of good quality. There was a question concerning the under predictions of high wind speeds by ECMWF. J. Schulz: Joerg Schulz gave a rundown on passive microwave wind speed retrieval methods. These methods include:
Wentz's algorithm appears to be the best. Some of the strengths and weaknesses of these methods were discussed. For example, the neural network method gives better results than the D-Matrix method for higher wind speeds, but are in disagreement with bouy measurements at low and high speeds. P. Taylor noted that bouy data might be biased low for strong winds. There are also validation problems with methods that use SSM/I data as input since within the comparison to buoys or ships there is a mismatch in space and time between those measurements. This part of the RMS can be the largest part of the whole error budget. No comparison of all these wind speed algorithms under identical conditions has been done. Comments: When using active sensor techniques, wind speeds should be calculated using a scalar average and direction calculated using a vector mean. For bouys, scalar averaging is typical. Is it feasible to obtain wind directions from a passive technique?
J. Curry: Could we combine active and passive retrievals? M. Freilich: Multiple active measurements are coming online soon. If we require daily or diurnal wind speeds, we need at least 4 broad-swath satellites (scatterometers) - only one is operational now. We have problems with resolution effects - what errors are we making at what scales (larger fields vs. points or pixels). We do not have a good definition of resolution effects. We really need the vector winds, not just wind speeds. During the report from the breakout group, D. Chelton stated that the best resolution we can obtain for a global wind data set using the available data is about 2° x 2° with a 5 day period. M. Bourassa was more optimistic, saying he thought a 0.5° x 0.5° twice daily global data set was possible. |
