Michael Owen and Dr. David Long, Electrical Engineering
The SeaWinds scatterometer aboard the QuikSCAT satellite has been functioning to measure radar backscatter over the ocean since its launch in 1999. The backscatter data from the scatterometer is used to infer wind speeds over the ocean and is vital to ocean wind measurement as the satellite covers nearly the entire ocean daily. Backscatter measurements made by QuikSCAT in the vicinity of land are contaminated by land as it is generally much more reflective than ocean. Land contamination of backscatter measurements causes wind speeds near the coast to be vastly overestimated and to be so inaccurate as to be useless. My research has focused on developing an algorithm for detection and removal of land contamination so that winds inferred from backscatter measurements are accurate regardless of location and wind speeds.
To understand why land contamination occurs and how it can be detected and removed requires a basic understanding of the operational principles of the QuikSCAT scatterometer. First, QuikSCAT measures backscatter by sending a radar pulse at the surface of the earth and measures how much reflects back to the satellite. Each pulse is further divided into smaller segments, called “slices”, using range and Doppler processing, which have a smaller footprint on the surface of the earth, meaning that they can be used for higher resolution wind retrieval, (2.5km x 2.5km wind cells). Although each slice is generally approximated by the most sensitive region of the slice (the main-lobe of the slice spatial response), the complete spatial response is much larger and is quite sensitive far from the main-lobe. Land contamination occurs where the main-lobe or the side-lobes of the antenna response lie partially or completely over land and the backscatter measurements for the slice become contaminated by the presence of land in the footprint rather than being solely the backscatter from the ocean.
Second, wind speeds and directions are only known from raw backscatter measurements after further processing called “wind retrieval.” Wind retrieval is biased by land contaminated measurements to cause false hurricane force wind speeds to be retrieved near the coast. To compensate for land contamination previous algorithms simply removed winds within 30km of the coast as in Fig. 1.
Working with my advisor, Dr. David Long, we developed a metric called the land contribution ratio (LCR) which is used to detect and remove land contaminated backscatter measurements prior to performing wind retrieval. To calculate the LCR we use the complete spatial response for each slice to calculate the amount of land (weighted by the spatial response) contributing to the backscatter measurement. The LCR for every slice near the coast is calculated and any slice with a LCR above a certain threshold is considered land contaminated and discarded prior to wind retrieval.
Wind retrieval from backscatter measurements is a non-linear function and determining error in wind speeds caused by land contamination is a complicated process. To determine the threshold for tolerable land contamination we performed Monte Carlo simulations to obtain simulated backscatter measurements with land contamination for computer generated wind fields. Wind retrieval was performed and the retrieved winds compared to the generated wind to determine the error caused by land contamination. The simulations indicated that land contamination removal thresholds using the LCR are functions of the local wind speed and direction, and the geometry of the satellite orbit. We tabulated the thresholds in a look up table and implemented the LCR calculations and thresholds in high resolution backscatter processing code.
Once land contaminated measurements are discarded there are coastal regions where no wind can be retrieved because there are no uncontaminated measurements. The distance that the nearest wind can be retrieved is a function of the LCR thresholds in the area and can be as far as 17km from the coast or as near as 2.5km as in Fig. 2.
In the wake of Hurricane Katrina where forecasts of both wind speeds and the storm direction were inaccurate, LCR processing of satellite scatterometer data can help to prevent similar catastrophes by providing accurate data near the coast under all conditions.
This research is ongoing and in the coming months we will continue to refine the LCR algorithm and work to validate the accuracy of coastal winds using independent wind data sets. A more complete report is currently being drafted for publication in IEEE Transactions on Geoscience and Remote Sensing