SeaWinds Scatterometer Wind Vector Retrievals within Hurricanes using AMSR and NEXRAD to Perform Corrections for Precipitation Effects: Comparison of AMSR and NEXRAD Retrievals of Rain
D. Weissman,S. Hristova-Veleva,P. Callahan
DOI: https://doi.org/10.1109/IGARSS.2006.912
2006-07-01
Abstract:The opportunity provided by satellite scatterometers to measure ocean surface winds in strong storms and hurricanes is diminished by the errors in the received backscatter (SIGMA-0) caused by the attenuation, scattering and surface roughening produced by heavy rain. Providing a good rain correction is a very challenging problem, particularly at Ku band (13.4 GHz) where rain effects are strong. Corrections to the scatterometer measurements of ocean surface winds can be pursued with either of two different methods: empirical or physical modeling. The latter method is employed in this study because of the availability of near simultaneous and collocated measurements provided by the MIDORI-II suite of instruments. The AMSR was designed to measure atmospheric water-related parameters on a spatial scale comparable to the SeaWinds scatterometer. These quantities can be converted into volumetric attenuation and scattering at the Ku-band frequency of SeaWinds. Optimal estimates of the volume backscatter and attenuation require a knowledge of the three dimensional distribution of reflectivity on a scale comparable to that of the precipitation. Studies selected near the US coastline enable the much higher resolution NEXRAD reflectivity measurements evaluate the AMSR estimates. We are also conducting research into the effects of different beam geometries and nonuniform beamfilling of precipitation within the field-of-view of the AMSR and the scatterometer. Furthermore, both AMSR and NEXRAD estimates of atmospheric correction can be used to produce corrected SIGMA-0s, which are then input to the JPL wind retrieval algorithm Introduction: The MIDORI-II mission, during 2003, carried five earthobserving sensors including the SeaWinds scatterometer and the Advanced Microwave Scanning Radiometer (AMSR). The latter's six frequency brightness temperatures are collected to derive atmospheric water-related parameters and to measure the sea surface temperature. The AMSR’s coverage was closely coincident and collocated with the scatterometer. It provided the opportunity to obtain the precipitation estimates necessary to evaluate the attenuation, volume backscatter and surface roughening caused by the raindrops within the scatterometer beam. One event of interest is Hurricane Isabel, which crossed the U.S. coastline on September 18 near a part of the North Carolina coast which is being continuously monitored by the NWS Morehead City NEXRAD radar. The NWS NEXRAD network supports the acquisition and utilization, in real-time, of three-dimensional rain reflectivity data (S-band) with high spatial resolution. The archived Level-II data files are characterized by an approximate one-degree antenna beamwidth and 1-kilometer range resolution. The data is collected with 360 azimuth sweeps at fixed elevation angles (0.5,1.45,2.4,3.35, etc). The duration of each sweep is about 6 minutes. This permits coincident timing with the spacecraft to within 4 minutes or less. Taking into account the earth curvature, at a distance of 150 km from the station, the center of the lowest elevation beam (directed at 0.5 above the horizon) is about 2,500 meters above sea level. Special algorithms have been developed to work with the NEXRAD reflectivity files and to convert the native observations into a Cartesian grid volumes. Using well known methods of interpolation, resolutions of 5,10 or 25 km in the horizontal directions are straightforward. Vertical resolution for these grid cells is 2 km. MIDORI-II encountered this same event within 4 minutes of the NEXRAD observations thereby providing an excellent opportunity to have the high resolution NEXRAD observe the atmospheric volume in which AMSR produces precipitation estimates. This is a vital factor since the typical rain spatial structure is on the order of a few kilometers, whereas the scatterometer and AMSR data are averages over 30 km sized areas. Our results shows how well the precipitation estimation techniques using the AMSR data perform in conditions where the NEXRAD observes appreciable variation of rain intensity (horizontal and vertical) within the much larger AMSR cell. Besides Hurricane Isabel, two other events, with different characteristics are being investigated. On August 20, near the Melborne, FL NEXRAD, a light wind but heavy rain condition had interesting effects on the SIGMA-0s. Hurricane Claudette, (July 15 2003), crossed the Texas coast near the Louisiana border. The hurricane was simultaneously observed by MIDORI-II and the Houston NEXRAD radar. Claudette's winds were not as violent as Isabel's, but they covered a large area in the Gulf of Mexico away from the coastline and had intense rainfall. The scatterometer swath, during these hurricane overpasses, spans a substantial range of rain intensities, wind magnitudes and directions permitting performance examinations across this wide distribution. In addition to comparing rain parameter estimates (rain rate and vertically integrated liquid water) obtained from AMSR to that obtained from NEXRAD, the two sensors can separately 0-7803-9510-7/06/$20.00 © 2006 IEEE 3540 derive atmospheric corrections (volumetric attenuation and rain backscatter) to the satellite scatterometer SIGMA-0. There are several steps in the process of providing atmospheric correction to the scatterometer signal and the NEXRAD 3-D data allows the separate evaluation of most of these. In addition, NEXRAD's high-resolution sampling of the 3-D precipitation volume allows taking into account the unique incidence angles of each of the scatterometer two beams. The SIGMA-0's, corrected separately based on either AMSR or NEXRAD estimates, can then be used in a modified JPL wind retrieval algorithm, to produce two sets of corrected wind magnitudes and directions. Comparison between AMSRand NEXRAD-corrected scatterometer winds will allow also evaluation of the impact of the observational geometry. Scatterometer rain effects The presence of rain during observations of the sea surface by orbiting Ku band (13.4 GHz) scatterometers, such as QuikSCAT and SeaWinds on MIDORI-II, usually results in the retrieval of winds that are erroneously oriented in a cross-track direction (at ~90 and ~270 degrees) and have higher speed than both buoy and global model winds suggest [1], [2]. The scatterometer signal that propagates through rain is impacted in three ways: the signal is attenuated by the rain, the cloud and the vapor in the atmosphere; the signal is augmented by the backscatter from the rain droplets; finally, the signal is augmented by the rain-induced roughening of the ocean surface (“splash”). Estimation of the near-surface wind velocity from scatterometer measurements is based on the assumption that variations in the measured power are solely due to variations in the backscattering cross-section (SIGMA-0) of the ocean surface that result from variations in the wind. It is, thus, very important to properly account for the three rain effects and to correct the SIGMA0 estimates before they are used to estimate the wind velocity. AMSR corrections: Passive microwave observations of the top of the atmosphere radiation have been proven to provide very valuable information about the sea surface temperature, the vapor and cloud amounts in the atmosphere, and the presence and the amounts of precipitating hydrometeors. We have developed a passive microwave precipitation retrieval algorithm for the purpose of providing atmospheric correction to be used by scatterometer wind retrieval algorithms for the SeaWinds instrument on board the short lived MIDORI-II satellite. This precipitation algorithm uses Advanced Microwave Scanning Radiometer (AMSR) observations that were closely collocated with the scatterometer observations. The algorithm addresses in a new way the issues of non-uniform beam filling and hydrometeor structure uncertainty. This involves the use of multiple retrieval databases, each representing a particular rain intensity and inhomogeneity regime. For each observational scene, the algorithm uses a specially developed Rain Indicator to determine the intensity and degree of rain homogeneity within the sensor's Field of View (FOV). With this information in hand, it selects the appropriate retrieval database to estimate a number of geophysical parameters: vertically integrated liquid water (or Liquid Water Path LPW), vertically integrated water vapor, near-surface wind speed, and sea surface temperature. Once the retrieval of the geophysical parameters is accomplished, we proceed to derive the atmospheric correction to the scatteromter observations as a function of these parameters. In doing so, two approaches could be adopted: the “physical” approach and the “empirical” one. They differ in how the geophysical information is used. The “physical” approach uses statistical relationships to directly estimate each of the three components of the atmospheric correction in the following sequence: LWP – RainRate – Volumetric Attenuation, Volumetric Backscatter by the rain. The empirical approach determines the atmospheric correction by developing relationships, as functions of the geophysical retrievals, between the observed SIGMA-0 and the model-wind-equivalent one. Hence, both approaches use the geophysical retrievals but differ in how the corrections are computed. The results: Applying the AMSR-based atmospheric correction to the scatterometer observations has resulted in significant improvement of the scatterometer winds in rain [3], [4]. Our current success shows the high potential of our AMSR-based geophysical retrieval algorithm and validates our approach. However, there is need of improvement as our corrections appear to be somewhat noisy. A good correction requires good estimates of attenuation, precipitation backscatter and even estimates of the rain-induced roughening of the ocean surface. It is, thus, very sensitive to the accuracy of the rain rate estimates. This prompts us to