JP1.7
Revealing the SeaWinds ocean vector winds under the rain using AMSR: Part II, the empirical approach

- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner
Tuesday, 31 January 2006
Revealing the SeaWinds ocean vector winds under the rain using AMSR: Part II, the empirical approach
Exhibit Hall A2 (Georgia World Congress Center)
Bryan W. Stiles, JPL, Pasadena, CA; and J. N. Huddleston, S. M. Hristova-Veleva, R. S. Dunbar, M. H. Freilich, B. A. Vanhoff, S. H. Yueh, S. V. Hsiao, G. Neumann, P. S. Callahan, R. W. Gaston, and W. Y. Tsai

Rain contamination is one of the most vexing problems for Ku-band scatterometer ocean wind data. With scatterometer data alone, correcting the effects of rain contamination is an exceedingly difficult if not insurmountable problem. Even flagging rain contaminated data can be problematic, as one often has to choose between flagging large amounts of good data or leaving significant amounts of contaminated data unflagged [1]. Wind vectors around storms are important for global vorticity analysis. Flagging even truly rainy areas as contaminated removes some of the most variable and interesting portions of the wind field from analysis. Fortunately, for the six months of the MIDORI II mission, one need not depend on scatterometer measurements alone. Coincident SeaWinds scatterometer and AMSR radiometer measurements were obtained. We used these measurements as inputs to a three-step rain correction strategy. 1) We retrieved three geophysical quantities; sea surface temperature (SST), vapor, and total liquid, from the multi-channel AMSR brightness temperatures using a physically based model. Wind speed was also retrieved but not used in rain correction. 2) We developed two complementary methods (physical and empirical) to quantify the impact of rain on scatterometer measurements as a function of the retrieved geophysical parameters and the SeaWinds antenna beam. 3) We corrected scatterometer measurements using each of the rain impact models and retrieved new wind fields from the corrected measurements.

The details of the method for retrieving SST, vapor, and liquid from AMSR brightness temperatures and the physical rain impact model are presented in Part I [2]. In Part II we describe the empirical rain impact model, contrast it with the physical model, and exhibit wind retrieval performance statistics for both models compared to the standard SeaWinds wind retrieval without rain correction.

The empirical model is computed from AMSR geophysical parameters, SeaWinds scatterometer measurements, and coincident NCEP winds. The NCEP winds are employed to estimate equivalent rain-free backscatter values. The rain induced attenuation and additive backscatter components are then regressed as functions of SST, vapor, liquid, and antenna beam. NCEP winds are only used in regressing the parametric model; once the model is computed, they play no role in estimating the correction to apply for each scatterometer measurement. Furthermore, constraints are applied to limit how much NCEP can influence the regression. The additive rain-induced backscatter term is constrained to be zero and the attenuation value is constrained to match the physical model for zero liquid.

The physical and empirical models have complementary advantages and disadvantages. The physical model is limited because several important components of the rain impact are not well understood, including surface scattering due to rain (splatter) and antenna beam-dependent differences in backscatter from the rain column. Other studies have found that the H polarization antenna beam has larger additive backscatter due to rain than the other (V polarization) beam. [3] Various explanations such as drop shape and ring waves from drops impacting the surface have been suggested, [3,4] but these effects are difficult to quantify. The empirical model, since it is regressed from the data, bypasses the limitations in physical understanding, but it has the disadvantage that systematic errors in the numerical wind product can creep into the model. The functional form of the empirical model has been chosen to reduce the impact of such errors, but it is impossible to remove them entirely.

Both rain correction methods show improvements in the retrieved winds. The two most important contributions are the removal of rain-induced speed biases and cross-swath directional preference.

[1] Huddleston, J. N., and B. W. Stiles, “A Multi-dimensional Histogram Rain Flagging Technique for SeaWinds on QuikSCAT.” Proc. of IGARSS Conference, Vol. 3, pp 1232-1234, Honolulu, 2000.

[2] S. M. Hristova-Veleva et al, “Revealing the SeaWinds ocean vector winds under the rain using AMSR. Part I: The physical approach," 14th Conference on satellite meteorology and oceanography., Atlanta, GA 2006.

[3] B. W. Stiles and S. H. Yueh, “Impact of Rain on Spaceborne Ku-Band Wind Scatterometer Data,” IEEE Trans. on Geoscience and Remote Sensing, Vol 40, No. 9, September 2002, pp. 1973-1983.

[4] Bliven, L. F., P. W. Sobieski and C. Craeye, “Rain generated ring-waves: measurements and modeling for remote sensing”, Int. J. Remote Sens., Vol. 18, No. 1, 221-228, 1997.