An a priori justification for the form of the cost function (J) is that the difference

1. High-Resolution Marine Wind Retrieval Using Synthetic Aperture Radar Rick Danielson, Michael Dowd Dalhousie University Luc Fillion, Harold Ritchie Canadian Meteorological Centre Radar ( ) Cross-Section (dB) Sea Level Pressure (hPa) Methodology Introduction Results Conclusions • The desired surface wind field is a • weighted combination of the SAR obs • and model winds that minimizes a cost • function J (assuming Gaussian errors) • We choose the denominators (errors in • backscatter and winds) to best fit the • ship/buoy winds : given wind errors of • 1.7 m/s, the best choice for SAR errors • appears to be 15% of the backscatter • (double that of Portabella et al. 2002). • As a result of the scatterometer wind • calibration effort (e.g. Stoeffelen 1998), • we have a relationship between wind • and backscatter (CMOD). • An a priori justification for the form of the • cost function (J) is that the difference • between the ship/buoy obs and either the • model or retrieved wind speeds is quite • Gaussian. This • also appears to • apply well to • differences in • backscatter, but • not as robustly • Scatterometers permit a retrieval of • marine winds at ~25-km resolution, but • that of synthetic aperture radar (SAR) is • much higher (~500m). However, a wind • calibration specific to SAR has yet to be • performed • If SAR wind errors could be quantified, • this would constitute an initial step • toward the assimilation of such data into • coastal models. Simple estimates of • these errors are derived here using a • variational approach • Use of SAR observations in a variational • context appears to improve the surface • wind analysis of a model. Monaldo et al. • (2001) further claim that 1-2 m/s RMS • wind speed errors can be achieved • In situ observations are also erroneous • (Stoeffelen 1998) because they are not • representative of large areas. In order to • avoid pseudobiases in SAR observations • all error types should be considered (i.e. • a SAR wind calibration is required) • Preliminary SAR obs errors are given • here using ship and buoy obs as a • reference. An equivalent weighting • scheme is used to produce these winds As the “Retrieved-Observed” bias is smaller, the winds have improved (but only insofar as ship obs are correct). Data CMOD Observation Operator Spatial Correlation References • Collocations of ship/buoy, model, and • SAR data are assembled using 400 • acquisitions in May-December, 2004. • Grids are processed to 1-km resolution. • The ship and buoy observations are our • reference dataset. Model winds are from • the GEM 15-km operational model • Ship/buoy obs are visually inspected and • SAR data is masked over land, sea ice, • and where winds are beyond 3 to 33 m/s. • Buoy observations within 90 minutes of • an acquisition are employed. These are • adjusted to the 10-m level and are • considered valid within 50 km. A total of • 7497 ship/buoy-model-SAR collocations • are obtained. • Bragg scattering from capillary waves • varies with wind speed, wind direction, • SAR beam angle, and other processes • The empirical relation used here is • CMOD5 (Hersbach 2003) • CMOD has been derived using ERS • scatterometer data (ERS is vertically • polarized). Radarsat-1 is horizontally • polarized, so we employ a polarization • correction (Vachon and Dobson 2000) Hersbach, H., 2003: An improved geophysical model function for ERS C-band scatterometry, ECMWF technical memorandum 395 (available at www.ecmwf.int/publications) Monaldo, F. M., D. R. Thompson, R. C. Beal, W. G. Pichel and P. Clemente-Colon, 2001: Comparison of SAR-derived wind speed with model predictions and ocean buoy measurements, IEEE Trans. Geosci. Remote Sens., 39(12), 2587-2600 Portabella, M., A. Stoffelen, and J. A. Johannessen, 2002: Toward an optimal inversion method for synthetic aperture radar wind retrieval, J. Geophys. Res., 107(C8), 3086, doi:10.1029/2001JC000925 Stoeffelen, A., 1998: Toward the true near-surface wind speed: Error modeling and calibration using triple collocation, J. Geophys. Res., 103(C4), 7755-7766 Vachon, P. W. and F. W. Dobson. 2000: Wind retrieval from RADARSAT SAR images: Selection of a suitable C-band HH polarization wind retrieval model. Can. J. Remote Sens., 26, 306-313 • Length scales of wind speed and • backscatter error covariance are about • 150 km and 100 km, respectively. Here, • the reference backscatter is defined by • ship and buoy winds and CMOD5.. Acknowledgements SAR data were obtained through the Integrated Satellite Tracking of Oil Polluters (ISTOP) Program and was provided by Radarsat International (RSI). The Canadian Centre for Remote Sensing made available code to manipulate this data. Bridget Thomas of Environment Canada assisted with the vertical adjustment of obs.