1 - OVERVIEW

Surface wind is a key parameter for the determination of many ocean-atmosphere interaction parameters such as air-sea latent and sensible heat fluxes and air-sea transfer rate of carbon dioxide, momentum flux and the wind stress on the surface layer of the ocean. Therefore a great deal of effort has been devoted to produce a gridded wind field using ERS-1 scatterometer-retrieved wind vectors over the globe. The ERS-1/2 and NASA scatterometers (Section 2 : ERS and NASA Scatterometer wind data) measure radar backscatter ⁰ from the ocean surface at a frequency of about 5.3 GHz (C-band) and 14 GHz (Ku-band), respectively. The scatterometer wind vectors are then estimated from the backscatter coefficients using semi-empirical models and inversion/dealiasing algorithms (see References). The ERS scatterometer provides wind vectors (wind speed and direction) at 50 km resolution with a separation of 25 km across a 500 km swath. NSCAT wind vectors are estimated over two 600 km swaths, separated by 300 km, with a resolution of 50 km. The accuracy of the ERS scatterometer wind speed and direction was evaluated by comparison with the National Data Buoy Center (NDBC), Tropical Atmosphere Ocean (TAO) and Japan Meteorological Agency (JMA) buoy wind measurements (Graber et al, 1997). The rms error is approximately 1.20 m/s for wind speed and 24° for wind direction. The accuracy of NSCAT wind speed and direction estimates are under considerations. However, the results presented during NSCAT CAL/VAL meeting in Hawaii, 11-15 November, 1997, indicated that NSCAT winds are in good agreement with in-situ observations. Furthermore, it was shown that ERS-2 and NSCAT winds are highly corellated. The rms values of wind speed and direction differences are about 1.10 m/s and 28°.

Where W, u, v are the scatterometer wind speed, eastward and northward wind components (at 10 m height), respectively. ,_x and _y indicate the stress magnitude and the stress components. is the density of surface air, C_D is the drag coefficient. was taken to be 1.225 kg/m³. The neutral stability form of C_D at 10 m is provided by Smith (1988) and is a function of observed wind speed.

Since wind estimated at a point can vary significantly over periods of a few hours, reconstruction of synoptic fields of surface winds on basin scales from discrete observations is difficult, without using an appropriate method. In this order we developed a statistical technique for the objective analysis of remote sensor wind data. This statistical interpolation is a minimum variance method related to the Kriging technique (section 3 : Estimation of gridded wind fields) widely used in geophysical studies. The analysis scheme is based on the determination of the spatial and temporal structure functions of wind and stress variables (magnitude, zonal and meridional components). The estimators of each variable sets were then determined and analyzed by the objective method. The latter was used to estimate weekly and monthly wind fields over the global oceans of the wind speed (W), the zonal wind component (u), the meridional wind component (v), the stress magnitude (), the zonal stress component (_x) and of the meridional stress component (_y) into 1° latitude by 1° longitude boxes. The standard errors of the above parameters were also computed at each grid point. The wind divergence and the wind stress curl were estimated from the grid scatterometer wind and stress. All these regular wind parameters are evaluated using a land and sea-ice mask which is derived from scatterometer measurements.

The accuracy of the objective method and of the resultant wind fields was evaluated by comparison with in-situ measurements and with meteorological and climatological wind analyses (section 4 : Validation). The aliasing problem which involves the aliasing in space and time domains, the multiple timescales and space scales, the oversampling and the undersampling, was investigated. The European Center for Medium-Range Weather Forecasts (ECMWF) gridded surface wind analysis was used as a control set. The scatterometer wind observations were simulated from the ECMWF analysis and weekly and monthly gridded wind fields estimated from the simulated and analysis winds were calculated. The differences between the two estimates did not exhibit any banded structures. However, they indicated that the variance of the kriging error is mainly related to the wind vector variability.

The comparisons with in-situ measurements were performed with the Tropical Atmosphere Ocean (TAO) buoy network (section 4 : Validation). It was found that the difference between the weekly scatterometer and buoy winds is within 2m/s for more than 95% of wind speed estimates and 80% for the zonal and meridional components. The bias values are 0.10 m/s, -0.10 m/s and -0.47 m/s for wind speed, zonal and meridional components, respectively. The corresponding rms values are 0.92 m/s, 1.64 m/s and 1.54 m/s.

To assess the global accuracy of the gridded wind fields calculated from scatterometer observation, comparison with ECMWF analysis and Hallerman and Roseinstein and Florida State University (FSU) wind stress climatologies were performed (section 4 : Validation, section 5 : Analysis) (Grima, 1997). For instance, the mean and standard deviation values of the difference between ERS and ECMWF wind fields analyses are 0.53 m/s and 1.15 m/s for the wind speed, 0.22 m/s and 1.34 m/s for the zonal component, 0.05 m/s and 1.26 m/s for the meridional component. However, large-scale differences were found in the Southern Hemisphere.