Wind Product
Get as PDF
| Below are typical examples of wind fields as produced in Soprano. Colors stand for wind speed intensity and black arrows for wind directions. |
|---|
|
|
As the wind blows over the sea surface, it generates gravity-capillary waves which increase small roughnesses at the surface. Those waves respond insantaneously to the wind speed and direction. As the radar backscatter results from interactions between the incidence elecromagnetic wave emitted by the radar and the small scales at the sea suface, it is directly related to the wind conditions at the time of the measurement.
A lot of work has been done to study the relationship between the normalized radar cross section (NRCS) and the wind thanks to co-locations between NRCS and buoys or numerical weather prediction (NWP) model. In general as wind speed increases, the NRCS increases up to a saturated point.
For a given NRCS and incidence angle, a large number of wind speed and direction pairs may be possible, leading to an ambiguous set of solutions. Unlike multi-antenna or rotating scatterometers, the SAR instrument points at a single direction. Wind direction cannot be directly retrieved from the SAR measurement. There are many approaches to solve this ambiguous problem. The most common is the so-called “scatterometry approach” where the wind direction is given by ancillary data and considered as exact. The default mode of the processing software used in Soprano relies on a more elaborate scheme (Bayesian approach) where an apriori wind (speed and direction) is given by ECMWF (or NCEP) NWP model outputs but each vectors components are assumed to be independent with probability density functions to exist modelled as Gaussian distributions. Possible errors on the prior wind vector are thus taken into account.
Error estimation in wind retrieval from SAR is dependant on NRCS, wind direction, incidence angle and on the algorithm used. The errors associated with the backscattered signal include noise from the instrument and the oceanographic and/or atmospheric condition. The spatial resoluion for wind esimate is mainly constrained by the speckle noise (i.e. scattering of coherent electromagnetic waves by the rough surface) and can be minimized by averaging a sufficient amount the data. In practice, for a Wide Swath mode SAR image, the resolution of the wind processed in Soprano is 975 m x 975 m. The relationship between wind forcing and roughness is disrupted by any oceanographic, atmospheric and/or anthropic actions which will impact the sea surface small wind generated waves. Rain changes the backscattered signal by damping the surface roughness. Oceanic features such as ocean fronts, internal waves, tidal currents, algal blooms, melt water, sea ice or oil spills will also interfere with the measurement. This is particularly the case at low wind speeds. For example oil spills change the surface roughness and thereby corrupt the wind estimate especially for low wind regimes.
It is important to note that SAR wind retrieval techniques are designed to retrieve the 10 m neutral equivalent wind. Commonly, equivalent neutral wind speed is defined as the mean wind speed that would be observed if there was neutral atmospheric stratification. Therefore, any unusual boundary layer wind conditions can interfere with the 10 m neutral wind assumption. This type of error can occur close to coasts where influences from local microclimates and topography can create unusual wind conditions which change rapidly over short distances. This source of uncertainty in wind retrievals is therefore particularly important to consider when the SAR is being used close to the coast, which is often cited as the area where it can be most useful. The Bayesian wind speed estimate achieves a root-mean square (RMS) error of 1.87 m/s for 2 to 20 m/s wind speed range, which is a 20% improvement over the classical scatterometry approach (i.e. wind inversion using wind direction given by independent ancillary wind information). This error is significantly reduced for average wind speeds (5 to 9 m/s). However the accuracy is reduced when one looks at very low or high wind speeds, the speed estimates are usually overestimated at low winds (e.g. bias approx 0.6m/s for wind ranging between 1 and 5 ms) and (sometime significantly) underestimated at high winds where the NRCS essentially saturates speeds. Therefore, any estimate of wind speed at 2 m/s or 20 m/s should be considered as20 m/s respectively. The influence from wind direction on the performance of the wind speed retrieval is very small, with a marginally improved accuracy for wind direction parallel to the flight path.
The Bayesian wind direction estimate reaches an RMS error of 30 degrees. This error is however still strongly influenced by the accuracy of the wind direction given by the ancillary source. There is no influence on the wind direction error from the wind speed or the actual wind direction with respect to the flight path. It is therefore true to say that the validity of the wind direction estimate is unrelated to the SAR data, but is a function of the source of the wind direction information. It is important to note that at present time the Bayesian prototype does not make use of the SAR wind streak signature to estimate the wind direction using a Fourier approach.
High winds
|
|
In case of hurricanes, in the Gulf of Mexico, the apriori wind information used is from NCEP GFS model because of its better time resolution (3 hours instead of 6 hours for ECMWF). As hurricanes are rapidly changing meteorological phenomena, the closer of observation is the prediction, the more accurate is the hurricane eye position and subsequently the better is the wind field from the SAR image. Color bar for wind speed indication is adapted to the particularly high values of wind speeds : it is extended up to 30 m/s. Below is an example with hurricane IKE ( ESA IKE web page, ESA web story on IKE). the image was acquired on September 11, 2008 around 03 h 00 UTC in the morning. In the roughness product we can see very well the eye of the hurricane and the wind field map reveals the high wind speeds as well as the winding of the wind directions around the eye. It must be noted that NRCS under extreme wind condition is still poorly known but an active research topic.