Cyclone Tracking with Ers-2 Scatterometer: Algorithm Performances and Post-processed Data Examples

Every year, Tropical Cyclones (TC) produce important damages on a high number of countries: the floods caused for the heavy rain, strong winds and bad sea conditions are producing human and economic losses. Satellite data can help the scientific comunity to study these events. In particular the processing of the backscattering measurements acquired over the ocean is giving important information on the wind field at sea level. The knowledge of the TC wind field structure can help the scientific community to better understand and better forecast these events. To meet the needs of the scientific community, the Product Control Service (PCS) at ESRIN has developed a post-processing procedure for the fast delivery (FD) products processed from the data acquired with the C-band Scatterometer (Scat) flown onboard the ERS satellites. The major skills of this post-processing are: the detection of a TC, the quality improvement of the retrieved wind field and the availability on a web in near "real-time"of a report about this TC and the corresponding reprocessed Scat FD products. The paper describes the techniques adopted to detect the TC and to improve the Scat wind field. It also presents statistic results about the detection skill of the method. INTRODUCTION The ERS Scatterometer is a three antennae radar working at C-band. This frequency is very sensitive to the sea surface roughness which could be related with the wind field by using an appropriate model. The returned echoes are stored onboard the satellite and downloaded to the ground station every orbit (roughly every 100 minutes). The processors installed at each ground station retrieve the backscattering coefficients (three sigma naught for three different azimuth angles) from the echoes with a resolution of 50x50 kilometers and sample on a grid of 25x25 kilometers. In addition, by using a wind retrieval algorithm based on the CMOD-4 empirical model [1] and an ambiguity removal algorithm, the wind vectors (speed and direction) are computed from each sigma naught triplet. Within three hours from sensing, backscattering coefficients and wind field estimations are provided to the international meteorological and oceanographic organizations via the World Meteorological Organization network called GTS. The European Centre for Medium-Range Weather Forecasts (ECMWF) has proven that the assimilation of Scat data into meteorological models can improve the forecast and the estimation of the TC’s position [2]. X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X Figure 1: Red points (nodes) are used to calculate the vorticity of the green node. However, wind information given in the FD product acquired over TC’s are showing two major problems. The first one is that the retrieved wind speed is highly underestimated for high wind speed like in TC’s. The second one is linked to the wind direction; the ambiguity removal algorithm has difficulties to choose the correct one when the wind structure is too complex. These two problems are linked to the calibration of the CMOD-4, which was not tuned for high wind speed (as we have in the TC), and with the performance of the ambiguity removal algorithm used at present in the ground processing. In this paper it is described a procedure used within the ESRIN PCS to detect the TC from Scat FD data and to upgrade the wind field. In the conclusion it is given a description of the report information available on the web. THE DETECTION OF A TC Dishtwal et al. in [5] estimate the TC centre using the methodology of high density Scat winds. They assume that there is a circular symmetry (high vorticity) in the inner core of the cyclone in the intense stage. As the speed contours are not perfect circles, the centre is found by using a minimisation approach. The method presented here uses the sigma naught properties acquired over TC’s to detect the centre of the cyclone. This is an advantage because the measurements are directly linked with a geophysical phenomena (sea roughness) rather than being the results of a geophysical model function inversion. The flow chart of the algorithm used to detect a TC is shown in Figure 2. The Vortex Integral The first step is the detection of the TC using the criteria of the high vorticity of a node. The vorticity of one node is calculated using the following formula: window the of dimension the is n angle look beam mid the is North wrt direction wind the is speed wind the is )) , ( ) , ( sin( ) , ( ) , ( )) , ( ) , ( cos( ) , ( ) , ( with )] , 1 ( ) , 0 ( [ )] 1 , ( ) 0 , ( [ )] 0 , ( ) 1 , ( [( )] , 0 ( ) , 1 ( [(