Wind Error Assessment with Triple and Quadruple Collocation Analysis
Ocean surface vector winds are measured by moored buoys and an increasing number of spaceborne scatterometers. These measurements are assimilated into NWP models on a routine base, and it is therefore important to know their error characteristics. The triple collocation method was introduced by Stoffelen in 1998 and applied to triplets of measurements by buoys, ERS scatterometer, and NWP forecasts in order to simultaneously obtain linear intercalibration coefficients and error variances of wind velocity components. The method has since been applied for assessing the accuracy of a number of scatterometers: ASCAT -A, -B, and -C scatterometers on board the MetOp series of satellites (Europe), SeaWinds on board QuikScat and RapidScat on ISS (USA), and OSCAT on Oceansat-2 and the scatterometer on ScatSat (India). These instruments have different design and therefore different processing, resulting in different error characteristics.
In 2016 the Indian ScatSat was launched in the same orbital plane as ASCAT-A and ASCAT-B but in a slightly lower orbit. As a result, ScatSat underpasses ASCAT-A and ASCAT-B at regular times, resulting in a large number of collocated measurements. Combined with NWP forecasts this yields large triple collocation data sets that enables one to study the error characteristics of the instruments in much more detail than is possible with buoy collocations.
Another possibility is to construct quadruple collocated measurements of buoys, ASCAT (A or B), ScatSat, and NWP forecasts. The triple collocation formalism can easily be generalized to any number of collocated measurements. A quadruple collocation analysis allows one not only to calculate the relative calibration coefficients and the error variances of the four systems involved, but also two additional error covariances. However, these are not easily interpreted as there are twelve possible ways to solve the quadruple collocation equations. The interpretation is facilitated by comparing the quadruple collocation results with those of triple collocation on subsets of the data set.
In this presentation some triple and quadruple collocation analyses will be presented and their interpretation will be discussed.
