| description abstract | Triple collocation is a powerful method to estimate the rms error in each of three collocated datasets, provided the errors are not correlated. Wave height analyses from the operational European Centre for Medium-Range Weather Forecasts (ECMWF) wave forecasting system over a 4-yr period are compared with independent buoy data and dependent European Remote Sensing Satellite-2 (ERS-2) altimeter wave height data, which have been used in the wave analysis. To apply the triple-collocation method, a fourth, independent dataset is obtained from a wave model hindcast without assimilation of altimeter wave observations. The seasonal dependence of the respective errors is discussed and, while in agreement with the properties of the analysis scheme, the wave height analysis is found to have the smallest error. In this comparison the altimeter wave height data have been obtained from an average over N individual observations. By comparing model wave height with the altimeter superobservations for different values of N, alternative estimates of altimeter and model error are obtained. There is only agreement with the estimates from the triple collocation when the correlation between individual altimeter observations is taken into account. The collocation method is also applied to estimate the error in Environmental Satellite (ENVISAT), ERS-2 altimeter, buoy, model first-guess, and analyzed wave heights. It is shown that there is a high correlation between ENVISAT and ERS-2 wave height error, while the quality of ENVISAT altimeter wave height is high. | |
