Any data set, irrespective of its source, should be used with a clear idea regarding the amount of error (or uncertainty) associated with it. In particular, the success of data assimilation systems depends heavily on the knowledge (or good estimates) of random errors of the model and the various data sets used in the assimilation. However, error values or even good estimates are usually either missing or not readily available. Traditionally, statistics regarding the difference between pairs of measurement (either instrumental or model) types are considered enough to provide a statement regarding the accuracy of the data. Although, such statistics are useful to indicate relative “quality” of various data sets, they do not provide any specific statement regarding the “absolute error” of any data set.
There are several methods for proper error estimation with various degrees of success. For example, triple collocation technique (TCT), which is a powerful tool for estimating random errors as far as the errors of various data sources are not correlated, was used to estimate data (and model) errors in a number of fields like surface winds, wave heights and soil moisture.
This work is an initial step towards the evaluation of wave height and wind speed error estimates. In particular, data from Jason-3 and Sentinel-3A (and possibly Sentinel-3B) will be used. The TCT will be used for that purpose. Other data to complement the triplets could be the in-situ (buoy and platform) measurements, ECMWF model forecasts, and data from other instruments like scatterometers.