Random, Environmental, and Representativeness Errors in Ocean Remote Sensing Versus In Situ Data: An Example of Wave Heights From Altimeters

Understanding the errors and uncertainties is important for the applications of oceanic remote sensing data products. In this study, an error modeling method is presented to decompose the error of ocean remote sensing data into three parts: random errors, environmental errors, and representativeness errors. Random errors refer to the zero-mean error uncorrelated to any external variable. Environmental errors are errors systematically dependent on some external or environmental variables. Representativeness errors are errors coming from the difference with the defined truth with respect to measured spatial–temporal ranges. These error components can be estimated with the collocations of only two systems if continuous observations are available from both systems. The methods of estimating and suppressing the magnitudes of these error components are also presented and are demonstrated using altimeter significant wave heights (SWHs) from the CCI-sea state dataset including 11 altimeter missions. For most altimeters, the three error components are within the same order of magnitude. After suppressing these error terms using the presented methods, the root-mean-square errors between altimeter and in situ measurements of SWHs can be decreased for 25%−45% for these altimeters. The understanding of the contributions of different error components in remote sensing data products can guide the direction to improve data accuracy since different error components need to be suppressed in different ways, and it also provides the reference for applications such as data assimilation of models.