The role of validation within MARSAIS
The motivation of the MARSAIS Project has been to promote and facilitate the use of synthetic aperture radar (SAR) data of the ocean by non-expert users, so that potential marine and coastal applications of SAR will be developed beyond experimental demonstrations and become operationally routine in fields such as offshore engineering, shipping operations, marine pollution control etc. If SAR ocean data products are to become widely accepted in operational oceanography and coastal management contexts, the users must have confidence in them. Discerning users of the ocean information products that are generated by the models and processes incorporated into the MARSAIS are concerned about the quality of those products, specifically their accuracy, applicability and reliability. It is for this reason that the MARSAIS Project, among other things, explicitly set out to address the issue of data quality. Here, we report on how calibration and validation questions have been addressed across the whole breadth of the Project.
The application of satellite data in an operational context should always be performed cautiously and critically, especially when the methodology is relatively young as for the marine uses of SAR. As applications of marine SAR develop it is important that users recognise the importance of validating SAR data products by comparison with independent measurements. Therefore the MARSAIS Project also set itself the goal of educating users to adopt a critical stance towards assessing the quality of the analytical tools and products being offered.
At the planning stage of the Project, it was assumed that work on calibration and validation would be largely a matter of abstracting all the quality assessment information emerging from the treatment of the different processes and applications of marine SAR throughout the rest of the MARSAIS activity. Early in the Project a report (Robinson and Ufermann, 2001) was produced which identified the specific quality assessment information requirements that should be met before models or procedures were incorporated within the MARSAIS prototype. It also characterised the validation data that should be made available to accompany the demonstration datasets.
However, as the project developed it became apparent that the status of calibration and validation was less mature in some marine SAR application areas than others. For some of the key applications of SAR data, such as measuring wind and waves, there exist well documented quality assessment references for the methods and reliable sources of independent validation data for the products. The same cannot be said for some of the other methods, such as those based on deriving the current gradients that modulate the surface roughness. While in these cases the analytical process models have been calibrated and verified, there have been insufficient independent trials of the methods to generate more than a few isolated independent data sets which validate the end products.
It is important to recognise at the outset that the main reason why some SAR applications are more difficult to validate independently than others is that the alternative means for measuring the ocean parameter either do not exist or are very difficult. Whereas there are established alternative methods for measuring wind and waves, and these allow the SAR measurements to be validated, there is no other way of easily measuring the propagation characteristics of internal waves, determining the current shear across ocean fronts, or for routinely monitoring surface slicks and films. In these cases the difficulty of providing independent validation represents the classic problem for any new remote sensing system that has unique capabilities. Such a problem is a consequence of a measuring technique's success rather than the reverse.