Evaluating Ocean Analyses and Forecasts

The usual process for developing a new ocean forecasting system, or significant upgrades to an existing system, involves a number of stages. Scientific developments will be tested individually to ensure that they are producing the expected change in the system. Once a number of developments are available, they will be put together into a new version of the system and this must then be thoroughly evaluated during the validation phase. This validation is usually done by means of the evaluation of a set of hindcasts of the system, where the system is run over a multi-annual period in the past. This tests that the overall changes to the system produce the expected improvements. Once the validation has been carried out, the system can be implemented operationally. At this stage it is important to continuously assess and monitor the accuracy of the system using a verification system. The results of both the validation and verification are useful for providing information to users of the system about the expected accuracy. User-specific evaluations can also be carried out to assess the suitability of the system for a given application.

A number of examples of evaluating ocean forecasting systems are given below taken from various sources (e.g. Ferry et al. 2007; Oke et al. 2008; Metzger et al. 2008, 2009; Storkey et al. 2010), providing illustrations of some commonly used methods. The advantages and shortcomings of each method are outlined.

0 0

Post a comment