Maybe the most unexpected and surprising result of this exercise is the large difference existing in the perception between users and experts. While users are always quite satisfied with the systems, experts are generally more critical, especially regarding those services related to the green ocean. These different perceptions are probably due to several factors. Experts are considering the quality by validating with observations, a process that can be more severe than the users' perception, which is based on application experience. On top of that, and probably more importantly, an approximate indication of the value of an EOV could be more than sufficient to allow the satisfactory practical use of a system. For example, a port pilot could be satisfied by knowing if wave heights will or will not be over a given threshold, but they are not affected if the waves have one value or another over that threshold because operations will be canceled independent of how much the variable is exceeding the threshold. On top of that, the experts know where to look to find the problems in the systems, and they do so by scrutinizing all variables, oceanographic situations, and geographical regions of the domain of the system, while the end-users commonly have a perspective limited into one particular aspect of the service. When the range of processes and results under scrutiny is larger, it is to be expected that more mishaps can be found.

To the trained eye, the evaluation of the experts is not surprising. It clearly shows that wave and sea level forecasting systems are quite mature and that more effort must be made with respect to the green ocean, probably at all points of the value chain, from observations to downstream services. Also, the progressive degradation of the system's performance as we come closer to the coast is obvious. This is most likely due to several factors that can be present on some occasions, such as the lack of an appropriate resolution, poor atmospheric forcing at that scale, the lack of river data, and the complexity and importance of the nonlinear effect and interactions between EOVs near the coast. Take, for example, a sea level that has a striking low mean quality value of only 5.7 at the coast. The degradation could be due to lack of coupling with waves near the coast or to poor bathymetry. The importance of this skill's degradation is reinforced with the analysis of the results from the applications survey, which demonstrates that it is precisely near the coast where ocean forecasting EOVs are more widely used.

It is also not surprising that in most cases, the lowest value for the experts is associated with quality and not with existence, timeliness, or accessibility. This demonstrates the relative maturity of the operational chains and the need to improve in all the aspects related to the prediction engines, including the acquisition of new data to be assimilated.

Nevertheless, there are some aspects that would require further investigation. It is striking that, when not satisfied, the experts tend to evaluate all aspects of the system in a negative way at the same time. There is no evident reason why biogeochemical or biological models behave so poorly in technical aspects such as timeliness. Maybe the difficulty in obtaining good, accurate solutions is biasing the criteria of the experts when evaluating these on more purely technical aspects. This fact will be further investigated in the future. Another potentially biased result is the qualification on the existence of systems. Both users and experts are replying based on their experience and, obviously, the geographical areas where they are developing or using systems, including at least the system(s) that they deal with. This situation can be different in another regions, unbeknownst to the participants. OceanPrediction DCC is developing a global map of existing OOFS that will illuminate this question.

In summary, during the last 3 decades, the evolution and improvement of ocean forecasting services has been dramatic, and this exercise has proven that the existing systems are useful and accurate in the opinion of the users. Nevertheless, the experts have clearly stated that further improvements are required, mainly when approaching the coast and for the systems dealing with the green ocean. The Ocean Decade Implementation Plan is giving us a framework to further develop the systems, and this is the objective of OceanPrediction DCC and its associated decadal actions, such as ForeSea, Ocean Best Practices, DITTO, CoastPredict, GEMS Ocean, and others. To measure the success of these initiatives, it could be interesting to repeat this exercise by the end of the decade.