html

ppt

Information is something we would like to know; for example, cod recruitment on Georges Bank. Data is something we measure; for example, wind speed on Georges Bank. In this context, knowledge acts to transform data into information, for example, strong northerly winds would lead to poor larval retention on Georges Bank. The main role of the GLOBEC program has been to provide knowledge, and our task is to identify the information this knowledge can provide from the data we have available.

Knowledge is an abstract concept. In an operational setting, knowledge is replaced by models. The GLOBEC NWA program has produced a wealth of models that could be appropriate for operational use. These models fall into three categories: conceptual, statistical, and dynamical. Conceptual models tend to be qualitative; however, they could be useful for identifying ecosystem indicators. Statistical models, for example the relationship between the NAO and slope water type, and the relationship between slope water and Calanus, are quantitative and are typically derived from historical data. Their main limitation is the potential for non-stationarity. For example, the relationship between slope water and Calanus seems to have become weaker during the 1990s. There is evidence that the presence of large numbers of herring during the 1990s reduced the Calanus population and masked the relationship with the slope water. Non-stationarity will be especially challenging in a warming world; however, statistical models could provide a way of detecting unusual changes and events in the system. Dynamical models are the gold standard. They attempt to represent completely the interrelationships among state variables. Physical models such as FVCOM are the quintessential dynamical models, although the various biological models (NPZ, copepods, larval fish) would also fall in this category. However, it is worth recognizing that all of these models employ some kind of underlying statistical model to parameterize relationships that are unknown (for example, copepod mortality) or are impractical to compute from first principles (e.g. surface heat flux). The main disadvantage of dynamical models is their complexity. They typically require huge amounts of data for initialization and boundary forcing, and given their complexity, they are hard to validate.

Data is simpler and can be classified based on time. Historical data, for example the GB Broadscale survey can be useful for validating dynamical models, but can't be used as an input into an operational model. At the opposite end sits the real-time observational programs, for example, GoMOOS buoys and the various satellite systems. These would be the ideal data source to drive an operational system. Their main limitation is that they collect largely physical data. Between these two extremes we have various on-going sampling programs. Many of the monitoring programs managed by NOAA, for example the CPR and trawl surveys, would fall in this category. Although these surveys are on going, there is a considerable delay, often several months, between when the samples are collected and the data becomes available. Identifying high-value data for priority analysis would be a valuable service that GLOBEC could provide.