at a variety of scales. For example, at relatively small scales: How does the consumption of intertidal herbivores by humans affect algal production? At relatively large scales: Is offshore production, as indicated by chlorophyll, related to the nesting success of seabirds? According to its title, the GEM plan takes the Gulf of Alaska as its scope. However, the central hypothesis of the plan—that natural and anthropogenic factors interact to influence biological productivity—could be addressed at a variety of scales in the Gulf of Alaska.
Building on the knowledge base. As a new research program is developed it can build on past work in three ways: (1) by continuing past work (extending the time frame), (2) by collecting information on unstudied variables (extending the intensity), or (3) by collecting information in unstudied locations (extending the spatial scale). The choice among these options requires that existing data be synthesized first. Many of the natural changes in the Gulf of Alaska are thought to cycle at intervals of several decades. Because little monitoring has been ongoing for such long periods, continuing past measurements may represent the most effective way of testing for variation at this temporal scale. Second, if two existing measurements show striking correlations, measuring new variables can be an effective way of testing the mechanisms of interaction among complex environmental factors. For instance, if ocean survival of salmon varies with phytoplankton production, then measuring forage fish abundance and demography could provide an intermediate food-web linkage. Finally, extending the spatial scale of measurements is important for determining the generality of hypotheses that have previously been tested only locally. This last choice in particular requires adequate synthesis of existing data; otherwise, it is impossible to ask whether existing patterns are general (because there are no existing patterns).
Management needs. Although GEM’s mandate is not resource management, most large science programs are justified in part by the usefulness of products provided for decision makers (Weisberg et al., 2000). Most management issues are fundamentally local because this is the scale of human impacts (barring atmospheric change); however, the precise locations where prior data would be useful can shift over time. For instance, baseline data in Prince William Sound would be useful if another oil spill occurred there but it would not address eutrophication in Cook Inlet. A broad geographic scope can improve the chances that long-term measurements remain relevant as management issues change.
Accessibility and cost. Cost is the basic limitation setting the tradeoff between intensity and scale of monitoring. One drawback of a large geographic scope is that tremendous resources are required simply to travel to research sites. Travel costs may be reduced if monitoring is carried out in local communities and if automated data collection is used for basic