Major Shifts in Species Composition and Ecosystem Structure

Cochairs: A. MacCall and P. Kremer

Participants: W. Graham, R. Haney, A. Hollowed, D. Mackas, M. Ohman, T. Powell, G. Rau, J. Rice, J. Schumacher, L. Shapiro, P. Smith, L. Welling, and E. Woehler


Time series of physical and biological measurements in eastern boundary currents (if not in the oceans in general) exhibit nonstationary properties: abrupt changes in descriptive parameters such as mean, variance, or phase relationships that are seemingly unpredictable and inconsistent with those from a preceding time period. Isaacs (1976), who referred to qualitative states as "regimes," described this problem as follows: "There are internal, interactive episodes locked into persistence, and one is entirely fooled if one takes one of these short intervals of a decade or so and decides there is some sort of simple probability associated with it....fluctuations of populations must be related to these very large alternations of conditions."

Qualitative shifts in physical properties, species composition, and ecosystem structure may exist on a variety of spatial and temporal scales, but in the California Current they are most obvious in long-term records of temperature and fish abundance. The historical temperature record from the Scripps Pier shows three prolonged periods of different mean annual temperature (Fig. 10). The middle period, extending from the early 1940s to the mid 1970s, is so much colder than the adjacent warm periods that the warmest years during the middle period (with the exception of the 1958-59 El Niño) are all below the average temperature of the warmer periods (MacCall and Prager 1988). Although the Scripps Pier record is local, SST observations from ships show the recent warming to encompass the entire northeastern Pacific rim, from the equator to the Aleutian Islands (Fig. 11). Again, the lack of cold anomalies since 1976 is striking. From an ecosystem perspective, temperature is only an easily measured proxy variable, and is related to a suite of physical conditions that in turn influence biological processes. Major biological shifts in the California Current ecosystem have been documented for the recent warm period, including a drop in zooplankton abundance and vigorous recovery of the previously depleted Pacific sardine (Barnes et al. 1992).

Despite our recognition that time series of physical and biological oceanographic variables often exhibit nonstationary properties, this concept is overlooked in actual practice. In many cases, nonstationary or qualitative shifts are ignored in order to simplify models to a tractable level. This is especially the case where spatial and temporal coverage are limited. Ambitious large-scale observation and modeling programs such as the World Ocean Circulation Experiment (WOCE) are necessarily still at the level of describing a single pattern of ocean circulation. Similarly, applied models of biological productivity, such as fishery management and marine mammal population models, are typically based on simplifying assumptions of constant reference points such as equilibrium-unexploited abundance.


The number of historically observed transitions is small, but it is reasonable to infer that qualitative shifts in state are typical of the California Current system and that they will continue to occur every few decades. Qualitative shifts in this and other eastern boundary currents pose a suite of challenging problems (Lluch-Belda et al. 1989).


Although the California Current is one of the best-studied regions of the world's oceans, our present understanding of its physical and biological oceanography is inadequate to explain or predict the qualitative shifts we have observed. With the exception of El Niño, these shifts have been overlooked, probably because studies have not included appropriate time and space scales to approach the problem quantitatively. Recent developments in the mathematics of dynamical systems (e.g., "chaos")have demonstrated that qualitative-state shifts can arise from rather simple nonlinear models (e.g., May 1986). With this realization, augmented by a new kit of conceptual "tools", the problem of qualitative shifts is now emerging as a legitimate and fundamentally important area of oceanographic investigation.

scripps sst

Figure 10. Mean annual sea-surface temperatures observed at Scripps Pier, La Jolla, California. Long-term averages are shown for three qualitatively different periods.

monthly sst anomalies

Figure 11. Anomaly of monthly mean sea-surface temperatures from ships of opportunity (Cole and McLain 1989). Contours are (+/-) 0.5 degrees Celsius, positive anomalies are shaded. "Coastal" is approximately 0 to 200 km from shore; "offshore" is approximately 200 to 600 km from shore.

Although qualitative shifts are most noticeable and best documented at large scales in space and time, there is a spectrum of conceptually related phenomena ranging in size down to much smaller, localized scales. For example, the zooplankton in one parcel of water may be dominated by crustaceans (e.g., copepods and euphausiids) while zooplankton in a nearby parcel may primarily comprise pelagic tunicates (salps and doliolids) . Physical circulation patterns have been documented to be important in determining the spatial distribution of the zooplankton, but we lack appropriate time series measurements to determine the mechanisms that cause qualitative difference in the zooplankton community. Zooplankton species compositions and abundances directly affect fisheries recruitment. The processes leading to localized alternative states may provide insights into mechanisms that are important at longer and larger scales. Although these short-term localized changes are more amenable to direct study than long-term ecosystem shifts, short-term studies should not be viewed as substitutes for long-term investigations. Research on state shifts must necessarily cover the full range of time and space scales.


Resource management policies that incorporate qualitative shifts in ecosystems have not yet been developed. Models commonly used for managing living marine resources assume a steady state, perhaps with allowance for environmental "noise." On the west coast of North America, this steady state assumption is clearly inappropriate even in the absence of global climate change: qualitative shifts in species composition and structure appear to be a property of the ecosystem. Many resources and industries may be at risk if qualitative ecosystem shifts result in inappropriate management expectations and responses. Eastern boundary currents are known for their spectacular fishery collapses such as those for the Monterey sardine and the Peruvian anchoveta. Such collapses seem to be an inevitable consequence of inadequate understanding of the resources and the ecosystems. The required knowledge consists of (1) improved resource management models based on understanding of qualitative-state shifts; and (2) improved capability to predict state shifts or to recognize them as early as possible after they have occurred. It is doubtful that adverse fluctuations in the stocks and related industries can be avoided, but if management were armed with the above knowledge and acted appropriately, it should be possible to reduce the severity and duration of the downturns and their resultant economic and social hardships.


Qualitative shifts in ecosystem state occur normally. Although there is no assurance that past states will recur under conditions of global warming, past ecosystem behavior is still our best source of clues about future states and dynamics. Logical induction leads us to the hypothesis that, at least in the California Current system, physical and ecosystem response to global change (whether or not the change is anthropogenically forced) may consist of abrupt changes in qualitative states (e.g., a step function) rather than the gradual change suggested by the smooth forcing function of increasing atmospheric greenhouse gases.

This possibility has profound implications not only for timely detection of the effects of global climate change, but also for planning appropriate societal responses. Therefore, investigations of the response of eastern boundary currents to global climate change must be designed to encompass these changes. We have referred to state shifts as "qualitative," because of their most noticeable properties, but it is nonetheless essential to describe and understand these phenomena quantitatively. GLOBEC-sponsored research on qualitative shifts will provide the understanding necessary to fully consider the effects of global climate change on the California Current system and similar eastern boundary currents.



Long time series of physical and biological observations exist from a number of sources for the California Current system (CCS). CalCOFI samples of fish larvae and larger zooplankton constitute a detailed 40-year time series (1951-91) for a major portion of the CCS. Catch records of commercially important fish have been kept for 75 years. Paleosedimentary records from anoxic basins provide data essential for documenting qualitative shifts on the time scale of decades and longer (see Section 3.3). In addition to directly examining material preserved in sediments, we need to identify proxy variables that reliably indicate physical and biological conditions.

Analyses of historical data will provide the basis for identifying shifts in ecosystem structure and suggest hypotheses for mechanisms or processes that may govern these shifts. These hypotheses can be tested within the context of the modeling and field effort or by comparing results with historical data not used previously to generate the hypotheses (e.g., cross-validation). They can also be validated or invalidated by future monitoring efforts.


The working group advocates two complementary, converging lines of modeling: process-oriented models and process-neutral models. Process-oriented models are particularly appropriate where processes are relatively well understood, as in physical oceanography. However, effects of some relatively well-known physical processes such as wind-induced turbulence may be better modeled by a "process-neutral" transfer function (wind speed cubed), which concisely describes the results of the process without explicitly modeling the process itself (surface wave dynamics, etc.). When less-well-understood processes must be included in a larger model, process-neutral models may be required. These may be drawn from a large family of models including probabilistic models (e.g., Markov models) and empirical transfer functions that may be nonlinear and incorporate appropriate time delays.

These process-neutral models form a natural beginning point for the evolution of more specific models tailored to the processes and mechanisms of the California Current. Process-oriented models follow naturally from neutral models as more information is gained . The process-oriented model then becomes available to replace the process-neutral model, depending on the modeling context. Ideally, interaction between construction and analysis of models and conduct of field research strengthens investigations in both areas.


Ecosystem shifts appear to be closely associated with changes in physical conditions. Better knowledge of the physical processes and characteristics of alternative system states is needed in the field of physical oceanography. Such knowledge would clearly help to explain qualitative shifts in the biology of the system. Improved circulation models (including patterns and effects of upwelling, advection, and transport) are needed. Beyond describing "average" conditions (the significance of which becomes questionable in view of nonstationarity and discontinuous qualitative shifts), we need descriptions of alternative physical states, and knowledge of mechanisms or processes that generate shifts between (and persistence within) those alternative states.


Biological systems are laden with many properties that applied mathematics has shown to generate complicated temporal and spatial behavior, and the phenomenon of qualitative biological shifts is a natural consequence. Some examples of these properties are nonlinear responses to physical and biological changes of the sorts often encountered in the recently developed field of "dynamical systems"; plasticity in trophic relationships among species (especially given individual development from larva through adult, spanning numerous trophic levels); effects of time lags; and continuous spatial (re)partitioning of populations. A mix of process-oriented and process-neutral models is necessary, and that mix will evolve with improved understanding.

Stability properties of ecosystems may arise from specific processes or mechanisms, but alternatively could arise from more general properties of the component physics, organisms, and ecological linkages. In the latter case, process-neutral models may guide subsequent research in several ways, including identification of dependencies that are likely to constrain system trajectories or maintain alternative states; identification of stable and unstable assemblages or configurations of the ecosystem; and determination of model sensitivity to assumed structure or parameter values.


Initially, field studies will concentrate on mechanisms governing qualitative-state shifts in systems small enough that transitions can be observed. This research would concentrate on many of the key questions listed above (e.g., predictability of qualitative shifts, comparative life histories of species, physical and biological precursors). The study of long-term qualitative shifts must rely on historical records, archived samples, and proxy or indicator variables. Efforts to model qualitative-state shifts will identify numerous features and processes requiring clarification and better understanding through field study, including identification of critical mechanisms or sensitive leverage points, and estimation of parameter values for use in models of key processes. Identification of specific areas of study is premature, but an ambitious GLOBEC field program is likely to result. Increased field activity should be anticipated as the program develops, and funding should vary accordingly.


Understanding the mechanisms and causes of long-term qualitative shifts in the California Current ecosystem poses a major intellectual challenge. If we are successful, our knowledge will provide a basis for monitoring and forecasting that would clearly benefit society. Even if prediction proves unfeasible, earlier recognition of qualitative shifts would also be beneficial. Therefore, a major product of this proposed GLOBEC research program is the development of a physical and biological diagnostic capability that could be implemented by an agency such as NOAA.

The proposed prediction and detection capability would be supported by a low-cost monitoring program. Design of that program will follow the research conducted under this GLOBEC program, but some aspects can be anticipated: presumably, existing climatological observation sets (e.g., information from ships of opportunity and coastal stations, atmospheric pressure fields) would be emphasized. A variety of satellite-based sensors may be expected to provide synoptic views of some variables of the coastal ecosystem. These could be supplemented by low-cost opportunistic biological samplers such as Hardy Continuous Plankton Recorders to obtain detailed information about shifts in species composition and distribution of zooplankton, with emphasis on possible indicator species. Other indicators of shifts (recruitment strengths, growth rates, physiological traits) may be extracted from routine biological monitoring of commercial fisheries at little incremental cost to ongoing sampling programs.

The combined monitoring and analysis could produce indexes of environmental indicators analogous to those published by the Department of Commerce for the U.S. economy. NOAA has already moved in this direction by initiating an annual compendium of environmental indicators. The program described here differs from the NOAA effort in two important ways. First, the indexes would be restricted to a better-defined system: the California Current off the west coast of the United States and Canada. Second, the indexes would be better focused, having been developed and selected on the basis of mechanisms and relationships identified by the research program.

The California Current is an ideal laboratory for developing such a predictive system. The background of knowledge and historical observations is among the best in the world, and provides a solid base from which to work. Successful effort in this system will point the way for similar programs in other coastal and oceanic systems.