A vital step in the U.S. GLOBEC program is to link the information developed from field studies on planktonic processes, to the much larger scales of global climate change, marine population space, and the rise and fall of living marine resources. Field studies on processes, by their nature, are limited in space and time, seldom being larger than 10,000 nm2, and lasting longer than five years, while major marine populations may occupy and migrate more than 100,000 nm2 and climate change evolves over several decades to hundreds of years. Linkage of the scales of process studies to large time and space scales requires modeling and retrospective studies. The objective of this report is to provide the rationale for a U.S. GLOBEC Retrospective Data Analysis Program or GLOBEC-Retrospective, for short.
Justifications for the other U.S. GLOBEC discipline-based-programs (e.g., biotechnology, acoustic and optic technology, and modeling) apply equally well to retrospective work: the need for substantial lead time; the special discipline and expertise required; and the integrative nature of the work. Clearly, retrospective work, if subsumed in other U.S. GLOBEC activities, will be subservient to the needs of those programs, narrowing the focus to specific sites and eliminating much of its integrative potential. An independent U.S. GLOBEC retrospective program should not be locality limited, but should extend beyond the spatial and parameter domain of individual U.S. GLOBEC field studies. Efforts within such a program may take on a basin-wide or coast-wide scope. Other efforts will involve comparative studies of multiple ecosystems on a global- scale. It is such work, in the long run, that will help integrate U.S. GLOBEC field work for the oceans in general, and lead to direct understanding of the impact of long-term climate change and climate variability on marine populations.
Satellite Data. These time series have the shortest period, with the longest of them, the Advanced Very High Resolution Radiometer (AVHRR) sea surface temperature (SST) series, approaching 15 years. The archived AVHRR data are not presently suitable for studies of interannual climate variability, since most of that variability is related to volcanic aerosols. There will soon be a reprocessed PATHFINDER global AVHRR data set, with 9 km spatial and twice-daily temporal resolution. These data will cover perhaps 10-13 years from about 1982 to the present. The Coastal Zone Color Scanner (CZCS) color data covering approximately seven years (1979-1986) is available globally with 4 km resolution and spotty temporal sampling. Both AVHRR and CZCS data suffer greatly from cloud contamination. AVHRR sensors will continue to collect data on SST into the future. The SeaWiFS color sensor, with similar sampling characteristics but greater sensitivity than the CZCS, will be launched in 1994 and collect data for five years. Other color sensors will be launched on European and Japanese satellites planned for the future. Altimeter data are available for about two years (November 1986--October 1988) from the Geosat exact repeat mission and will also be available for a decade starting in 1992, with two sensors (ERS-1 and TOPEX/POSEIDON) operating at present and more planned.
Although satellite-sensed data are not extensive in time, they are extensive in space and in the level of detail provided. Thus, they cannot be used to investigate variability of a given ecosystem over periods greater than about ten years. Satellite data, however, can be used to investigate how similar ecosystems in different regions of the world ocean respond to differences in the magnitudes and patterns of the various forcing fields within the present environmental variability. This may substitute present spatial variability for future temporal variability. As an example, Bakun (1990) has hypothesized an increase in upwelling favorable winds over eastern boundary currents as a response to global warming. The present eastern boundary currents (e.g., California, Benguela, Peru-Chile, Canary) experience a range in the strength of large-scale local wind forcing, and differences in forcing by basin-wide currents (e.g., west wind drift and equatorial currents). Comparative studies, using satellite-sensed data, between the different eastern boundary currents can quantify the degree to which changes in the relative strengths of the different forcing fields affect the structure of the current systems, the amount of mesoscale structure within the larger circulation, and the biological response to the mesoscale structure (using ocean color). Such studies often generate more focussed questions and hypotheses, which can then be tested by model studies or used to direct more efficient field work. The net result may be the development of new, satellite-based indices of biologically relevant physical processes, which can be applied in an ongoing, operational monitoring fashion into the future.
In Situ Data at Points and Gridded Products. In situ data at fixed points include records of tide gauge stations, measured winds from buoys or coastal sites, shoreline SST and salinity data. Some of these data extend back 100 years. Another source of in situ data is the Comprehensive Ocean-Atmosphere Data Set (COADS) from merchant ships (Woodruff et al., 1987). The ship observations are at points which are not exactly repeated, although they may be nearly repeated along frequently used ship tracks. The data (available from the National Center Atmospheric Research (NCAR)) consist of surface ocean and atmosphere data from ships and buoys, extending back more than 100 years. Besides the quality controlled original observations, monthly means of the most useful data, on a two-degree grid, are also available (and more commonly used), making the data set more like the gridded data sets described below. The Master Oceanographic Observations Data Set (MOODS), is a compilation of subsurface profile data from CTDs, BTs and XBTs, much of it extending back into the 19th century. The data are stored at the Fleet Numerical Oceanography Center (FNOC). These data are extensive in coastal areas, but unwieldy in the form available from NCAR. Gridded surface physical fields, such as the FNOC pressure and winds, NMC and European Winds, gridded SST from NOAA, and others extend back in time 10 to 50 years. Both NMC and ECMWF are currently performing reanalysis of extended periods of meteorological data. This reanalysis will remove the effects of changes in analysis [and forecast] in the analyses. The effect of changing data distribution will still be present in the new analyses and should be considered when making use of these data. In addition, there are long time series of other ocean and atmospheric indices-examples are the El Niño-Southern Oscillation (ENSO), Pacific-North American (PNA), North Atlantic Oscillation (NAO), Aleutian Low (ALI), and coastal upwelling (Bakun) indices.
Complementary biophysical in situ survey data also exist, which, in the case of California Cooperative Oceanic Fisheries Investigations (CalCOFI), extend back about 40 years. This research consortium began in 1947 to examine the physical and biological oceanographic causes for the collapse of the U.S. west coast Pacific sardine population. Data collected during the CalCOFI program include meteorology, hydrography, plankton abundance and fisheries catches. Although the physical data from CalCOFI surveys are available (and are probably included in the MOODS data set), much of the biological time series information, other than fish larvae and plankton volumes, still remains in the collection jars-unidentified and uncounted. These samples desperately need their contents examined, enumerated, and made widely available in digital form. Similar biological data sets, perhaps not so long in duration as CalCOFI, are available for the Atlantic Ocean and perhaps the Indian Ocean. The continuous plankton recorder (CPR) has been used extensively to document the seasonal, interannual and interdecadal patterns of zooplankton from both the North Sea and the NE Atlantic, and for a shorter period for the NW Atlantic (see for example, Colebrook, 1991). Plankton populations on the continental shelf of the northeastern U.S. were collected during 1977 to 1987 by the Marine Resources Monitoring, Assessment and Prediction (MARMAP) program. Plankton biomass and abundance of zooplankton taxa from the coarse-mesh nets are reported in various reports, but the fine-mesh samples remain unexamined in jars. Recently, talks have begun between U.S. and Russian scientists which might result in the extensive Indian Ocean plankton data sets collected by Russian researchers becoming more widely available for the examination of longer-term trends and their potential relation to climate.
Saying that these extensive arrays of data are available from various sources obscures the fact that any research project that intends to compare multiple data sets will spend a great deal of time acquiring, checking, and organizing the data. In addition, in the case of unexamined, archived biological samples, individuals must be identified and counted. All of this must occur before the data series can be examined individually for trends or periodicity, compared with other data series, or examined in new time and space scales.
Animal Population Data. Fisheries data, and marine bird and mammal censuses, provide extensive time series-some going back over a hundred years. A special property of the long historical records of fish, mammal and bird abundances, is that they may provide not only a history of variation in population size, but also may provide estimates of life table variables that control population growth (growth, mortality, maturity, and reproduction). Ultimately, these population variables are the ones that must be linked to physical forcing variables to forecast the effects of climate change on marine animal populations. In this regard, time series of growth and survival of the early planktonic larvae of fishes and other living marine resources are a rare, and particularly valuable asset for U.S. GLOBEC because of the programs focus on planktonic stages. The CalCOFI program is by far the richest source of such information in the world, but even here such data exist for only two species (sardine and anchovy) out of the hundreds of possible species. Conversion of a larval abundance time series to stage specific survival would be a valuable contribution to GLOBEC. In the North Atlantic, fisheries catch records for many stocks are available since the 1940s, and for some stocks (e.g., Arcto-Norwegian cod) since the mid-1800s. Patterns of Norwegian cod catches, but not magnitude have been estimated from as early as 1570. Likewise, environmental conditions from that early are known from temperature and drifting ice records. However, due to changes in the operation of specific fisheries, catch statistics are not the same as population size and neither is an estimate of fisheries recruitment (the parameter that we would really like to know).
Paleoecological Data. Since the focus of U.S. GLOBEC is on animal population dynamics, the most pertinent paleoecological time series are those with high temporal resolution (1-10 year sample interval) and, in particular, those that cover the last few hundred years where prehistoric marine ecosystem data can be merged with observational data. Traditional marine paleoecological studies consider processes occurring over geologic time scales, with temporal resolutions of hundreds of years, and are of little interest to U.S. GLOBEC. Marine sediments in a few locations preserve the annual to decadal signals needed for animal population work, allowing detailed histories to be reconstructed. Examples of such locations are the varved sediments of the Santa Barbara Basin off southern California, Saanich Inlet (British Columbia), the Gulf of California (Mexico), Cariaco Basin (Venezuela), and possibly slope waters off Peru and Chile.
However, in order to fully exploit the potential of using varved sediments to study high resolution climate change and its impact on marine biota, it is important to know 1) how and during what season individual laminae are formed, 2) how the seasonally varying input of both biogenic and terrigenous material relates to changing environmental conditions, and 3) what proxies, of both biological and physical processes, are preserved in the varves. Time series sediment trapping, combined with hydrographic measurements and remotely sensed observations of surface ocean conditions provide an ideal means to investigate these varve-related questions (Thunell, et al., in press). In addition, Deuser (1987) has clearly demonstrated the usefulness of long term sediment trapping programs for studying interannual variability in biological processes. Such sediment trap studies have been initiated in the Gulf of California (by Thunell & Baumgartner) and Santa Barbara Basin (by Thunell), and will hopefully begin in Cariaco Basin in late 1994.
Most of the high-resolution paleoclimatic information for the past one to two millennia has come from terrestrial sources, particularly from tree-ring studies (see Fritts, 1976; Lough and Fritts, 1985; and many others) and glacial ice cores (Thompson et al., 1986). Marine records developed from coral growth ring rings (Cole and Fairbanks, 1990; Druffel, 1985; Lea et al., 1989) and varved sediment deposits (Baumgartner et al., 1992) are incomplete. We still have only a vague notion of the high-frequency variability (decadal through centennial scales) in pelagic populations and ocean climate, in large part because of the lack of concentrated research on appropriate sediments at these time scales. The Santa Barbara Basin and the Gulf of California are now yielding detailed interdecadal descriptions of pelagic fish populations through the analysis of rates of fish scale accumulation in these sediments (Baumgartner et al., 1992; Holmgren and Baumgartner, in press). This information will soon be accompanied by other proxy data sets such as the interannual and decadal scale variability in the planktonic foraminifera from both the Santa Barbara Basin and the Gulf of California, and the carbon and oxygen isotopes in their shells (Baumgartner and Herguera, personal communication; Thunell and others, personal communication). Work elsewhere indicates a well preserved high- resolution fish scale record off southwestern Africa (Shackleton, 1986; and R. Johnson, personal communication) accompanied by a well-preserved foraminifera record for which the isotopes have been analyzed (Herbert, 1987). Many more potential sites exist in the coastal ocean which need to be explored. Development of additional high resolution paleoecological time series data will probably require funding by the U.S. Global Change Research Program through NOAA or NSF GLOBEC sources.
Physical Time Series Analysis. A wealth of physical data exists that has not been used to any great extent to address climate questions relevant to marine ecosystems. Such work will require detailed reexamination in space and time of existing information and the development of new and more powerful indices of physical forcing events of pivotal significance to marine animals. For example, indices of physical processes that are more closely linked to regional transport of larvae, or planktonic production are needed. Other examples are indices of mesoscale processes such as the amount of mesoscale structure and the amount of eddy kinetic energy and how these might relate to secondary production and fish recruitment. As mentioned above, such indices might be constructed from existing satellite IR and color images and from satellite altimetry (wave number spectra, structure functions, etc.). Besides extending these indices into the future, if these indices can be linked (i.e., correlated) to longer existing time series of wind or sea level variability, proxy time series of physical variability in the ocean can be constructed that extend back in time. Development of proxy time series for biological, as well as physical environmental variability is also needed, based on easily measured and widely available parameters. Studies to understand the linkages between the proxies and the biological and physical variables they represent will ultimately lead to a greater understanding of the mechanisms controlling the variability in the biological parameters.
A second, and highly important, topic in time series analysis is the identification of underlying trends in physical processes attributable to climate change, that may be obscured by local, interannual and interdecadal variability. An example is Bakun's (1990) analysis which suggests that climate change (warming) may result in increasing upwelling-favorable wind speeds along the California coast.
Approaches to these questions include the application of new statistical approaches to existing time series, examination of regional and temporal patterns, use of state-space and other statistical methods to describe long-term, non-stationary and non-linear trends and patterns in variability.
Hypothesis Testing Using Time Series Data. Retrospective comparisons of the dynamics of marine animal populations can be used to test a variety of hypotheses using the comparative method. Comparisons of the same or similar species in different ecosystems can be used to assess population responses to different forcing variables (Bakun & Parrish, 1982). Alternatively, comparisons can be made of recruitment success of different species within the same ocean basin to identify patterns in the responses of species to basin-wide environmental changes (Hollowed, et al., 1987; Hollowed and Wooster, in press; Wooster and Hollowed, in press; Cohen et al., 1991).
Time series analysis can also be used to evaluate hypotheses regarding physical forcing events. For example, the hypothesis that an optimal level of physical forcing exists for recruitment (Cury and Roy, 1989) which balances the contravening effects of high productivity and offshore transport of larvae, might be tested by linking life stage population matrices (Caswell, 1989) to an appropriate environmental time series. Clearly, such hypothesis testing using retrospective analysis not only contributes to the understanding of marine ecosystem response to climate change but also provides important baseline information to guide future U.S. GLOBEC field work.
New primary time series data could be developed using a variety of approaches. An obvious first step is to expand the species base by identifying additional fish and plankton species occurring as remains or fossils in existing sediment cores and in preserved samples in survey archives. Changes in species composition or relative abundances may indicate shifts in the distribution of planktonic organisms in response to physical forcing (Baumgartner et al., 1985; Lange et al., 1990; Sautter and Thunell, 1991; and Thunell and Sautter, 1992). Second, chemical tracers in sediments or plant and animal remains might be used to describe changes in the intensity of coastal upwelling and to characterize circulation patterns and intensity (e.g. delta-18O and delta-13C of forams: Thunell and Sautter, 1992; Cd/Ca of forams: Van Geen, personal communication; radiocarbon age differences, Southon et al., 1990).
There are other questions which must be answered to properly evaluate models. In particular, the question of predictability should be addressed. Predictability is measured by considering the same system in two slightly different states and allowing the system to evolve for a forecast period. The difference between the final states of this system relative to the initial difference determines the predictability of the system. Highly predictable systems will have a final difference which is similar to the initial difference. Unpredictable systems may have a much larger final difference than the initial difference. The basic goal of ecosystem modeling is to predict, given current conditions and defined processes, the state of a system in the future. Ecosystems are complex and therefore are most certainly difficult to model. Still, there may be bulk measures of the state of the system which can be accurately predicted. Why is predictability important to model evaluation? Because there is no value in comparing a model to the ecosystem on the basis of an unpredictable measure of the system state. From a broad perspective, we need to understand the ecosystem function and dynamics to determine which aspects can be predicted. Ecosystem studies are dramatically different from physical systems, in that the complexity of the interactions in ecosystems prevent all but the most simple (and hopefully, most important) processes of the system from being parameterized. Because of this difference, the definition of predictability must be broadened to include the sensitivity of the system to perturbation in the system description in addition to changes in the state of the system.
Model development may be improved by using retrospective data sets to set (or provide limits on) little-known parameters. This must be done with care since errors in model formulation or unconsidered processes may lead to incorrect parameter estimates. One especially troublesome aspect of this process of model development is the possibility that the value of a determined parameter may be a function of the climate. Retrospective data sets provide the best means of elucidating this aspect since they extend over the largest possible climate variation and therefore may identify the invalid assumption that a parameter is a constant.
Other federal agencies support various facets of retrospective ecological data analysis, but few encourage marine ecosystem research. Terrestrial systems are the focus of DOE, DOI, EPA and USDA. Exceptions are NSF (GLOBEC and LMER), ONR (Ocean Ecosystem Dynamics) and NASA (through satellite missions and the Earth Observing System (EOS)). Many of these programs concentrate on primary production rather than taking a more integrated view of ocean ecosystems. More importantly, these exceptions devote their resources to observation -- the combination of model studies and retrospective analysis are given low priority.
Thus, a substantial gap in support for retrospective biophysical research exists. No single agency funds marine research that integrates physical and biological variability as derived from long term historical data sets, within a modeling framework. That gap must be filled by a U.S. GLOBEC-Retrospective program. It is U.S. GLOBEC's responsibility to support such work, because GLOBEC is the marine ecosystem component of the U.S. Global Climate Change Research Program.
"The first step in any investigation is to determine what we know and what we don't ..." (U.S. GLOBEC, 1991).It is the responsibility of U.S. GLOBEC to integrate our present knowledge of physical oceanography with the known population biology of marine organisms. This requires assessment of what we think we know plus examination of long term physical and biological variability contained in historical records. From the beginning U.S. GLOBEC has recognized the importance of historical records, since one of the criteria for the selection of U.S. GLOBEC field study sites is "a considerable historical database on the distribution and abundance of target species, their physiology and ecology, local climate, and fluid dynamics". Such retrospective data sets "will aid not only in planning research, but also in model verification." (Peterson, 1991). We advocate that the analysis of historical data should be used to accomplish integration and comparison of multiple ocean ecosystems, in parallel with modeling efforts. A U.S. GLOBEC-Retrospective initiative should occupy it's place in the GLOBEC "tool box" along with other integrative tools-modeling, biotechnology, acoustic and optical technology. All of these U.S. GLOBEC components aid us to design, support and integrate the U.S. GLOBEC regional field studies and ultimately to extend our understanding of the processes responsible for climate change and its effect on marine populations.
Bakun, A. and R.H. Parrish. 1982. Turbulence, transport, and pelagic fish in the California and Peru Current systems. Calif. Coop. Oceanic Fish. Invest. Rep. 23:99-112.
Baumgartner, T.R., A. Soutar, and V. Ferreira-Bartrina. 1992. Reconstruction of the history of Pacific sardine and northern anchovy populations over the past two millennia from sediments of the Santa Barbara Basin, California. Calif. Coop. Oceanic Fish. Invest. Rep. 33:24-40.
Baumgartner, T.R., V. Ferreira-Bartrina, H. Schrader, and A. Soutar. 1985. A 20-year varve record of siliceous phytoplankton variability in the central Gulf of California. Marine Geology v.64:113-129.
Caswell, H. (ed.). 1989. Matrix population models: Construction, analysis and interpretation. Sinauer Associates, Inc., Sunderland, Mass., 328p.
Cohen, E. B., D. G. Mountain, and R. O'Boyle. 1991. Local-scale versus large-scale factors affecting recruitment. Can. J. Fish. Aquat. Sci., 48, 1003-1006
Cole, J. E. and R. G. Fairbanks. 1990. The southern oscillation recorded in the delta-18O of corals from Tarawa atoll. Paleoceanography, 5, 669-683.
Colebrook, J. M. 1991. From seasons to decades in the plankton of the north-east Atlantic, pp. 29-46 in Kawasaki, T., S. Tanaka, Y. Toba and A. Taniguchi, [Eds], Long- term Variability of Pelagic Fish Populations and their Environment. Pergamon Press, Oxford.
Cury, P. and C. Roy. 1989. Optimal environmental window and pelagic fish recruitment success in upwelling areas. Can. J. Fish. Aquat. Sci., 46, 670-680.
Deuser, W.G. 1987. Seasonal variations in isotopic composition and deep-water fluxes of the tests of perennially abundant planktonic foraminifera of the Sargasso Sea: Results from sediment-trap collections and their paleoceanographic significance. J. Foram. Research 17(1):14-27.
Druffel, E. M. 1985. Detection of El Niño and decade time scale variations of sea surface temperature from banded coral records: implications for the carbon dioxide cycle, in E. T. Sundquist and W. S. Broecker, [Eds], The Carbon Cycle and Atmospheric CO2: Natural Variations Archean to Present, American Geophysical Union Geophysical Monograph, 32, 111-122.
Fritts, H. C. 1976. Tree-Rings and Climate. Academic Press, New York.
Herbert, R.S. 1987. Late Holocene climatic change: The Little Ice Age and El Niño from planktonic foraminifera in sediments off Walvis Bay, South West Africa. Joint Geol. Surv. / University of Cape Town (South Africa), Mar. Geosci. Unit. Bull. No. 18, 45p.
Hollowed, A.B., K.M. Bailey, and W.S. Wooster. 1987. Patterns in recruitment of marine fishes in the northeast Pacific Ocean. Biol. Oceanogr. 5:99-131.
Hollowed, A. B., and W. S. Wooster. (in press) Decadal-scale variations in the eastern subarctic Pacific: B. Response of northeast Pacific fish stocks. Can. Spec. Publ. Fish. Aquat. Sci.
Holmgren-Urba, D. and T.R. Baumgartner. In press, 1993. A 250- year history of pelagic fish abundances from the anaerobic sediments of the central Gulf of California. Calif. Coop. Oceanic Fish. Invest. Rep. 34:
Lough, J. M. and H. C. Fritts. 1985. The southern oscillation and tree-rings: 1600-1961. J. Clim. Appl. Met., 24, 952-966.
Lange, C.B., Burke, S.K., and W.H. Berger. 1990. Biological production off southern California is linked to climatic change. Climatic Change 16:319-329.
Lea, D. W., E. A. Boyle, and G. T. Shen. 1989. Coralline barium records temporal variability in equatorial Pacific upwelling. Nature, 340, 373-376.
Peterson, C.H., ed. 1991. GLOBEC: Global Ocean Ecosystems Dynamics. Initial Science Plan. Joint Oceanographic Institutions, Inc., Washington, D.C., 93p.
Sautter, L.R. and R.C. Thunell. 1991. Planktonic foraminiferal response to upwelling and seasonal hydrographic conditions: sediment trap results from San Pedro Basin, Southern California Bight. J. Foraminiferal Res. 21(4):347-363.
Shackleton, L.Y. 1986. An assessment of the reliability of fossil pilchard and anchovy scales as fish population indicators of Namibia. Joint Geol. Surv. / University of Cape Town (South Africa), Mar. Geosci. Unit. Bull. No. 17, 141p.
Southon, J.R., D.E. Nelson, and J.S. Vogel. 1990. A record of past ocean-atmosphere radiocarbon differences from the northeast Pacific. Paleoceanography 5(2):197-206.
Thompson, L. G., E. Mosley-Thompson, W. Dansgaard, and P. M. Grootes. 1986. The Little Ice Age as recorded in the stratigraphy of the tropical Quelccaya ice cap. Science, 234, 361-364.
Thunel1, R.C. and L. Sautter. 1992. Planktonic foraminiferal faunal and stable isotopic indices of upwelling, In: The Evolution of Upwelling Systems since the Early Miocene. Geological Society Special Publication #64:77-91.
Thunell, R., C. Pride, E. Tappa, and F. Mueller-Karger. In press. Varve formation in the Gulf of California; insights from time series sediment trap sampling and remote sensing. Quaternary Science Reviews.
U.S. GLOBEC. 1991. Theory and modeling in GLOBEC: a first step. Joint Oceanographic Institutions, Inc., Washington, D.C., 9pp.
Woodruff, S. D., R. J. Slutz, R. L. Jenne, and P. M. Steurer. 1987. A comprehensive ocean-atmosphere data set. Bull. Am. Meteor. Soc., 68, 1239-1250.
Wooster, W. S. and A. B. Hollowed (in press) Decadal scale variations in the eastern subarctic Pacific: A. Winter ocean conditions. Can. Spec. Publ. Fish. Aquat. Sci.