chapter 8. field

Chapter 8. FieldMonitoring Programs

Introduction:Field Assessment of MarinePollution Effects:The Agony and the Ecstasy

Donald F. BoeschLouisiana Universities Marine ConsortiumChauvin, Louisiana 70344

Documented and quantified measures of marine pollution ef-fects in the field i.e., in nature! are ultimately the most conclu-sive and, therefore, meaningful measures of pollution. Resolutionof doubts concerning the realism of experimental approaches in thelaboratory and in simulations of various scales--microcosms, meso-cosms and field experiments � relies implicitly on field data. How-ever, the ecstatic appeal of such defi~itive resolution is temperedby agonizing limitations.Not the least of these limitations is that field monitoring is in-herently retrospective, allowing effects of an activity to be ob-served, but not predicted. Rolf Hartung discusses this dichotomybetween prospective, experimental approaches and retrospective,observational approaches: there are distinct limitations to both,and clearly they can be complementary. Hartung endorses a holis-tic approach, integrating a variety of prospective and retrospectivemethods. No one would seriously disagree with that recommenda-tion. But how does one implement it? The impediments to such in-tegration are cultural as well as conceptual. I will leave these is-sues for the broader synthesis to follow. Rather, l will here outlinesome other sources of agony confronting field assessments in orderto set the stage for the papers to follow, which, I believe, castsome rays of optimism on the briars of our passion for field study.

Specifically, field assessments have generally been criticizedbecause of l! generally weak or ineffective design tbis is usually aretrospective evaluation!; �! the difficulties in relating observa-tions to specific causes; �! the difficulty in gauging the broaderimportance of an observed modification to an ecosystem cornpo-nent; and �! the difficulty regulators have in codifying the prod-ucts of even the best field studies in the context of existing regula-tions, and frequently even in subsequent actions.

644 / Field Monitoring Programs

Field assessments have been criticized as basical ly unscientificin that they involve only observation and not hvpothesis te,ting'Many field assessment programs have not been clearly pl~oned taddress explicit questions. Moreover, many field programs werenot based on an appreciation of the dominant sources of naturalspatial and temporal variation, sornetirnes resulting in "controls»that are in vastly different habitats. Once a monitoring prograrhas begun, there is a reluctance to change the design because ofthe need for consistency, which may simply perpetuate initial mis-takes. ALL too frequently, there is an implicit assumption that ifenough biological and chemical parameters are measured at thesame place and time, truth will mysteriously "~ge from statis-tical analyses, liberally applied to bless the proceedings.

Roger Creen provides sound advice regarding the need to iden-tify the purpose, question, hypotheses and underlying model for thedevelopment of sampling and experimental designs. He assertsthat biologically defined objectives should dominate and determinethe statistics.

The variability of nature is legend. In simple terms, variationsin biological populations and rates of processes in time and spaceconstitute statistical noise from which the pollution effects mustbe deciphered. However, some understanding of the causes, or atleast the patterns, of natural variation must be gained in order toproperly design sampling. It is not sufficient to treat natural pat-terns merely as statistical variation. Nonetheless, there is alwaysthe Lingering doubt whether causes other than the pollutant or hu-man disturbance factor in question were responsible for an ob-served biological change or difference. The variability of naturalsystems and the resultant difficulty in relating observations to spe-cific causes have brought into dispute the validity of the baselineand monitoring approach Gray 1976!. Such concern was one rea-son, for example, for the abrupt termination of the Bureau of LandManagement's BLM! extensive baseline studies of continental shelfenvironments prior to oiL and gas development Burroughs L981!The lack of focus and simplistic assumptions of the BLAM baselinestudies were widely criticized by the scientific community, al-though, in fairness, the prematurely standardized design of t"estudies was based on the collective recommendations of the scien-tific community � a scientific "tragedy of the cornrnon".

Notwithstanding its inherent limitations, the environmentalmonitoring approach remains our ultimate "ground truth," and b't-ter understanding of the causes of natural variations in space andover long " periods is urgently required. Therefore Srnayda pr~sents the results of his unusualLy long-term study of the plankt~~community in Narragansett Bay. Plankton communities are n«or'ously highly variable, and, for this reason, have often been «s-counted as a useful subject of applied environmentaL monito~i~gSrnayda's study illustrates that temporal variability of phytop»nk

Introduction: Field Monitoring Programs / 645

ton is indeed the usual, but can be related to env roo environmental factorsoperational over various time scales. Determinatio f ' f lbiological alterations due to human effects depends on such under-standing coupled with suitable experimentation.

The quantitative determination of human effects on a biologicalpopulation is the objective of most field assessments. This wouldbe sufficient if the underlying philosophy were that any resultia any resu Ing

biological change is undesirable and, therefore, should dictate fur-ther control or elimination of the pollutant or activity. This is notpractical, however, particularly given the current policy revisionsthat allow waste disposal as a legitimate use of the ocean. Oneneeds to know to what degree these effects will reduce or interferewith other resources or uses. As marine environmental scientists,we need to be prepared to answer in quantitative terms the "sowhat" question and thus contribute to an assessment of whether theef f ect constitutes "unreasonable degradation." Unfortunately, veryfew marine environmental research programs are designed to an-swer this question. Sponsors consequently become upset with theanalyst when they are told that their discharge is resulting in rnor-talities of larvae, for example, and demand conclusions that thiseffect is insignificant to the adult populations. However, theyshortsightedly may not have specified a design or allowed flexibil-ity adequate to address even the spatial extent of the effect. Theanalyst is often forced to borrow data and assume certain relation-ships that significantly reduce the quantitative rigor of the assess-rnent Boesch l982!.Robert Livingston's studies of coastal ecosystems of the FloridaGulf Coast illustrate the necessary complexity of multifaceted re-search designed to place changes in biological populations in abroader ecosystem context. Alan Mearns and Thomas O' Connorseek to place documented alterations in the common perspective ofspatial scale. Furthermore, Mearns and O' Connor are able to dem-onstrate relationships between the magnitude of waste inputs andthe spatial scale of effect, which allow feedback to waste rnaoage-ment decisions. This addresses, in part, the question of signifi-cance and also the fourth limitation l initially identified � the useof results in decision-making.Although field assessment approaches require general irnprove-rnent and continued development, disparate currency and timing ofobservations and regulations are at least as much the fault of theregulatory institutions and mechanisms as of researchers. Theecologist should not be blamed for the fact that environments de-teriorate in complex and subtle ways as a result of human activi-ties.


REFERENCES

Boesch, D.F. 1982. Ecosystem consequences of alterations of ben-thic community structure and function in the New York Bightregion. ln: G.F. Mayer ed.!, Ecological Stress in the New YorkBight: Science and Management, pp. 543-568. Estuarine Re-search Foundation, Columbia, S.C.

Burr oughs, R.H. 198l. OCS oil and gas: Relationships between re-source management and environmental research. Coast. ZoneManage. 3. 9: 77-88.

Gray, 3. l976. A,re marine base-line surveys worthwhile? NewSci. 70: 219-221.

Some Guidelines for the Design ofBiological Monitoring Programs in theMarine Environment

Roger H. GreenOepartment of ZoologyUniversity of Western OntarioLondon, Canada N6A 517

INTRODUCTION

Statistics in environmental studies are overused and are oftensuperficial. They have been too much used as an after-the-factsalvage operation, and as a window dressing of respectability fordemonstrating the "significance" of studies that were not designedto test any clearly formulated hypotheses in the first place. %hatis needed is ~ariori design of environmental studies validatedby some preliminary sampling of experimentation, so that the re-sults � effectively displayed � can speak for themselves.

With hypothesis-testing statistics in particular, most of the sta-tistics � including I! estimates of the replication needed to obtainspecified power in the test and �! description of multivariatestructure to provide a basis for choice of variables and good de-sign � should be done on preliminary data. ln the final report goodgraphic display should suffice to show what is going on. In descrip-tive studies, on the other hand, statistics can play a major role inthe final report, and multivariate statistical methods are often ap-propriatee.

NONSTATISTICAL CoiNSIDERATIONS

Many authors have emphasized that in the design, of any studythere shouM be a logical flow of purpose question ~ hypoth-eses ~ model ~ sampling or experimental! design ~ statis-t>cal analysis ~ tests of hypotheses ~ interpretation and presen-

Designing Monitoring Programs / 609

What does constitute change of a kind and of a magnitude tocause concern? Who decides? The politician, the manager, theecologist? Certainly it is one of these; it should not be the profes-sional statistician, who if competent will tell the ecologist consult-ing him to begin by clearly stating his objectives, questions and hy-potheses. Then and only then will he proceed to advise on designand analysis. If, however, ecologists claim � as many of us do thesedays--to be both competent ecologist and competent statisticianrolled into one professional person, then they assume a responsibil-ity to decide what the meaningful questions are for a given envi-ronmental problem as well as to design the study properly and toanalyze the results and present them effectively. I think that weoften do not realize what we take on when we agree, and even pas-sionately argue, that we are capable of performing this dual rolefor society. For one thing we cannot blame poor results on a ~aiveecological conception of the problem, or on poor design, insuffi-cient sampling effort, failure to do preliminary sampling, and thelike. I would argue strongly that the role of the statistical!y com-petent ecologist, difficult as it may be, is vital to environmentalresearch and management.In attempting to answer the question "How big a change is eco-logically significant?" one has to come back to the question "Onwhat time scale?" Is the concern one of short term ugliness andthe fouling of a few very visible and lovable animals such as birdsand seals? Or is it long term deterioration of the environment insome sense? When gradual, long term, perhaps irreversible envi-ronrnental impacts are not tied in with any sudden spectacularmess, it is most impossible to generate concern or research money.Perhaps the only solution to this problem is to have a core of inter-nationally respected environmental scientists who receive steadyyear-to-year funding, and who have a mandate to study the prob-lems they believe to be most important in the long run. Such anapproach to research funding is not at all in fashion these days; in-stead we have short term impact studies" which must generateimmediate results and fast decisions, Eutrophication, heavymetals, PCBs, oilspills--each comes and goes in pulses of urgencyand financial support. Never mind that we lack a basic understand-ing of these natural communities and ecosystems. This may repre-sent good politics, but it is certainly not a sound way to conductenvironmental science.For that matter what is environmental deterioration? Individu-al oil spill.s may cause short term messiness and deaths af some an-ima|s usually not of whole populations!, medium term increases ofbiological productivity in the region, and few demonstrable long'term effects. A more important concern � as with nuclear powerstation accidents � may not be the average or typical incident atall> but the very improbable and very worst ones. Most importantof all is the cumulative effect on the world's oceans of a century of


average incidents. If these things are what matters, then what roleis there for short term environmental studies in general, or sophisticated statistical analyses in particular? Statistical estimation inthe tails of distributions is notoriously imprecise and often biased,and here we are estimating for distributions whose parameters areunlikely to remain constant over time.

At the least we must convince the politicians and managersthat significance of impact effects represented by the numeratoreffect in an F-test, for examp}e! must be tested against a, meaningful null hypothesis error term the denominator in the F-test!.would argue that the appropriate error term is usually among-yearvariation in the unimpacted situation. The common practice of us-ing replicate sample observations within the impact and controlarea for this purpose is particularly inappropriate. This is an es-tirnate of sampling error, not of meaningful naturaL variability inthe system. Can we possibly convince those who approve and fundour studies that years of work, enough years to provide adequateamong-years error degrees of freedom for robust tests, are notonly legitimate but necessary? It will not be easy, but we canpoint to a trade-off: more time and money wilL be needed for agiven study but any rejection of the null hypothesis concLudingthat there is an impact effect! will be much more meaningful.Trivially "significant" impact effects, judged against inappropriateerror terms with excessive pseudoreplication, will preoccupy thepoliticians and managers less frequently.

Within-year temporal variation can be of great importance e.g., high discharge of nutrients, pollutants and sediment Loadfrom rivers into coastal areas in the spring! but this importance re-lates to the numerator in the test the impact effect! rather thanto the denominator the null hypothesis error!. Time of year can,and should, be controlled or stratified in the design e.g., springrunoff time only, or each month!. Use of yearly averages is rarelygood practice. There is no way to stratify or control among-yearvariation, and there will not be unless predictable long term cyclesin natural phenomena of interest to us are convincingly demon-strated.

What are Appropriate Criterion aod Predictor Variables?Criterion variables should represent a direct, robust linkage to

the question and the hypotheses. There are three possible situa-tions. In some cases, the criterion variable is defined as part ofthe question, for exampLe, "Are lobsters in the vicinity of spoiLsites picking up PCBs or heavy metals?" The design and executionmay be difficult but it is at least clear what the subject of studyis. Of course the question of temporal and spatial scale remainsFor example, over what length of time the maximum lengt»f Li«of a lobster?! and what area how far do lobsters disperse~! mayaccumulation of heavy metals occur? Against what temporal scale

Designing Monitoring Programs / 6%i

o f variation in "natural" heavy metal concent t'concentrations wiII anyapparent increased levels be judged?

In other situations ecologists are presented thwi apparentlyval ue-f ree questions, f or example, "Does the impacte impact cause achange of any kind? If so, describe it." In such a case thsuc a case t ere are

choices to be made but without any ~ariori rationale for doing so.guch studies are easily trivialized, in that some change of somekind can probably be demonstrated. The difficult and completelynonstatistical job of deciding what really matters has only beenpostponed and passed on to the person who receives the results ofthe study. Unfortunately, when the managers finally do decidewhat is important socially, economically and politically � in relationto a given problem � it usually turns out to be something that wasnot optirnally treated by the sampling design or the choice of cri-terion variables.

These "Describe any change that occurs" problems often lendthemselves to multivariate approaches. It is analogous to the situ-ation in numerical taxonomy where any variation among specimensis of potential interest. A recommended strategy �neath andSokal 1973! is to use a large and representative sample from theuniverse of possible variables, assuming that there will be little in-formation in omitted variables that is not contained in = correlat-ed with = predicted by! included variables. A similar strategy maybe employed by the environmental biologist faced with this kind ofproblem. He may choose as a criterion variable set an assemblageof organisms that is easily and reliably sampled {e,g., the benthiccommunity in freshwater or marine studies!. Then field samplingcan be visualized as a random sample from that universe of vari-ables, using numerical abundance, biomass, percentage cover orwhatever measure seems most appropriate. There is usually highredundancy among taxa in the information they contain about spa-tial and temporal structure of both natural and impacted communi-ties Kaesler et al. 1974; Green 1979!. Despite the passionate be-liefs of devotees of particular measures e.g., biomass as opposedto numerical abundance!, there is also high redundancy among dif-ferent measures of taxonomic variables. For example, in a recentstudy of marine benthic species molluscs, crustaceans, polychaetesand others! in demersal fish diet, it was found that 9496 of the spa-tial and temporal variation described by measures of numericalabundance, wet weight and frequency of occurrence was containedin one principal component. The best = containing the most irdor-mation about all measures and variables! measure, numerical abun-dance, contained 9096 of the total information Macdonald andGreen, in press!.The process of choosing the best variable in the above sense!,then the next best, and so on, can be done by one of several rnulti-variate "dimension reduction" techniques. The most efficient re-duction in the sense af the greatest percentage of the variation


being described by the fewest variables! will be accompl}shed byeigenvalue-eigenvector technique such as principal componentsanalysis. Another approach, which is usually only slightly less efficient and has the advantage of retaining the identities of the original variables species, say! is to use a variable subset selection al-gorithm such as that proposed by Orloci {L973!. Another approachto dimension reduction is to use diversity indices or cumulativespecies abundance curves, but such criterion variables are notclearly and logically related to any generalizable environmentalreality. They are also not robust empirical indicators of any im-portant "environmental health" correlates of biological systems, inthat they are strongly influenced by naturally varying environmental parameters. Finally, they carry no information about speciesidentities. See Green L979! for further discussion of diversity in-dices.

Most often ecologists are handed problems that fall betweenthese extremes of a! criterion variable explicitly defined as partof the question, and b! unspecified criterion variables, which canbe chosen on purely statistical grounds without ~ariori constraints.For example we may be asked, "Does the power plant effluentcause deterioration of environmental quality in that estuary?"These in-between problems are the most difficult ones becausethey imply that the ecologist can and wiLL make nonstatistical judg-ments about what the concern is, judgments that have ethical andpolitical consequences and must precede choice of criterion varia-bles. They must also, of course, precede choice of appropriatespatial and temporal scales for the sampling and experimental de-sign.! The ecologist usually accepts this responsibility, often im-pLicitly rather than explicitly, but in my experience does not al-ways effectively discharge it. This is where the greatest weaknessin environmental studies lies, rather than in the statistical after-math. Multivariate approaches can also be useful here, for exam-ple, in selecting a subset of the variables that are most informativeabout the natural spatial and/or temporal variation of the commun-ity as described above! and then selecting from those the variablesthat previous studies or laboratory toxicological experiments haveshown to be most sensitive to the impact in question. ideally onewants I! socioeconomically meaningful" variables that are �!sensitive to the impact and �! relatively invariant in the unimpact-ed naturaL" situation. However, criteria �! and �! may often becorrelated, in that "sensitive" species may be sensitive to envir'on-mental variation whether natural or impact related.

What of the predictor variables? They must represent the p«s-ence/absence the latter being a control! of the impact, or thetensity of the impact. Also they must represent any importantnon-impact environmental variables that might influence the «L-terion variable s!. Ideally the sampling design should hold constantany such extraneous environmental variables, or if that is not pos-


sible then it should explicitly include them as controlled levels of afixed effect e.g., as depth strata offshore from I drom a power plant de-velopment!, Less desirable is their inclusion as uncontrolled varia-bles measured along with the criterion variables during the sarn-pling; they will not then be Model I fixed effects, and their is noguarantee that the distribution of their values will represent a bal-anced design. If the result is a statistically significant impact ef-fect, but level of impact is correlated with an uncontrolled eviron-rnental variable, then interpretation of the result will be difficult,Ignoring potentially important environmental heterogeneity, andnot including measures of it at all, is of course the worst approach.

SUMMARY

My theme has been that statistical methodology rests on know-ing what the question is, and that environmental biologists worrytoo much about statistical methodology and too little about whatthey are monitoring for. To play the difficult role of statisticallycompetent ecologist as honestly, as ethically and as effectively aspossible, we must first choose our level of concern. Let those whowish to fight all change do so; we cannot afford the loss of energy,of time, or of credibility that such a pointless and unending fightrequires. For example, we cannot man the trenches in the face ofevery threatened species extinction. Neither should we equate allenvironmental change with loss of environments. In thousands ofyears most of North America's ecosystems will inevitably bedifferent from what they are now, just as long term naturalprocesses have made them very different now from what they werein the late Pleistocene. To influence the eventual result of largescale processes of change patterns over large areas changing overlong times! is a reasonable and attainable ambition for ecologists;to stop them, or even to totally control them, is not.

Related to choosing our level of concern is the necessity of be-ing aware that the time and space scales of greatest importance toour economic and political masters do not usually coincide withwhat an ecologist sees as being of greatest importance to the envi-ronment. In most environmental studies the concern, the moneyand the deadlines are roughly on the scale of the life of a govern-ment, the length of a term in office, or the period of a researchgrant. Too many ecologists accept such time scales willingly. Thequestion "What is a simple, easily and cheaply determined index ofdamage to ecosysterns?" is very much like the question "Have youstopped beating your wife?" The proper response in both cases isthat the implied terms of reference are unacceptable. Perhaps thetraining of ecologists requires more emphasis on the large temporaland spatial scale arts and sciences, such as geology, geography, his-tory and even the better quality science fiction, and less emphasis

654 / Field 'Monitoring Programs

on ever more sophisticated statistical methods for fine tuning answers to temporally and spatially coarse-grained questions,must learn to look at environmental problems on a realistic timescale. On the scale of a human generation or, at the other ex-treme, on the scale of geological processes, the importance of human impact on the environment is relatively slight. It is on thetime scale of half-centuries to millenia that the most importantchanges take place. The role of paleoecologists in environmentalstudies certainly should increase. Rhoads and Lutz �959! provide astate-of-the-art review of methods and examples of application tomarine organisms and environments.

I believe that ecologists are responsible, as a matter of profes-sional ethics, for looking beyond the wording of the problem as it isgiven to us, to the real environmental problem on the time scalethat really matters. %'e must argue for the importance of longterm studies, to obtai~ reliable estimates of "baseline" variation onthe among-years time scale and to determine the long term effectsof impacts on complex systems. We should also insist that planningof short term studies include time and money for publication of theresults in the refereed primary literature, so that it can be easilyaccessed by others later on. In the long term, this will providegenuine replication, though admittedly of a qualitative nature, ofgiven types of impact {e.g., of impoundments, of ocean dumping, ofheated effluents! and their lasting biological effects.

There is no ~e kind of study that is the answer. We need themix: the long term sophisticated study at one locatio~, the broadgeographic survey over a short time period, and spatially and tem-porally restricted "impact studies" But the assessment of naturalvariation over years, and of long term effects of impact on naturalecosystems, must be included in the mix. Maturity is the accep-tance that same things require long term solutions. We should hopethat ecologists, politicians and the public will acquire the maturityto support studies that will yield rewards only on a time scale ofdecades or longer. The attitude of the last decade or two, that allcan be solved by impact studies of I or 2 year duration, may-withluck � be an immature stage we will outgrow.

%e must say that any single variable, whether it is a diversityindex or the first principal component or the incidence of lesions orany other measure, cannot represent more than one thing that isgoing on in the system. It is a mathematical impossibility for it torepresent more than that, and loosely coupled marine ecosystemscannot be effectively described by the single most important thingthat is going on. Certainly techniques such as principal compo-nents analysis or nonrnetric multidimensional scaling canshould more often be used on preliminary data to determine theminimum dimensionality for adequate representation of the systemunder study. However, no ecologist should encourage a nan-biolo-gist manager or politician who wants to search for that elusive


goal, the single variable that can be easily and inexpensively deter-mined and that represents human impact on natural ecosystems.

Finally, we must insist that any statement of statistical signifi-cance include a statement in simple English--not statistical jar-gon! of what null hypothesis error term is being used in the con-trast. If it is claimed that the impact is "significant," then "rela-tive to what" must also be stated and justified. Year-to-year nat-ur al variation? Replicate sampling error? Something else? With-out that information the result cannot be evaluated.

REFERENCES

Green, R.H. 1979. Sampling Design and Statistical Methods forEnvironmental Biologists. 3ohn Wiley and Sons, New York.

Kaesler, R.L., j. Cairns, 3r. and 3.S. Crossman. 1974. Redundancyin data from stream surveys. Water Res. 8: 637-602.

Macdonald 3.S. and R.H. Green. In press. Redundancy in environ-mental variables. Can. 3. Fish Aquat. Sci.

Orloci, L. 1973. Ranking characters by a dispersion criterion.Nature 2WQ: 371-373.

Rhoads, D.C. and R.A. Lutz. 1950. Skeletal Growth of AquaticOrganisms. Plenum Press, New York.

Sanders, H.L. l977. Evolutionary ecology and the deep-sea ben-thos. In: The Changing Scenes in Natural Sciences l776-197,pp. 223-293. Spec. Publ. 12, Academy of Natural Sciences,Philadelphia, Pa.

Sneath, P.H.A. and R.R. Sokal. 1973. Numerical Taxonomy: ThePrinciples and Practice of Numerical Classification. W. H.Freeman, San Francisco, Calif.

Holistic Assessments of PollutionEffects: The Interplay of Prospectiveand Retrospective Studies

Rolf HartungSchool of Public HealthThe University of MichiganAnn Arbor, Michigan 48109

Assessments of the impact of pollutants in any system can bedivided into two major categories. The first, prospective or pro-tective assessment, tries to predict the immediate effects of acertain pollutant dose rate on organisms, as well as the eventualeffects on an ecosystem. The basic goal is to prevent adverse ef-fects by establishing acceptable release rates of chemicals subjectto control. Characteristically, protective assessments compensatefor any perceived errors in the estimation of "safe" exposure ratesby the use of safety factors, uncertainty factors and/or conserva-tisrn built into the assessment methodology. The compensation isusually one-sided; that is, poor data result in the use of greater un-cer tainty factors.

The second category, retrospective assessment, tries to predictthe impact of a known ongoing exposure, or the residual effects ofan exposure that has been terminated only recently. This categoryactually embraces a continuum from obvious, major impact to mi-nor or even undetectable impact. In situations of obvious impact,cause-effect relationships are not really in doubt, and thus themajor goal of the assessment may be to determine the qualitativeand quantitative aspects of the impact. At the other end of thecontinuum are situations where it may be known that certain pollu-tants are being released into the environment, but elucidation of aneffect or a cause-effect relationship is difficult. Ideally, findingsfrom retrospective studies can be used to assess the validity ofpredictions made as part of protective assessments.

The types of experimental designs used in these two assessmentapproaches can be very different, and by necessity the types of er-


rors that can be made are also very different. Each assessmentcategory is plagued by different gaps in knowledge and unde,sta�ding. In prospective or preventive assessment we use many laboratory tests which are readily executed using generally accepted statistical methodologies. These tests are understood in terms oftheir chemical and physical principles, their superficial biology,biochemistry, genetics, etc. The one exception is when we try toanalyze the low end of the dose/response curve; in this, we are at-ternpting to evaluate minimal responses in test species after longperiods of exposure, and to utilize this information to predict whatmight happen to other species exposed to even lower doses forlonger periods of time. Since relationships among various speciessensitivities and exposures are often poorly understood, the predic-tive reliability of such analysis can be disappointingly poor.

5imilarly, retrospective studies that attempt to evaluate eco-system responses are usually hampered by a lack of understandingof the functioning of the ecosystem as a whole Most studies thatseek to evaluate impacts on ecosystems are based on an identifica-tion of the species present and an estimate of the number andjormass of the individuals representing each species. The impliedhope is that fluctuations in the arrays of individuals per specieswill reflect something about the structure or functioning of theecosystem as a whole. It is usually perceived that an ecosystem isa level of organization that exhibits more phenomena than the pop-ulations of the species that make up the ecosystem. Thus any anal-ysis based purely on species enumerations and abundances is likelyto miss aspects of ecosystem structure that are as yet only poorlyunderstood. Understanding in this area may be particularly diffi-cult to come by.

While many of us enjoy and feel an affinity to coastal marinelife, we are nevertheless poorly equipped visitors in a very foreignenvironment. %'e are not even equipped to register the inputs thatare meaningful to individual marine organisms! How do we sensethe low frequency pressure waves that fish experience throughtheir lateral line system? How do we distinguish among the"smells" in ocean currents as the salmon may do to find their an-cestral home? How can a terrestrial semi-urban human understandthe grand design and functioning of a marine ecosystem? The' hu-rnan, including the marine biologist, is blind to some of the veryimportant stimuli that shape behavior at the individual level, thespecies level and certainly at the ecosystem level. Thus our inter-pretations regarding marine ecosystems are Likely to represent acurious mixture of truths, half-truths and wishful thinking- One ofthe biggest problems may well be that we often do not knowwhether we are dealing with truths or falsehoods. Our systems forcollecting data and analyzing them may incorporate misconcep-tions that inherently invalidate the data, irrespective of samplingdesign or analytical precision.

Prospective and Retrospective Studies j 659

Let us examine some of the problem areas of predictive assess-ments in greater detail. In almost ail instances this type of assess-ment is based on laboratory studies involving single chemicals orperhaps single process streams tested on single species. The com-plex environmental situation, involving many individual compoundswhere duration of exposure and concentrations change continuous-ly, is essentially never studied for predictive purposes. This is sofor the the simple reason that the complexity of the varying pa-rarneters transcends our capabilities to analyze them properly.Another problem is that we are able to study only a few models,often under highly artificial conditions, as predictors for all thespecies present in the ecosystem. The species most commonlyselected for such studies have not been chosen for their ecologicalimportance. 0 am not sure that we can objectively gauge ecologi-cal importance. Pcological predominance might be a better andmore testable concept.! Species have commonly been selected fortheir adaptability to laboratory conditions, and for unstated subjec-tive preferences on the part of the investigator. Characteristical-ly, many phyla are not represented in the testing scheme.

Predictive assessments for ecosysterns tend to contain more un-certainties than predictive assessments for the protection of hu-man health. The latter assessments characteristically examine theresponses of several species whose physiological and biochemicalrelationships to humans are backed by a vast amount of previouslygathered information and experience. The responses are then ex-trapolated to the most likely response in the human species. %'ehave spent comparatively vast resources in our own self-interestfor the simple task of drawing conclusions from several species to-wards ourselves. The resources required to provide relatively reli-able predictive and protective assessments of ecosystem responsesmay be even greater, unless we are willing to tolerate far largeruncertainties for the protection of ecosysterns.

In the case of retrospective assessments we find ourselves in asituation similar to that of the human epidemiologist. %'hile themethodology establishes associations, it cannot readily establishcause-effect relationships, and often has to rely on controlled lab-oratory experiments to establish those relationships. Dose levelsand time courses of exposure are usually only poorly defined.Chemical analyses are usually few and far between.

The choice of analytical methodology is more often governe ydb

expediency or regulatory demands than by toxicological considera-tions. For example, metal analyses usually do not differentiate be-tween the chemical forms of the metal i.e., whether the metal isionic, chelated, adsorbed onto particles, or existing in mineral formin suspended particles!. These various forms can have quite differ-ent dose effects when absorbed by the exposed organism. The formin which an organic chemical is present in the environment also issignificant. An organic chemical may occur inur in true solution in


water, or as rnicelles, or dissolved in lipids inside other organismsor adsorbed onto particles. Again, the bioavailability of these var,.ous forms differs.

Some of the same problems are found in experimental situationsin laboratory testing for predictive purposes. For instance, metalsare often added in the form of a soluble salt, but depending ontreatment conditions, they end up as an unresolved mix of ionicmetal, hydroxides, carbonates, oxides and organic complexes.highly questionable whether the responses of organisms exposed toa mix of metal forms in nature are comparable with responseswhen the organisms are exposed to an unknown mix of metal formsresulting from experimental expediencies.

Similar problems pertain to organic chemicals. Monitoring datausually measure total concentrations for each xenobiotic. Thistotal is usually a mix of dissolved, adsorbed and bioconcentratedchemicals. During experimental testing, such chemicals are com-monly dispersed into the test chamber in solvent and surfactant ve-hicles. The final solution contains a mixture of dissolved and mi-

cellar xenobiotics. The biological availability of the xenobiotics inthe test chamber is again likely to be very differert from that innature.

One dilemma is that our present analytical techniques cannotdetermine various chemical and physical forms at very low concen-trations. However, the state of the art of analytical chemistry ismuch further advanced than present routine environmental analy-ses would lead one to suspect. In this respect, regulatory require-ments have stifled the introduction of state-of-the-art techniques.

The majority of laboratory tests employ single species. Suchtests fail to explore interactions between species, such as symbio-sis, parasitism and predation. A number of experiments have beenconducted on species interactions, usually simple two-species in-teractions. Such studies have a Long way to go before they can beverified in the field. Since each species normally is expected to in-teract with many other species, multi-species tests might be moreapplicable as an intermediate step of experimental complexity-research technique that seeks to bridge the gap between single-species tests and the environment as a whole employs microcosmsand rnesocosms. Problems still unresolved in these systems includescaling and species selection for tests with specific purposes-

As problems are resolved there should be a convergence be-tween the results of prospective and restrospective assessments.In summary, the validity of such assessments will strongly dependupon a proper understanding of I! the environmental dynamics ofthe pollutants, �! appropriate chemical analyses, �! knowledge ofecosystem structure and function, �! knowledge of the responsesof individuals, populations and ecosystems to specific dose ratesfor specific durations, �! knowledge of interactions of chemicalsacting on biota at various levels of organization. Not all of the~

Prospective and Retrospective Studies / 66l

problem areas seem to be susceptible to early solution, and forsome our understanding is pathetically poor. But l believe it is pos-sible to move towards a more holistic analysis of our present pollu-tion problems in the marine environment and elsewhere, althoughwe seem to be missing opportunities to do so. On the way towardsthat goal we may even find that our early attempts to reconcilepredictive and retrospective assessments of pollution impacts willincrease our understanding of ecosystem functions that currentlyescape our vision.

Variations and Long-Term Changes inNarragansett Bay, A Phytoplankton-Based Coastal Marine Ecosystem:Relevance to Field Monitoring forPollution Assessment

Theodore J. SmaydaGraduate School of OceanographyUniversity of Rhode IslandKingston, Rhode island 02881

INTRODUCTlON

Natural variation is the essential baseline against which any ef-fects of man-induced changes in the ocean are to be detected,measured and predicted. A major problem in monitoring anthropo-genic effects on life in the sea, however, is separation of these ef-fects from those induced by natural climatic and hydrographicchanges in the environment. Fishery biologists discovered this longago. Natural variability, fluctuations and change are probably theleast understood of the major characteristics of marine ecosys-terns, notwithstanding numerous descriptive and process-orientedstudies and, more recently, mechanistic numerical modeling of thestructure and functioning of marine ecosystems. Unfortunately, asLonghurst et al. �972! have reported, natural fluctuations in cer-tain animal stocks have already been attributed incorrectly to theeffects of pollutants. Presumably similar errors have been madewith regard to other components of the food web and environrnent-al properties.The natural variations which characterize plankton cornmuni-ties in terms of species composition, abundance, dynamics and tro-phic structure largely have been ignored by pollution analysts. Thisis so partly because the widely scattered literature not reviewedhere! indicative of such variation is usually not applicabie to pollu-


tion assessment, either because the data tiirie series was too sho~t the normal time series duration is usually one year or less! or relevant variables were not measured. For exainpie, plankton coinmunity structure is often analyzed without atterition to nutrientsMonitoring� therefor e, usually cannot separate anthroppgenic effects from those due to natural climatic and hydrographic variabiiity.

Some insight into long-terin variability is available from an ongoing 23-year process-oriented time series �959-198l!, based onweekly sampling, for the unpolluted waters of lower NarragansettBay, Rhode Island. This data set probably the most extensive con-tinuous record of pJankton dynamics in the coastal waters of theUnited States! illustrates that significant short-term and long-termvariations and changes characterize ph>toplankton abundance, spe-cies com positi on, seasonal cycles and pr i m ar y production and,based on Jess extensive data, nutrients and zooplankton abundance.A synopsis of the pertinent results from this data set is presentedhere.

Narragansett Bay �1 30' 5, 7l 20' 4!, a well-mixed, shallow es-tuary 451 km and 9 rn in mean depth, extends inland approximate-ly 40 km from its connection with Rhode Island and Fetlock IslandSounds. It is a phytopiankton-based ecosystem, Quantitativeweekly samples were collected at 0, 4 and 8 rn depths at a perma-nent station located �l'34'07" 8, 71 23'3l" W! in the lower Baynear its entrance, where the mean depth is 8 m. Weekly measure-ments included temperature, salinity, phytoplarikton species com-position and abundance as cell numbers, chlorophyll, ATP-carbon>phosphate, ammonia, nitrate, silicate, water column extinction co-efficient, incident radiation, and zooplankton species compositio~and abundance as dry weight. Some of these observations are sum-marized here. Also measured, but not discussed here, were zoo-plankton abundance as carbon and nitrogen, ctenophore abundanceand dry weight, phytoplankton primary production, phytoplanktoncarbon growth rate and generation time d '!, turnover of the»-trogen and phosphorus nutrient pools, nitrate reductase and alka-line phosphatase.

Climatologic ChangesSignif icant long-term climatological changes have occurred

over the 23-year time series since 1959, with measurable effectson river runoff and in situ conditions of temperature and irradi-

Plankton Variability and Monitoring / 665

ance, factors which influence plankton dynamics. Figures 1 and 2show the progressive decrease in annual mean wind speed and inci-dent light intensity. Note the significant decline in annual meanincident light based on daily measurement!. The years since 1973can be characterized as the "dark" decade, This long-term patternof decreasing irradiance available for photosynthesis is also evidentin the annual mean in situ irradiance levels calculated from the in-

cident irradiance and water column extinction coefficient Figure3!. Note the considerable year-to-year i.e., short term! variationswithin the iong-term pattern. This is a general characteristic ofthe data set. Temperature shows the opposite trend Figure 0!. Asignificant warming period began in 1969, but note the relativelycool period during 1976-1978. Precipitation and river runoff havealso varied, elevated river flow generally characterizing the yearssince l972.

Clearly, monitoring programs must sort out short-term andlong-term variations in both community dynamics and environ-mental conditions caused by environmental properties i,e., mixing,light, temperature and runoff! which are naturally variable and un-der climatologic influence. These can exhibit considerable interan-nual variations and significant Long-term trends. Moreover, statis-tically significant correlations were found between annual changesin phytoplankton abundance and annual temperature and light con-ditions. That is, the phytoplankton are indeed responsive to clima-tologic variability in these two key growth factors.

NutrientsNutrient data, although less extensive, also exhibited signifi-

cant variations and trends. Annual mean phosphate levels exhibit-ed significant year-to-year changes with an inter esting pattern Figure 5!. A "phosphate-rich" year was followed usually by a"phosphate-poor" year. For example, the mean annual phosphateconcentration of about 15 mg-at m > in l971 decreased to 7 mg-atm in 1973, then increased the following year. A similar patternis evi ent or ni ra e,'d f 't te although the data suggest that nitrate levels

Fi ure 6!. Whilehave increased significantly since the early 1960s {Figure . i evariations in nutrient levels are also under g'un er direct climatologicregulation through precipitation and runoff, nutrient levels obvi-ously are influenced significantly by the utilization and recyclingmechanisms characteristic of marine ecosyslevels and oscillations are u~der the dual reg ulation of climate andtrophodynamics. This characteristic po pases a rticularly great

problem to monitoring programs which at p' h attem t to sort out the eu-

trophic links and signals from the naturally variable ones.

PhytoplanktaaFi ure 7 shows the significant year-to-year variar variations about

threefold! in mean annual diatom abundance in g<gure s ows 'n lower Narra ansett

11. 5

11. 0

10.5

o 10.0

9.5

F% 9. 0

8.5

8.019741969

roe

19641959 1979

380

360

1%9 19791969 1974

Figure 2. Annual mean incident Light intensity measured at New-port, R.l. Dashed line represents 21-year mean.

Figure l. Annual mean wind speeds registered at Theodore FrancisGreen Airport, Providence, R.l. Dashed line represents mean windspeed for 1959- 980.

70

66

62

5

58

H 54

4fso

19e41959 l969

YEAR

l974 1979

Figure 3. Annual mean in situ irradiance as ly day ' in lowerNarragansett Bay calculated from

I = lo l-e !kz

12 ~ 0

11. 0

10.5

10.019741959 1964 1969 l979

YERR

Figure 0. Annual mean sea-surface temperature in NarragansettBay- Dashed line is 2l-year mean.

where lo is incident irradiance, k the extinction coefficient ra !,z the depth 8 rn! of the mixed layer which extends to the bottom.Dashed line is 21-year mean.

15

]$

l3

iz

LO

f 974 19793 9S4

Figure 5. Annual mean phosphate concentrations in the v ater col-umn of lower Narragansett Bay. Dashed line is mean for all years.

32

24K

20L

tDx I6

12

8

l 9791 959 1%4 2%9

YEAR

l974

Figure 6. Annual mean nitrate concentrations in the water columnof lower Narragansett Bay. Dashed line is mean for all years.


pay. A clear long-term trend is not evident, but, as pointed out,these fluctuations and annual mean temperature levels are relatedstatistically.

The interannual variations and annual successions in the occur-rence and abundance of phytoplankton species have been subjectedto Principal Components Analysis and Stepwise Discriminant Anal-ysis SDA!. Multivariate analyses were also applied to the light,temperature and nutrient data to establish trends in these environ-mental variables which might relate to successional patterns, tern-poral variability and change in phytoplankton. These multivariatestatistical analyses of the occurrence patterns of the phytoplank-ton species provide a numerical approach towards the identifica-tion of species groupings and their relationships to each other andto environmental conditions, and facilitate evaluation of the extentto which the taxonomic structure of the phytoplankton assemblagesmay have changed in Narragansett Bay during the 23-year timeseries. Approximately lo00 sample dates involving V9 significantphytoplankton species using log-transformed cell counts! havebeen anaLyzed in 37 different treatments, two of which are shown.

Figure 8 illustrates the mean SDA values for each year between1959 and L980. The position of these mean values is a rel.ativemeasure of the similarity of species occurrence patterns withineach yearly group. The proximity of the points representing l959-l965 indicates a high degree of repetition in species compositionand succession during these years. Note that 1969 is distinctly sep-arate fr orn this cluster, and marks the beginning of a different phy-toplankton pattern in the early 1970s, which continued to evolve in1980. Notwithstanding this trend, a sequence of 5-year cycles pri-or to l975 is evident. The years 1959-l963, l960-L968 and l970-1970 represent interannual periods of greater phytoplankton sirni-larity, each period characterized by a similar 5-year pattern ofvar iabili ty.A similar analysis based on the five most abundant species givesa very different pattern, however Figure 9!. A distinct separationbetween the 1960s and l970s is not evident, nor the sequence ofthe three 5-year cycles prior to 1975. This indicates that theinterannual differences in yearly phytoplankton successional se-quence patterns in Narragansett Bay are primarily due to yearlydifferences in the frequency and magnitude of occurrence of theless abundant species of the 09 species in Figure 8. This result isconsistent with other field and experimental observations that themajor phytoplankton species simply become more abundant uponexposure to eutrophication. Hence, the search for marine ecosys-tern phytoplankton equivalents of the "miner's canary" as indicatorsof incipient or more advanced stress, prior to permanent ecosystemchange, has not been a very productive approach. However, thisfinding that most of the annual variability in phytoplankton assem-blage composition and response in Narragansett Bay was due to

40

35KC3

30Cl

25

20

1974 1'9791964 1969

YEAR

p i SCrimih4ol Aft'/yf i %49 happ

'l959- i98Q

~62IIII

60 459

64

T 0- 4.0 2.0-3.0 -2 0

Figure 8. Stepwise discriminant analysis of annual mean abundanceof 49 species of phytoplankton in lower Narragansett Bay from1959 through 1980. Ordinate and abscissa represent canonical vari-ables CV1 and CV2, respectively.

Figure 7. Annual mean diatom abundance as iG cells m in lowerNarragansett Bay. Dashed line is 21-year mean.

Flankton Variability and Monitoring / 671

variations of the less abundant species requires assess t f hires assessment of theextent to which it reflects episodic species-introductions or vari-able responses of the resident community. Such assessment would

preliminary to more detailed evaluation of the applicability ofthe "indicator species" and "early warning community" concepts tomarine pollution assessment. Should this variability primarily rep-resent the response of the resident community, then in situ moni-toring programs should take special care to assess the less abun-dant species, a group usually ignored.

Interannual variability also characterized the total zooplanktonassemblage which exhibited a 9 to 10 year cycle and was due pri-mari}y to variations in the copepod component The rneroplankton-ic component remained quite similar from year to year, however.

Thus, in addition to long-term changes in plankton and selectedenvironmental variables characteristic of the Narragansett Baytime series, both the phytoplankton and zooplankton communitiesexhibited cycles in their structure and dynamics. Phytoplanktonassemblages appear to cycle in 5-year units and zooplankton in LO-year units. That is, the annually varying phytoplankton assem-blages and their dynamics are generally more similar to each otherduring a given 5-year cycle than to those in other 5-year cycles.The associated characteristic of this cyclical behavior is a precipi-tous annual shif t into a new cycle, a phenomenon most notably ex-hibited by the 1969 year-class Figure 3!. Such cyclical variabilityduring long-term environmental change complicates significantlythe evaluation of short-term = ephemeral! and long-term environ-mental ef fects of marine pollution.

Environrneotal CorrelationsTheoretically, measurement of the relative importance of envi-

ronmental parameters in regulating phytoplankton succession andcommunity change Can be evaluated from the correlations occur-ring between a successional rate parameter and the rates of changeof various environmental factors Smayda 1980!. We have appliedLewis' �978! summed difference index g! of succession rate, de-fined as

as = Ei d bi t!/B t!! Idtwhich is estimated over the short time interval as

os= Z,- ! b; t,!/B«,]- , ! ! !]t -t

where b- t! is the abundance of the ith species at time t, and B t! isthe size of the community at time t. This index is insensitive tochanges in community size unless there are accompanying changesin relative abundance of the species. A value of zero obtains for a


shrinking or expanding community, if all species are changing atidentical rate. The entire narragansett Bav phytoplanktonseries of weekly observations from 1959 through 198' was a�alyzed.

Correlation of the 22-Year time series succession rate indiceswith environmental variables revealed the follow ing. The succession rate index was positively and significantly correlated withlight and temperature in the 1960s, but not with nitrate and silicate concentrations. However, in the 1970s ~s v as positively andsignificantly correlated with nitrate and silicate concentrations,but not with light and temperature. o was also positively and sig-nificantly correlated with phosphate concentrations during the en-tire time series. These correlations reveal that the apparent rela-tive importance of an environmental variable, or combination ofregulatory variables, in the regulation of phytoplankton can changeover time.

Of the other variables, o was positively correlated only withthe rate of change in zooplankton biomass per unit time: I ~ dryweight/ ~t I. It was not correlated with the abundance of totalzooplankton, total copepods, Acartia hudsonica and Acartia tonsa h d i p pd!, d ph~~Ml~dwith their rates of change with time. Neither was the successionrate index correlated with phytoplankton production rates ex-pressed in various ways, unlike that found for a tropical lake Lewis1978!.

Multivariate statistical analyses clearly showed that the shift inyearly patterns of light and temperature from the 1960s to thel970s Figures 2-0} was primarily manifested during the wintermonths, and that winter temperatures were rTiore repetitive duringthe 1960s than during the 1910s. Independently, the successionrate index, 4, correlated positively and significantly with light andtemperature during the 1960s, but not during the l.970s. Multivari-ate statistical analyses also established that variations in nitrateconcentrations were particularly pronounced, and exhibited dis-tinctly different behavior between the 1960s and 1970s. Silicate,temperature, phosphate and light, in that order, followed in thedegree to which they exhibited interannual variability. Interest-ingly, % was significantly and positively correlated with riitra«and silicate levels during the 1970s but not during the l960s.fact that os always correlated with phosphate concentrations, a nu-trient rarely limiting to phytoplankton growth in Narragansett >ay~based on experimental evidence Smayda l974; Hitchcock andSmayda 1977!, is notable and poses a problem: it suggests that +correlates with nonlirniting variables. If this is correct, then theappropriate interpretation of the results is the converse, namely~that nutrients, and not temperature or light, were limiting mo«often! in the 1960s, and that in the l970s nutrients were no long«limiting and, thus, light and temperature were. Monitoring pro-


grams without benefit of experimental data are no less vulnerableto such potential misidentification of cause-and-effect combina-tions ~

Nonetheless, the significant finding is that long-term variationand changes in phytoplankton assemblages are evident upon statis-ticai analysis, and that two independent analyses suggest that thesewere accompanied by long-term changes in light, temperature andnutrients. Although statistical correlations were found, these arepresumed to reflect parallel trends rather than cause-and-effectfeatures. The data have also been subjected to rigorous statisticalanaLysis, correlating between the variables when expressed interms of their annual, quarterly and monthly means, and the week-ly levels. These analyses confirmed that the relative importanceof a given variable in regulating plankton dynamics varies consider-ably between seasons and years. Thus, plankton dynamics areunder multivariate environmental regulation. While the growthfactors remain the same, their importance and factor-interactioneffects in regulating in situ growth are variable. This compromisesin situ monitoring programs, which usually cannot evaluate factorinteractions and which usually focus on a limited number of vari-ables because of the incorrect assumption that they exert an un-varying and similar regulatory role.

Phytoplankton-Zooplankton-Phosphate InteractionsIt has been emphasized that lower Narragansett Bay appears to

be an unstressed ernbayment characterized by significant naturalvariability in its plankton dynamics and environmental properties.W'e have some data relevant to this and, especially, to the problemsof eutrophication, assimilative capacity and in situ biological rnoni-toring. Between l973 and l979, the annual mean phytoplankton

d zooplankton biomass levels progressively increased. he rela-tionship between these annual mean standing stocks between 19 73and 198L is especially provocative Figure 10!. There is a strong,highly signi icant rf' t r = 0.62! direct relationship between annualmean zooplankton predator! and phytoplankton prey! a un ance,h h f flows the classic yield-dose response. Moreover, it sug-wic o ows ec

s that there isan ~~nial mean ca~ryi~g capaciton in lower Narragansett Bay of about L.O g dry weight m ~ anthat annual mean phytoplankton standing stocks exceeding about p

g C m ~ represent surplus production not accompanied byincreased zooplankton bio

These variations and build-up in annual mean p ytop anted with temperature, light inten-

ass as carbon were not ce or silicate, but correlated positive y wiannual phosphate level Figure ll!. The mar e y ain Figure ll represents 1979; exclu g podin this int yie s a ig

significant correla iation coefficient of r = 0.67; it woul! iven this rovocative and un-t"e f978 datum were also dropped.! Given is p

piSCriminant Anolysis 5 Spy,

CV I + CV2 = 69 /~ ygrigS9-Idea

0.5

-0.5

-0,5 0,5 i.0

I.O

o.e

Q

o,s

v

A TP-CAR80k' g m-z

Figure 10. Relationship between annual mean standing stocks ozooplankton dry weight and phytoplankton carbon in lower Narra-gansett Bay for the years 1973-1981. Dashed kine, drawn by eye~goes to origin.

Figure 9, Steps ise discrirninant analysis of annual mean abundanceof five most dominant phytoplankton species in lower Narrangan-sett Bay. See legend, Figure S.

4 3 2 7 8 3 10 11 12 13 14 ]SANNUAL NEHN I'HG5PVRTE HG-R7/N f

Figure 11. Relationship between annual mean phytoplankton car-bon and annual mean phosphate concentrations in Lower Narragan-sett Bay for 1973-1981. Correlation coefficient r = 0.67 highlysignif icant! excluding aberrant l979 datum.

676 / Field Monitoring P r ogram s

expected correlation indicating that annual variatiors in phosphateinput or recycling regulated annual mean phytoplankton aridby, zooplankton biomass levels, the potential sources of phosphoruswere evaluated. The annual mean phosphate levels were inverselyrelated statistically significant! to the volume of inflow fromBlackstone River, which accounts for about ~J% of he fluvial dis-charge inta Narragansett Bay. The other nutrients ~ere not correlated with river flow.! Despite this correlation, there was nocorrelation with the flux of phosphate in this flow, nor was watercolumn phosphate correlated with the benthic flux of phosphate,either solely or when combined ~ith the river flux. This suggestedthat since the annual mean phosphate levels represent only the re-sidual remaining after uptake, the rates of phosphorus recyclingand amounts found in the plankton needed to be considered.

These analyses revealed that the annual mean phosphate con-centrations present in the phytoplankton + zooplankton + watercolumn have progressively increased since l973. That is, a phos-phorus build-up has been occurring in lower narragansett Ray with-out any evidence for an increase in dissolved phosphate levels Fig-ure 5!. A positive statistically significant correlation r =- 0.65! oc-curs between the annual river flux of phosphorus, from v hich theannual mean dissolved levels were subtracted i.e., it representsunused phosphorus!, and phytoplankton biomass converted to phos-phorus on the basis of the ratio between phosphorus and chloro-phyll. Moreover, a highly positive and statistically significant cor-relation r = 0.68.! occurs between the annual mean phosphorus fluxfrom rivers and benthic processes and the annual mean phytoplank-ton + zooplankton biomass converted to phosphorus.

These correlations implicate riverine influxes of phosphorus asa factor contributing to the apparent increased buildup of planktonbiomass in Narragansett Bay over the last decade. However, asemphasized earlier, climatological trends during this period haveresulted in lower in situ light levels Figure 3! and higher tempera-ture Figure 0!. Statistically, temperature and diatom abundanceare positively correlated, while in situ irradiance and total phyt~plankton abundance are negatively correlated. lt must also be ern-phasized that there is no nutrient evidence for an increasing eutro-phication of lower Narragansett Bay, yet increased biomass requires increased nutrient availability, and riverine inputs are irnpli-cated. And so we are left with a dilemma. Js there a progressiv~eutrophication occurring in lower Narragansett Bay which is n«detectable through nutrient assessment? Alternatively, have theobserved long-term climatologic trends, with which phytoplanktonabundance is correlated, simply altered the rates of nutrient recy-cling and other processes significant to p}ankton dynamics? lf t"elatter, then the observed annual variations and long-term increasein plankton abundance might simply reflect the natural range ofeutrophication f ertility} characteristic of lower Narragansett Bay


CONCLUSIONS

Variability is an intrinsic characteristic of climate, hydrographyand plankton communities. !ts documentation here for Narragan-sett Bay illustrates, therefore, a general property of marine eco-systems and environments rather than a unique situation. The re-sults revealed correlations between climatological variations andplankton processes, and they revealed that the correlative factorswere operational on a time scale of a year, 5 years and even 10years. Moreover, the relative importance of growth factors variedwith time. The results even suggested that there may be an on-going eutrophication process in lower Narragansett Bay hiddenwithin the patterns of normal variability, modified by climatologictrends, and without any obvious signals of the kind sought in rnoni-toring programs.

It is well known among marine ecologists that plankton and en-vironrnental surveys are powerless to identify cause and effect; thelatter can be identified only by experimentation. Surveys are,after all, mere descriptions of a phenomenon; and biological rnoru-toring is a survey, nothing more. Monitoring as usually appliedwithin a pollution assessment context serves only to give warningor documentation of apparent change, and can indicate neither thecauses of such change, nor the relative importance of the variousfactors and trophic processes contributing to the change. It maybe argued that the role of biological monitoring is truly descrip-tive � to document change. Upon evidence of change, however, amanager must decide whether the change is the result of a pollu-tant. This requires data interpretation which carries with it the

dneed to sort out changes due to natural variability. But, as pointeout, biological monitoring never deals with the problem of naturalvariability; the time series surveys are too short. Thus, biologicalmonitoring is inadequate in two key methodological procedures:the survey itself arid quantitative procedures.

A cess-oriented in situ study involving rate measurements isneeded to help establish causes and effects of oscillations in eenvi-

ronmental variables and associated plankton dyd namics. It follows,

t en, t a in si uh h t 'n situ monitoring as traditionally carried out cannotmeet i s asic j't b objectives namely, to discover whether c angdue to anthropogenic or natural causes, whether it wi on ytemporary transient dysfunction or disruption, pand to redict ensu-

ing changes. n assessment ro-What revised scientific form should pollution ass pgrams take to provide answers to such basic q uestions? First, one

back to first principles and incorporate the approach ofmust gp c o surve in techniques must becontemporary marine ecology. The surveying ec qquantitative and statistically sound, and th yhe surve must be carriedout on the relevant properties over a suffi ' p'cient time eriod. Sec-

ond, key processes must be measured by means of ex rirnents us-pe


ing suitable experimental design and analytical equipment. Thirdmeso-microcosm experiments using entrained communities whichcan be manipulated are needed. ln this procedure, the influence ofselected perturbants on selected communities, or community corn-ponents, including their influence on key rates processes! androutes transformations!, is tracked under controlled conditionsThis "hybrid" between the classical monitoring approach, in situprocess-oriented studies, and meso-microcosm experiments, 1 aiTiconvinced, is not merely suitable for scientific evaluation of problems of pollution, it is essential. All three activities must cpmplement each other, rather than be carried out piecemeal. An approach short of this represents a flawed conception of how marinesystems work, and ignores the functional holism that exists be-tween climate, habitat and biotic structure.

hCKNOVLEDGhAENTS

Deborah French assisted with the data processing, developedcomputer programs and, because of my absence at sea, deliveredmy paper at the conference on Meaningful Measures of Marine Pol-lution Effects in April, 1982, at Pensacola Beach, Florida. Dr.Deneb Karentz, and Ms. Ellen Deason carried out multivariate sta-tistica1 analysis on the phytoplarikton and zooplankton, respective-ly. Ms. Blanche Coyne typed the manuscript. This study was su~ported by National Science Foundation Grants 68-1500, 71-0056,OCE-76-22563, and Department of Commerce {NOAA! Grant No.NASOR A-D00064.

age

Hitchcock, G.L. and T3. Srnayda. 1977. Bioassay of lower Narra-gansett Bay waters during the 1972-1973 winter-spring diatombloom using the diatom Skeletonema costatum. Limnol. Ocean-ogr. 22: 132-139.

Lewis, %.M., 3r. 1978. Analysis of succession in a tropical phytoplankton and a new measure of succession rate. Amer. Nat112: 40l-414.

Longhurst, A., M. Colebrook, 3. Gulland, R. Lebrasseur, C. Lo«n-zen and P. Smith. 1972. The instability of ocean populations.New Sci. Vol. 54: 1-4.

Smayda, TZ. 1974. Bioassay of the growth potential of the sur-face water of lower Narragansett Bay over an annual cycle us�g 4 1; Th 1;; p A �- U l3-li.

Limnol. Oceanogr. l9: 889-901.


Smayda, T.3. l980. Species succession. In: I. Morris ed.!, ThePhysiological Ecology of Phytoplankton, pp. 093-570. BlackwellScientific Publ., Oxford, England.

chapter 8. field

Documents