Multivariate analysis of experimental marine ecosystems
Helgol~nder wiss. Meeresunters.
Multivariate analysis of experimental marine ecosystems
C. A. O V I A T T 0
. T . PERF
S. W . N I x o N 0
0 Graduate School of Oceanography, University of Rhode Island; Kingston, Rhode Island, USA , and Environmental Research Laboratory, U.S. Environmental Protection Agency; Narragansett , Rhode Island , USA
Twelve replicate 150-1 laboratory microcosms were developed using whole water samples and natural benthic communities from Narragansett Bay, Rhode Island (USA). The microcosms were scaled to the bay in terms of salinity, temperature, light input, volume pelagic community to area of benthic community, density of macrofauna, turbulent mixing, and flushing time. The microcosms were self maintaining during a six month study period with over 35 species of phytoplankton and 30 species of macro-invertebrates. Some 25 species of meroplankton entered tahe microcosms and successfully colonized the benthic communities. Zooplankton were present in all life stages. Atter an initial study of replication among the 12 ranks, the microcosms were perturbed with 3 levels of treated urban sewage for a three month period. Three microcosms were maintained at each level with 3 tanks remaining as controls. At the end of 3 months, the sewage input was terminated and the response of the systems was followed for an additional 2 months. Both time series data and multivariate statistical analysis of over 10 different parameters indicated that the replication of the microcosms was adequate to show the effects o f experimental treatments. Control microcosms were generally Within the range of variation expected in Narragansett Bay. Moreover, ttle results suggested that the microcosms responded to the gradient of sewage input in a manner similar to that of the bay. During the two month period aider the sewage was discontinued, all of the microcosms became increasingly similar, though the tanks that had been subjected to higher levels of sewage remained distinct. It was apparent throughout the study that comparisons of microcosms and natural systems must account for the large variation characteristic of each. For this, and other reasons, multivariate statistical techniques appear to provide a powerful tool for experimental ecosystem analysis.
I N T R O D U C T I O N
E c o s y s t e m
a n a l y s i s
- h o l i s m
a n d
r e d u c t i o n i s m
An interesting .dilemma faces anyone t r y i n g to do ecosystem level research,
whether it is in the marine enviro.nment or elsewhere. O n one hand, there is a
general acceptance t h a t the holistic principle applies to living systems
, and t h a t n a t u r a l communities are more, or at least something other than the
sum of their parts
(Odum, 1971; Patten, 1971; Mann, 1972, 1975)
. If so, the d a t a
gathered by traditional reductionist methods may not be very helpful in making
predictive statements about the behavior of complex marine ecosystem or their
probable response to perturbations of various kinds. On the other hand, natural ecosystems
are usually large, unreplicated, without environmental control, and therefore
protected to some degree from tampering by well meaning researchers. While it has
occasionally been possible to overcome these problems in doing experimental ecosystem
work on land
(e. g. Likens et at., 1970; Odum, 1970)
, it has seldom been practical with
natural marine ecosystems. As a result, our knowledge of marine ecology has
developed in bits and pieces from many individual studies of the behavior and dynamics of
phytoplankton, zooplankton, etc. We can only hope that the best data base that has
evolved under this reductionist approach can be synthesized in some useful way
through the mechanistic modeling projects now underway (
Kremer & Nixon, 1975; and numerous others). T h e m i c r o c o s m m e t h o d
In seeking a way to carry out practical yet holistic ecosystem research, a number
of workers have used small-scale living models or microcosms of larger natural
systems that can be maintained and manipulated in the laboratory. The rationale and
potential of the microcosm approach in general have been described by
, and with specific reference to coastal marine waters by
. A review of the use of multi-species cultures and microcosms, especially in the
marine environment, has been prepared by Levandowsky (in press). At the present
time, at least five major marine microcosm studies are underway, including projects at
Kiel Bay, Germany (yon Bodungen et al., 1976; Smetacek et al., 1976), at Loch Ewe,
Scotland (Davies et al., 1973), at Narragansett Bay, Rhode Island (Knauss et al.,
1976), at Sanich Inlet, British Columbia (Takahashi et al., 1975), and at Kaneoke Bay,
Hawaii (S. V. Smith, personal communication). In spite of the impressive efforts
involved in these, and other microcosm studies, the methodology of microcosms (or, in
some cases, mesocosms) is still evolving. For example, it is far from clear how
microcosms may best be scaled to the "real world" in terms of energy inputs for
biological production and physical mixing or what the effects of larger animal
exclusion may be. There is also the possibility of serious artifacts from the lack of spatial
heterogeneity and small size in microcosms, as well as from the "wall effects" of high
surface to volume ratios characteristic of enclosed communities. As with numerical
simulation models, there appears to be no generally agreed upon criteria for evaluating
the credibility of living models or for comparing the behavior of microcosms with the
"real" systems they represent. In a sense, microcosm research is still an iterative
process in which methodological and conceptual problems of experimental design are
as much a part of the study as data collection and analysis.
With these limitations and challanges in mind, however, the microcosm approach
may yet prove to be a powerful tool for marine ecosystem research and management.
This paper reports results from the first in a series of perturbation and recovery
experiments using microcosms designed as analogs of Narragansett Bay, Rhode Island.
The experiments were designed to examine the r e p l i c a b i l i t y of marine microcosms, to compare the behavior of the microcosms w i t h t h a t of a n a t u r a l marine bay, and to evaluate the use of the microcosms as a tool for studying the larger system. M E T H O D S
D e s i g n
6.5 C-24.2 C (maintained in a water bath to + 1.5 C Narragansett Bay)
9 to 16 hours (adjusted monthly to mat& the natural system)
The microcosms were stirred with 0.14 m=, 1.8 cm plastic mesh
paddles at 32 rpm. The paddles were rotated for 30 see., then
stopped for 6 sec., then reversed for 30 sec. in a continuous
cycle. The dissolving rate of hard crystalline sugar bails, our
measure of relative turbulence, was -0.150 g/rain (o = 1.005,
N = 5) in the microcosms compared with -0.174 g/rain (o =
0.004, N = 6) in the bay
167 cm~. Maintained in opaque box cores immersed in each
microcosm. A vacuum pump system flowed microcosm water
over the benthic community at 0.73 l/rain, a flow fast enough
to prevent the Og concentration over the benthos from dropping
by more than 0.1 mg/1
* At surface of the tanks. Calculations of light intensity at mid-depth in the lower West
Passage of Narragansett Bay during this period ranged from 14.2-'21.1 ly/day (z = 5 m,
K = 0.55 m-1, surface light from Epply Laboratories, Newport, RI). A mixture of high
intensity fluorescent and incandescent bulbs was used to give a spectral composition similar
to that found at mid-depth in the bay.
Analysi.s of marine ecosystems
generally high salinity ( > 20 ~ and little vertical stratification. The average depth
is about 10 m. While there is little organic input to the bay from marshes or
macrophytes, there is a clear eutrophication gradient from north to south that results from
large inputs of primary and secondary sewage in the upper bay (Fig. 1). In general,
however, the water quality at the location of the laboratory in the lower bay is
E x p e r i m e n t a l
d e s i g n
The 12 microcosms were filled in mid March, 1975, using hand bucketed water
samples (to prevent damage to plankton) collected over a 6 hour flood tide from the
surface of the lower West Passage of the bay (Fig. 1). For the next 7 days, the
microcosms were throughly intermixed to insure a uniform distribution among the
tanks. A heterotrophic benthic community was included in each microcosm in a flow
Fig. 1: Location of the microcosm laboratory on the lower West Passage of Narragansett Bay,
Rhode Island (USA). The lettered stations were used to collect field data on the spatial
variability of plankton and nutrients in the bay over an annual cycle in 1971--1972
through (vacuum pump) dark box immersed in the tank
(Perez et al., 1976)
community used in the work discussed here was collected near the middle of
Narragansett Bay in an area with unpolluted silt-clay sediments dominated numerically by
the bivalve Nucula annulata. The sediments were screened (0.75 mm) to separate the
macrofauna, which were then placed in equal numbers in each benthic box. Material
scrubbed from the sides of the pIastic tanks and collected on the bottom during
cleaning was added to the benthos three times a week.
Aflcer the intermixing period, all 12 of the microcosms were monitored for 15
days to determine the replication of the systems. The parameters
methods used during this period and throughout the study are summarized in Table 2.
At the end of this period, 9 of the microcosms were perturbed with 3 different levels
of treated urban sewage collected from the largest treatment plant on the Providence
River (Fig. 1). S e w a g e i n p u t s
Since one of the objectives of this study was to see if the behavior of the micro
cosms approximated that of the larger natural system, it seemed appropriate to find
Successive washings through 80/~m screen with
mechanical disruption of detrital lumps using a
needle and 12 x dissecting scope. Animals were
pipetted onto a grid, counted, and major groups
Analyzer Model 1100, Carlo Erba
nitrite & nitrate
Plankton: chlorophyll a ATP
Particle size frequency (0-50 ~)
( > 0.75 mm)
sediment C H N
only at end
out if the microcosms which resembled the "clean" lower West Passage of the bay
would develop the characteristics of the more eutrophic upper regions of Narragansett
Bay when they were exposed to a sewage gradient similar to that found in the n a t u r a l system. I n i t i a l l y large quantities (4 1, 10 1 and 20 1) of sewage were added to tripliTable 3 Composition of the treated urban sewage added to the Narragansett Bay microcosms
Farrington and Quinn (1973)
, mean (range); samples collected from the same treatment
plant as the sewage used in the experiments
** Mean of triplicate samples analyzed by Earl Davey, U. S. Environmental Protection
Agency, Narragansett, R. I
~/ ...... t
care microcosms to simulate an ammonia gradient f r o m a sewage outfall of 10, 25 and
50/~g at N/1. The Providence River which has ammonia concentrations as high as 31 #g
at N/1 (~ = 14, N = 52) and which receives 0.3 ~/0 sewage by volume per day provided
the basis for choosing this gradient and a logarithmic series (0, 0.01, 0.1 and 1.0 ~ of
sewage by volume for new input water added three times a week. Even though the
sewage addition caused only a small decrease in salinity (4.0 ~ fresh water was
added proportionally to the lower sewage dose tanks to make the change constant for
all treatments. The sewage for the entire experiment was collected at one time,
analyzed (Table 3), separated into aliquots of the volume necessary to add to the bay
water that was added to the tanks on three days each week (Table 1) and frozen.
Later~analysis of the frozen sewage showed no appreciable change in its composition over time. The sewage input was maintained for 102 days (4 April-15 July) before it was terminated and the recovery of the systems ~ollowed for the next 69 d a y s (16 J u l y - 2 3 Sept.).
RESULTS A N D D I S C U S S I O N R e p l i c a t i o n o f t h e m i c r o c o s m s
For 15 days a~er the initial intermixing, the 12 tanks remained similar with
respect to all of the parameters measured (Figs 2, 3; Table 4). Once the systems
were perturbed with sewage, the tanks began to diverge and show the effect of the
experimenta! treatment. However, replication within each treatment remained strong
for another 15-20 days. A t that point, one replicate in each of the treatments and in
the control set began to diverge markedly with respect to ammonia (Fig. 3). The same
tanks later developed extremely high nitrate and nitrite levels. It is difficult to account
for this anomalous behavior with respect to nitrogen. It apparently had nothing to do
Replication of the 12 marine microcosms just prior to the addition of urban sewage. Samples were collected 14 days atter intermixing of the microcosm tanks
with the sewage input, since one of the control tanks had the highest nitrogen level.
Measurements of the nitrogen fluxes from the benthic boxes were not significantly
higher in these tanks. Moreover, the total biomass of infauna present in the sediment
(,-" 250 mg dry weight) could not have provided enough nitrogen to produce the
measured increases even if it had completely decomposed9 The same is true of the
amount of nitrogen in the pore waters of the sediments. The C / N ratio and nitrogen
content of the sediments in the anomalous tanks were not significantly different from
the other tanks at the end of the experiment9 At this point, the source of the additional
nitrogen remains unknown 9 Replication within treatments of the other parameters
continued to be adequate to separate the effects of the various sewage levels9
M u l t i v a r i a t e
a n a l y s i s o f m i c r o c o s m
a n d r e s p o n s e
r e p l i c a t i o n
The long, 170 day time series plots of individual parameters in the microcosms are
useful in documenting the behavior of one part of the system and in analyzing the
mechanisms that may operate in regulating the structure and function of the
community. They also give a simple visual impression of the variation of each parameter
within each treatment over time. However, one of the main reasons for using
microcosms is to study the response of whole ecosystems rather than their parts. Even with
the best reasonably attainable replication, the noise of measurement and sampling
errors and the day to day variation of the systems themselves make it almost
impossible to use the time series data alone to separate one system or microcosm from
88.4% VARIATION EXPLAINED
MEDfUM -3 u)
4I AXIS I 5I
another or to separate one system in its present state from itself at some other state.
Analyses are needed which incorporate a number of parameters simultaneously in
arriving at an objective statistical description of ecosystem state or conditions. O u r
feeling is that a number of multivariate statistical techniques, including canonical
(Blackith & Reyment, 1971)
and correspondence analysis (David et al.,
1973) are especially appropriate for use with experimental ecosystem studies that
produce large amounts of data on a large number of parameters. The application of
these techniques to ecological data from natural systems has already been described
( C h a r d y et al., 1976)
The relative variation within and among treatments in the microcosms during the first 30 days a~er sewage additions were begun is clearly evident in a canonical variates analysis of a data set ttlat included ammonia, chlorophyll a, numbers of diatoms, numbers of flagellates, pelagic A T P , and the numbers of particles in three
40% VARIATION EXPLAINED
]EGINNING OF EXPERIMENT
Fig. 6: Correspondence analysis of data on the species composition and abundance of the
phytoplankton in the experimental microcosms during the three month addition. Numbers
represent samples in individual microcosms while the letters represent particular species of
diatoms (D) or dinoflagellates (DF) or flagellates (F) that were associated with the various
samples. All of the tanks were similar at the beginning of the experiment before sewage was
added (see Fig. 4 for numbers of individual microcosms)
size ranges (Fig. 4). The outlying point in each of the replication triangles represents
the individual microcosm in each treatment that developed the anomalous ammonia
values discussed earlier. Since the tanks with high nitrogen levels did not appear to
differ with respect to any of the other parameters (Fig. 2), it is evident that the
multivariate analysis will reflect the extreme behavior of any one parameter. A
similar canonical variates analysis on a data set that included all of the nutrients and
chlorophyll a from the entire three month period during which the sewage inputs were
maintained again showed that the replication remained adequate to show treatment
effects along axis one and anomalous ammonia values along axis 2 (Fig. 5).
The data used in canonical variates analysis came from relatively aggregated parameters such as total chlorophyll a. Since the treatments could easily be separated
at this level, it was somewhat surprising to find that variations in the more detailed
measurements of p h y t o p l a n k t o n species abundance prevented a clear separation of
treatment on the basis of the total p h y t o p l a n k t o n community structure or on the basis
I0 ...... ~
, " ~ ". /
of "characteristic species" (Fig. 6). The correspondence analysis of taxonomic data did
clearly show initial similarity of all of the microcosms. It also reflected some tendency
for flagellate groups, especially Prorocentrurn, to be more abundant in the higher
sewage treatments (Figs. 6 and 7). H o w e v e r , the results suggest that the large amount
of effort involved in the t a x o n o m i c analysis at l o w e r trophic levels did not provide
data that were particularly sensitive to perturbations of this kind. It is not clear
whether this result comes about because individual species are relativelyinsensitive to the
perturbation or because the large amount of variation inherent in less aggregated ,data,
including the seasonal change from diatoms to flagellates, makes it more difficult to
measure the response of individual species in the context of a total plankton community.
C o m p a r i s o n
w i t h
t h e
n a t u r a l
s y s t e m
One of the prevailing myths in marine ecology is that the goal of microcosm
studies is to develop an exact replica in miniature of some particular natural system.
Such a goal is neither attainable or necessary. What is desired, however, is to develop
a microcosm which is generally similar to the larger system in terms of its trophic
structure, the major features of its taxonomic composition, its level o f productivity,
its rates of material cycling, and in its response to perturbations. The microcosm
community should also be self-maintaining over a time span that is appropriate for the
processes being studied. It is probable that neither the microcosms described here nor
those of any other study yet reported have fulfilled all of these criteria. It is hard to
k n o w this for sure since, as mentioned earlier, there is no generally agreed upon
S I , I ~ E 9 ?
criteria for the success of microcosms with respect to any of these goals. N o r is it clear
that any one of them is of greater or lesser importance than the others for any
particular study. I t is clear, however, that comparisons of microcosms and natural
systems must consider the variation of each. For example, if one compares the time
series data from the control microcosms in this study with measurements of the input
water collected from the laboratory dock, it appears that there were often appreciable
differences (e.g. Figs 2 and 3). However, there is no o n e chlorophyll or nutrient value
for Narragansett Bay. Instead, there is a range of values that one might normally expect
to find in different parts of the bay at different times of the year (Fig. 8), and it is
against this background of natural variation that the behavior of the microcosm should
be evaluated. With the exception of the anomalously high nitrogen levels in one tank,
the control microcosms generally fell within the range of values characteristic of the
lower West Passage of Narragansett Bay. A canonical variates analysis of nutrient and
chlorophyll data from the microcosms and from 13 stations located throughout the bay
confirmed this intuitive analysis and indicated that the microcosms responded to
sewage inputs by developing characteristics similar to those found along a gradient
from the Providence River to the mouth of Narragansett Bay (Fig. 9).
The microcosms remained throughout the study as plankton based systems with
grazing food chains containing over 30 forms of phytoplankton and zooplankton. The
benthic communities contained a healthy assemblage of macrofauna similar to that
-5 - 6I
o Microcosms t
n ~ e d ~ ~ l O
80.2% Voriotion exploi
found in Narragansett Bay. While the benthos was composed of 10 species of macro
invertebrates at the beginning of the experiment, the diversity increased through the
natural seeding of meroplankton in the input water to 35 species by the end of the
study. Benthic biomass increased from 2-4 fold and measurements of oxygen uptake
and ammonia flux from the sediment indicated that total benthic metabolism was
similar to that found in the bay
(Nixon eta]., 1976)
E f f e c t
o f s e w a g e
The addition of sewage to the microcosms brought about an immediate increase in inorganic and organic nutrients (Fig. 3) which was followed by clear increases in chlorophyll, particulate ATP, suspended particles and cell numbers (Fig. 2). In general, the time series data show higher values and greater variation for all these
Composition and metabolism of the benthic community in the experimental microcosms from mid - March to mid - September. Sewage inputs
were maintained from April through mid - July
Treatment (volume of sewage)
* Most abundant species were Arnpelisca abdita, NucuIa annulata and Mulinia lateralis
** Most abundant species in all treatments were Nucula annulata, Tharyx acutus, and Polydora ligni
*** Major genera were Terschellingia, Odontophora, and Sabatieria
parameters in the microcosms with higher sewage input. There was also a tendency for
the phytoplankton in the sewage tanks to contain a larger percentage of flagellates
(Fig. 7). However, as shown in Figure 6, changes in the specific taxonomic composition
and relative abundance of the plankton as a result of the sewage addition could not be
demonstrated with correspondence analysis. The same was true for the taxonomic
composition, growth and metabolism of the benthos, where the only statistically
significant change appeared to be an increase in the carbon and nitrogen content of the
surface sediments in the high sewage microcosms (Table 5).
TREATMENT 82.6% variation a
Fig. 10: Canonical variates analysis of chlorophyll a, nutrients, pelagic ATP, and particle
counts in the experimental microcosms during three months with sewage input compared with
the same analysis during the two months following termination of the sewage input.
Substantial recovery of lower treatment levels is apparent in overlap of the replication triangles
Fig. 11: Canonical variates analysis of chlorophyll, ammonia, nitrate, and nitrite in the
microcosms and in Narragansett Bay during the recovery period. Good agreement with lower bay
stations is indicated by all treatment levels for this data set
The time series data show declines in cell counts, chlorophyll, and pelagic A T P
that began before the sewage was stopped (Fig. 2). Marked fluctuations in these, and
other parameters became particularly apparent and are also characteristic of N a r r a
gansett Bay in the warmer months. The high seasonal variability obscures clear trends
in the single parameter time series data. However, canonical variates analysis of
nutrient values, chlorophyll a, pelagic ATP, and particle counts during the three month
sewage addition and the two month recovery period did show differences between the
sampling periods (Fig. 10). Aflcer the sewage was stopped, the replication triangles all
began to overlap to some degree, even though seasonal changes had influenced the
general condition and variability of all the microcosms. A reduced data set from the
microcosms which included chlorophyll, ammonia, nitrate and nitrite was compared
with the 13 bay stations during the recovery period using canonical variates analysis
(Fig. 11). All treatment microcosms and two controls overlap mid and lower bay
stations indicating good recovery for these parameters. Control microcosm 1, by
contrast, was positioned high on Axis 2 due to very high values of nitrate and nitrite,
apparently derived from earlier high ammonia values. Recovery of the microcosms
that had been subjected to the high sewage inputs was still far from complete with
respect to all parameters, even though the recovery period was almost twice as long
as the flushing time of the systems.
Acknowledgements. We are grateful to Alina Froelich, Betty Buckley, Steve Hale, Jona
than Garber, Jack Kelley, Sybil Seitzinger and to our EPA colleagues Andrea Hurtt, Sue Cheer,
Earl Davey and Barbara Guida for help with the extensive laboratory analysis. M. Gayle Kraus performed the meiofaunal analyses. Robert Marrero identified and counted phytoplankton species. BjSrn Malmgren and Wendell Hahm assisted with the multivariate analyses. This research was supported by a grant from the U.S. Environmental Protection Agency, R-803 143.
L I T E R A T U R E C I T E D
Abbott , W. , 1966 . Microcosm studies on estuarine waters. I. The replicability of microcosms . J. Wat. Pollut. Control Fed . 30 , 258 - 270 .
Bertalanffy , L. yon , 1968 . General system theory . Braziller , New York, 289 pp.
Beyers , R. J. , 1963 . A characteristic diurnal metabolic pattern in balanced microcosms . Pubis Inst. mar. Sci. Univ. Texas 9 , 19 - 27 .
Blackith , R. E. & Reyment , R. A. , 1971 . Multivariate morphometrics . Acad . Press, London, 412 pp.
Bodungen , B. yon, yon Brockel, K. , Smetacek , V. & Zeitzschel , B. , 1976 . The planktontower. I. A structure to study water/sediment interactions in enclosed water columns . Mar. Biol . 34 , 369 - 373 .
Chardy , P. , Glemarec , M. & Laurec , A. , 1976 . Application of inertia methods to benthic marine ecology: practical implications of the basic options . Estuar. coast, mar. Sci. 4 , 179 - 205 .
Cheer , S. , Gentile , J. H. & Hegre , C. S. , 1974 . Improved method for ATP analysis . Analyt. Biochem . 60 , 102 - 114 .
Cooke , G. D. , 1971 . Aquatic laboratory microeeosystems and communities . In: The structure and function of freshwater microbial communities . Ed. by J. Cairns . Polytech. Inst., Blackburgh, VA., 47 - 85 .
David , M. , Campiglio , C. & Darling , R. , 1974 . Progress in R- and Q-mode analysis: correspondence analysis and its application to the study of geological processes . Can. J. Earth Sci . 11 , 131 - 146 .
Davies , J. M. , Gamble , J. C. & Steele , J. H. , 1975 . Preliminary studies with a large plastic enclosure . In: Estuarine research. Ed. by L. E. Cronin . Acad. Press, New York, 1 , 251 - 264 .
Farrington , J. W. & Quinn , J. G. , 1973 . Petroleum hydrocarbons and fatty acids in waste water effluents . J. Wat. Pollut. Control Fed . 45 , 704 - 712 .
Knauss , J. A. , Griffin , J. J. , Nixon , S. W. , Pilson , M. E. Q. & Smayda , T. J. , 1976 . A facility for the experimental analysis of coastal marine ecosystems . Tech. prog. Rep. U.S. envir. Protect. Agency , 124 pp.
Kremer , J. N. & Nixon , S. W. , 1975 . An ecological simulation model of Narragansett Bay - the plankton community . In: Estuarine Research. Ed. by L. E. Cronin . Acad. Press, New York, 1 , 672 - 690 .
Levandowsky , M. , 1977 . Multispecies cultures and microcosms . In: Marine ecology. Ed. by O. Kinne. Wiley-Interscience, London, 3 ( 3 ) 1399 - 1458 .
Likens , G. E. , Bormann , F. H. , Johnson, N. M. , Fisher , D. W. & Pierce , R. S. , 1970 . Effects of forest cutting and herbicide treatment on nutrient budgets in the Hubbard Brook watershed ecosystems . Ecol. Monogr . 40 , 23 - 47 .
Mann , K. H. , 1972 . The analysis of aquatic ecosystems . In: Essays in hydrobiology. Ed. by R. B. Clark & R. J. Wooton . Univ. of Exeter, Exeter, 1 - 14 .
- - 1975 . Relationship between morphometry and biological functioning in three coastal inlets of Nova Scotia . In: Estuarine Research. Ed. by L. E. Cronin . Acad. Press, New York, 1 , 634 - 644 .
Menzel , D. W. & Corwin , N. , 1965 . The measurement of total phosphorus in seawater based on the liberation of organically bound fractions by persulfate oxidation . Limnol. Oceanogr . 10 , 28O - 282 .
Nixon , S. W. , Oviatt , C. A. & Hale , S. S. , 1976 . Nitrogen regeneration and metabolism of coastal marine bottom communities . In: The role of terrestrial and aquatic organisms in decomposition processes . Ed. by J. M. Anderson & A. Macfaden . The 17th symposium of the British Ecological Society. Blackwell , Oxford, 269 - 283 .
Odum , t-I. T. (Ed.), 1970 . A tropical rain forest, a study of irradiation and ecology at E1 Verde, Puerto Rico . Div. Tech. Inf., U.S. Atomic Energys Commn, 9 chapters .
- - 1971 . Environment, power and society . Wiley-Interseience, New York, 331 pp.
Patten , B. C. , 1971 . Coda. In: Systems analysis and simulation in ecology . Ed. by B. C. Patten . Acad. Press, New York, 1 , 583 - 584 .
Perez , K. , Morrison , G. , Oviatt , C. , Nixon , S. & Buckley , B. , 1976 . The importance of physical and biotic scaling to the experimental simulation of a coastal marine ecosystem. Helgol~inder wiss . Meeresunters . 80 , 144 - 162 .
Smetacek , V. , Bodungen , B. yon, Bro&el, K. yon & Zeitzschel , B. , 1976 . The planktontower . II. Release of nutrients from sediments due to changes in the density of bottom water . Mar. Biol . 34 , 373 - 378 .
Solorzano , L. , 1969 . Determination of ammonia in natural waters by the phenolhypochlorite methods . Limnol. Oceanogr . 14 , 799 - 801 .
Steele , J. H. , 1974 . The structure of marine ecosystems . Blackwell , Oxford, 128 pp.
Strickland , J. D. H. & Parsons , T. R. , 1968 . A practical handbook of seawater analysis . Bull. Fish. Res. Bd Can . 167 , 1 - 311 .
Takahaski , M. , Thomas , W. H. , Siebert , D. L. R. , Beers , J. , Koeller , P. & Parsons , T. R. 1975 . The replication of biological events in enclosed water columns . Arch. Hydrobiol . 76 , 5 - 23 .
U.S. Army Corps of Engineers, 1959 . Contamination dispersion in estuaries, Narrangansett Bay . Hydraulic Model Investigation . Misc. Pap. U.S. Army Eng. Waterways Exp. Sm 2 , 1 - 332 .
Walsh , J. J. , 1975 . A spatial simulation model of the Peruvian upwelling ecosystem . Deep Sea Res . 22 , 201 - ' 236.
Graduate School of Oceanography University of Rhode Island Kingston, Rhode Island 02881 USA