Professional Documents
Culture Documents
1 s2.0 0304380094901430 Main
1 s2.0 0304380094901430 Main
Kenneth H. Reckhow
Center for Environmental Toxicology and Risk Assessment, School of the Environment,
Duke University, Durham, NC 27708-0328, USA
(Received 20 August 1992; accepted 16 March 1993)
ABSTRACT
1. I N T R O D U C T I O N
for management decisions in the belief that it will yield more accurate,
detailed predictions. Cost of application and difficulty in understanding
the model and results appear to have ultimately led to a mixed view
among members of this group.
4. Successful case studies, in which a mechanistic surface water quality
model other than QUAL2E has aided decision making (and yielded
predictions that have been shown to be fairly accurate) appear to be
rare.
tion that most parameters were known to within + 20%, (2) without a term
for model error (Halfon did identify problems with the model), and (3) with
no measurements of mirex loading (loadings were "back-calculated").
Halfon set the _+20% parameter error level based on his sense of the
literature, and used it as the full range (max-min) on the uncertainty in the
parameters. This seems low (given natural variability, measurement error,
and bias due to model mis-specification) for many of the model terms,
particularly when used as an estimate of range. Unfortunately, Halfon's
analysis and thoughtful discussion are constrained by data and model
limitations, many of which he identifies.
Additional results presented by Beck (1987) in his thoughtful and thor-
ough review paper indicate that the error levels cited in these studies are
typical of those reported elsewhere. He concludes "Most of the evidence
suggests that the current models of water quality, in particular, the larger
models, are easily capable of generating predictions to which little confi-
dence would be attached." Yet despite this view, there appears to be the
belief among many modelers that since the world is complicated, so too
must be our simulation models if they are to be accurate.
In a previous section, models were characterized as either mechanistic or
empirical. In truth, simulation models exist on a continuum on a mechanis-
tic (or empirical) axis, and the degree of mechanism attributed to a given
model is apt to vary among individuals and among disciplines. For example,
among water quality simulation modelers, WASP4 (Ambrose et al., 1988) is
universally considered to be a mechanistic model. However, environmental
chemists and aquatic biologists are apt to view the mathematics for the
individual mechanisms as simplifications of the actual processes, and often
as empirical simplifications, particularly when first-order reactions are
employed. This does not mean that the simplifications are inappropriate.
Rather, it means that "correct process description" should not be cited as
the basis for selection of a "mechanistic" model.
Beck's conclusions concerning the problems with prediction error magni-
tude for the larger models raises interesting questions. Given current
knowledge and available data, at what point does the level of detail in the
model exceed the modeler's ability to produce an acceptable model? Can
we model the transport and fate of every organism in a waterbody? Can we
model the transport and fate of each species of phytoplankton in a
waterbody? Can we model the transport and fate of major phytoplankton
groups (e.g., greens, blue-greens, diatoms) in a waterbody? Can we model
the transport and fate of overall phytoplankton biomass in a waterbody? A
similar series of questions could be posed about s p a c e / t i m e detail in a
simulation model.
An interesting analogy to this is found in physics in the area of statistical
8 K.H. RECKHOW
4. P A R A M E T E R D E T E R M I N A T I O N IN M E C H A N I S T I C M O D E L S
(3) limited available data to fit and evaluate models, and (4) limited
resources to collect additional data. For these reasons, model coefficients
and reaction rates in the models generally characterize actual processes
and are not (at least initially) intended to be empirically-fitted constants.
Since the parameters of mechanistic models are intended to describe
real processes, it may be assumed that an experimental study of a particu-
lar process can yield a parameter estimate that can be directly inserted into
the model. In some cases, it is acknowledged that the rate or coefficient in
the model is affected by other conditions in the waterbody (e.g., turbu-
lence), and thus adjustments must be made to the experimentally-based
value. However, if the model truly is a complete mechanistic description of
the system of interest, then adjustment should be unnecessary. This view is
reflected in Halfon's (1984) statement concerning EXAMS (Burns and
Cline, 1985): "Since the EXAMS model is based on documented formulas,
it does not require calibration."
However, given the relative simplicity of all simulation models in com-
parison to the complexity of nature, it seems reasonable to question the
legitimacy of any "mechanistic" mathematical description of surface water
quality. Further, given data limitations and scientific knowledge limitations,
it seems reasonable to question even the goal to strive for a model that
need not be calibrated. The correctness of model structure, the knowledge
of the model user, and the availability of experimental and observational
evidence all influence parameter choice for mechanistic models. Unfortu-
nately, too often knowledge and data are extremely limited, making choice
of parameters and choice of important model processes guesswork to a
distressingly large degree. The examples presented in the next two sections
are not re-assuring with respect to these two issues: (1) scientific support
for the selection of model parameters, and (2) scientific support for the
specification of appropriate model functional relationships.
5. AN EXAMPLE OF PARAMETER SELECTION APPROACHES: PHYTOPLANK-
TON SE'Iq'LING VELOCITY
C 2 = zooplankton concentration ( g / m 3)
Fza = fractional feeding preference
O t h e r examples are quite similar; a common alternative approach is that
phytoplankton settling is sometimes treated as a velocity term with an areal
loss (see Chapra and Reckhow, 1983, Chapter 14):
phytoplankton settling ( m a s s / t i m e ) = t:IAC 1. (2)
To understand some of the problems with the current approach for
p a r a m e t e r determination in mechanistic surface water quality models, it is
useful to examine this process further. For that purpose, "phytoplankton
settling velocity" provides a good example. Phytoplankton, or algae, are
important in aquatic ecosystems, and thus one or more phytoplankton
compartments are found in most mechanistic surface water quality models
c o n c e r n e d with nutrient enrichment. Phytoplankton settling is one of the
key mechanisms for removal of phytoplankton from the water column.
Stoke's law provides the starting point for the mathematical characteri-
zation of phytoplankton settling. Few models, however, employ Stoke's law;
instead a simple constant settling velocity (in units of l e n g t h / t i m e ) expres-
sion is commonly used. To apply a model with this settling velocity term, a
m o d e l e r must either measure phytoplankton settling directly, or select a
representative value from an other study. Since field m e a s u r e m e n t of
phytoplankton settling is a difficult task, use of literature-tabulated values
is standard practice.
Probably the most thorough listing of suggested values for phytoplankton
settling velocity is Bowie et al. (1985), which presents a two-page table of
reported values, by algal type (see Table 1). Bowie et al. note that under
quiescent conditions in the laboratory, phytoplankton settling is a function
of algal cell radius, shape, density, and special cell features such as gas
vacuoles and gelatinous sheaths. For natural water bodies, water turbu-
lence can be quite important. In two- or three-dimensional models with
hydrodynamic simulation, turbulence is accounted for in the model equa-
tions; in zero- or one-dimensional models, the effect of turbulence on
phytoplankton settling must usually be incorporated into the choice of
settling velocity.
WATER QUALITY MODELING AND UNCERTAINTY ANALYSIS 11
TABLE 1
Phytoplankton settling velocities (Bowie et al., 1985) a
phytoplankton; rather, they are simulation model studies, and the value for
phytoplankton settling velocity listed in Table 1 is the value chosen for the
model. In some of the references checked, little or no basis was provided
for the choice. When a rationale for choice was given, it was usually to
adopt or adjust the few values presented in the literature from experimen-
tal studies, or to adopt a value from another modeling study. In one way or
another, it appears that virtually all of the values presented in Table 1 have
some dependency on the early experimental work of Smayda and Boleyn
(1965) and other work by Smayda.
Unfortunately, evaluation studies of simulation models have provided
little insight on good point estimates for this parameter. Observational data
on surface water quality are almost always inadequate for testing functional
relationships and assessing parameter choices. Typical observational data
sets are noisy, with few measurements of each of only a few variables. In
the case of phytoplankton settling velocity, observational data are apt to
consist of pfiytoplankton cell densities at various dates, times, and areal
locations, but probably not depths. Since phytoplankton are also removed
from the water column through consumption by higher food chain organ-
isms, the observational data do not permit separate identification of the
removal mechanisms.
Given this situation, modelers have relied almost exclusively on the few
experimental studies in the laboratory and their judgment concerning
adjustments to these values. For one-dimensional models without explicit
modeling of hydrodynamics, the chosen value may be as much as an order
of magnitude higher than the laboratory values. Two- or three-dimensional
models with hydrodynamics may incorporate the unadjusted laboratory
value. After early modeling studies presented chosen values, these values
were sometimes adopted in subsequent studies without comment (in effect,
"default" values were identified). Thus, there is probably much less infor-
mation in the columns of Table 1 than implied by the number of values
reported.
In summary, the choices for phytoplankton settling velocity appear to be
based on ad hoc adjustments to a few values measured under controlled
conditions. There is virtually no field confirmation of choices made for
parameters individually (as opposed to collectively). This situation is fairly
typical of the state-of-the-art in surface water quality simulation modeling.
6. F R O M SCIENTIFIC K N O W L E D G E , E X P E R I M E N T A T I O N , A N D O B S E R V A T I O N
TO M O D E L SPECIFICATION: B I O D E G R A D A T I O N OF PCBS
V = system volume (m 3)
c = total contaminant concentration ( g / m 3)
While it has been shown that most SOCs (like PCBs; polychlorinated
biphenyls) are biodegradable under controlled conditions (e.g., in labora-
tory experiments), there is often a great deal of uncertainty concerning the
magnitude, and even the existence, of biodegradation occurring naturally in
waterbodies. As Madsen (1991) notes, " I n situ biodegradation of organic
compounds is very difficult to prove." For example, as Chapra and Reck-
how (1983) observe, "many laboratory degradation studies are performed
using microbial populations that have been acclimated to a particular
organic compound. Whether such species would tend to behave similarly in
the lake environment is questionable and makes it difficult to apply some
of the laboratory work to natural situations." Unfortunately, Madsen
(1991) concludes, " T h e problems of extrapolating from laboratory data to
the field have never been solved despite decades of debate." Even with
these difficulties, biodegradation is frequently included as a transformation
process when models such as WASP4 are used for SOC transport and fate.
In a study of the scientific basis for parameter determination in simula-
tion models, Wolpert et al. (1993) looked at biodegradation and other
processes affecting PCBs in surface waterbodies. Their review of the
mechanistic modeling literature indicated that for the most part, even in
modeling studies with the most thorough databases, the source for the
biodegradation rate constant either was not specified or was based on a
synthesis of literature values. In other words, there did not appear to be in
situ evidence for a particular biodegradation rate.
In their own modeling work on PCB transport and fate in the upper
Hudson River (NY), Wolpert et al. (1993) stated "It is believed that there
14 K.H. R E C K H O W
the mixture of PCBs present, it is hard to see how to use the reported rates
to predict biodegradation rates in the Hudson River bed with its unknown
mixture of PCB congeners unless each congener is modeled separately and
it is known how different ratios of congener concentrations affect the
biodegradation rates. Clearly, congener-specific degradation rates for all
possible mixtures are unavailable. Our ability to model individual con-
geners is also limited by uncertainty about the initial congener composition
of the river loadings and of the observed water column and sediment
samples. Other reasons for suspecting the laboratory data may not reflect
field conditions accurately include:
• It is likely that laboratory growing conditions are more conducive to
bacterial populations than are river conditions. Populations of PCB-de-
grading bacteria could well be orders of magnitude smaller in the river
than in the laboratory;
• The experiments do not include enough measurements over time to
verify a stationary biodegradation rate constant, as assumed in [the]
model;
• Environmental conditions in the sediment must be expected to differ
from those in the laboratory - for example, the ratio of carbon dioxide
to hydrogen is a particularly sensitive environmental parameter for
bacteria and can vary from place to place in the sediment."
Initial reaction to the uncertainty regarding in situ biodegradation of
PCBs is apt to be one of concern. This may be followed quickly by the
observation that, in fact, PCBs are 209 compounds; thus, why expect that a
single mechanistic model will be adequate for transport and fate? Perhaps
so, yet:
• PCBs are among the most studied SOCs in the environment; thus,
scientific knowledge, experimental evidence, and observational data
should provide a "best case" for the state-of-the-art in SOC transport
and fate modeling,
• Decisions (including the decision to do nothing) are being made concern-
ing PCBs in the aquatic environment, and models should be considered
for decision support.
Certainly, we must have an understanding of cause-effect relationships
to make reliable predictions of pollutant transport and fate. Predictions
based on statistical relationships alone may work for interpolation in
well-studied situations (e.g., aggregate response for lake eutrophication),
but they may fail to account for important mechanisms in new situations.
Yet, uncertainties in the mechanistic transport and fate models should
cause us to search for other ways to provide predictive scientific analyses
for decision support. A suggested alternative is presented below.
16 K.H. RECKHOW
toring is evaluated in terms of the value function (e.g., "Including the cost
of experimentation or monitoring, is the expected utility higher or lower
with additional information?"). In addition, a decision theoretic approach
to environmental risk assessment illustrates the critical role of uncertainty
analysis in the interaction between modelers and managers discussed
above.
8. A CONCLUDING THOUGHT
ACKNOWLEDGMENTS
REFERENCES
Ambrose, R.B., Jr., Wool, T.A., Connolly, J.P. and Schanz, R.W., 1988. WASP4, A
Hydrodynamic and Water Quality Model - Model Theory, User's Manual, and Program-
mer's Guide. U.S. Environmental Protection Agency, EPA/600/3-87/039.
Beck, M.B., 1987. Water Quality Modeling: A Review of the Analysis of Uncertainty. Water
Resour. Res., 23: 1393-1442.
Bedard, D.L., Unterman, R., Bopp, L.H., Brennan, M.J., Haberl, M.L. and Johnson, C.,
1986. Rapid Assay for Screening and Characterizing Microorganisms for the Ability to
Degrade Polychlorinated Biphenyls. Appl. Environ. Microbiol., 51: 761-768.
Bowie, G.L., Mills, W.B., Porcella, D.B., Campbell, C.L., Pagenkopf, J.R., Rupp, G.L.,
Johnson, K.M., Chan, P.W.H., Gherini, S.A. and Chamberlin, C.E., 1985. Rates, Con-
W A T E R Q U A L I T Y M O D E L I N G AND U N C E R T A I N T Y ANALYSIS 19
stants, and Kinetics Formulations in Surface Water Quality Modeling. U.S. Environmen-
tal Protection Agency, EPA/600/3-85/040.
Brown, L.C. and Barnwell, T.O., Jr., 1987. The Enhanced Stream Water Quality Models
QUAL2E and QUAL2E-UNCAS: Documentation and User Manual. U.S. Environmen-
tal Protection Agency, EPA/600/3-87/007.
Burns, L.A. and Cline, D.M., 1985. Exposure Analysis Modeling System, Reference Manual
for EXAMS II.U.S. Environmental Protection Agency, EPA-600/3-85-038.
Chapra, S.C. and Reckhow, K.H., 1983. Engineering Approaches for Lake Management.
Vol. II: Mechanistic Modeling. Butterworths, Boston, 492 pp.
Chen, C.W. and Orlob, G.T., 1972. Ecologic Simulation for Aquatic Environments. Office
of Water Resources Research, US Dept. of Interior, Washington, DC.
Di Toro, D.M. and van Straten, G., 1979. Uncertainty in the Parameters and Predictions of
Phytoplankton Models. Working Paper WP-79-27, International Institute for Applied
Systems Analysis, Laxenburg, Austria.
Halfon, E., 1984. Error analysis and simulation of mirex behaviour in Lake Ontario. Ecol.
Modelling, 22: 213-252.
Jorgensen, S.E., 1980. Lake Management. Pergamon Press, Oxford, 167 pp.
Madsen, E.L., 1991. Determining in situ biodegradation - facts and challenges. Environ.
Sci. Technol., 25: 1663-1673.
Morgan, M.G. and Henrion, M., 1990. Uncertainty. Cambridge University Press, Cam-
bridge, 332 pp.
Reckhow, K.H., 1992. The importance of scientific uncertainty in decision making.
Manuscript, Duke University Center for Environmental Toxicology and Risk Assess-
ment.
Reckhow, K.H. and Chapra, S.C., 1979. Error analysis for a phosphorus retention model.
Water Resour. Res., 15: 1643-1646.
Reckhow, K.H., Coffey, S.C., Henning, M.H., Smith, K. and Banting, R., 1992. Eutromod:
Technical Guidance and Spreadsheet Models for Nutrient Loading and Lake Eutrophi-
cation. Duke University School of the Environment, Durham, NC.
Reif, F., 1965. Fundamentals of Statistical and Thermal Physics. McGraw-Hill, New York,
651 pp.
Rhee, G.-Y., Bush, B., Brown, M.P., Kane, M. and Shane, L., 1989. Anaerobic biodegrada-
tion of polychlorinated biphenyls in Hudson River sediments and dredged sediments in
clay encapsulation. Water Res., 23: 957-964.
Risk Assessment Forum, 1992. Framework for Ecological Risk Assessment. U.S. Environ-
mental Protection Agency, EPA/630/R-92/001.
Scavia, D., Canale, R.P., Powers, W.F. and Moody, J.L., 1981. Variance Estimates for a
Dynamic Eutrophication Model of Saginaw Bay, Lake Huron. Water Resour. Res., 17:
1115-1124.
Schnoor, J.L., Sato, C., McKechnie, D. and Sahoo, D., 1987. Processes, Coefficients, and
Models for Simulating Toxic Organics and Heavy metals in Surface Waters. U.S.
Environmental Protection Agency, EPA/600/3-87/015.
Smayda, T.I. and Boleyn, B.J., 1965. Experimental observations on the floatation of marine
diatoms. Part I: Thalassiosira naria, T. rotula and Nitzschia seriata. Limnol. Oceanogr.,
10: 499-510.
Thomann, R.V. and Mueller, J.A., 1987. Principles of Surface Water Quality Modeling and
Control. Harper & Row, New York, 644 pp.
van Straten, G., 1983. Maximum likelihood estimation of parameters and uncertainty in
20 K.H. RECKHOW
phytoplankton models. In: M.B. Beck and G. van Straten (Editors), Uncertainty and
Forecasting of Water Quality. Springer Verlag, Berlin.
Wolpert, R.L., Steinberg, L.J. and Reckhow, K.H., 1993. Bayesian decision support using
environmental transport and fate models. In: C. Gatsonis, J.S. Hodges, R.E. Kass and
N.D. Singpurwalla (Editors), Case Studies in Bayesian Statistics. Springer, New York.