Population Level Impacts of Cooling Water Withdrawals on Harvested

Feb 20, 2007 - Population Level Impacts of Cooling Water Withdrawals on Harvested Fish Stocks. Stephen C. Newbold * and Rich Iovanna. U.S. EPA, Nation...
0 downloads 8 Views 151KB Size
Environ. Sci. Technol. 2007, 41, 2108-2114

Population Level Impacts of Cooling Water Withdrawals on Harvested Fish Stocks STEPHEN C. NEWBOLD* AND RICH IOVANNA U.S. EPA, National Center for Environmental Economics, 1200 Pennsylvania Avenue Northwest, Washington, DC

Trillions of gallons are withdrawn every year from U.S. rivers, estuaries, lakes, and coastal waters to cool the turbines of power plants and other equipment in manufacturing facilities. In the process, large numbers of aquatic organisms die from entrainment into the plant or impingement against the outer portion of the intake structure. In this paper, we develop a generalized age-structured population model with density dependent survival of sub-adult age classes, and we use the model to perform a screening analysis of the effects of entrainment and impingement for 15 harvested fish stocks off the California and Atlantic coasts. Stock sizes are estimated to be depressed by entrainment and impingement by less than 1% in 10 of the 15 cases considered, between 1 and 3% in two cases, and between 20 and 80% in three cases. A variety of sensitivity analyses are conducted to evaluate the influence of several sources of model and parameter uncertainties.

1. Introduction Trillions of gallons are withdrawn every year from U.S. rivers, estuaries, lakes, and coastal waters to cool the turbines of power plants and other equipment in manufacturing facilities. In the process, large numbers of aquatic organisms are killed by entrainment into the plant or by impingement against the outer portion of the intake structure. Section 316(b) of the Clean Water Act requires that the “location, design, construction and capacity of cooling water intake structures shall reflect the best technology available for minimizing adverse environmental impact” from entrainment and impingement (1). “Adverse environmental impact” has been interpreted broadly by EPA (2), but in this paper, we focus exclusively on the population level impacts of entrainment and impingement at a large geographic scale for commercially or recreationally harvested fish stocks. When conducting ecological risk assessments, it is common for analysts to focus on effects at the level of the organism, for example, changes in average reproductive or survival rates through particular stages of a species’ life cycle, the frequency of developmental abnormalities, indicators of reproductive viability such as egg shell thickness, etc. However, there is growing interest in evaluating ecological impacts at higher levels of biological organization as well (3, 4), in particular the population level (5, 6). Population modeling can help determine the relative importance of impacts on different life stages by translating them into * Corresponding author phone: 202-566-2293; fax: 202-566-2338; e-mail: [email protected]. 2108 9 ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 41, NO. 7, 2007

impacts on adult abundances and harvest levels, and it can help make comparisons across species by expressing the effects of changes in entrainment and impingement rates as percentages of the baseline or maximum abundance or biomass for each species. Thus, population modeling can help put raw data on entrainment and impingement lossess the number of eggs, larvae, and other life stages killed by cooling water withdrawals every yearsinto a broader ecological context. In this paper, we describe a general modeling framework for evaluating the population-level impacts of cooling water withdrawals, and we apply the framework to 15 harvested fish stocks. The population model we use is a generalized age-structured model written in discrete time, and the model can incorporate density dependent survival in one or more of the sub-adult life stages. We use life history parameters, collected previously by EPA, for evaluation of the 316(b) Phase II rulemaking (1, 7), and we add information on reproductive rates and historic harvest levels for each species. This model is broadly similar to several previous models used to estimate the population level impacts of entrainment and impingement on fish stocks (see especially refs 8 and 9). However, we develop a calibration approach that allows us to use independent data on harvest levels and aggregate entrainment and impingement losses over a large geographic scale to estimate two key life history parameters and lifestage-specific entrainment and impingement mortality rates simultaneously for each stock. In addition, we apply our model to more fish stocks (15) than have been analyzed in the past, and we perform a variety of sensitivity analyses to check the robustness of our results. Most observers cite overfishing followed by excessive nutrient loads as the leading causes of degradation of aquatic ecosystems overall (e.g., refs 10, 11). However, previous studies by several researchers have implicated cooling water withdrawals as a potentially important source of stress in some aquatic systems (8, 12-15). Much of the previous research on the impacts of entrainment and impingement has focused on estimating “conditional mortality rates,” i.e., the “fractional reduction in year-class abundance due to [entrainment or] impingement, provided that density-dependent mortality is low during the period in which impingement occurs” (16). However, a number of previous studies have estimated population-level effects of entrainment and impingement (i.e., changes in long-term abundance or harvest levels) on particular fish stocks in localized areas (see Table 1). As part of an early conference on the impacts of cooling water withdrawals on fish populations (8), Christensen et al. (17) developed Beverton-Holt- and Ricker-style models to assess the impacts of entrainment and impingement on the striped bass population in the Hudson River, and Swartzman et al. (18) reviewed eight population models developed to predict effects on striped bass in the Hudson River and the Delaware and Chesapeake Bays and winter flounder in Long Island Sound. A series studies by Jensen and others (14, 19, 20) used population models to investigate the impacts of cooling water withdrawals on several fish species in the Great Lakes. A large set of loosely coordinated studies on the population-level effects of entrainment and impingement were conducted in the late 1970s during the course of adjudicatory hearings over mitigation of power plant impacts on the Hudson River (12). More recently, Nisbet and others analyzed the potential impacts of entrainment and impingement on fish populations off the coast of southern California (13, pp 129-144 in 21). Another series of entrainment and 10.1021/es060812g Not subject to U.S. copyright. Publ. 2007 Am. Chem.Soc. Published on Web 02/20/2007

TABLE 1. Estimated Population Level Impacts of Cooling Water Withdrawals on Fish Populations from Previous Studies authors (citation)

location

estimated pop. level impacts

species

notes

Christensen et al. (17)

Hudson River

striped bass

-3.5 to -30%

based on an assumed 10% conditional mortality rate from entrainment and impingement.

Swartzman et al. (18)

Hudson River, Chesapeake & Delaware Bays, Long Island Sound

striped bass, winter flounder

-0.9 to -12.4%

review of eight models, based on various assumptions about conditional mortality rates from entrainment and impingement.

Jensen and Hamilton (20)

Lake Erie

yellow perch

-1.7%

reduction in fishable biomass. only one power plant considered.

Jensen (19)

Lake Erie

yellow perch

-1.98%

minor variation on (20).

Jensen et al. (14)

Lake Michigan

alewife

-2.86%

rainbow smelt yellow perch

-0.76% -0.28%

measured as changes in equilibrium biomass.

Savidge et al. (37)

Hudson River

striped bass

-3 to -40%

stylized models, parametrized to cover biologically plausible range of values.

Lawler (38)

Hudson River

striped bass

+3 to -21%

found that populations could increase (by up to 3%) accounting for temperature increases from cooling water outflows.

Goodyear (39)

Hudson River

striped bass

-16 to -62%

potential reductions in landings by 2015.

Nisbet et al. (13)

Coastal southern California

various/unspecified

> or < proportion ofimmatures killed

strategic models. Cited estimates of entrainment and impingement mortality rates on the order of 1 - 10%.

Barnthouse et al. (15)

Delaware Estuary

Weakfish

-3%

other species analyzed showed lower impacts.

impingement studies was commissioned to inform EPA’s work toward drafting regulations under Section 316(b) of the Clean Water Act (22). Barnthouse (23) reviewed the ecological modeling studies that informed the Hudson River case in the late 1970s and advancements since that time, and Barnthouse et al. (2002) investigated the impacts of the Salem Generating Station on a variety of fish species in the Delaware Estuary. These previous applications notwithstanding, the reliability of population models for estimating the effects of cooling water withdrawals on fish stocks is still uncertain. Van Winkle (24) summarized what seems to be a common view: “...the ability to prove what level of compensation is occurring for a particular species at a particular stage at a particular site to the satisfaction of a regulator or court is extremely difficult and is likely to remain that way in the future.” Many of the inherent uncertainties of population modeling discussed by the researchers cited above and summarized by Van Winkle remain relevant for our application. However, our view is that more empirical research on this topic is needed despite (or because of) these inherent uncertainties and can be useful to decision-makers. Specifically, we believe that the overall distribution of results across all stocks and sensitivity analyses presented below, while not definitive, provides the most comprehensive picture of the potential population-level impacts of cooling water withdrawals developed to date.

2. Methods 2.1. Model Structure. The population model developed here is designed to use previously collected biological information on the life history characteristics of commercially and recreationally harvested fish species known to be affected by

cooling water withdrawals (Apprndix Dl in 1), plus a limited amount of extra information on harvest levels from NOAA and estimated recruitment rates at low abundance from the fisheries literature. In this section, we lay out the mathematical structure of the model, and in the next section we show that (a simplified version of) the model has a single positive and locally stable equilibrium state. (All parameters will be defined as they appear and are listed alphabetically with their definitions in Appendix 1 of the Supporting Information.) The model allows for as many life stages as needed to accommodate the available information on life history characteristics and entrainment and impingement losses for each species. Life stages are indexed by k and years are indexed by t. For life stages where mortality is treated as density independent, the total per capita mortality rate, Zk, is an additive function of mortality from “natural” causes (all causes other than entrainment and impingement and fishing), Mk, entrainment and impingement, Ek, and fishing, Fk:

dNk 1 ‚ ) -Zk ) -Mk - Ek - Fk dt Nk

(1)

Solving the differential equation in eq 1 and integrating over the duration of a life stage gives the familiar exponential growth function:

Nk+1,t+1 ) Nk,te-zk

(2)

(Here and in all updating equations below the t + 1 subscript on the left-hand side would be a t for stages less than 1 year in duration.) Density-dependence typically is believed to VOL. 41, NO. 7, 2007 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

2109

operate most strongly in the sub-adult life stages for most fish species, though the precise stage and mechanism(s) often are uncertain (9). We have developed this model to be flexible enough to incorporate density-dependent survival into any one or more of the stages prior to the first age of reproduction. In the baseline model runs reported in Section 2.4 we assume that density-dependence operates in the larval stage through the stage just prior to reproductive maturity. For the densitydependent stages, per capita natural mortality is modeled as a linear function of abundance (as in the standard logistic model) within the stage:

dNk 1 ‚ ) -(Mk + bNk) - Ek - Fk dt Nk

(3)

In this case, integrating over the duration of a life stage gives (pp 63 and 76 in 21)

Nk+1,t+1 )

-Zk b - (Zk / Nk,t + b)eZk

(4)

and

Lk,t ) Nk(1 - e-Zk)Ek/Zk

(eq 8 is Baranov’s catch equation, ref 26, and eq 9 is directly analogous. Different expressions apply to the density dependent life stages, analogous to eq 4. In the applications in Section 2.4 we use a numerical integration approach to calculate harvest levels and entrainment and impingement losses for the density dependent stages; see the Supporting Information for details.) 2.2. Local stability analysis. We expect the model to be well behaved because of its close relationship to the logistic, but before applying it in empirical exercises, we would like to confirm that there is a single positive equilibrium and that it is locally stable. We can investigate local stability using a simplified version of the model. Consider a population with only two life stages, sub-adults and adults (here denoted by subscripts S and A), where survival from the sub-adult stage to the adult stage is density dependent:

NS,t+1 ) RNA,te-ZA

Because individuals that survive a year in the terminal life stage remain in that stage, we have

NK,t+1 ) NK-1,te-ZK-1 + NK,te-ZK

(5)

Next, the number of eggs (the first life stage, k ) 1) at the beginning of year t + 1 is the sum of eggs laid by individuals in all reproductive life stages that survive to the end of year t:

NA,t+1 )

-ZS b - (ZS/NS,t + b)eZS

∑w f N k k

k,te

-Zk

(6)

k)1

where R is the average number of eggs laid per pound of spawning adults, wk is the average weight in pounds of individuals in life stage k, and fk is an indicator variable equal to 1 for reproductive life stages and 0 otherwise. Thus, reproduction is modeled as a pulse that occurs at the end of the year, and fecundity is assumed to be a function of the weight of spawning adults rather than age per se. The annual transition matrix for this model when the population size is small (i.e., ignoring density dependence) and excluding entrainment and impingement and fishing mortality is: where sk ) e-Zk and all other parameters are as

NA,t+1 )

K

Ht )

∑ w N (1 - e k

k

-Zk

)Fk/Zk

(8)

k)1

2110

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 41, NO. 7, 2007

+ NA,te-ZA

(11)

-ZS b - (ZS/(RNA,te-ZA) + b)eZS

+ NA,te-ZA (12)

Equation 12 retains the key features of the more general model outlined in the previous section, but allows us to investigate local stability in a more straight forward manner. Setting NA,t+1 ) NA,t, we find that there are two equilibrium states, N ˆ A ) 0 and

N ˆA )

ZSeZA(R - eZS(eZA - 1))

(13)

bR(eZS - 1)(eZA - 1)

For the population to persist the reproductive rate must be larger than a minimum value determined by the combined influences of mortality in both the sub-adult and adult stages; specifically, R > Rmin ) eZS(eZA - 1). Differentiating eq 12 and evaluating the resulting expression at the equilibrium and simplifying, we find that the eigenvalue is

λeq ≡ defined above. This is a standard Leslie matrix, and is useful for estimating the population growth rate, the stable age distribution, and population stability characteristics and return times (9, 25). The dominant eigenvalue of this matrix calculated at low abundance, here labeled λ0, is the intrinsic population growth rate and will play a key role in calibrating the model, as described in Section 2.3 below. Finally, the total harvested biomass in year t, Ht, and the number of individuals killed by entrainment and impingement in each density independent age class in year t, Lk,t, are

(10)

R sub-adults per adult are produced at the end of each time period, and survival of adults is density independent. We focus on the adult stage by combining eqs 10 and 11 to give a single equation:

K

N1,t+1 ) R

(9)

(

)

∂NA,t+1 eZS(eZA - 1)2 -ZA |N ˆA ) 1 + e ∂NA,t R

(14)

All terms are positive, so λeq > 0. Plugging the expression for Rmin into eq 14 gives λeq(Rmin) ) 1. It also is apparent upon inspection that if R > Rmin, then λeq < 1: λeq(R > Rmin) ) e-ZA + (Rmin/R)(1 - e-ZA) < 1. (Another way of seeing this is to note that ∂λeq/∂R < 0. Thus, as R grows larger than Rmin, λeq will diminish.) This establishes that for viable populations λeq will be between 0 and 1 always, which is the condition for local stability: after a small perturbation the population will return to the equilibrium geometrically (pp 111-118 in 27). λeq also provides an estimate of the return times to the equilibrium from small temporary perturbations in N ˆ A: the time required to achieve an X% return to equilibrium is approximately TR ) ln(1 - X/100%)/ln λeq (9). In the applications to follow, we estimate changes in both N ˆ A and

95% return times due to elimination of entrainment and impingement mortality for each fish stock. 2.3. Calibration. To implement the model, values for Mk, Fk, wk, and the recruitment rate at low abundance for each species are taken from government reports and the primary literature. R, b, and all Ek’s are estimated by calibration. To calibrate R, we use a basic result from matrix algebra: R is set such that the dominant eigenvalue of the annual transition matrix calculated at low abundance, λ0, matches the intrinsic population growth rate (e.g., see ch. 3 in ref 28). We estimate the intrinsic population growth rate using values for the recruitment rate at low abundance and the life cycle mortality schedule for each species. Specifically, R is set such that the average annual growth rate, NA,t+1/NA,t, equals λ0, where NA,t ) a˜NA,t-am + NA,t-1e-MA, a˜ is the recruitment rate at low abundance (29), am is the age of maturation in years, and MA is the adult mortality rate. To ensure a stable age distribution, the first 25 years of the series are excluded when calculating the average annual growth rate. Next, the parameter b, which controls the strength of density-dependence in the natural mortality rate among the sub-adult life stages and determines the carrying capacity for the species, is calibrated so the equilibrium harvest level predicted by the model matches the average historic harvest level. Finally, the Ek’s are calibrated so the entrainment and impingement levels predicted by the model match the observed levels of annual entrainment and impingement losses for each life stage. After the model is calibrated to baseline conditions, policy scenarios can be simulated by changing the Ek’s according to the expected reductions in entrainment and impingement rates from the management options under consideration. The model can then be used to forecast the population forward in time and will reequilibrate with a new steady-state stock size, harvest level, and set of entrainment and impingement losses conditional on the new (hypothetical) entrainment and impingement mortality rates. In this paper, we simulate the total elimination of entrainment and impingement for each stock. This places an upper bound on the potential population increases from controlling cooling water withdrawals. 2.4. Applications. In this section, we apply the population model using estimates of annual entrainment and impingement losses, fishing mortality rates, and harvest levels for 15 fish stocks in California and the Atlantic regions. These species constitute all of the commercially and recreationally harvested species known to be exposed to cooling water withdrawals for which sufficient life history information has been compiled for EPA’s 316(b) Phase II regulatory assessment. These parameters were collected from a variety of sources including the peer reviewed literature, reports by regulated facilities, and FishBase, a publicly available database containing ecological information on hundreds of fish species worldwide (1, 30, 31). We also use estimates of entrainment and impingement losses previously calculated by EPA, mainly from reports produced by regulated facilities to satisfy permitting requirements. Only a small subset of all facilities monitor entrainment and impingement losses, so losses at non-reporting facilities were estimated by extrapolation based on the ratios between the average annual intake flow rates at the reporting and non-reporting facilities in the same regions. Because of the potential for measurement error in the extrapolated entrainment and impingement losses, we conduct a sensitivity analysis for each species using entrainment and impingement levels 50% less and greater than the baseline estimates. Reproductive rates for each species were adapted from Myers et al. (29), who used a meta-analysis approach based on a standardized Ricker model to estimate recruitment rates at low abundance, a˜, for 57 fish species. By calibrating R such that λ0 is consistent with a˜ and the life cycle mortality

schedule for each species, as described above, we are able to use all available life-stage-specific data on natural mortality, fishing mortality, and entrainment and impingement losses while ensuring that the model is consistent with the best available information on each species’ maximum reproductive rate. To evaluate the sensitivity of the results to uncertainty in a˜, we calculate the mean and standard deviation of the distribution of percent population increases from eliminating entrainment and impingement, assuming a normal distribution for ln(a˜) with a mean and standard deviation as reported in Myers et al. (29), with all other parameters fixed at their baseline values. Historic data on commercial and recreational harvest levels were obtained from NOAA’s Annual Commercial Landings Statistics and Marine Recreational Fisheries Statistics. NOAA collects and maintains data on harvest levels by month and region across the U.S. for all major harvested species. Data on commercial harvests are available for all years since 1950, and data on recreational harvests are available for all years since 1981. Because recreational harvest data are not available prior to 1981, we impute values for recreational harvests using a simple linear regression between recreational and commercial harvests between 1981 and 2002. We use the average total (commercial plus recreational) harvest level during the span of years when entrainment and impingement loss data were collected to calibrate the model under baseline conditions for each species. To evaluate the importance of uncertainty in the appropriate average historic harvest level used to calibrate the model for each species, we conduct a sensitivity analysis using the average historic harvest levels (between 1970 and 1990 for most stocks) plus and minus 1.5 standard deviations. Table 2 lists the key life history parameters and average historic harvest levels used for calibrating the population model for each stock.

3. Results and Discussion Table 3 shows the main results, including the estimated entrainment and impingement mortality rates for the subadult life stages (ES); 95% return times (TR) and adult equilibrium population sizes (N ˆ A) under baseline conditions, without entrainment and impingement, and without entrainment and impingement or fishing; and percent changes in adult equilibrium stock sizes from elimination of entrainment and impingement (%∆N ˆ A). Estimated entrainment and impingement mortality rates are highly variable across stocks, spanning several orders of magnitude. However, for most stocks elimination of entrainment and impingement mortality alone leads to virtually no change in population stability as measured by the 95% return time. The only exceptions are for the striped bass stocks, whose return times are modestly reduced. On the other hand, the model predicts that most stocks will become less stable (return times will increase) if fishing mortality is eliminated. Thus, the effects of reduced mortality on population stability are ambiguous. These results are consistent with Newbold and Iovanna (32), who showed (using a simplified model based on eq 12) that a reduction in sub-adult mortality generally will not decrease stability but a reduction in adult mortality may decrease stability if the reproductive rate is high enough relative to the combined sub-adult and adult mortality rates (also see refs 13, 33). Estimated changes in stock sizes are less than 1% in 10 of the 15 cases, between 1 and 3% in two cases, and between 20 and 80% in three cases. In contrast to the mostly negligible estimated effects on population stability, these changes correspond closely to the estimated entrainment and impingement mortality rates. The simple correlation coefficient between the sub-adult entrainment and impingement mortality rates and the percent changes in equilibrium adult abundances is r ) 0.99. This close correspondence is most VOL. 41, NO. 7, 2007 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

2111

TABLE 2. Key Life History Parameters and Average Historic Harvest Levels (∼1970-1990) Used to Calibrate the Population Model for Each Stock MS

stock

American shad, California American shad, Atlantic Anchovy, California Atlantic cod, Atlantic Atlantic croaker, Atlantic Atlantic herring, Atlantic Atlantic mackerel, Atlantic Atlantic menhaden, Atlantic Pollock, Atlantic Scup, Atlantic Silver hake, Atlantic Striped bass, California Striped bass, Atlantic Summer flounder, Atlantic Winter flounder, Atlantic

10.907 10.907 9.521 11.617 12.297 9.884 12.986 6.950 11.926 9.337 12.630 9.860 13.095 7.038 9.436

MA 0.870 0.870 0.700 0.233 0.399 0.200 0.381 0.450 0.200 0.383 0.400 0.320 0.213 0.200 0.302

˜a

R (eggs/lb)

am (years)

H (metric tons/ year)

1.586 1.586 1.886 4.751 7.289 2.891 3.722 3.337 1.597 4.023 2.024 1.495 1.551 1.948 1.637

4.864 × 4.864 × 105 1.611 × 106 4.025 × 107 1.141 × 108 1.690 × 106 8.277 × 107 1.074 × 105 3.628 × 105 3.532 × 106 1.108 × 107 4.229 × 104 1.030 × 106 3.651 × 103 3.087 × 105

4 4 1 1 1 1 1 3 4 2 2 4 4 2 3

9.36 1 207.08 2 041.59 39 243.22 3 266.71 41,961.55 2 715.49 280 903.66 11 130.53 4 389.76 12 572.31 158.59 3 483.21 6 483.15 10 994.24

105

TABLE 3. Estimated Entrainment and Impingement Mortality Rates for Sub-adult Life Stages, 95% Return Times, Equilibrium Population Sizes, and Percent Changes in Population Sizes on Elimination of Entrainment and Impingement Mortality for 15 Harvested Stocks off the California and Atlantic Coasts of the U.S.

stock

American shad, California American shad, Atlantic Anchovy, California Atlantic cod, Atlantic Atlantic croaker, Atlantic Atlantic herring, Atlantic Atlantic mackerel, Atlantic Atlantic menhaden, Atlantic Pollock, Atlantic Scup, Atlantic Silver hake, Atlantic Striped bass, California Striped bass, Atlantic Summer flounder, Atlantic Winter flounder, Atlantic

ES 10-4

1.764 × 9.818 × 10-3 4.635 × 10-4 1.673 × 10-6 5.695 × 10-1 1.395 × 10-4 2.723 × 10-4 1.431 × 10-2 2.760 × 10-6 9.753 × 10-6 4.441 × 10-5 1.381 × 10-1 1.775 × 10-1 1.876 × 10-3 4.565 × 10-3

TR (95% return time)

N ˆA

without ent. without ent. and imp. or baseline and imp. fishing

baseline

without ent. and imp.

2.409 × 1.077 × 106 2.603 × 109 3.694 × 108 3.504 × 107 2.453 × 109 3.203 × 107 4.406 × 108 1.634 × 107 2.698 × 107 1.211 × 108 2.146 × 105 2.515 × 106 1.660 × 107 3.936 × 107

2.410 × 1.089 × 106 2.606 × 109 3.694 × 108 6.288 × 107 2.453 × 109 3.204 × 107 4.535 × 108 1.634 × 107 2.698 × 107 1.211 × 108 2.790 × 105 3.090 × 106 1.671 × 107 3.956 × 107

7.81 7.95 7.37 6.65 7.95 12.72 7.81 6.79 13.73 3.47 6.51 13.01 14.02 10.84 11.28

7.81 7.95 7.37 6.65 7.95 12.72 7.81 6.79 13.73 3.47 6.51 11.85 13.15 10.84 11.28

apparent for Atlantic croaker and the striped bass stocks. These three stocks have the highest estimated entrainment and impingement mortality rates and the highest estimated percentage changes in equilibrium stock sizes. The magnitude of the estimated changes in equilibrium abundances for these stocks may seem surprising at first since, even in these cases, the estimated entrainment and impingement mortality rates are relatively small fractions of the natural mortality rates (compare the ES values in Table 3 to the corresponding MS values in Table 2). However, it turns out that, for a wide range of biologically plausible life history parameter values, the proportional changes in equilibrium stock sizes are on the same order of magnitude as the sub-adult entrainment and impingement mortality rates. (This can be shown via algebraic manipulation of eq 13. See Appendix 3 in the Supporting Information for details.) Table 4 shows the sensitivity of the estimated populationlevel impacts to several key structural and parameter uncertainties, and Figure 1 summarizes the sensitivity analysis results. The first column of numbers in Table 4 contains the baseline estimates of population changes carried over from Table 3. To assess the sensitivity of these estimates to uncertainty about when in the life cycle density dependence operates most strongly, we re-estimated the model under two extreme assumptions for each stock: first where density dependence operates only in the egg stage (k ) 1), and second where density dependence operates only in the 2112

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 41, NO. 7, 2007

7.37 7.37 7.37 15.32 10.41 15.61 8.82 7.81 17.06 8.53 9.25 12.72 21.83 16.34 17.06

104

104

without ent. and imp. or fishing

%∆N ˆA

2.896 × 0.032 1.406 × 106 1.103 9 2.753 × 10 0.091 7.216 × 1008 0.000 8.414 × 107 79.444 3.078 × 109 0.011 3.835 × 107 0.020 2.652 × 109 2.934 8.912 × 107 0.000 6.418 × 107 0.001 2.683 × 108 0.005 4.940 × 105 30.009 1.139 × 107 22.853 5.375 × 107 0.684 1.344 × 108 0.508 104

last stage prior to sexual maturity (k ) km - 1). In general, the earlier in the life cycle density dependence operates, the larger the estimated impact of entrainment and impingement on the adult population. This occurs because density dependent survival will partially offset the detrimental effect of entrainment and impingement of previous life stages but not later stages. If density dependence operates late enough in the life cycle and entrainment and impingement only or predominantly affects earlier life stages, then cooling water withdrawals may have a negligible effect on adult abundance (13). The next two columns show the sensitivity of the results to uncertainty in the recruitment parameter adapted from Myers et al. (29). The results are relatively robust to uncertainty in this parameter. The mean of the estimated distribution is nearly identical to the baseline estimates, and for most species, the standard deviation is a small fraction of the baseline estimates. The next two columns show the sensitivity of the results to the estimated overall entrainment and impingement losses. Since average annual entrainment and impingement losses for each species were estimated using data from only a few facilities and extrapolated to the remainder of the facilities in each region, there is potential for substantial measurement error in this variable. For most species, the predicted population increases are nearly proportional to the extrapolated entrainment and impingement losses. The final two columns show the sensitivity of the results to the average historic harvest levels used to

TABLE 4. Sensitivity of Estimated Population Level Effects to Key Model and Parameter Uncertaintiesa sensitivity to density dependent stage stock

%∆N ˆA

k)1

American shad, California American shad, Atlantic Anchovy, California Atlantic cod, Atlantic Atlantic croaker, Atlantic Atlantic herring, Atlantic Atlantic mackerel, Atlantic Atlantic menhaden, Atlantic Pollock, Atlantic Scup, Atlantic Silver hake, Atlantic Striped bass, California Striped bass, Atlantic Summer flounder, Atlantic Winter flounder, Atlantic

0.032 1.103 0.091 0.000 79.444 0.011 0.020 2.934 0.000 0.001 0.005 30.009 22.853 0.684 0.508

0.034 1.184 0.119 0.026 79.645 0.048 0.106 2.978 0.001 0.004 0.011 51.137 23.456 0.690 1.218

a

k ) km - 1 0.006 0.028 0.051 0.000 1.452 0.001 0.000 1.437 0.000 0.000 0.001 6.258 0.816 0.474 0.058

sensitivity to ln(˜a) mean

st. dev.

0.0324 1.105 -0.005 79.492 0.011 -2.939 0.001 0.001 0.005 30.233 22.956 0.688 0.517

0.001 0.026 -0.006 0.236 0.002 -0.055 0.000 0.000 0.000 2.099 0.808 0.027 0.067

sensitivity to estimated ent. and imp. losses 0.5L0 0.016 0.552 0.046 0.002 41.709 0.005 0.010 1.459 0.000 0.001 0.003 14.624 11.470 0.342 0.255

sensitivity to average historic harvest levels -σH

1.5L0 0.049 1.654 0.137 0.007 116.212 0.016 0.030 4.424 0.001 0.002 0.008 45.665 33.770 1.028 0.762

1.657 1.628 0.159 0.007 164.186 0.017 0.038 3.386 0.001 0.002 0.007 55.423 126.698 0.869 0.753

+σH 0.016 0.834 0.064 0.004 52.502 0.008 0.014 2.588 0.000 0.001 0.004 20.580 12.614 0.564 0.384

Note: “--” cell entries indicate that the population was not viable at the low end of the ln(a¯) sampling distribution.

FIGURE 1. Estimated range of potential population level impacts of cooling water withdrawals from 316(b) Phase II regulated facilities on 15 fish stocks in California and the Atlantic regions. Each bar shows the baseline estimate, and whiskers show the range of effects estimated effects from the sensitivity analyses shown in Table 4. calibrate the model for each stock. We suspect that annual harvest levels are measured relatively accurately compared to the other model inputs (with the caveat that pre-1981 estimates are relatively more uncertain owing to our extrapolation from commercial to recreational harvest levels as noted above), but there is substantial uncertainty about what averaging period is most appropriate for each species and about the possible structural changes in economic or ecological conditions in each fishery that may have occurred over this time span. The differences in the estimated values of %∆N ˆ A from this sensitivity analysis typically are of the same order of magnitude as those from uncertainty in the entrainment and impingement losses, but somewhat larger. Long term forecasts of dynamic ecological phenomena generally are highly uncertain, but forecasts that report useful information sometimes are possible (34). Because of the unavoidable data limitations and simplifying assumptions involved in this analysis, we have presented a range of estimates based on a variety of assumptions about key structural uncertainties (when in the life cycle density dependence operates) and parameter uncertainties (recruit-

ment rates at low abundance, entrainment and impingement losses, and associated average harvest levels). Ideally, longer and more complete time series data for all exposed life stages from all facilities and all species of concern would be available. But until a standardized protocol for collecting entrainment and impingement data is implemented on a consistent basis at more facilities, policy evaluations must be based on the best available data, however incomplete. As with any modeling exercise, the usefulness of the results depends on both the confidence placed on the model predictions and the level of accuracy that the decision maker deems acceptable. With these caveats in mind, we view this as a “screeninglevel” assessment (e.g., refs 35, 36) to generate plausible orderof-magnitude estimates of population level effects on fish stocks at a large geographic scale. It should be possible to reduce the uncertainty in these estimates using more detailed site-specific information, but screening assessments such as this can be used to provide preliminary policy guidance and to identify species and regions on which to focus future research and protection efforts. Based on the results for the 15 stocks analyzed here, the effects of cooling water withdrawals appear to be minor for most harvested fish stocks but may be severe for a few. (However, that the overall impacts for some stocks appear minor at the large geographic scales considered here does not rule out the possibility of higher impacts in localized areas.) Estimated changes in equilibrium stock sizes if cooling water withdrawals are completely eliminated were estimated to be less than 1% in 10 of the 15 cases but between 20 and 80% in three cases. Further, the results appear robust for those stocks with apparently minor impacts but are more uncertain for those stocks with apparently severe impacts. No stock for which the baseline results suggest entrainment and impingement impacts are negligible show significantly larger effects under any of these sensitivity analyses shown in Table 4, but the estimates for the three most severely affected stocks span a wide range, from negligible or minor to what may amount to the brink of collapse. In light of these sensitivity analyses, our overall impression remains that the population level impacts of cooling water withdrawals may be negligible for many fish stocks but could be severe for a few.

Acknowledgments All information on entrainment and impingement losses used here but not appearing in USEPA (1) were provided by the VOL. 41, NO. 7, 2007 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

2113

EPA Office of Water. However, the authors alone are responsible for the use and interpretation of the data and any remaining errors. The views expressed in this paper are those of the authors and do not necessarily represent those of the U.S. EPA. No official Agency endorsement should be inferred. Note Added after ASAP Publication. The assignment of copyright for the version published ASAP on February 20, 2007 was in error; the corrected version was published ASAP March 12, 2007.

Supporting Information Available A pdf file containing an alphabetical list of all parameters used in the population model with definitions, graphs of historic harvest data from NOAA used to calibrate the model for each species, and further details on the relationship between the estimated sub-adult entrainment and impingement mortality rates and changes in equilibrium adult abundances; an Excel file on life history parameters for Atlantic menhaden; and text files and Matlab code containing all input data and programs needed to reproduce the results presented in the main text. This material is available free of charge via the Internet at http://pubs.acs.org.

Literature Cited (1) USEPA. Regional Analysis Document for the Final 316(b) Phase II Existing Facilities Rule; U.S. Environmental Protection Agency, Office of Water: Washington, DC, 2003. (2) National pollutant discharge elimination systemsFinal regulations to establish requirements for cooling water intake structures at phase II existing facilities; Fed. Regist. 2004, 69, 41576-41693. (3) Risk Assessment Forum. Framework for Cumulative Risk Assessment; U.S. Environmental Protection Agency: Washington, DC, 2003. (4) Risk Assessment Forum. Generic Ecological Assessment Endpoints (GEAEs) for Ecological Risk Assessment; U.S. Environmental Protection Agency: Washington, DC, 2003. (5) Landis, W. G. Population is the appropriate biological unit of interest for a species-specific risk assessment. SETAC Globe 2002, 3, 31-32. (6) Munns, W. R. J.; Beyer, W. N.; Landis, W. G.; Menzie, C. What is a population? SETAC Globe 2002, 3, 29-31. (7) USEPA. Case Study Analysis for the Proposed Section 316(b) Phase II Existing Facilities Rule; USEPA Office of Water: Washington, DC, 2002. (8) Proceedings of the Conference on Assessing the Effects of Power Plant Induced Mortality on Fish Populations; Van Winkle, W., Ed.; Pergamon: Elmsford, NY, 1977. (9) DeAngelis, D. L.; Svoboda, L. J.; Christensen, S. W.; Vaughan, D. S. Stability and return times of Leslie matrices with densitydependent survival: applications to fish populations. Ecol. Modell. 1980, 8. (10) Vitousek, P. M.; Mooney, H. A.; Lubchenco, J.; Melillo, J. M. Human domination of Earth’s ecosystems. Science 1997, 277, 494-499. (11) Jackson, J. B. C.; Kirby, M. X.; Berger, W. H.; Bjorndal, K. A.; Botsford, L. W.; Bourque, B. J.; Bradbury, R. H.; Cooke, R.; Erlandson, J.; Estes, J. A.; Hughes, T. P.; Kidwell, S.; Lange, C. B.; Lenihan, H. S.; Pandolfi, J. M.; Peterson, C. H.; Steneck, R. S.; Tenger, M. J.; Warner, R. R. Historical overfishing and the recent collapse of coastal ecosystems. Science 2001, 293. (12) Science, Law, and Hudson River Power Plants: A Case Study in Environmental Impact Assessment; Barnthouse, L. W., Klauda, R. J., Vaughan, D. S., Kendall, R. L., Eds.; American Fisheries Society: Bethesda, MD, 1988; Vol. 4. (13) Nisbet, R. M.; Murdoch, W. M.; Stewart-Oaten, A. Consequences for adult fish stocks of human-induced mortality on immatures. In Detecting Ecological Impacts: Concepts and Applications in Coastal Habitats; Schmitt, R. J., Osenberg, C. W., Eds.; Academic Press: San Diego, CA, 1996; pp 257-277. (14) Jensen, A. L.; Spigarelli, S. A.; Thommes, M. M. Use of conventional fishery models to assess entrainment and impingement of three Lake Michigan fish species. Trans. Am. Fish. Soc. 1982, 111, 21-34. 2114

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 41, NO. 7, 2007

(15) Barnthouse, L. W.; Heimbuch, D. G.; Anthony, V. C.; Hilborn, R. W.; Myers, Sci. World J. 2002, 2, 169-189. (16) Barnthouse, L. W.; Van Winkle, W. Analysis of impingement impacts on Hudson River fish populations. Am. Fish. Soc. Monogr. 1988, 4, 182-190. (17) Christensen, S. W.; DeAngelis, D. L.; Clark, A. G. Development of a stock-progeny model for assessing power plant effects on fish populations. In Proceedings of the Conference on Assessing the Effects of Power Plant Induced Mortality on Fish Populations; Van Winkle, W., Ed.; Pergamon Press: Elmsford, NY, 1977. (18) Swartzman, G.; Deriso, R.; Cowan, C. Comparison of simulation models used in assessing the effects of power-plant-induced mortality on fish populations. In Proceedings of the Conference on Assessing the Effects of Power Plant Induced Mortality on Fish Populations; Van Winkle, W., Ed.; Pergamon: Elmsford, NY, 1977. (19) Jensen, A. L. Impact of a once-through cooling system on the yellow perch stock in the western basin of Lake Erie. Ecol. Modell. 1982, 15, 127-144. (20) Jensen, A. L.; Hamilton, T. A. Application of a conventional fishery model for assessment of entrainment and impingement impact. Environ. Biol. Fishes 1982, 7, 181-185. (21) Gurney, W. S. C.; Nisbet, R. M. Ecological Dynamics; Oxford University Press: New York, 1998. (22) Dixon, D. A.; Zammit, K. D. Preface to the supplemental issue. Environ. Sci. Policy 2000, 3, vii-viii. (23) Barnthouse, L. W. Impacts of power-plant cooling systems on estuarine fish populations: the Hudson River after 25 years. Environ. Sci. Policy 2000, 3, S341-S348. (24) Van Winkle, W. A perspective on power generation impacts and compensation in fish populations. Environ. Sci. Policy 2000, 3, S425-S431. (25) Leslie, P. H. On the use of matrices in certain population mathematics. Biometrika 1945, 33, 183-212. (26) Ricker, W. E. Computation and interpretation of biological statistics of fish populations. Fish. Res. Board Can., Bull. 191. 1975. (27) Murdoch, W. W.; Briggs, C. J.; Nisbet, R. M. Consumer-Resource Dynamics; Princeton University Press: Princeton, NJ, 2003. (28) Roughgarden, J. Primer of Ecological Theory; Prentice Hall: Upper Saddle River, NJ, 1998. (29) Myers, R. A.; Bowen, K. G.; Barrowman, N. J. Maximum reproductive rate of fish at low population sizes. Can. J. Fish. Aquat. Sci. 1999, 56, 2404-2419. (30) Quinlan, J. A.; Crowder, L. B. Searching for sensitivity in the life history of Atlantic menhaden: inferences from a matrix model. Fish. Oceanogr. 1999, 8, 124-133. (31) Froese, R.; Pauly, D.; version (09/2004) ed., 2004. (32) Newbold, S. C.; Iovanna, R. Potential effects of cooling water withdrawals on fish populations and aquatic ecosystems. Ecol. Appl. 2006, in preparation. (33) Mertz, G.; Myers, R. A. An augmented Clark model for stability of populations. Mathematical Biosciences 1996, 131, 157-171. (34) Clark, J. S.; Carpenter, S. R.; Barber, M.; Collins, S.; Dobson, A.; Foley, J. A.; Lodge, D. M.; Pascual, M.; Pielke, R. J.; Pizer, W.; Pringle, C.; Reid, W. V.; Rose, K. A.; Sala, O.; Schlesinger, W. H.; Wall, D. H.; Wear, D. Ecological forecasts: an emerging imperative. Science 2001, 293, 657-660. (35) Hill, R. A.; Chapman, P. M.; Mann, G. S.; Lawrence, G. S. Level of detail in ecological risk assessments. Mar. Pollut. Bull. 2000, 40, 471-477. (36) USEPA. The Role of Screening-Level Risk Assessments and Refining Contaminants of Concern in Baseline Ecological Risk Assessments; USEPA, Office of Solid Waste and Emergency Response: Washington, DC, 2001. (37) Savidge, I. R.; Gladden, J. B.; Campbell, K. P.; Ziesenis, J. S. Development and sensitivity analysis of impact assessment equations based on stock-recruitment theory. Am. Fish. Soc. Monogr. 1988, 4, 191-203. (38) Lawler, J. P. Some consideration in applying stock-recruitment models to multiple-age spawning populations. Am. Fish. Soc. Monogr. 1988, 4, 204-215. (39) Goodyear, C. P. Implications of power plant mortality for management of the Hudson River striped bass fishery. Am. Fish. Soc. Monogr. 1988, 4, 245-254.

Received for review April 4, 2006. Revised manuscript received January 9, 2007. Accepted January 12, 2007. ES060812G