*3.2. Determination of the Probability of Survival Sj*

An important part of the EAM model is the estimation of natural survival rate, Sj. The original Goodyear [7] paper uses a complex method based on fecundity and cumulative catch data segregated into separate length classes. Others have segregated the data into day-age classes, cohorts (groups of fish spawned within a similar time frame or of like age) or life stages (yolk-sac or prolarvae, post yolk-sac larvae, juveniles, *etc.*).

Another method, and one used locally in the Salem PSEG 316(b) demonstration as part of their NPDES permit application report [10], is to compute Sj from the mortality rate of each successive life stage through to the adult stage. A limitation of this approach is the varying efficiencies of sampling used to estimate populations of successive lifestages of species, but the survival through an individual life stage can be computed as:

$$f\_l = \exp^{-z\_l T\_l} \tag{7}$$

where *fl* is the probability of survival through life stage *l* (fraction survived at the end of the life stage), and *l* = an index for the individual life stage; e.g., *l* = 1 is eggs, *l* = 2 is larvae, *l* = 3 is juveniles *etc.* The *zl* is the natural mortality rate (1/day) for life stage *l*, and *Tl* is the length (days) of life stage *l*. *Sn <sup>j</sup>* the product of the probability of survival from life stage *l* through all successive life stages to the adult life stage, then can be given by:

$$S\_{\prime\prime}^{n} = \prod\_{l=j}^{n} f\_{l} \tag{8}$$

where "*n*" is the final life stage. Equation (8) can also be written as:

$$\mathbf{S}\_{\rangle}^{u} = \mathbf{exp}^{\sum\_{l=j}^{s} T\_{l}} \tag{9}$$

The sum within Equation (9) goes from life stage j through all life stages to the adult stage, n, even though entrainment does not take place beyond a certain life stage. This varies by species, but was assumed to be through the post-yolk larval stage. For the computations in Equation (6) only the densities up through the entrainable life stages should be used.

#### *3.3. Percent of Population Entrained*

Entrainment impact assessment compares the number of fishes entrained through and lost to the intake against the population at risk. A common limitation in extrapolating the IP losses to equivalent adults in many of these analyses is the lack of good fisheries data for the broader system feeding IP into the region potentially affected by the intake. In fact, locally, comments made in the Salem report [10] were that the EAM was not used extensively because there were no fishery data or other information available for comparative purposes. This limitation can be overcome, in part, using the extended model with weekly regional sampling. Examination of Equation (6) shows that the adult population within the sampled regions that could have survived in the absence of the entrainment from the ballast intake over the different life stages is:

$$N\mathbf{a}\_r = \sum\_i \sum\_k \left\{ \sum\_j D\_{i,j,k} \mathbf{S}\_j^n \right\} V\_{i,k} \tag{10}$$

Thus, the percent entrainment of equivalent adults from the full population of adults that might have been entrained from those regions becomes:

$$\%Entrained = \bigvee\_{Na\_T} \* 100\tag{11}$$

The percent entrained is determined only from the potential adults that could have been entrained from the surveyed regions, and not to the total fisheries population produced for the whole waterbody that may include un-sampled regions. Thus, in that case, Equation (11) will give a larger percentage entrained than if the total fisheries of the whole waterbody (entire waterbody sampled) population were known.

#### *3.4. Entrainment Mortality*

The above formulation includes the implicit assumption that all the species are entrained at the beginning of the life stage, in that it does not account for natural mortality that occurs on that life stage. This assumption, however, is invalid *per se*, as both the natural mortality and entrainment mortality occur simultaneously. The assumption of all the entrainment occurring at the beginning of the life stage is therefore conservative; it overestimates the effect of entrainment mortality on the population by not accounting for the natural mortality that would have reduced that same population over the period for which they were entrained. To decrease this overestimate, an entrainment function is required. This entrainment function depends on the entrainment mortality.

Entrainment mortality is defined as the rate of entrainment from each region due to the intake operations and is treated similar to the natural mortality. For instance, if there is no natural mortality, then the survival rate for species can be estimated by a formula similar to the one given by Equation (9). Equation (12) below shows the entrainment mortality which depends only on the intake operations and the regional hydrodynamics.

$$\text{Entrainment Mortality} = \frac{\underline{Q}\_{\text{lu}}}{V\_T} \tag{12}$$

where *QIn* is the intake flow rate and *VT* is the total volume of the waterbody disturbed. In cases where the waterbody is divided in to "k" regions, the entrainment mortality is given by Equation (13).

$$\text{Entrainmen t Mortality} = \frac{\underline{Q}\_{\text{lu}}}{V\_k} \ast F\_{i,k} = m\_{i,k} \tag{13}$$

where *Fi,k* is defined as the ratio of the number of individuals entrained at the intake from region "k" during week i to the total number of individuals entrained from all regions and is shown in Equation (14).

$$F\_{i,k} = \frac{E\_{i,k}}{E\_n} \tag{14}$$

The survival function due to the entrainment is then given by Equation (15).

$$S\_{i,f,k}^{m} = \exp^{\sum\_{l=j}^{s} r\_l} \tag{15}$$

The number of equivalent adults entrained then can be expressed by Equation (16).

$$\mathbf{Method 2: } Na = \sum\_{i} \sum\_{k} \left\{ \sum\_{j} D\_{i,j,k} S\_j^m (1 - S\_j^m) \right\} FE\_{i,k} V\_{i,k} \tag{16}$$

#### *3.5. EPRI Adjusted Entrainment Survival Function*

The above formulation applies the entrainment mortality at the end of the life stage, and thus has tendency to under-predict the entrained population. When day-age class population size and survival functions by day-age class are unavailable, the Electric Power Research Institute [8] suggests that the entrainment mortality should be applied assuming that half the natural mortality

has occurred (e.g., the middle of the life stage). At this point, the IP population at each life stage has decreased due to the natural mortality and thus more accurately attributes losses due to entrainment. When the entrainment mortality is applied at this half-life-stage, the survival function due to entrainment is given by EPRI [8]:

$$S\_{\ll,k}^{\ast} = \mathbb{Z}^{\star} S\_{\ll}^{n} \ast \exp^{-(1+\ln(S\_{\ll}^{a}))}\tag{17}$$

The number of equivalent adults entrained then can be expressed by Equation (18).

$$\textbf{Method 3: } Na = \sum\_{i} \sum\_{k} \left\{ \sum\_{j} D\_{i,j,k} S\_j''(\mathbf{l} - \mathbf{S}''\_{~j}) \right\} FE\_{i,k} V\_{i,k} \tag{18}$$

#### *3.6. Exponent, Inc.-Adjusted Entrainment Survival Function*

Exponent, Inc. was commissioned by the Center for Liquefied Natural Gas (CLNG) to investigate and critique the technical work adopted to estimate the entrainment impacts in Open Loop Vaporization (OLV) systems, which were commonly proposed in warmer open water (offshore) LNG revaporization designs. Exponent, Inc. found that the assumption of half natural mortality used in EPRI's method still makes the entrainment mortality conservative since the natural mortality survival function is an exponentially decreasing term. This results in an overestimation of the entrainment impacts. To further improve the entrainment survival function, Exponent, Inc. suggested an alternate relationship which defines the survival function at the end of each life stage as given by Equation (19) [9]:

$$S^{\star\star}\_{\ \ i,k} = \frac{S^n\_{\ \ j} \left\{ \frac{S^m\_{\ i,k}}{m} + T\_{\ \ i} - \frac{1}{m} \right\}}{S^n\_{\ \ j} \ast S^m\_{\ i,k} \left\{ \frac{1}{k+m} - \frac{1}{k \ast S^n\_{\ \ j}} \right\} - \left\{ \frac{1}{k+m} - \frac{1}{k} \right\}} \tag{19}$$

The survival function is continuous within each stage as opposed to at the half-life stage. This further corrects the estimate of survivability by discretizing the continuous effects of natural and entrainment mortality into smaller steps (life stages). The number of equivalent adults entrained then can be expressed by Equation (20).

$$\mathbf{Method 4: } Na = \sum\_{i} \sum\_{k} \left\{ \sum\_{j} D\_{i,j,k} S\_j^{\ast} (1 - S^{\ast \ast}\_{\ \ \ \ \cdot}) \right\} E\_{i,k} V\_{i,k} \tag{20}$$

#### **4. Case Study**

The case study used to demonstrate the entrainment computations and study the possible fish population-level impacts is vessel ballasting in an estuarine system. The site is located in the Delaware River Estuary where concerns were raised about the potential negative effects on the population of ichthyoplankton from entrainment during ballast water intake at the terminal berth and also the entrainment risk to sturgeon presented by the ballast intake. A 138,000 m3 LNG carrier, the design vessel for this study, will withdraw approximately 8 million gallons [11] of ballast water while at berth. The operations assume that one LNG carrier will call at the project every two to three days, resulting in approximately 150 ship calls per year or 12 per month. For the purposes of this study, the design intake rate of 660,430 gallons per hour and the design reballasting period of 12 h were selected. Figure 2 shows the site location in New Jersey near the state lines of Pennsylvania and Delaware.

#### **Figure 2.** Site location.

#### *4.1. Ichthyoplankton*

Striped bass, white perch, clupeidae (river herring and American shad), and bay anchovy were the species and groups of concern since they consistently dominate the IP population in the project area. Mean IP densities between the months of April and July were available for the Delaware River in the project area. These density data were collected by PSEG during the springs of 2002 [12] and 2003 [13]. The density data were separated into different zones along the length of the Delaware River and, for purposes of this case study, it was assumed that these densities did not vary vertically and were composites of all trawls in each zone; however, the hydrodynamic model was divided into three vertical zones: epilimnion, metalimnion and hypolimnion. Only four zones (8–11) around the import terminal were considered as they covered a region of 25 miles downstream and 30 miles upstream (see Figure 3). The entrainable population within these four zones is shown in Table 1. IP is assumed to be present only during the months of April through July, so annual losses are calculated as a sum of these four months.

**Figure 3.** Project location and IP data zones.

**Table 1.** Entrainable IP by PSEG zone (number of organisms per 1000 ft3 of water).


**Figure 4.** Ballast intake extent of influence.

A hydrodynamic model and a water quality transport model were developed to evaluate the percentage of water entrained during a single ballast water intake at the project location. The numerical modeling protocol established in Edinger and Kolluru [14] to estimate the entrainment of ichthyoplankton due to power plant cooling water intakes along a body of water was used as the basis for this study. Edinger and Kolluru divided the domain into several regions and used a numerical hydrodynamic model and a numerical transport model to simulate the entrainment of dye from each of these regions. The dye does not react or decay. The percentage of the dye entrained is used as a proxy to represent the amount of ichthyoplankton entrained. The model was based on the Delft3D [1,2] modeling system. The grid was created around the LNG terminal with grid sizes varying between 50 m and 200 m in the *x*-direction (along the river) and 65 m and 100 m in the *y*-direction (across the river). The grid extends all the way down to the Delaware Bay on the southern side and to Philadelphia on the northern side with 21,669 surface cells. The vertical resolution of the grid is chosen as 1 m. The model was run under flood and ebb conditions at the import terminal. Figure 4 shows the extent of influence of the ballast intake as modeled.

The percentage of water entrained during each ballast water intake cycle during the flood condition is shown in Table 2.


**Table 2.** Percent of each water layer entrained by ballasting—flood condition.

These percentages were further used to estimate the entrainment potential for the four species based on their lifestage duration and natural mortality as shown in Tables 3 and 4.


**Table 3.** IP stage durations.



Overlaying the water capture fraction with the IP densities, life stage durations and natural mortality, the EAM model can calculate the effective entrainment. All four methods described previously were examined to compare the predictions of the different methods. Table 5 summarizes the results and shows the predicted number of equivalent adults lost annually due to the intake operations. Methods 1 and 2 provide estimates of the likely upper and lower bound of the adults lost, while Methods 3 and 4 tend to give results reflective of the actual numbers likely to be lost. Table 6 shows the equivalent number of adults lost as a percentage of total equivalent adults available within the entrainable zone. The projected number of adults lost are very small (maximum of 0.12%) and thus suggest that the intake operations will likely have very little effect on the regional fish populations, if any.


**Table 5.** Total number of equivalent adults lost annually during flood and ebb condition ballast water intake operations using the four equivalent adult methods available in GEMSS-EAM.

**Table 6.** Total number of equivalent adults lost annually as percentage of total equivalent adults available in the entrainable zone (average of flood and ebb condition).


The Methods 1 and 2 are clearly bounding approaches because Method 1 applies the entrainment mortality at the beginning of the life stage (largest population size within the life stage) whereas Method 2 applies entrainment mortality at the end of the life stage (smallest population size within the life stage). These two assumptions lead to Method 1 over predicting and Method 2 under predicting the actual losses. The remaining two models attempt to apply the entrainment mortality at a more reasonable point (within the life stage). Method 3 uses the midpoint of the life stage whereas Method 4 applies it continuously along with the entrainment mortality (net mortality which is a result of exponentially decaying population from entrainment and natural mortality individually). It is therefore important to bind the estimated impacts using these approaches along with considering the Method 4 as a more realistic estimate.

### *4.2. Zone of Influence—Sturgeon Effects*

To address sturgeon entrainment, the zone of influence from the operation of the ballast water intake was studied using a second hydrodynamic model, GEMSS (Generalized Environmental Modeling System for Surface waters). A second model was applied to capture the level of detail required to evaluate the active swimming of fish as opposed to the passive entrainment of IP. The first step of the modeling was to create a high resolution near field grid around the ballast intake. The high resolution grid was created around this intake location with grid sizes varying between 10 m and 250 m in the *x*-direction (along the river) and 15 m and 300 m in the *y*-direction (across the river). The grid extends about 1 km on either side of the terminal along the river and covers the entire cross section with 1850 surface cells. The vertical resolution of the grid is chosen as 0.5 m. Based on the design information, the ballast intake of diameter 0.6 m (600 mm) and location of 3.7 m above the keel and 57 m forward of the stern was assumed. The high resolution model did not separate the two tidal cases as with the low resolution model earlier. The entire tidal cycle was combined into one model simulation to obtain cumulative effects. Some adults are long enough to become impinged if captured sidelong across the intake. Other mobile stages (adults and juveniles) can become entrained due to size or body orientation when they encounter the intake. These intake dimensions were closely followed in setting up the LNG carrier in the model domain as shown in Figure 5.

**Figure 5.** Carrier representation on the model domain.

Shortnose sturgeon have typical mean swimming speeds of 8.1 cm/s to 34.0 cm/s. These swimming speeds (and also burst speeds) are size dependent [15]. McCleave [16] found that the mean swimming speeds for shortnose sturgeon vary between 0.07 and 0.37 body lengths/s. Limited information is available about the swimming speeds for Atlantic sturgeon. Shortnose sturgeons are endangered species due the historical fishing and Atlantic sturgeon are similarly protected due to their resemblance to the shortnose. Entrainment of these sturgeons due to intake operations in the Delaware River is of concern. Sturgeons, and other species, as well can attain peak swimming speeds of up to 10–22 lengths/s [15] under increased opposing flows. However, during these increased swimming speeds (burst speed), respiration increases which results in exhaustion decreasing the distance traveled. Thus, they are expected to escape high flow conditions if the zone of this high flow is less than their endurance. The swimming speeds for these sturgeons were obtained from Amaral and Sullivan [17] and are summarized in Table 7.


**Table 7.** Swimming speeds.

The ballast water intake was located at a depth as shown in Figure 5. The high resolution model was run for two cases. Case 01 was run with the intake turned on whereas the Case 02 was run without the intake. These two cases were needed to determine the change in hydrodynamic conditions attributable to the intake operations. At this depth, in the near field, the horizontal velocity was predominantly towards the intake. There is a drastic difference in the horizontal velocities between the two cases. To better understand the effects due to the intake operation alone, consider Figure 6, which shows the velocity difference (with intake—without intake). The difference was very high close to the intake (~35 cm/s) but this difference rapidly decreased with distance from the intake.

A northwest-southeast plane (Slice AB) passing through the ballast water intake, as shown in Figures 5 and 6, was chosen to study the vertical and northwest-southeast directional flow and zone of influence. The hydrodynamics of the region close to the intake is mostly dominated by the intake momentum as compared to the tidal influence. As we move slightly away from this region a small tidal variation in the circulation is seen under different tidal phases. The difference in the flow conditions between the two cases is shown in Figure 7. The flows are high and directed towards the intake with a defined circulatory flow pattern. The effects of the intake, though smaller, can be seen even close to the bottom (~6 m below the intake). The difference in the velocity magnitude between the two cases is as high as 50 cm/s (Figure 7). The differential velocity vectors clearly show a new eddy type circulation pattern introduced during the up estuary flow condition. The incremental velocity magnitude close to the bottom ranges between 0.5 cm/s and 6 cm/s. The zone of influence is about 5 m–6 m in the vertical direction and about 50 m in the horizontal direction. After around 50 m in the horizontal direction at the intake level, the difference in net velocity magnitude drops to 0.1% (<0.5 cm/s as compared to 50 cm/s at the intake cell).

**Figure 6.** Horizontal velocities at intake for each case.

The mouth of the intake experiences the maximum increased flow (increase of 30–50 cm/s) which results in similar intake velocities of 30–50 cm/s (Figure 8). These velocities at the intake when the ballast water intake is active are within the mean swimming speeds of both Atlantic and shortnose sturgeon. The burst speeds are much higher than the intake velocities and thus can allow them to escape the zone of influence. The burst swimming speeds cannot be sustained for a long duration (Table 7). However, at these speeds (e.g., 90 cm/s) fish can move up to 18 m (90 cm/s ×

20 s) to 54 m (90 cm/s × 60 s) which will move them out of the zone of influence (50 m). Even though the burst swimming speeds are not sustainable, the extent of the zone of influence is small to allow the fish to quickly escape. However, there may be instances when the fish are close to the intake and may be entrained due to the smaller net velocity (swimming speed—velocity towards intake). Only a volume of 13 m3 was above the lowest burst swimming speed of 40 cm/s, a very small region.

#### **5. Summary and Conclusions**

#### *5.1. Ichthyoplankton and Equivalent Adults*

Using these data, the GEMSS-EAM was run for each of the four months in question (April, May, June and July), assuming all ballast intake occurs while the ship is at berth. Methods 1 and 2 provide estimates of the likely upper and lower bound of the adults lost, while Methods 3 and 4 tend to give results reflective of the actual numbers likely to be lost. The projected numbers of adults lost are very small (maximum of 0.12%) and thus suggest that the intake operations will likely have very little effect on the regional fish populations. Use of all four methods provides confirmation bounds for these quantifications that rely heavily on lifestage information which is hard to predict. A range of equivalent adults lost, therefore, is an appropriate choice. Additionally, sensitivity analysis should be performed to assess the uncertainty potential of these estimates.

It should be noted that there is always some risk associated with the withdrawal of ballast water by the ship intake. An expectation of no risk should not be a criterion for evaluating such impacts as the ecosystem, under natural conditions, has the ability to recover from small impacts. The risk evaluation completed as part of the study shows that the impacts are minimal, and within the ecosystem's ability to recover from.
