Reference Document Home | Publications Home
CONTENTS
Executive Summary
Introduction
Model Overview
Model Inputs
Results
Analysis of Hatchery and State of Recovery
Model Diagnostics and Sensitivity Analyses
Conclusions
Acknowledgements
References Cited

Northeast Fisheries Science Center Reference Document 13-09

Dam Impact Analysis Model for Atlantic Salmon in the Penobscot River, Maine

Julie L. Nieland1, Timothy F. Sheehan2, Rory Saunders2, Jeffrey S. Murphy2, Tara R. Trinko Lake2, and Justin R. Stevens3

1NOAA, National Marine Fisheries Service, Northeast Fisheries Science Center, 166 Water Street, Woods Hole, MA 02543
2NOAA National Marine Fisheries Service, Northeast Regional Office,
Maine Field Station, 17 Godfrey Drive, Suite 1, Orono, ME 04473
3NOAA National Marine Fisheries Service, Northeast Fisheries Science Center,
Maine Field Station, 17 Godfrey Drive, Suite 1, Orono, ME 04473

Web version posted December 2, 2013

Citation: Nieland JL, Sheehan TF, Saunders R, Murphy JS, Trinko Lake TR, Stevens JR. 2013. Dam impact analysis model for Atlantic salmon in the Penobscot River, Maine. US Dept Commer, Northeast Fish Sci Cent Ref Doc. 13-09; 524 p. Available from: National Marine Fisheries Service, 166 Water Street, Woods Hole, MA 02543-1026, or online at http://nefsc.noaa.gov/publications/

Information Quality Act Compliance: In accordance with section 515 of Public Law 106-554, the Northeast Fisheries Science Center completed both technical and policy reviews for this report. These predissemination reviews are on file at the NEFSC Editorial Office.

AcrobatDownload complete PDF/print version

Executive Summary

The Dam Impact Analysis (DIA) Model is a population viability analysis that was developed to help better understand the impacts of dams on the production potential of Atlantic salmon (Salmo salar). Dams have been identified as a major contributor to the historic decline and current low abundance of salmon in the Gulf of Maine Distinct Population Segment, which was first listed as endangered in 2000 and then expanded in 2009 to include Atlantic salmon in all rivers from the Androscoggin River north along the Maine coast to the U.S.-Canada border. The DIA Model specifically simulates the interactions of Atlantic salmon and 15 hydroelectric dams in the Penobscot River watershed in Maine.

The modeling approach incorporates life stage-specific information for Atlantic salmon to simulate the life cycle of Atlantic salmon in the Penobscot River. Most model inputs were considered to be random variables, and Monte Carlo sampling from probability density functions was used to create multiple realizations of population trajectories over time. All DIA Model iterations were run for 50 years, roughly ten generations of fish, and 5,000 iterations were run for each simulation. The DIA Model was built in Microsoft Excel with the @Risk add-on.

The DIA Model can be used to compare alternative scenarios of changes in future abundance and identify critical parameters and information needs for recovery efforts. The predicted abundance and distribution of adults and number and proportion of smolts killed due to the effects of dams were reported for several modeling scenarios. The DIA Model simulations are not meant to predict absolute abundance, distribution, or mortality, but rather are meant to project the relative changes under different modeling scenarios. The modeled population of Atlantic salmon in the Penobscot River decreased in abundance and distribution when DIA Model inputs were set at the base case values, whereas abundance increased and Atlantic salmon remained distributed throughout the Penobscot River watershed when marine and freshwater survival rates were increased appreciably in a recovery scenario. The production potential of Atlantic salmon was also more affected by the operational characteristics of mainstem dams than tributary dams in the Penobscot River watershed because mainstem dams tend to impact access to multiple upstream tributary dams. Sensitivity analyses were performed on all input values to determine which model inputs had the greatest impact on the results. The DIA Model results revealed that recovery of Atlantic salmon is most sensitive to marine survival and downstream dam passage survival rates.

1 Introduction

The Gulf of Maine Distinct Population Segment (GOM DPS) of Atlantic salmon is listed as endangered under the U.S. Endangered Species Act (65 Federal Register 69469, November 17, 2000; 74 Federal Register 29344, June 19, 2009). Dams have been identified as a major contributor to the historic decline and current low abundance of salmon in the GOM DPS (NRC 2004; Fay et al. 2006). To better understand the impacts of dams on the production potential of Atlantic salmon, a tool was developed to simulate the interactions of Atlantic salmon and dams, particularly hydroelectric dams in the Penobscot River watershed. The Penobscot River watershed was chosen as the area of study for several reasons. In recent years, approximately 75% of all U.S. Atlantic salmon returns have come from the Penobscot River (USASAC 2011). Also, multiple hydroelectric dams, which reduce migration success for downstream migrating smolts and upstream migrating adults, are located on both mainstem and major tributary reaches. Fifteen Federal Energy Regulatory Commission (FERC)-licensed dams were focused on because these dams are located within designated Atlantic salmon critical habitat (74 Federal Register 39003, August 10, 2009) or currently occupied Atlantic salmon watersheds.

Predicting the future viability of an endangered or threatened species is a vital part of planning management and recovery actions (NRC 1995), and population models are important tools for assessing management strategies and evaluating risks to these species (Morris and Doak 2002; McGowan and Ryan 2009; McGowan and Ryan 2010). A life history modeling approach was undertaken because a large amount of life stage-specific information is available for Atlantic salmon. Life history models can provide biological realism but may require many assumptions regarding the various inputs. Population viability analysis (PVA) is a stochastic life history model for predicting changes in population abundance given uncertain biological parameters (Beissinger 2002).

PVAs vary greatly in their complexity. A simple PVA quantitatively estimates information related to population growth and extinction probabilities for a single population (Dennis et al. 1991). A simple PVA is a stochastic exponential growth model of population size, which is equivalent to a stochastic Leslie-matrix projection with no density dependence. More complex PVA approaches account for a wider range of life history characteristics, such as age distribution, juvenile survival rates, adult survival rates, habitat limitations or degradation, age-specific fecundity, and migration rates (Beissinger 2002). One such life-cycle model, SalmonPVA, was developed for the GOM DPS (Legault 2004). The SalmonPVA is a state-space model structured to represent GOM DPS Atlantic salmon life history characteristics. Results from these more complex PVA models can be used to explore the potential effects of management actions in light of unknown future conditions, variability of input data, and assumptions made when designing the model (Legault 2005). The more complex approach, such as was applied within the SalmonPVA, may provide information to decision makers related to an array of management measures available (Samson 2002).

The Dam Impact Analysis (DIA) Model was built in Microsoft Excel with the @Risk add-on and was developed as a state-space model that is similar in structure to the SalmonPVA but representative of Penobscot River Atlantic salmon life history characteristics. Most DIA Model inputs were specified as random variables with known probability density functions, and Monte Carlo sampling was used to simulate many iterations of the Penobscot River population of Atlantic salmon forward in time. The DIA Model projections of future abundance can identify critical information requirements for recovery efforts. Specifically, the DIA Model was developed to assess the relative impacts of hydroelectric facility operations within the Penobscot River watershed on the production potential of the Penobscot River Atlantic salmon population. The DIA Model simulations do not predict absolute abundance, but instead project the relative change in abundance and distribution under different modeling scenarios.

This document describes the DIA Model developed for Atlantic salmon in the Penobscot River watershed. A full description of the chronology of the model development (Table 1.1) and modeling approach is presented and all input values, distributions, and assumptions are outlined.

2 Model Overview

An overview of the DIA Model is provided below. A schematic outlining the life stages modeled, additions and subtractions to the population and other metrics effecting the population was developed (Figure 2.1). A full description of all model inputs is provided in Section 3. The DIA model describes the dynamics of a population spatially distributed over 15 sections, hereafter referred to as production units (PUs; Figure 2.2). The linkages among PUs are defined by the physical configuration of the Penobscot drainage and accessibility (i.e., dams).

The initial distribution of salmon in the DIA Model is based on the mean annual number of two sea-winter (2SW) female returns captured at the trap above Veazie Dam during 2002-2011, which equaled 587 fish. These fish were randomly assigned among the PUs according to an underlying multinomial distribution based on the amount of salmon habitat available in each PU (see Section 3.1). Production potential was zero in PUs that could not be accessed due to lack of upstream dam passage, and, therefore, no adults were seeded into these PUs. For all subsequent calculations, the numbers of Atlantic salmon were rounded, rather than binomially assigned, to maintain whole numbers of fish and to minimize computational time.

For each DIA Model iteration, the 2SW females in year 1 were multiplied by the fecundity rate to estimate the number of eggs produced in that same year (see Section 3.2). The number of eggs was then multiplied by the egg to smolt survival rate to estimate the number of two-year old smolts produced in year 4 (see Section 3.3). If the number of smolts in a PU exceeded the production potential cap, then the number of smolts was reduced to the maximum allowed for that PU to ensure that projections remained biologically reasonable. The carrying capacity assigned to each PU defines the maximum potential population size and induces a spatially explicit density dependence in the PU set comprising the Penobscot River watershed. Smolts surviving from the egg stage were considered wild-origin fish. Additionally, the option was available to have hatchery-origin smolts "stocked" into each PU (see Section 3.4). All smolts (hatchery- and wild-origin) then migrated downstream from their initial PU, through subsequent downstream PUs and over dams, to Verona Island. As fish migrated through PUs, the number of surviving smolts in a PU was multiplied by the distance-specific in-river survival rate (i.e., 1 - in-river mortality rate raised to the distance traveled; see Section 3.5). To simplify modeling, smolts were assumed to travel only half the length of their initial PU because fish could start their migration from a variety of locations within the PU (e.g., the furthest point upstream, the furthest point downstream). As smolts migrated downstream through subsequent PUs, they traveled the distance from the point of entry to the point of exit (e.g., fish from PU 7 would travel the distance from Milo Dam to Howland Dam in PU 4). The in-river mortality rate was applied to each PU-specific group of smolts as they migrated downstream through each subsequent PU, until reaching the northern tip of Verona Island.

Smolts exiting a PU had to traverse a dam to enter into a downstream PU. To account for dam-related mortality, the number of smolts above each dam was multiplied by the correlated draws from the dam-specific cumulative distribution functions of total hydroelectric project survival to estimate the number of smolts remaining after passing each dam (see Sections 3.6.1 and 3.6.2). Smolts that started their migration in PU 9, or further upstream, could travel through the Mainstem or Stillwater Branch of the Penobscot River (see Section 3.6.3). The number of smolts reaching PU 9 was multiplied by the Stillwater Branch path choice to estimate the number of smolts that migrated through that branch. The remaining smolts migrated through the mainstem. Smolts continued to migrate downstream through subsequent PUs, encountering in-river and dam-related mortality, until the survivors reached the estuary at Verona Island.

At Verona Island, an option was available to apply an indirect latent mortality rate to account for the negative effects on survival from passing multiple dams (see Section 3.7). An indirect latent mortality rate was calculated for smolts originating in each PU, based on the number of dams that fish from each PU passed.

Although wild- and hatchery-origin smolts were treated the same during downstream migration (i.e., subjected to the same in-river mortality rates, smolt survival probabilities at dams, and indirect latent mortality rates), hatchery-origin smolts typically experience lower survival than wild-origin smolts (see Section 3.8). Hence, a survival discount was applied to hatchery-origin smolts to estimate the total number of wild-equivalents before they migrated out to sea. The remaining number of wild-equivalent smolts was halved to convert the number to wild-equivalent female smolts, which was needed to estimate the number of adult female returns. These wild-equivalent female smolts were considered post-smolts as they migrated beyond Verona Island, and the total number of female post-smolts in year 4 was multiplied by the marine survival rate to estimate the number of 2SW females that returned in year 6 (see Section 3.9).

Maine Atlantic salmon return to their natal river to spawn with high fidelity (estimated straying rates 1-2%; Baum 1997). However, homing to the Penobscot River was assumed to be 100% in the DIA Model, and the proportion of 2SW females that attempted to migrate upstream to each PU equaled the proportion of wild-equivalent female smolts that originated from each PU. Within the Penobscot River watershed, homing to natal PUs is less than 100%, and straying of adults is incorporated by randomly assigning a target PU based on estimated straying rates (see Section 3.10). Adults then migrated upstream from Verona Island and encountered dams as they attempted to migrate to their targeted PU (see Section 3.11). Upstream dam passage rates dictated the proportion of adults that were able to pass each dam. 2SW females that were unable to pass a dam died, returned to sea, or migrated to a different downriver PU to spawn (see Section 3.12). Adults that successfully passed dams continued to migrate upstream through all upriver PUs, until they reached their desired PU. No in-river mortality factor was applied, as freshwater mortality in free flowing stretches of river is assumed to be low for adult Atlantic salmon. In years when hatchery-reared smolts were stocked, 150 2SW females were removed from the migrating population for hatchery broodstock purposes just after passing the Veazie Dam (see Section 3.4). Hatchery broodstock were removed in a way that each PU contributed adult spawners in proportion to their adult returns (except PUs 13 and 14 because adults that returned to these PUs did not pass Veazie dam). The 2SW females that reached their desired PU spawned and produced eggs in that same year (i.e., year 6). This entire process was then repeated for nine more generations (one generation equaled 5 years).

All fish were tracked according to their PU of origin. The adult portion of the Atlantic salmon life cycle focused on 2SW females because the vast majority of females return as 2SW fish and egg production is one of the limiting factors for this population (USASAC 2011). The smolt life stage focused on age-2 fish because the majority (>85%) of naturally-reared Atlantic salmon smolts from Maine, and specifically the Penobscot River, migrate to the ocean as age-2 fish, with smaller proportions of both age-1 and age-3 juveniles present (USASAC 2011). Although kelts play a vital role in the life history of Atlantic salmon, this life stage was not included in the DIA Model due to limited quantitative information for model inputs and the limited number of kelts present.

A cohort of fish and its descendants were tracked through the life stages. Inputs were year- and iteration-specific random draws from distributions to incorporate stochastic variation into the model. All DIA Model iterations were run for 50 years, which equaled ten plus generations of fish, and 5,000 iterations were run for each simulation. All model iterations were run with @Risk.

3 Model Inputs

3.1 Production Units

The DIA Model was built for the Penobscot River watershed comprising 15 sections, or PUs (Table 3.1.1; Figure 2.2). The upstream boundary of each PU was either the headwaters of a tributary or a FERC-licensed hydroelectric dam. The downstream boundary of each PU was a hydroelectric dam, except in PU 14, where the downstream boundary was the northern tip of Verona Island. Using dams as PU endpoints meant that Atlantic salmon could not enter or exit a PU without attempting to pass a dam, with the exception of PU 14. This scheme helped further delineate the salmon-dam interactions in the model.

Total network length, longest segment length, and partial segment length were distances calculated to describe each PU (Table 3.1.1). Total network length represents the sum of all perennial stream kilometers within a particular PU. Longest segment length represents the longest straight path distance that a fish could swim within a PU. Partial segment length represents the distance that a fish would swim when traversing from one PU to another (e.g., fish from PU 2 would travel the distance from Mattaceunk Dam to West Enfield Dam in PU 3; Figure 2.2). PUs can have no partial segment length (e.g., PU 15), one partial segment length (e.g., PU 2), or two partial segment lengths (e.g., PU 4). The longest segment lengths and partial segment lengths were also used to calculate in-river mortality (see Section 3.5).

Each PU has the potential to support a different number of fish based on available habitat. Our measurement unit for Atlantic salmon is a habitat unit (HU) equal to 100 m2. The number of Atlantic salmon HUs was calculated for each PU using a model which estimated spawning and rearing habitat (Table 3.1.2; Wright et al. 2008). The number of Atlantic salmon HUs was used as a measure of production potential (i.e., the number of Atlantic salmon each PU could produce), and the proportional production potential (i.e., proportion of HUs in a PU compared the total habitat units for the drainage) was used to seed adults as well as to limit the number of smolts in each PU.

The model was seeded with 2SW females that were randomly assigned among the PUs according to an underlying multinomial distribution based on the proportion of HUs in each PU (Table 3.1.2). PUs 1, 7, 8, and 11 were not allotted any HUs because adults were unable to access them due to lack of upstream dam passage. Therefore, no 2SW females were allocated to these PUs.

The number of smolts in each PU was limited with a production potential cap, which was the maximum number of smolts allowed per HU (i.e., 10 smolts per 100 m2; Table 3.1.2). The cap of 10 smolts per 100 m2 is greater than the commonly accepted production potential of three smolts per 100 m2 in the Penobscot River (Meister 1962) but was implemented to prevent biologically unrealistic outputs from being produced via stochastic sampling.

PU 1, which is the West Branch of the Penobscot River above Medway, is different than the other PUs. Medway does not have upstream or downstream passage, so no fish are able to access this PU. Also, no anadromous Atlantic salmon are stocked in PU 1, so no juveniles are produced and no smolts migrate through this PU en route to PU 2 (Figure 2.2). Although PU 1 was built into the DIA Model, this PU did not contribute to the Atlantic salmon population. PU 1 was included in the model because the West Branch was historically important Atlantic salmon habitat and could be recognized as a potential component of Atlantic salmon recovery efforts in the Penobscot River in the future.

3.2 Eggs per Female

Adult female Atlantic salmon spawn at various ages, and typically older females produce more eggs. In the DIA Model, a fecundity rate was applied to the number of 2SW females in a year to estimate the number of eggs that would be produced the same year.

The number of eggs produced per female Atlantic salmon was estimated using fecundity data for Penobscot River sea-run female Atlantic salmon, spawned at Craig Brook National Fish Hatchery during 1997-2010 (Denise Buckley, U.S. Fish and Wildlife Service, personal communication). The data were derived primarily from 2SW females, but a small number of older females were also spawned each year. A distribution was fit to the average number of eggs per female in each year by using a combination of characteristics of the data (e.g., discrete distributions were not considered for values that could be treated as continuous) and goodness of fit tests. The data were best described by a normal distribution with µ = 8,304 and σ = 821 (Figure 3.2.1). Year- and iteration-specific values were drawn from this distribution for base case fecundity values in all DIA Model simulations.

3.3 Egg to Smolt Survival

Atlantic salmon spend the first years of their lives in rivers, from the time they are eggs until they migrate to the ocean as smolts. Atlantic salmon go through several life stages during this time: egg, fry, parr, and smolt. The DIA Model did not calculate the number of fish at all of these life stages. Instead, an egg to smolt survival rate was applied to the number of eggs in a year to estimate the number of smolts that would survive three years later (i.e., age-2 smolts) and be available to initiate a downstream migration to the ocean.

The egg to smolt survival rate was calculated based on the methods of Legault (2004). Egg to fry, fry to parr0+, parr0+ to parr1+, and parr1+ to smolt survival rates were obtained from the literature and were combined using a method that would account for uncertainty in each study. In order to be combined, studies for a particular life stage were standardized to the same time interval. The standardized mean, minimum, and maximum values were used to generate a triangular distribution for each study. The triangles were added together to form a new survival rate distribution for that life stage. This probability distribution function was converted to a cumulative distribution function, and the 10th and 90th percentiles were used as the limits of a uniform distribution. The uniform distribution was used to describe the uncertainty in survival for each life stage. Instream survival studies described in Legault (2004) were augmented with more recent studies.

The egg to fry survival rate came directly from a study of GOM DPS Atlantic salmon (Jordan and Beland 1981) instead of using the objective process described above. The uniform distribution for survival of 15 to 35%, covered most other estimates of survival in the literature (see Table 2 in Legault 2004), and was thought to best represent egg to fry survival of Atlantic salmon in Maine (Legault 2004). Two additional studies were excluded because they were not considered representative of Atlantic salmon survival in Maine (Table 3.3.1; Dumas and Marty 2006; Flanagan et al. 2008).

The fry to parr0+ survival rate was derived using the objective process described above, with the standard time period of two months. Seven studies were included, resulting in a uniform distribution ranging from 31 to 60% (Table 3.3.2; see Table 3 in Legault 2004; Figure 3.3.1). Other studies were excluded because they were not considered representative of Atlantic salmon survival in Maine for various reasons. One study had extremely low survival (Coghlan and Ringler 2004). Another study had a wide range of survival and did not report a mean survival rate (Coghlan et al. 2007). The duration of one study could not be determined (Raffenberg and Parrish 2003). Two studies (Aprahamian et al. 2004; Millard 2005) had multiple survival rate estimates, and these estimates were averaged for each study after standardizing the time period so that neither study would have undue influence on the overall calculation of survival for this life stage. The seven studies which were included had mean standardized survival rates ranging from 40.3 to 59.2% (Egglishaw and Shackley 1973; Egglishaw and Shackley 1980; Gardiner and Shackley 1991; Orciari et al. 1994; McMenemy 1995; Aprahamian et al. 2004; Millard 2005).

The parr0+ to parr1+ survival rate was derived using the objective process described above, with the standard time period of twelve months. Eight studies were included, resulting in a uniform distribution of survival ranging from 13 to 56% (Table 3.3.3; see Table 4 in Legault 2004; Figure 3.3.2). One study was excluded because survival was parsed out by season (Letcher et al. 2002). The eight studies which were included had mean standardized survival rates ranging from 11.3 to 51.0% (Meister 1962; Egglishaw and Shackley 1980; Kennedy and Strange 1980; Kennedy and Strange 1986; Gardiner and Shackley 1991; Orciari et al. 1994; Cunjak et al. 1998; Aprahamian et al. 2004).

The parr1+ to smolt survival rate was derived using the objective process described above, with the standard time period of nine months. Five studies were included, resulting in a uniform distribution ranging from 17 to 50% (Table 3.3.4; see Table 5 in Legault 2004; Figure 3.3.3). One study was excluded because the life stage of the fish was unclear (Letcher et al. 2002). The five studies which were included had mean standardized survival rates ranging from 16.8 to 45.8% (Meister 1962; Myers 1984; Orciari et al. 1994; Cunjak et al. 1998; John F. Kocik, NOAA's National Marine Fisheries Service, personal communication).

Combining the minimum and maximum values across these life stages produced a possible range from 0.10 to 5.88% for the egg to smolt survival rate, with a mean of 1.31% (Table 3.3.5). The egg to fry, fry to parr0+, parr0+ to parr1+, and parr1+ to smolt distributions were each sampled 10,000 times, and the life stage survival values from each iteration were multiplied together to calculate an egg to smolt survival rate. The sum of random values from the egg to fry, fry to parr0+, parr0+ to parr1+, and parr1+ to smolt distributions was approximately normal by the central limit theorem, and egg to smolt survival could be expressed as the sum of the natural logs of each survival rate (Hilborn and Walters 1992; Legault 2004). This meant that the distribution of egg to smolt survival approximated a lognormal distribution (Figure 3.3.4). These data were fitted with a lognormal distribution with µ = 1.31%, minimum = 0.10%, and maximum = 5.88% for the base case egg to smolt survival distribution (Figure 3.3.5). The 90% confidence interval encompasses survival values between 0.5 and 2.4%, which coincides with the general perception that egg to smolt survival should be around 1-2% (Legault 2004). Year- and iteration-specific values were sampled for all DIA Model simulations.

3.4 Hatchery Stocking

Hatchery-origin fry, parr, and smolts are stocked annually into the Penobscot River to supplement wild production with the goal of recovery of the Atlantic salmon population in the Penobscot Bay Salmon Habitat Recover Unit. The DIA Model allowed for smolt-stocking, as more than 90% adult returns to the Penobscot River have originated from smolt stocking (USASAC 2011). Within the DIA Model, hatchery smolts were stocked and proceeded through the downstream migration and ocean migration with their wild conspecifics.

Smolt stocking could be turned on or off on a yearly basis in the DIA Model. When smolt stocking was turned on, a total of 550,000 smolts were stocked, to mimic the approximate number stocked annually. Smolts were distributed throughout the watershed according to the mean proportion stocked in each PU during 2003-2012 (Table 3.4.1; USASAC 2011; Justin Stevens, NOAA's National Marine Fisheries Service, personal communication). In years when stocking was turned on, 150 2SW females were removed above Veazie Dam from the upstream migrating population of adults to fulfill the broodstock requirements. If 150 or fewer 2SW females were present above Veazie Dam, all of the fish were removed for hatchery broodstock. A total of 550,000 smolts were stocked annually regardless of the number of 2SW females removed for broodstock as broodstock shortages were assumed to be made up from backup broodstock sources. If smolt stocking was turned off, no broodstock were collected, and all 2SW females that successfully ascended the Veazie Dam fishway proceeded upriver.

3.5 In-river Mortality

Emigrating smolts are subjected to varying levels of in-river natural mortality as they migrate from their rearing habitat to the ocean. To incorporate this dynamic into the DIA Model, a distribution of mortality estimates per km was generated from telemetry studies conducted within the Penobscot River.

A network array of telemetry receivers was deployed throughout the Penobscot River, and groups of both wild- and hatchery-origin smolts were tagged and released at various locations throughout the drainage in 2005 and 2006 (Holbrook et al. 2011) and again in 2009 and 2010 (Joseph Zydlewski, U.S. Geological Survey, Maine Cooperative Fish and Wildlife Research Unit, personal communication). Estimates of mortality per km between successive telemetry unit/array pairs for each year- and origin-specific release group were derived from mark-recapture model outputs performed in Program MARK (White and Burnham 1999). Only fish that survived to the first receiver/array were included to eliminate potential bias associated with tagging-related mortality. Mortality estimates for successive telemetry unit/array pairs that spanned a hydroelectric facility were excluded because dam-related mortality was accounted for in Section 3.6.1. A total of 64 estimates of in-river mortality per km were available. Eleven of these estimates were removed from the analysis due to concerns that they were biased by tagging-release effects, the river segment being too small (<1 km long), or the river segment being flanked by two dams. The resulting dataset included estimates ranging from 0.0 to 2.8% loss per km migrated. These estimates were calculated from river segments that were between one and 20 km long. A cumulative frequency distribution was created from the data (Figure 3.5.1), and 34.6% of the distribution represented a 0.0% mortality per km.

The DIA Model applied year- and iteration-specific values from the in-river mortality distribution, which meant the same mortality per km value was used for all PUs in a year. To avoid the unlikely scenario of 35% of the iterations having 0% mortality per km, a new in-river mortality distribution was developed for use in the DIA Model. This new in-river mortality distribution was created using a sub-model. A total of 500,000 smolts were proportionally distributed across all PUs, according to the production potential of each PU, in the sub-model. No smolts were stocked into PU 1, as this PU was excluded from the DIA Model due to the lack of upstream access into this system. Smolts were not stocked into PU 11 (Stillwater Branch) to simplify the simulation by not requiring an input variable for path choice between the Mainstem and Stillwater branches. PU-specific in-river mortality values were based on random draws from the cumulative distribution in-river mortality estimates described above. To calculate the number of surviving smolts entering each downriver PU, the PU-specific in-river mortalities were subtracted from one and raised to the distance travelled within a PU for each group of smolts (Table 3.1.1). Smolts in the sub-model were stocked in the middle of a PU, and the number of smolts surviving from the PU in which they were stocked was based on half the longest segment length of that PU. Smolts were assumed to have traveled the entire length of subsequent PUs (i.e., partial segment length; Table 3.1.1). The survivors after PU 14 were summed, and an estimated mortality rate per km was calculated as the proportion of smolts that survived divided by the total distance smolts migrated. A total of 10,000 iterations were performed, and the resulting mortality per km distribution was best described by a beta distribution with shape parameters α2 = 11.245 and α2 = 9.8007, minimum = zero, and maximum = 0.00038077 (Figure 3.5.2). This distribution was fit by using a combination of characteristics of the data and goodness of fit tests. Year- and iteration-specific values were sampled from this new distribution for base case in-river mortality rates in all DIA Model simulations.

3.6 Downstream Dam Passage Survival Rates

3.6.1 Desktop Survival Analysis

The Penobscot River Basin has been extensively developed for hydroelectric power generation. Approximately 123 dams are located in the Penobscot River watershed, and 31 of these dams operate under a FERC hydropower license or exemption (Tara Trinko Lake, NOAA's National Marine Fisheries Service, personal communication). However, the DIA Model focused only on 15 FERC-licensed dams within designated Atlantic salmon critical habitat (74 Federal Register 39003, August 10, 2009) or occupied Atlantic salmon watersheds.

Hydroelectric dams are known to impact Atlantic salmon through various mechanisms, such as habitat alteration, fish passage delays, and entrainment and impingement (Ruggles 1980; NRC 2004). Site-specific survival studies are available for some hydroelectric facilities in the Penobscot Basin (as summarized by Fay et al. (2006) and Holbrook (2009)). However, the limitations of currently available data are significant. As the DIA Model was designed to understand the impacts of these FERC-regulated dams on the productivity of the Penobscot River Atlantic salmon population, an accurate description of the total mortality associated with each of these facilities was required. Given the paucity of field data to describe these effects, Alden Research Laboratory, Inc. (hereafter referred to as Alden) was contracted to estimate current smolt survival rates at 15 FERC-regulated dams on the Penobscot River, based on site-specific project data (e.g., turbine type, revolutions per minute, head, presence of fishways), fish characteristics, and hydrological records. The factors to be considered were to cover both direct and indirect mortality effects attributable to dam passage as well as delayed mortality based on available literature.

Two types of mortality effects were incorporated within the DIA Model: direct and indirect. Direct mortality is the result of a lethal injury that occurs during passage through turbines, over fishways, or through fish bypasses and leads to death during passage or shortly thereafter (Amaral et al. 2012). An example of direct mortality would be a lethal injury from blade strike. Indirect mortality may occur through a variety of mechanisms such as predation (that may be attributable to reduced migration speed or turbulence at a dam), disease (that may be more likely to occur as a result of sub-lethal injury such as scale loss), and the additive effects of stress and injury associated with passing one or multiple dams. The effects of indirect mortality may be felt during or immediately post-dam passage or sometime thereafter at a later state of migration. Indirect mortality was segregated into two discrete factors for the DIA Model: cumulative and latent. Indirect cumulative mortality can occur when passage through turbines, over spillways, and through bypasses results in injuries such as scale loss, lacerations, bruising, eye or fin damage, or internal hemorrhaging (Amaral et al. 2012). Although indirect cumulative mortality is likely fairly low, this mortality may increase after fish pass multiple dams. An indirect cumulative mortality factor was incorporated into smolt survival rate estimates at each of the 15 hydroelectric dams that were modeled (Amaral et al. 2012). Indirect latent mortality is believed to occur early in the marine phase of the salmon's life history and is discussed further in Section 3.7.

The route that a salmon smolt takes when passing a dam is a major factor in its likelihood of survival. A fish that passes through a properly designed downstream bypass has a better chance of survival than a fish that goes over a spillway, which, in turn, has a better chance of survival than a fish swimming through the turbines. Facility-specific characteristics were obtained and used by Alden to estimate flow-specific total project smolt mortality estimates based on flow-specific turbine, spillway, and bypass mortality estimates with an additional indirect cumulative mortality rate applied (i.e., mortality due to predation and sub-lethal injuries during passage). The probability of all possible flow conditions was estimated in discrete cubic feet per second (cfs) increments at all modeled facilities (Amaral et al. 2012). Cumulative flow probability distributions were generated for each modeled facility (Figure 3.6.1.1) and were used in combination with the total project smolt survival estimates (Figure 3.6.1.2) to generate year- and iteration-specific estimates of smolt survival at each of the 15 dams in the DIA Model, as described in Section 3.6.2. Flow probabilities, and hence total project smolt survival, was not calculated for approximately 0.5% of the flow probability at each of the modeled facilities due to the very low probability of occurrence at the extreme upper and lower cfs bins. These missing probabilities for extreme cfs bins were accounted for by subtracting the sum of the flow probabilities from one, dividing the missing probability in half, and assigning the halves to two new cfs bins, one on each end of the flow probability distribution. The total project smolt survival in each new flow bin was set equal to the survival at the adjacent cfs bin provided by Alden. Although ad hoc, results are likely robust to these probabilities for rare events. A full description of the Alden procedures can be found in Amaral et al. (2012).

The Upper Dover Dam was an exception to the above outlined procedures. The total project survival for this facility was set to 92.15% for each year and iteration of the DIA Model. No turbine entrainment occurs at this facility, as the project is not presently operating. Also, a downstream bypass is not available for smolts to utilize. As such, all migrating smolts must pass the facility via the spillway, which has a set 97% survival rate. Additionally, an estimated 5% indirect cumulative mortality rate (i.e., 95% survival), due to sub-lethal injuries, increased stress, and disorientation, was applied to all smolts migrating past any facility (Amaral et al. 2012). The total project survival of 92.15% for the Upper Dover Dam was calculated as the product of the spillway and the indirect cumulative survival rates.

Alden updated smolt survival estimates for Milford, Great Works, Stillwater, and Orono dams due to a change in the flow allocation to the Stillwater Branch of the Penobscot River. These updates were not used for the analyses reported in this document as they were provided after all DIA Model runs were performed, but the new smolt survival estimates are available for future use (Amaral et al. 2012). The updated smolt survival estimates would not alter the results appreciably as the survival estimates are very similar to the previous estimates (Table 3.6.1.1).

3.6.2 Downstream Passage Correlation

Survival of smolts migrating past hydroelectric facilities is generally positively correlated with river flow. Downstream migrating smolts typically have two or three routes by which they can traverse a hydroelectric facility: a downstream bypass (if available), over the spillway, or through the turbines. Under low flow conditions, more flow is proportioned to the turbines and less flow is proportioned to the downstream bypass and the spillway, thereby increasing the proportion of smolts passing through the turbines. Passing through the turbines generally results in increased mortality and injury rates compared to passing via a downstream bypass or the spillway. Conversely, under high flow conditions, a greater proportion of the flow, and, therefore, downstream migrating smolts, passes through the downstream bypass and spillway where smolt survival is typically higher.

Alden estimated probability of flow and total project smolt survival for all possible flow conditions in discrete cfs increments for 15 FERC-regulated hydroelectric facilities on the Penobscot River (see Section 3.6.1). Within the DIA Model, year-specific random draws from the facility-specific cumulative probability of flow relationships (Figure 3.6.1.1) were used determine the flow levels and subsequent total project smolt survival estimates for each facility (Figure 3.6.1.2). These estimates were used to calculate the number of smolts that survive at each facility as they migrate downstream to the ocean. Within the Penobscot River, if one facility is experiencing high flows and consequentially high smolt survival, all facilities are likely experiencing relatively high flows and high smolt survival. Therefore, a mechanism was needed to correlate total project smolt survival across all facilities within each year and to incorporate the variation in flow documented within the drainage.

Flow data from 24 current and historic monitoring sites within the Penobscot River watershed were accessed through the USGS National Water Information System (http://waterdata.usgs.gov/nwis). Available flow data spanned from the lower reaches of the system to the headwaters, including all major tributaries. Careful review of the available data resulted in 19 sites being removed from the analysis because of a lack of contemporary data, the location within the drainage was not applicable to the DIA Model, or the data series consisted of a single year. Continuous flow data were available for the remaining five sites (USGS gauge 1029500 - East Branch Penobscot River at Grindstone, USGS gauge 1030500 - Mattawamkeag River near Mattawamkeag, USGS gauge 1034000 - Piscataquis River at Medford, USGS gauge 1031500 - Piscataquis River near Dover-Foxcroft, and USGS gauge 1034500 - Penobscot River at West Enfield) for the period 1935-2010. The smolt migration occurs within the months of April through June, so a correlation analysis was run on the mean April - June flow for each site (Table 3.6.2.1). The minimum correlation coefficient (r) = 0.831, maximum = 0.981, and µ = 0.901, suggest that flow within the Penobscot drainage was highly correlated and, therefore, high flow and high smolt survival at one facility should correspond with high flow and high smolt survival at all facilities within the drainage.

As stated above, a year-specific cumulative probability of flow common to all facilities was drawn from a uniform distribution bounded by zero and one. A year- and facility-specific random error drawn from a uniform distribution bounded by ± 0.1695 was added to these year-specific cumulative probabilities. Each year- and facility-specific probability sum was constrained from zero to one. For each of these year- and facility-specific probability sums, corresponding flow rates and smolt survivals were obtained from facility-specific relationships between the cumulative probability of flow rates (Figure 3.6.1.1) and total project smolt survival (Figure 3.6.1.2). This method of combining a year-specific random variable with a year- and facility-specific random variable ensured that smolts experienced similar relative flows among all dams. As an example, if a year-specific cumulative flow probability of 0.40 was drawn, the resulting year- and facility-specific probability sums would range from 0.2305 to 0.5695 (i.e., 0.40 ± 0.1695), with an approximate mean of 0.40. The range of the uniform distribution used for the year- and facility-specific random errors (i.e., ± 0.1695) was specified so that the mean correlation of the subsequent flow rates among all dams equaled 0.901, which approximated the actual correlation of flows for dams in the Penobscot drainage. In a few instances, the distance between neighboring hydroelectric facilities was small enough that flow conditions at the up-river dam were likely identical to the lower dam. In these cases, the same year- and facility-specific random error was used for both dams to match to the cumulative distribution. This occurred with four pairs of facilities: Great Works and Milford, Orono and Stillwater, Brown's Mills and Dover Upper, and Milo and Sebec. Year- and iteration-specific smolt survival estimates were selected in this manner for all DIA Model simulations.

3.6.3 Downstream Path Choice

A unique feature of the Penobscot River is the Stillwater Branch (i.e., Stillwater River). The Stillwater Branch is an approximately 17-km long side channel of the Penobscot River that begins at river km 47 (measured from the top of Verona Island), runs along the north and western sides of Orson and Marsh Islands, and rejoins the mainstem at river km 58.5, upriver of Veazie Dam (Figure 2.2). Smolts originating upriver of the Stillwater Branch have the option of migrating via the Stillwater Branch or the mainstem. Differential survival is likely experienced by smolts migrating through these two routes due to differences in local environs and the presence of multiple hydroelectric facilities. Smolts that migrate via the mainstem encounter 2 dams: Milford and Great Works. (Great Works Dam was still operating at the time the DIA Model was built but was removed in 2012.) Smolts that migrate via the Stillwater Branch encounter 3 dams: Gilman Falls, Stillwater and Orono. Gilman Falls serves to control Stillwater head pond height and was not included within the DIA Model as this dam is assumed to have a minor negative effect on downstream migrating smolts due to the presence of a natural bypass channel adjacent to the dam and the lack of hydroelectric production capacity. However, Milford, Great Works, Stillwater, and Orono dams do have the potential to significantly affect downstream migrating smolts and have been shown to have varying levels of total project smolt survival (Figure 3.6.1.2). Additionally, previous telemetry investigations have shown that the proportion of the smolts accessing the Stillwater Branch varies annually (Holbrook et al. 2011). To accurately assess the impacts that hydroelectric facilities may have on migrating smolts in the Penobscot River, the option of migrating down the Stillwater Branch or mainstem was incorporated into the DIA Model.

As previously mentioned (Section 3.5), a network array of telemetry receivers was deployed throughout the Penobscot River and groups of both wild- and hatchery-origin smolts were tagged and released at various locations throughout the drainage in 2005 and 2006 and again in 2009 and 2010. Release group-specific (2005 and 2006) and origin-specific (2009 and 2010) estimates of Stillwater Branch use were calculated (Holbrook et al. 2011; Joseph Zydlewski, U.S. Geological Survey, Maine Cooperative Fish and Wildlife Research Unit, personal communication). Stillwater Branch use estimates (n = 6) were fitted to a triangular distribution with a minimum value = 4.4%, a most likely value = 25.9%, and a maximum value = 25.9% (Figure 3.6.3.1). This distribution was fit by using a combination of characteristics of the data and goodness of fit tests. A cumulative frequency distribution was developed from 5,000 random draws from the triangular distribution (Figure 3.6.3.2). The proportion of smolts that accessed the Stillwater Branch during their migration was determined via a random draw from the cumulative frequency distribution. Smolts that migrated through the Stillwater Branch were subjected to in-river mortality and mortality associated with the Stillwater and Orono dams. All remaining smolts migrated via the mainstem and were subjected to in-river mortality and mortality associated with the Milford and Great Works dams. Random draws for Stillwater Branch use were correlated with the total project survival estimates according to the methods detailed in Section 3.6.2. Year- and iteration-specific Stillwater Branch use estimates were selected in this manner for all base case DIA Model simulations.

3.7 Indirect Latent Mortality

Additional dam-related mortality that occurs in the early marine phases of the salmon's life history has been previously discussed (Budy et al. 2002; Schaller and Petrosky 2007; Haeseker et al. 2012). This additional dam-related mortality has been identified by a number of different names such as cumulative mortality, latent mortality, and the hydrosystem-related delayed mortality hypothesis. Hereafter, this additional dam-related mortality is referred to as indirect latent mortality. Indirect latent mortality is defined as mortality that occurs in the ocean and estuary after exiting the hydrosystem but is related to the fish's earlier experience within the hydrosystem (Budy et al. 2002). This mortality is due to effects of stress and injury over the course of passing one or multiple dams (Budy et al. 2002; Schaller and Petrosky 2007; Haeseker et al. 2012). Some indirect latent mortality may occur within a hydropower system (Budy et al. 2002), but the cumulative in-river effects are difficult to separate from direct and indirect cumulative mortality that occur at or near individual dams. The DIA Model contained an option to apply an indirect latent mortality rate at Verona Island. This rate was calculated for smolts originating in each PU and was based on the number of dams that fish passed. An indirect latent mortality was applied to smolts at a rate of 10% per dam passed.

Although indirect latent mortality has been demonstrated in other river systems (Budy et al. 2002; Schaller and Petrosky 2007; Haeseker et al. 2012), effectively quantifying this mortality, including in the Penobscot River, has been challenging, mainly because of difficulties directly measuring mortality after fish have left the river system. Due to the number of hydroelectric dams that are currently in the Penobscot River watershed, even a small indirect latent mortality rate can have a large effect on the number of smolts (and consequently 2SW females) in the population. An indirect latent mortality rate of 10% per dam is within the range of estimates for this mortality type developed from individual studies in the Snake River and lower Columbia River basins (Deriso et al. 1996; Schaller and Petrosky 2007).

3.8 Hatchery Discount

Although hatchery- and wild-origin smolts experience the same kinds of mortality, hatchery-origin smolts typically experience lower survival than wild-origin smolts, and so a discount was applied to hatchery-origin smolts to estimate the number of wild-equivalents before they migrated out to sea.

To estimate a hatchery discount, survival rates of wild- and hatchery-origin fish were obtained from the literature. Studies were included or excluded from the hatchery discount calculation with some subjectivity, and the decisions to include or exclude them are described below (Table 3.8.1).

Studies of wild- and hatchery-origin Atlantic salmon were used to estimate the relative difference in survival between hatchery and wild fish from the smolt to adult life stages. Studies were excluded because they were not considered representative of Atlantic salmon in the Penobscot River watershed for various reasons. Studies were excluded if survival rates were not given (De Leaniz et al. 1989; Fleming et al. 1997; Einum and Fleming 2001; Salminen et al. 2007). Other studies were excluded because their study design made the survival rates inapplicable for the hatchery discount (e.g., life stages outside of smolt to adult stages were included, adult Atlantic salmon were captured at sea rather than in the river, survival of wild and semi-wild fish were compared instead of wild and hatchery fish; Jonsson et al. 1991; Jonsson and Fleming 1993; Jonsson 1997; Jonsson et al. 2003; Jokikokko et al. 2006; Peyronnet et al. 2008; Kallio-Nyberg et al. 2011). The data points that were included (n = 17) had wild to hatchery survival ratios ranging from 1.18 to 8.20% (Jonsson et al. 1991; Crozier and Kennedy 1993; Jonsson and Fleming 1993; Jonsson et al. 2003; Jutila et al. 2003; Kallio-Nyberg et al. 2004; Saloniemi et al. 2004; Jokikokko et al. 2006; Peyronnet et al. 2008; Kallio-Nyberg et al. 2011).

A distribution was fit to the included wild versus hatchery survival ratios by using a combination of characteristics of the data and goodness of fit tests. The data were best described by a log logistic distribution, with γ = 1, β = 1.4271, α = 1.9922, and maximum = 12 (Figure 3.8.1). Year- and iteration-specific values were drawn from this distribution for base case hatchery discount values in all DIA Model simulations. The proportion of hatchery smolts at Verona Island (after the indirect latent mortality rate was applied) was divided by the year- and iteration-specific hatchery discount to estimate the number of wild-equivalent smolts.

3.9 Marine Survival

U.S. Atlantic salmon spend approximately one half of their life in the marine environment. To account for this, the DIA Model estimated the number of female post-smolts that successfully emigrated to Verona Island at the upper-most reaches of Penobscot Bay, and a marine survival distribution was applied to this population to estimate the number of 2SW female returns that would successfully migrate to Greenland and back to Verona Island over the course of the following two years. These 2SW females would then be available to migrate upstream en route to their natal spawning grounds.

Although the marine survival phase has received increased attention in recent times, an accurate assessment of marine survival for the Penobscot River salmon population is not available. Counts of adult returns divided by the total number of smolts stocked into the Penobscot River can be used as a surrogate for the marine survival rate, and these data are available from 1969 through the present. However, these are not accurate estimates of marine survival because they incorporate mortality of smolts in freshwater (i.e., stocking, in-river, and dam-related mortality). Marine survival estimates do exist for the Narraguagus River, a small coastal Gulf of Maine river located approximately 105 km northeast of Penobscot Bay, but the estimates are from a short time series (1997-present) that only includes data from a period of low marine productivity (Chaput et al. 2005). Finally, the DIA Model focused on 2SW female returns, and none of the existing datasets provide sex-specific estimates of marine survival. As such, a new 2SW female-specific marine survival distribution was generated from available data from the Penobscot River, which aimed to remove the freshwater mortality factors.

To estimate a 2SW female marine survival distribution, the number of female smolts at Verona Island had to be estimated first. Year-specific estimates of the number of smolts stocked into the Penobscot River during 1969-2008 (USASAC 2011) were halved to approximate the number of stocked female smolts and then multiplied by the proportion of smolts that survived to Verona Island to adjust for mortality during the freshwater portion of the migration. Smolt survival to Verona Island was estimated from five years (2005, 2006, 2009, 2010, and 2011) of telemetry studies conducted within the Penobscot River (Joseph Zydlewski, U.S. Geological Survey, Maine Cooperative Fish and Wildlife Research Unit, personal communication). Seventeen estimates were obtained from hatchery- and wild-origin groups released at six different locations, and the means were fitted to a beta distribution with shape parameters α1 = 4.1923 and α2 = 1.8648, minimum = zero, and maximum = one (Figure 3.9.1). The distribution was fit by using a combination of characteristics of the data and goodness of fit tests. Year-specific values were sampled from this distribution to estimate the number of female smolts that would survive from stocking to Verona Island.

Estimates of 2SW adults returning to the Penobscot River were obtained for return years 1971-2010 (Figure 3.9.2; USASAC 2011). These estimates represented all 2SW returns and, as the DIA Model focused on 2SW female returns, needed to be discounted accordingly. Sex statistics were available for the Penobscot River from 1978 to 2011 (Figure 3.9.3; Maine Department of Marine Resources fishway trap database, 2010 version). During 1978-1999, sex statistics were based on field determinations made at the adult trap. Starting in 2000, fish collected for broodstock were individually tagged in the field and brought to the hatchery, where their sex could accurately be determined during spawning. The 2000-2011 data are considered more accurate because sex determinations made in the field early in the season, prior to sexual dimorphism, are difficult. When converting the 2SW adult returns to female 2SW returns, the year-specific sex ratio estimates were used for 2000-2010, and the 2000-2010 mean ratio was used for all years prior.

Year-specific 2SW female marine survival rates were calculated by dividing the estimated number of 2SW female returns by the estimated number of female smolts at Verona Island. A total of 10,000 iterations were run, where the number of female smolts that would survive from stocking to Verona Island was a stochastic process (as described above). The maximum survival was capped at 25%, which was exceeded in less than 0.05% of the iterations. The resulting 1971-2010 median values were fitted to an inverse gaussian distribution with μ = 0.006265, shape parameter λ = 0.0068723, and a shift of 0.00000813424 (Figure 3.9.4). Year- and iteration-specific values were sampled from this distribution for base case marine survival rates in all DIA Model simulations.

3.10 Straying

Adult Maine Atlantic salmon have been shown to have a high degree of river-of-origin homing, with rates of 98-99% in hatchery-release studies (Baum 1997). However, the in-river migration behavior and the effect of this behavior on reach-level productivity are poorly understood. Within-river homing behavior and its effect on distribution of spawning adults is postulated as being driven by habitat (i.e., temperature, flow, and substrate) (Kocik and Ferreri 1998), the presence of conspecifics (i.e., pheromone cues), and environmental cues (Fleming 1996). Atlantic salmon have a strong tendency to return to river reaches where they have been reared. Saunders (1967) estimated a homing rate of 70% for naturally-reared smolts in the upper Miramichi, NB, Canada. Similarly, Heggberget et al. (1988) showed adult Atlantic salmon returned with very high affinity (μ = 87%) to areas they had selected as spawning grounds when artificially displaced. Evolutionarily, in-river homing is logical as the success of an individual's rearing would provide selection for the local habitat characteristics, and returning adults provide this selective advantage to future progeny. However, limited levels of straying also benefits salmon populations by allowing for plasticity in habitat use in response to varying population levels (i.e., balancing density dependent effects) and the opportunity to colonize new habitat as well as the prevention of genetic bottlenecking (Heggberget et al. 1988).

Estimated in-river homing rates and straying patterns were developed to more accurately model the spatial distribution of Atlantic salmon production in the Penobscot River watershed. PU-specific homing rates and straying patterns were developed through an assessment of all available pertinent data and information including various Atlantic salmon behavioral studies conducted within the Penobscot (Power and McCleave 1980; Shepard 1995; Gorsky 2005; Gorsky et al. 2009; Holbrook et al. 2009; Douglas B. Sigourney, U.S. Geological Survey, Maine Cooperative Fish and Wildlife Research Unit, personal communication), fishway trap data from throughout the drainage (Maine Department of Marine Resources fishway trap database, 2010 version), and Expert Panel recommendations made on the topic (NMFS 2012).

Estimates of PU-specific homing rates and straying patterns could not be developed based on the behavioral studies and fishway trap data for two primary reasons. First, the available data were not representative of the entire drainage as some PUs had no information from which to draw conclusions. Second, the patterns observed within the various datasets could not be delineated into behavioral effects versus effects confounded by upstream passage issues. Estimates of PU-specific homing rates and straying patterns should be based on behavioral patterns only and need to be free from influences of upstream passage issues as these affects are included within the Upstream Dam Passage Inefficiency dynamics (Section 3.12).

A set of logical rules was developed to assist with estimating PU-specific homing rates and straying patterns by using the specific study results combined with the Expert Panel opinions and local knowledge (Table 3.10.1). The logical rules are as follows:

Exceptions to these logical rules are as follows:

The actual rates of homing and straying for returning Penobscot Atlantic salmon are likely determined by a combination of biotic and abiotic factors, but a dataset of homing rates and straying patterns with dam passage factors removed was needed for the DIA Model. Because observational data from the Penobscot was considered biased, model rates were based on logical concepts, field data, expert opinions, and biological theory. The PU-specific homing rates and straying patterns described above were the best available information for use in the DIA Model.

3.11 Upstream Dam Passage Survival Rates

3.11.1 Veazie, Great Works, Milford, and All Other Dams

After spending several years feeding in the ocean, adult Atlantic salmon return to rivers to spawn. As stated in Section 3.6.1, a large number of dams are located within the Penobscot River watershed, and Atlantic salmon must attempt to pass these dams on their upstream migration to their spawning grounds. The DIA Model also addressed upstream passage dynamics at 15 of those dams. The calculation of upstream dam passage was dependent upon each dam.

Numerous telemetry studies have been conducted within the Penobscot River that focused on evaluating upstream passage of adult Atlantic salmon. These studies were conducted in 1987-1990, 1992, and 2002-2006 and have provided estimates of upstream passage at Veazie, Great Works, and Milford dams (Holbrook et al. 2009). Veazie estimates ranged from 0.4210 to 0.9840, with μ = 0.6485 and σ = 0.1907, Great Works estimates ranged from 0.1190 to 0.9440, with μ = 0.6730 and σ = 0.2783 and Milford estimates ranged from 0.6670 to 1.0000, with μ = 0.8993 and σ = 0.0958. These data were used to generate cumulative frequency distributions (Figures 3.11.1.1, 3.11.1.2, and 3.11.1.3). To avoid using outliers from these datasets, minimums and maximums were placed on each of the cumulative distributions, using μ ± σ to calculate the limits (Table 3.11.1.1). Year- and iteration-specific values were randomly drawn from these cumulative distributions for base case upstream dam passage rates in all DIA Model simulations.

Four dams (i.e., Medway, Milo, Sebec, and Orono) do not have any upstream passage facilities, meaning adults are not able to access the PUs above these dams (i.e., PUs 1, 7, 8, and 11), and so upstream passage was set to zero (Table 3.11.1.1). No adults were seeded in these PUs (because of the lack of upstream access). Subsequently, no smolts originated in them, and no 2SW females would home to them. However, a small proportion of adults were allowed to attempt to stray to these PUs (see Section 3.10) although their attempts would be unsuccessful due to the lack of passage at the facilities at the lower boundary of the PU. These adults would then die, return to the ocean un-spawned, or stray and spawn in a downstream PU (see Section 3.12).

Upstream passage estimated for the eight remaining modeled dams (i.e., Mattaceunk, West Enfield, Dover Upper, Brown's Mills, Howland, Lowell, Stillwater, and Frankfort) were not available. Generalized estimates were used in previous modeling efforts (USFWS 1988) and were adopted here. A uniform distribution was developed for the eight remaining dams using µ ± σ (i.e., 0.92 ± 0.0325) as the upper and lower limits of the distributions (Table 3.11.1.1). Year- and iteration-specific values were sampled from the uniform distributions for the base case upstream dam passage rates in all DIA Model simulations. Adults that were not able to pass a dam died, returned to sea, or went to another PU (see Section 3.12).

3.11.2 Upstream Path Choice

As stated in Section 3.6.3, the Stillwater Branch presents a unique situation in the Penobscot River. Fish have the option to migrate through the Stillwater Branch or the mainstem. Whereas smolts were able to migrate downstream through either the Stillwater Branch or the Mainstem in the DIA Model, all adult spawners that attempted to migrate upstream of PU 12 were forced to migrate through the mainstem. This was because Orono Dam, which is the downstream endpoint of PU 11 and the Stillwater Branch, has no upstream fish passage facilities.

No adults were seeded in PU 11 (because of the lack of upstream access). Subsequently, no smolts originated in PU 11, and no 2SW females would home to PU 11. A small proportion of adults attempted to stray to PU 11 (Section 3.10). However, given the lack of upstream passage, all adults were diverted to the mainstem.

3.12 Upstream Dam Passage Inefficiency

Few, if any, upstream fishways provide safe, timely, and effective passage for 100% of migratory fish, including Atlantic salmon. Although multiple studies have been conducted in the Penobscot River to measure the effectiveness of fishways at various hydroelectric facilities, very little data are available concerning the fate of adult Atlantic salmon that are unsuccessful in locating or negotiatingupstream fishways at dams.

Within the DIA Model, the fate of adult salmon that were unsuccessful in passing an individual dam needed to be defined to more accurately model the spatial distribution of Atlantic salmon production in the Penobscot River watershed. In the absence of site-specific data, NMFS convened an expert panel, consisting of state, federal, and private sector biologists and engineers with expertise in Atlantic salmon biology and behavior at fishways, to address the issue. Specifically, the Expert Panel was asked if Atlantic salmon that are unsuccessful in locating and negotiating upstream fishways at the 15 hydroelectric projects modeled in the DIA Model die, return to the ocean un-spawned, or stray and spawn in downstream reaches. Through best professional judgment, the Expert Panel reached consensus regarding the fate of adult Atlantic salmon that are unsuccessful at locating and negotiating upstream fishways at the 15 hydroelectric projects modeled in the DIA Model (Table 3.12.1). Hydroelectric projects upstream of the first impassable dam on the West Branch of the Penobscot River were not evaluated by the group (e.g., dams upstream of Medway). A full description of the discussions and decisions reached are detailed in NMFS (2012).

The Expert Panel recognized that no upstream fishway is 100% effective and concluded that a baseline 1% mortality is likely at all fishways for fish that do not successfully pass (Table 3.12.2). Mortality estimates were increased for specific facilities due to a variety of reasons, such as a high percentage of fallback at a dam and, therefore, a high percentage of re-ascent and failure, the possibility of poaching-related mortality caused by migration delays, mortality due to a lack of thermal refuge for delayed adults, and the possibility of predation, mainly by seals, at the lower river dams. The logic behind assigning specific proportions of fish to return to the ocean un-spawned were related to proximity of the facility to the ocean and increased handling at the fishway trapping facility at Veazie Dam. The proportions of fish confined within the various downstream PUs after unsuccessfully attempting to ascend a particular fishway were determined by consensus within the Expert Panel.

Within the DIA Model, adult returns must pass at least one dam en route to their spawning grounds, with the exception of fish destined for PU 14. Some percentage of these fish will not successfully pass each facility according to the upstream dam passage survival rates (see Section 3.11). These unsuccessful fish will die, return to the sea unspawned, or be redirected to a downstream PU according to the proportion detailed in Table 3.12.1.

4 Results

The DIA model was run under two different scenarios: with the base case inputs (see Section 3) and with increased freshwater (i.e., egg to smolt) and marine survival rates. A scenario was run with increased survival rates (i.e., two times the base case freshwater survival and four times the base case marine survival) to simulate a recovering population of Atlantic salmon. The model run that used the base case survival rates is referred to as the "Base Case" scenario, and the model run that used the increased survival rates is referred to as the "Recovery" scenario. Five thousand iterations were run for both the Base Case and Recovery scenarios, and each iteration was run for 50 years (i.e., 10 generations).

The reported results include estimated total adult abundance, distribution of adults, and total number and proportion of smolts killed by dams. These metrics were chosen to help monitor the reproduction, numbers, and distribution of the Atlantic salmon population in different scenarios. Total adult abundance was recorded as the median number of 2SW females across all PUs. For each of three areas of the Penobscot River watershed, the distribution of adults was recorded as the proportion of iterations when at least one 2SW female was present. The three areas of the Penobscot River watershed were the Upper Penobscot (i.e., above West Enfield Dam, PUs 1-3), the Piscataquis (i.e., the Piscataquis River watershed, PUs 4-8), and the Lower Penobscot (i.e., below West Enfield Dam, PUs 9-15) (Figure 2.2 and Figure 4.1). PUs were grouped into these areas because of natural break points in the Penobscot River (i.e., the upper part of the mainstem and tributaries, a large primary tributary, and the lower part of the mainstem) and to avoid spurious results from stochasticity at the PU level. Total number of smolts killed by dams was recorded as the median number of smolts killed during emigration due to direct and indirect cumulative mortality associated with dam passage across the 15 modeled hydroelectric dams. Total proportion of smolts killed by dams was recorded as the median proportion of smolts killed during emigration due to direct and indirect cumulative mortality associated with dam passage across the 15 modeled hydroelectric dams. The total number and proportion of smolts killed by dams did not include mortality due to indirect latent mortality.

4.1 Base Case

Adult abundance and distribution decreased in the Base Case scenario. The median number of 2SW females declined from generation 1 to generation 2, and varied without trend in subsequent generations (Table 4.1.1; Figure 4.1.1). The proportion of iterations when at least one 2SW female was located in each Penobscot River watershed area equaled one in generation 1 for all three areas (Table 4.1.2; Figure 4.1.2). The proportion of iterations remained at one for the Lower Penobscot in generations 2-10 but declined from generations 1 to 4 in the Piscataquis and Upper Penobscot.

The number of smolts killed decreased in the Base Case scenario, whereas the proportion of smolts killed remained constant. The median total number of smolts killed during emigration due to direct and indirect cumulative mortality associated with dam passage was the highest in generation 1, declined from generations 1 to 2, and varied without trend in generations 2-10 (Table 4.1.3; Figure 4.1.3). The median total proportion of smolts killed during emigration due to direct and indirect cumulative mortality associated with dam passage remained constant at 0.11 for all generations (Table 4.1.4; Figure 4.1.3).

4.2 Recovery

Adult abundance increased and adult distribution remained near one in all three areas in the Recovery scenario. The median number of 2SW females increased from generation 1 and reached a plateau by generation 7 (Table 4.1.1; Figure 4.1.1). The proportion of iterations when at least one 2SW female was located in each Penobscot River watershed area equaled one or was close to one in generations 1-10, for all three areas (Table 4.1.2; Figure 4.1.2).

The number of smolts killed increased slightly overall in the Recovery scenario, whereas the proportion of smolts killed remained constant. The median number of smolts killed declined from generations 1 to 2 and then increased in subsequent generations (Table 4.1.3; Figure 4.1.3). The median total proportion of smolts killed during emigration due to direct and indirect cumulative mortality associated with dam passage was 0.10 or 0.11 for all generations (Table 4.1.4; Figure 4.1.3).

4.3 Summary

Adult abundance, adult distribution, and the number of smolts killed decreased overall in the Base Case scenario, whereas adult abundance increased, adults remained distributed throughout the Penobscot River watershed, and the number of smolts killed increased overall in the Recovery scenario. The median number of 2SW females decreased in the Base Case scenario because survival rates were too low to sustain the initial number of adults. The median number of 2SW females increased in the Recovery scenario because the increase in marine and freshwater survival rates enabled the population to grow. The proportion of iterations when at least one 2SW female was located in each Penobscot River watershed area equaled one in all three areas in generation 1 because of the PUs where adults were seeded in this generation (see Section 3.1). In the Base Case scenario, the proportion of iterations when at least one 2SW female was present in the Lower Penobscot equaled one in all generations because returning adults did not have to pass as many dams (no dams for PU 14) to access this area of the watershed. The proportion of iterations was less than one in the Piscataquis and Upper Penobscot because too few 2SW females were able to pass the dams in the Lower Penobscot to enter these areas. The number of adults was also depleted before entering the Piscataquis or Upper Penobscot because 150 2SW females were removed above Veazie Dam to fulfill hatchery broodstock requirements. The proportion of iterations when at least one 2SW female was present was high in all areas and generations of the Recovery scenario because survival rates were high and more 2SW females returned (enough to fulfill hatchery broodstock requirements and to have a large number left to attempt to pass dams) and produced smolts, which tried to home to their natal PU when they returned as adults. The number of smolts killed in the Base Case scenario decreased after generation 1 because spawning 2SW females were seeded throughout the Penobscot River watershed in generation 1, but low return rates in subsequent generations resulted in fewer spawners and, therefore, fewer smolts being produced. The number of smolts killed in the Recovery scenario increased because more smolts were available and attempted to migrate downstream. More smolts were available because of higher survival rates.

5 Analysis of Hatchery and State of Recovery

The DIA Model was used to run scenarios to test the affects of stocking of hatchery-reared smolts, freshwater and marine survival rates, and dams on the Penobscot River population of Atlantic salmon and was divided into these three parts.

Each part of the analysis included five scenarios to test the impact of dams on the Atlantic salmon population (Table 5.1). The first scenario was run as the base case scenario for dams (i.e., all dams turned on). The second scenario incorporated the proposed changes to the Penobscot River watershed that are included in the Penobscot River Restoration Project (PRRP; Trinko Lake et al. 2012). These changes include removing Veazie and Great Works dams and decommissioning and building a bypass around Howland Dam. This second scenario was represented in the DIA Model as all dams turned on with the exception of downstream and upstream passage rates at Veazie, Great Works, and Howland dams being set to one (i.e., all smolts and adults successfully pass). Although 100% survival was assumed at Howland Dam after implementation of the PRRP, this assumption was likely overly optimistic and a small amount of take will still occur. The third scenario was run with all dams turned off (i.e., all smolts and adults successfully pass) with the exception of Medway, which no adults or smolts were allowed to pass. The fourth and fifth scenarios grouped dams by whether they were located in the mainstem or a tributary (Table 5.1). The mainstem of the Penobscot River begins at the confluence of the East Branch and West Branch of the Penobscot River (Baum 1983). Only dams that are physically on the mainstem Penobscot (i.e., below the confluence of East Branch and West Branch and not impounding a tributary) are considered mainstem dams. In the fourth scenario, dams on the mainstem were turned off and dams on a tributary were turned on. In the fifth scenario, dams on the mainstem were turned on and dams on a tributary were turned off.

The reported results for each scenario include estimated total adult abundance, distribution of adults, and total number and proportion of smolts killed by dams, as recorded in Section 4. Number and proportion of smolts killed by an individual dam were also reported in the third part of this analysis. Number of smolts killed by an individual dam was recorded as the median number of smolts killed during emigration due to direct and indirect cumulative mortality associated with dam passage at each one of the 15 modeled hydroelectric dams. Proportion of smolts killed by and individual dam was recorded as the median proportion of smolts killed during emigration due to direct and indirect cumulative mortality associated with dam passage at each one of the 15 modeled hydroelectric dams. All DIA model iterations were run for 50 years (i.e., 10 generations), and 5,000 iterations were run for each scenario.

5.1 Part 1 - Hatchery On Base Case

Adult abundance and distribution decreased in all five scenarios in part 1. The median number of 2SW females declined from generation 1 to generation 2, and varied without trend in subsequent generations (Table 5.1.1; Figure 5.1.1). The proportion of iterations when at least one 2SW female was located in each Penobscot River watershed area equaled one in generation 1 for all three areas and remained at one for the Lower Penobscot in generations 2-10 (Table 5.1.2; Figure 5.1.2). The proportion of iterations declined from generations 1 to 3 in the Piscataquis and Upper Penobscot and varied without trend in subsequent generations. Adult abundance and the proportion of iterations in the Upper Penobscot and Piscataquis were lowest in the scenario with all dams turned on and highest in the scenario with all dams turned off. Adult abundance and the proportion of iterations were similar in the scenarios with all dams turned on and with mainstem dams turned on and tributary dams turned off.

The number and proportion of smolts killed differed between the scenarios in part 1. The median total number of smolts killed during emigration due to direct and indirect cumulative mortality associated with dam passage declined from generations 1 to 2, and varied without trend in generations 2-10 in all scenarios except the one with dams turned off (Table 5.1.3; Figure 5.1.3). In the latter scenario, the number of smolts killed equaled zero in all generations. The median total proportion of smolts killed during emigration due to direct and indirect cumulative mortality associated with dam passage remained constant in all generations at different values for each scenario (Table 5.1.4; Figure 5.1.3). The number and proportion of smolts killed were lowest in the scenario with all dams turned off and highest in the scenario with all dams turned on. The number and proportion of smolts killed were similar in the scenarios with all dams turned on and with mainstem dams turned on and tributary dams turned off.

5.2 Part 2 - Hatchery Off Base Case

Adult abundance and distribution decreased in all five scenarios in part 2. The median number of 2SW females decreased to zero by generation 6 in all scenarios (Table 5.2.1; Figure 5.2.1). The proportion of iterations when at least one 2SW female was located in each Penobscot River watershed area equaled one in generation 1 for all three areas and declined to zero by generation 10 in all areas and scenarios (Table 5.2.2; Figure 5.2.2).

The number and proportion of smolts killed differed between the scenarios in part 2. The median total number and proportion of smolts killed declined to zero by generation 6 in all scenarios except the one with dams turned off (Table 5.2.3 and Table 5.2.4; Figure 5.2.3). In the latter scenario, the number and proportion of smolts killed equaled zero in all generations.

5.3 Part 3 - Hatchery Off Recovery

Adult abundance increased overall and adult distribution remained at or near one in all five scenarios in part 3. The median number of 2SW females decreased from generation 1 to generation 2 and increased in subsequent generations in the scenarios with all dams turned on and with mainstem dams on and dams in tributaries turned off (Table 5.3.1; Figure 5.3.1). The number of 2SW females increased from generation1 to generation 10 in the scenarios with the implementation of the PRRP, all dams turned off, and mainstem dams turned off and dams in tributaries turned on. Adult abundance was lowest in the scenario with all dams turned on and highest in the scenario with all dams turned off. Adult abundance was similar in the scenarios with all dams turned on and with mainstem dams turned on and tributary dams turned off. The proportion of iterations when at least one 2SW female was located in each Penobscot River watershed area equaled one or was close to one in generations 1-10 for all three areas in all scenarios (Table 5.3.2; Figure 5.3.2).

The total number and proportion of smolts killed differed between the scenarios in part 3. The median total number of smolts killed during emigration due to direct and indirect cumulative mortality associated with dam passage decreased overall in the scenarios with all dams turned on and with mainstem dams turned on and dams in the tributaries turned off, increased overall in the scenarios with the implementation of the PRRP and with mainstem dams turned off and dams in tributaries turned on, and equaled zero in all generations in the scenario with all dams turned off (Table 5.3.3; Figure 5.3.3). The median total proportion of smolts killed during emigration due to direct and indirect cumulative mortality associated with dam passage declined in the scenarios with all dams turned on, with the implementation of the PRRP, and with mainstem dams turned on and dams in tributaries turned off (Table 5.3.4; Figure 5.3.3). The proportion of smolts killed remained low in the scenarios with mainstem dams turned off and dams in tributaries turned on and with all dams turned off.

5.3.1 Individual Dam Impacts

The number and proportion of smolts killed at individual dams differed between the scenarios in part 3. The number and proportion of smolts killed at Medway, Sebec, and Milo dams equaled zero in all generations for all scenarios because fish are not able to access habitat above these dams. In general, higher numbers of smolts were killed at dams that were located on the mainstem of the Penobscot River and close to the river mouth. The numbers of smolts killed in generation 10 were higher at these dams because fewer 2SW females were able to pass dams that were higher in the watershed in each subsequent generation. Hence, more fish spawned in and migrated out from lower PUs and were killed by dams that were lower in the watershed in later generations. The proportion of smolts killed at each dam depended on the characteristics of each individual dam.

In the scenario with all dams turned on, the median number of smolts killed during emigration due to direct and indirect cumulative mortality associated with dam passage decreased from generation 1 to generation 3 or 4 at all dams except at Frankfort Dam, where the number of smolts killed increased in every generation (Table 5.3.1.1; Figure 5.3.1.1). The number of smolts killed in generation 10 was highest at Veazie, Great Works, Frankfort, and Milford dams. In general, the number of smolts killed was lower at dams that were located farther from the mouth of the Penobscot River or on a tributary rather than the mainstem of the Penobscot River. Frankfort Dam was an exception to this rule because it is the closest hydroelectric dam to the mouth of the Penobscot River. The median proportion of smolts killed during emigration due to direct and indirect cumulative mortality associated with dam passage remained constant at dams that were closer to the mouth of the Penobscot River and decreased at dams that were farther from the river mouth (Table 5.3.1.2; Figure 5.3.1.2). The proportion of smolts killed in generation 10 was highest at Great Works Dam.

In the scenario with the implementation of the PRRP, the median number and proportion of smolts killed during emigration due to direct and indirect cumulative mortality associated with dam passage equaled zero at Veazie, Great Works, and Howland dams because passage and survival were set to one at these dams due to their removal as part of the PRRP (Table 5.3.1.3 and Table 5.3.1.4; Figure 5.3.1.3 and Figure 5.3.1.4). The median number of smolts killed increased overall at the dams that were closer to the mouth of the Penobscot River and decreased overall at dams that were farther from the river mouth. The number of smolts killed in generation 10 was highest at Milford Dam. The median proportion of smolts killed during emigration due to direct and indirect cumulative mortality associated with dam passage remained constant or decreased slightly at all hydroelectric dams in the Penobscot River watershed. The proportion of smolts killed in generation 10 was highest at Mattaceunk Dam.

In the scenario with all dams turned off, the median number and proportion of smolts killed during emigration due to direct and indirect cumulative mortality associated with dam passage equaled zero in all generations. Passage and survival were set to one at all dams in this scenario. Therefore, no smolt mortality occurred.

In the scenario with mainstem dams turned off and dams in tributaries turned on, the median number and proportion of smolts killed during emigration due to direct and indirect cumulative mortality associated with dam passage equaled zero at Veazie, Great Works, Milford, West Enfield, and Mattaceunk dams because passage and survival were set to one at these dams (Table 5.3.1.5 and Table 5.3.1.6; Figure 5.3.1.5 and Figure 5.3.1.6). The median number of smolts killed increased overall at all tributary dams where fish had access to the habitat above the dam. The number of smolts killed in generation 10 was highest at Howland and Orono dams. Although Orono Dam is closer to the mouth of the Penobscot River, fewer smolts were killed at this dam because the lack of upstream passage at Orono Dam likely inhibited adults from returning, spawning, and producing smolts in PU 11. Plus, only a proportion of the smolts migrate downstream through the Stillwater Branch. The median proportion of smolts killed during emigration due to direct and indirect cumulative mortality associated with dam passage remained constant at all tributary dams where fish had access to the habitat above the dam. The proportion of smolts killed in generation 10 was highest at Brown's Mills Dam.

In the scenario with mainstem dams turned on and tributary dams turned off, the median number and proportion of smolts killed during emigration due to direct and indirect cumulative mortality associated with dam passage equaled zero at Frankfort, Orono, Stillwater, Lowell, Howland, Brown's Mills, and Dover Upper dams because passage and survival were set to one at these dams (Table 5.3.1.7 and Table 5.3.1.8; Figure 5.3.1.7 and Figure 5.3.1.8). The median number of smolts killed decreased overall at all mainstem dams. The number of smolts killed in generation 10 was highest at Veazie Dam. The median proportion of smolts killed during emigration due to direct and indirect cumulative mortality associated with dam passage remained constant at all mainstem dams except Mattaceunk dam, which decreased. The proportion of smolts killed in generation 10 was highest at Great Works Dam.

5.4 Summary

In the DIA Model, stocking of hatchery-reared smolts sustained the Penobscot River Atlantic salmon population when freshwater and marine survival rates were at base case values. In part 1, when the hatchery component of the model was turned on, adult abundance and distribution declined in the first two and three generations, respectively, but lower levels were maintained throughout the rest of the times series. In part 2, when the hatchery component of the model was turned off, adult abundance and distribution declined to zero within a few generations. The numbers of smolts killed decreased in parts 1 and 2 because low return rates of 2SW females after generation 1 resulted in fewer spawners and, therefore, fewer smolts being produced and killed.

In the DIA model, a two-fold increase in freshwater survival and a four-fold increase in marine survival were able to sustain and increase the population of Atlantic salmon in the Penobscot River watershed when no smolts were stocked, as shown in part 3. Adult abundance increased and adult distribution equaled or was close to one when freshwater and marine survival rates were increased and the hatchery component of the model was turned off in part 3. In contrast, adult abundance and distribution decreased in part 1 and declined to zero in part 2. In part 3, the number of smolts killed declined initially but increased at the end of the time series because higher survival rates led to more adults and, therefore, more smolts being in the watershed and attempting to migrate downstream at the end of the time series. In contrast, the numbers of smolts decreased in part 1 and declined to zero in part 2.

The numbers and locations of dams that were turned on in the DIA model affected the population of Atlantic salmon. In parts 1, 2, and 3, adult abundance and distribution were lowest in the scenarios with all dams turned on and highest in the scenarios with all dams turned off (i.e., 100% passage of adults and smolts). Adult abundance and distribution were higher in the scenarios with mainstem dams turned off and dams in tributaries turned on than in the scenarios with mainstem dams turned on and dams in tributaries turned off. Adult abundance and distribution were also higher in the PRRP scenarios than in the scenarios with mainstem dams turned on and dams in the tributaries turned off but were lower in the PRRP scenarios than in the scenarios with mainstem dams turned off and dams in the tributaries turned on. The results of the scenarios with only some dams turned off imply that dams in the mainstem of the Penobscot River are more detrimental to the DIA Model population of Atlantic salmon than dams in the tributaries. This likely occurred because most Atlantic salmon have to attempt to pass dams in the mainstem to reach the ocean or their natal PU, whereas fewer Atlantic salmon migrate through and encounter dams in the tributaries. Aside from the scenarios with all dams turned off, fewer smolts were killed when more dams were turned on and when dams in the mainstem of the Penobscot River were turned on. This occurred because survival rates were higher in the scenarios with fewer dams on overall and fewer dams in the mainstem turned on. Higher survival rates led to more adults and smolts produced in these scenarios, enabling more smolts to be killed. No smolts were killed in the scenarios with all dams turned off because passage and survival were set at 100% for smolts.

6 Model Diagnostics and Sensitivity Analyses

The DIA Model was evaluated using model diagnostics and sensitivity analyses (Table 6.1). The model diagnostics (Sections 6.1 and 6.2) examined the appropriate number of model iterations to run for each scenario and the stability in results for a given number of iterations (Legault 2004). The sensitivity analyses (Sections 6.3-6.22) examined which model inputs had the most influence on model results (McCarthy et al. 1996; Cross and Beissinger 2001) and were performed by holding model inputs at the base value while changing one input at a time. The model diagnostics and sensitivity analyses were also run with Base Case and Recovery scenarios (see Section 4). The number of iterations used in the model diagnostics runs depended on the scenario that was being tested, whereas 5,000 iterations were run for all sensitivity analysis scenarios. Each model diagnostic and sensitivity analysis iteration was run for 50 years (i.e., 10 generations). The reported results include estimated total adult abundance, distribution of adults, and total number and proportion of smolts killed by dams and were recorded as in Section 4.

6.1 Number of Iterations

When performing Monte Carlo simulations, the appropriate number of model iterations to use must be found by trial and error. This can be done by conducting trials using different numbers of model iterations and comparing the variability in the results. Conducting more simulations produces more consistent results but takes more computation time. The Base Case and Recovery scenarios were run with 100, 500, 1,000, 5,000, and 10,000 iterations.

Each result (i.e., adult abundance, distribution of adults, and total number and proportion of smolts killed by dams) was similar for all numbers of iterations tested among the Base Case scenarios and among the Recovery scenarios (Tables 6.1.1-6.1.8; Figures 6.1.1-6.1.6). As expected, results varied most when the model was run with 100 iterations and least when the model was run with 10,000 iterations. The decrease in variability seemed especially noticeable when 1,000 or more iterations were run. The variability between results from using 5,000 iterations and 10,000 iterations was considered minimal, and the additional time to run 10,000 iterations compared to 5,000 iterations did not seem justified. Therefore, 5,000 iterations was used as the standard for all Base Case and Recovery scenarios.

6.2 Model Stability

The model was run five times each under the Base Case and Recovery scenarios to look at the variability in the results between runs with 5,000 iterations. This was a second test to ensure the results would be stable when using 5,000 iterations.

Each model result was similar among the five Base Case scenario runs and among the five Recovery scenario runs (Tables 6.2.1-6.2.8; Figures 6.2.1-6.2.6). The variability between runs was considered acceptable. Hence, 5,000 iterations were confirmed as the standard for all Base Case and Recovery scenarios.

6.3 Production Potential Cap

The number of wild smolts that originated in each PU was limited with a production potential cap, which was the maximum number of smolts allowed per HU (i.e., 10 smolts per 100 m2; Table 3.1.2). The production potential cap represented the number of wild smolts that the habitat could support. No other density-dependent effects were included in the model. Ten smolts per 100 m2 was used as the base input value, and sensitivities were run at values of 0.25, 0.5, 2, and 4 times the base value (i.e., 2.5, 5, 20, and 40 smolts per 100 m2, respectively).

6.3.1 Base Case

Adult abundance and distribution decreased in all five Base Case scenarios. These results were not sensitive to the production potential cap in the Base Case scenarios. The median number of 2SW females declined from generation 1 to generation 2, and varied without trend in subsequent generations (Table 6.3.1.1; Figure 6.3.1.1). The proportion of iterations when at least one 2SW female was located in each Penobscot River watershed area equaled one in generation 1 for all three areas (Table 6.3.1.2; Figure 6.3.1.2). The proportion of iterations remained at one for the Lower Penobscot in generations 2-10. The proportion of iterations for the Upper Penobscot and Piscataquis were similar, declining from generation 1 to 3 and varying without trend in subsequent generations for all five scenarios.

The number of smolts killed decreased in all five Base Case scenarios, whereas the proportion of smolts killed remained constant. The number and proportion of smolts killed were not sensitive to the production potential cap in the Base Case scenarios. The median total number of smolts killed during emigration due to direct and indirect cumulative mortality associated with dam passage was the highest in generation 1, declined from generation 1 to generation 2, and varied without trend in generations 2-10 (Table 6.3.1.3; Figure 6.3.1.3). The median total proportion of smolts killed during emigration due to direct and indirect cumulative mortality associated with dam passage remained constant at 0.11 for all generations (Table 6.3.1.4; Figure 6.3.1.3).

6.3.2 Recovery

Adult abundance increased and adult distribution remained near one in all five Recovery scenarios. The median number of 2SW females increased from generation 1 and reached a plateau by generation 10 (Table 6.3.2.1; Figure 6.3.2.1). The plateau occurred at the lowest abundance and earliest generation when the production potential cap was the lowest (2.5 smolts per 100 m2) and occurred at the highest abundance and latest generation when the production potential cap was the highest (40 smolts per 100 m2). The proportion of iterations when at least one 2SW female was located in each Penobscot River watershed area equaled one or was close to one in generations 1-10, for all three areas (Table 6.3.2.2; Figure 6.3.2.2).

The number of smolts killed differed among the Recovery scenarios, whereas the proportion of smolts killed remained constant among the scenarios. In the scenario with a production potential cap of 2.5 smolts per 100 m2, the median total number of smolts killed during emigration due to direct and indirect cumulative mortality associated with dam passage declined from generation 1 to generation 2, and varied without trend in subsequent generations (Table 6.3.2.3; Figure 6.3.2.3). In all other scenarios, the number of smolts killed declined from generation 1 to generation 2 and then increased in subsequent generations. In generation 10, the fewest smolts were killed in the scenario with a production potential cap of 2.5 smolts per 100 m2, and the most smolts were killed in the scenario with a production potential cap of 40 smolts per 100 m2. The median total proportion of smolts killed during emigration due to direct and indirect cumulative mortality associated with dam passage equaled 0.10 or 0.11 for all generations (Table 6.3.2.4; Figure 6.3.2.3).

6.4 Eggs per Female

The number of eggs produced per 2SW female was drawn from a normal distribution with μ = 8,304 and σ = 821 (Figure 3.2.1). This distribution was used as the base input, and sensitivities were run at values of 0.25, 0.5, 2, and 4 times the base.

6.4.1 Base Case

Adult abundance and distribution decreased overall in all five Base Case scenarios. The median number of 2SW females declined from generation 1 to generation 2 in all scenarios, varied without trend in subsequent generations in the scenarios with 0.25, 0.5 and 1 times the base eggs per female rate, and increased in subsequent generations in the scenarios with 2 and 4 times the base eggs per female rate (Table 6.4.1.1; Figure 6.4.1.1). Adult abundance was lowest in the scenario with 0.25 times the base and highest in the scenario with 4 times the base. The proportion of iterations when at least one 2SW female was located in each Penobscot River watershed area equaled one in generation 1 for all three areas (Table 6.4.1.2; Figure 6.4.1.2). The proportion of iterations remained at one for the Lower Penobscot in generations 2-10. The proportion of iterations for Piscataquis and Upper Penobscot were similar, declining from generation 1 to 3 and varying without trend in subsequent generations for all five scenarios.

The number of smolts killed decreased in all five Base Case scenarios, whereas the proportion of smolts killed remained constant. The median total number of smolts killed during emigration due to direct and indirect cumulative mortality associated with dam passage was the highest in generation 1, declined from generation 1 to generation 2, and varied without trend in generations 2-10 (Table 6.4.1.3; Figure 6.4.1.3). The median total proportion of smolts killed during emigration due to direct and indirect cumulative mortality associated with dam passage was 0.10 or 0.11 for all generations (Table 6.4.1.4; Figure 6.4.1.3).

6.4.2 Recovery

Adult abundance increased in all five Recovery scenarios, whereas adult distribution differed by scenario. The median number of 2SW females increased from generation 1 and reached a plateau by generation 10 (Table 6.4.2.1; Figure 6.4.2.1). The plateau occurred at the lowest abundance and earliest generation in the scenario with 0.25 times the base and occurred at the highest abundance and latest generation in the scenario with 4 times the base. The proportion of iterations when at least one 2SW female was located in each Penobscot River watershed area equaled one in generation 1 for all three areas and remained at one for the Lower Penobscot in generations 2-10 for all scenarios (Table 6.4.2.2; Figure 6.4.2.2). The proportion of iterations for the Upper Penobscot and Piscataquis declined from generation 1 to 2 and increased in subsequent generations. In the Upper Penobscot and Piscataquis, the proportion of iterations was the lowest in the scenario with the lowest eggs per female rate (i.e., 0.25 times the base) and equaled or was near one in all other scenarios.

The number of smolts killed differed among the Recovery scenarios, whereas the proportion of smolts killed remained constant among the scenarios. In the scenario with 0.25 times the base eggs per female rate, the median total number of smolts killed during emigration due to direct and indirect cumulative mortality associated with dam passage declined from generation 1 to generation 2 and varied without trend in subsequent generations (Table 6.4.2.3; Figure 6.4.2.3). In the scenarios with 0.5, 1, and 2 times the base, the median number of smolts killed declined from generation 1 to generation 2 and increased in subsequent generations. In the scenario with 4 times the base, the median number of smolts killed increased in all generations. The median total proportion of smolts killed during emigration due to direct and indirect cumulative mortality associated with dam passage was 0.10 or 0.11 for all generations (Table 6.4.2.4; Figure 6.4.2.3).

6.5 Egg to Smolt Survival

The survival rate from the egg to smolt life stages was drawn from a lognormal distribution with μ = 1.31%, minimum = 0.10%, and maximum = 5.88% (Figure 3.3.5). This distribution was used as the base input, and sensitivities were run at values of 0.25, 0.5, 2, and 4 times the base.

6.5.1 Base Case

Adult abundance and distribution decreased overall in all five Base Case scenarios. The median number of 2SW females declined from generation 1 to generation 2 in all scenarios, varied without trend in subsequent generations in the scenarios with 0.25, 0.5 and 1 times the base egg to smolt survival rate, and increased in subsequent generations in the scenarios with 2 and 4 times the base egg to smolt survival rate (Table 6.5.1.1; Figure 6.5.1.1). The number of 2SW females was lowest in the scenario with 0.25 times the base and highest in the scenario with 4 times the base. The proportion of iterations when at least one 2SW female was located in each Penobscot River watershed area equaled one in generation 1 for all three areas and remained at one for the Lower Penobscot in generations 2-10 (Table 6.5.1.2; Figure 6.5.1.2). The proportion of iterations for the Upper Penobscot and Piscataquis were similar, declining from generation 1 to 3 and varying without trend in subsequent generations for all five scenarios, and were the highest when the egg to smolt survival rate was the greatest (i.e., 4 times the base) and the lowest when the egg to smolt survival rate was the least (i.e., 0.25 times the base).

The number of smolts killed decreased in all five Base Case scenarios, whereas the proportion of smolts killed remained constant. The median total number of smolts killed during emigration due to direct and indirect cumulative mortality associated with dam passage was the highest in generation 1, declined from generation 1 to generation 2 (base times 0.25, 0.5, and 1) or generation 3 (base times 2 and 4), and varied without trend in subsequent generations (Table 6.5.1.3; Figure 6.5.1.3). The median total proportion of smolts killed during emigration due to direct and indirect cumulative mortality associated with dam passage was 0.10 or 0.11 for all generations (Table 6.5.1.4; Figure 6.5.1.3).

6.5.2 Recovery

Adult abundance increased in all five Recovery scenarios, whereas adult distribution differed by scenario. The median number of 2SW females increased from generation 1 and reached a plateau by generation 10 (Table 6.5.2.1; Figure 6.5.2.1). The plateau occurred at the lowest abundance and earliest generation in the scenario with 0.25 times the base and occurred at the highest abundance and latest generation in the scenario with 4 times the base. The proportion of iterations when at least one 2SW female was located in each Penobscot River watershed area equaled one in generation 1 for all three areas (Table 6.5.2.2; Figure 6.5.2.2). The proportion of iterations remained at one for the Lower Penobscot in generations 2-10 for all scenarios. The proportion of iterations for Piscataquis and Upper Penobscot declined from generation 1 to 2 and varied without trend in subsequent generations. The proportion of iterations in the two latter areas was lowest in the scenario with 0.25 times the base and was at or near one in the scenarios with 2 and 4 times the base.

The number of smolts killed differed between the Recovery scenarios, whereas the proportion of smolts killed remained constant between the scenarios. In the scenario with 0.25 times the base eggs per female rate, the median total number of smolts killed during emigration due to direct and indirect cumulative mortality associated with dam passage varied without trend in all generations (Table 6.5.2.3; Figure 6.5.2.3). In the scenarios with 0.5 and 1 times the base, the median number of smolts killed declined from generation 1 to generation 2, and varied without trend in subsequent generations. In the scenarios with 2 and 4 times the base, the median number of smolts killed declined from generation 1 to generation 2 and then increased in subsequent generations. The median total proportion of smolts killed during emigration due to direct and indirect cumulative mortality associated with dam passage was 0.10 or 0.11 for all generations (Table 6.5.2.4; Figure 6.5.2.3).

6.6 In-river Mortality

The in-river mortality rate was drawn from a beta distribution with shape parameters α1 = 11.245 and α2 = 9.8007, minimum = zero, and maximum = 0.00038077 (Figure 3.5.2). This distribution was used as the base input, and sensitivities were run at values of 0.25, 0.5, 2, and 4 times the base.

6.6.1 Base Case

Adult abundance and distribution decreased in all five Base Case scenarios. These results were not sensitive to the in-river mortality rate in the Base Case scenarios. The median number of 2SW females declined from generation 1 to generation 2, and varied without trend in subsequent generations (Table 6.6.1.1; Figure 6.6.1.1). The proportion of iterations when at least one 2SW female was located in each Penobscot River watershed area equaled one in generation 1 for all three areas (Table 6.6.1.2; Figure 6.6.1.2). The proportion of iterations remained at one for the Lower Penobscot in generations 2-10. The proportion of iterations for Piscataquis and Upper Penobscot were similar, declining from generation 1 to 3 and varying without trend in subsequent generations for all five scenarios.

The number of smolts killed decreased in all five Base Case scenarios, whereas the proportion of smolts killed remained constant. The number and proportion of smolts killed were not sensitive to the in-river mortality rate in the Base Case scenarios. The median total number of smolts killed during emigration due to direct and indirect cumulative mortality associated with dam passage was the highest in generation 1, declined from generation 1 to generation 2, and varied without trend in generations 2-10 (Table 6.6.1.3; Figure 6.6.1.3). The median total proportion of smolts killed during emigration due to direct and indirect cumulative mortality associated with dam passage remained constant at 0.11 for all generations (Table 6.6.1.4; Figure 6.6.1.3).

6.6.2 Recovery

Adult abundance increased and adult distribution remained near one in all five Recovery scenarios. The median number of 2SW females increased from generation 1 and reached a plateau by generation 10 (Table 6.6.2.1; Figure 6.6.2.1). The plateau occurred at the lowest abundance when the in-river mortality was the highest (base times 4) and occurred at the highest abundance when in-river mortality was the lowest (base times 0.25). The proportion of iterations when at least one 2SW female was located in each Penobscot River watershed area equaled one or was close to one in generations 1-10, for all three areas (Table 6.6.2.2; Figure 6.6.2.2).

The number of smolts killed during emigration due to direct and indirect cumulative mortality associated with dam passage increased overall in the Recovery scenarios, whereas the proportion of smolts killed remained constant. In all scenarios, the median number of smolts killed during emigration due to direct and indirect cumulative mortality associated with dam passage declined from generation 1 to generation 2, and increased to generation 10 (Table 6.6.2.3; Figure 6.6.2.3). The median total proportion of smolts killed during emigration due to direct and indirect cumulative mortality associated with dam passage was 0.10 or 0.11 for all generations (Table 6.6.2.4; Figure 6.6.2.3).

6.7 Marine Survival

The marine survival rate was drawn from an inverse gaussian distribution with μ = 0.006265, shape parameter l = 0.0068723, and a shift of 0.00000813424 (Figure 3.9.4). This distribution was used as the base input, and sensitivities were run at values of 0.25, 0.5, 2, and 4 times the base. Because this sensitivity analysis was performed on the marine survival rate, the Base Case and Recovery scenarios are different than most of the other sensitivity analyses. The Base Case scenarios were run with the base freshwater survival rate, and the Recovery scenarios were run with freshwater survival increased by two times the base value, as in the other sensitivity analyses. Unlike the other sensitivity analyses, five marine survival values were tested in both the Base Case and Recovery scenarios.

6.7.1 Base Case

Adult abundance and distribution differed between Base Case scenarios. The median number of 2SW females decreased in the scenarios with 0.25, 0.5, 1, and 2 times the base and increased in the scenario with 4 times the base (Table 6.7.1.1; Figure 6.7.1.1). The proportion of iterations when at least one 2SW female was located in each Penobscot River watershed area equaled one in generation 1 for all three areas and remained at one for the Lower Penobscot in generations 2-10 (Table 6.7.1.2; Figure 6.7.1.2). The proportion of iterations for Piscataquis and Upper Penobscot were similar in each scenario, with the proportion being the lowest (close to zero) in the scenario with 0.25 times the base and highest (close to one) in the scenario with 4 times the base.

The number of smolts killed differed between Base Case scenarios, whereas the proportion of smolts killed remained constant. The median total number of smolts killed during emigration due to direct and indirect cumulative mortality associated with dam passage was the highest in generation 1, declined from generation 1 to generation 2, and varied without trend in generations 2-10 in all scenarios except 4 times the base (Table 6.7.1.3; Figure 6.7.1.3). In the latter scenario, the number of smolts killed increased after generation 2. The median total proportion of smolts killed during emigration due to direct and indirect cumulative mortality associated with dam passage remained constant at 0.11 for all generations (Table 6.7.1.4; Figure 6.7.1.3).

6.7.2 Recovery

Adult abundance and distribution differed between Recovery scenarios. The median number of 2SW females decreased in the scenarios with 0.25, 0.5, and 1 times the base and increased in the scenarios with 2 and 4 times the base (Table 6.7.2.1; Figure 6.7.2.1). The proportion of iterations when at least one 2SW female was located in each Penobscot River watershed area equaled one in generation 1 for all three areas and remained at one for the Lower Penobscot in generations 2-10 (Table 6.7.2.2; Figure 6.7.2.2). The proportion of iterations for Piscataquis and Upper Penobscot were similar in each scenario, with the proportion being the lowest (close to zero) in the scenario with 0.25 times the base and highest (close to one) in the scenario with 4 times the base.

The number of smolts killed differed between Recovery scenarios, whereas the proportion of smolts killed remained constant. The median total number of smolts killed during emigration due to direct and indirect cumulative mortality associated with dam passage was the highest in generation 1, declined from generation 1 to generation 2, and varied without trend in generations 2-10 in all scenarios except 4 times the base (Table 6.7.2.3; Figure 6.7.2.3). In the latter scenario, the number of smolts killed increased after generation 2. The median total proportion of smolts killed during emigration due to direct and indirect cumulative mortality associated with dam passage remained constant at 0.11 for all generations (Table 6.7.2.4; Figure 6.7.2.3).

6.8 Initial Number of Adults

The model was seeded with 587 2SW females in generation 1, which was the mean annual number of 2SW female returns captured at the trap above Veazie Dam during 2002-2011. This value was used as the base input, and sensitivities were run at values of 0.25, 0.5, 2, and 4 times the base value (i.e., 147, 294, 1,174, and 2,348, respectively).

6.8.1 Base Case

Adult abundance and distribution decreased in all five Base Case scenarios. The median number of 2SW females decreased at different rates from generation 1 to generation 2 in the five scenarios, but abundance was approximately the same in all scenarios by generation 10 (Table 6.8.1.1; Figure 6.8.1.1). The proportion of iterations when at least one 2SW female was located in each Penobscot River watershed area equaled one in generation 1 for all three areas and remained at one for the Lower Penobscot in generations 2-10 (Table 6.8.1.2; Figure 6.8.1.2). The proportion of iterations for Piscataquis and Upper Penobscot declined from one in generation 1 to approximately 0.5 by generation 10 in all five scenarios.

The number of smolts killed decreased in the Base Case scenarios, whereas the proportion of smolts killed remained constant. The median total number of smolts killed during emigration due to direct and indirect cumulative mortality associated with dam passage was the highest in generation 1, declined from generation 1 to generation 2, and varied without trend in generations 2-10 (Table 6.8.1.3; Figure 6.8.1.3). The median total proportion of smolts killed during emigration due to direct and indirect cumulative mortality associated with dam passage equaled 0.10 or 0.11 for all generations (Table 6.8.1.4; Figure 6.8.1.3).

6.8.2 Recovery

Adult abundance increased overall and adult distribution equaled or was close to one in all Recovery scenarios. The median number of 2SW females increased from generation 1 to generation 10 in all scenarios except when starting abundance equaled 2,348 2SW females (Table 6.8.2.1; Figure 6.8.2.1). In the latter scenario, adult abundance decreased from generation 1 to 2 and then increased. By generation 10, median adult abundance was close to the same number in all scenarios. The proportion of iterations when at least one 2SW female was located in each Penobscot River watershed area equaled one in generation 1 for all three areas and remained at one for the Lower Penobscot in generations 2-10 (Table 6.8.2.2; Figure 6.8.2.2). The proportion of iterations for Piscataquis and Upper Penobscot decreased from generation 1 to generation 2 but equaled or was near one in generations 3-10 in all scenarios except when starting abundance equaled 2,348 2SW females. The proportion of iterations for Piscataquis and Upper Penobscot equal or was close to one in all generations in the latter scenario.

The number of smolts killed differed between Recovery scenarios, whereas the proportion of smolts killed remained constant. The median total number of smolts killed during emigration due to direct and indirect cumulative mortality associated with dam passage increased overall when initial adult abundance was 147, 294, and 587 but decreased when initial adult abundance was 1,174 and 2,348 (Table 6.8.2.3; Figure 6.8.2.3). The median total proportion of smolts killed during emigration due to direct and indirect cumulative mortality associated with dam passage equaled 0.10 or 0.11 for all generations (Table 6.8.2.4; Figure 6.8.2.3).

6.9 Hatchery Stocking

In the base hatchery input, the hatchery was turned on, meaning 550,000 smolts were stocked and 150 2SW females were removed above Veazie Dam to fulfill stocking requirements. Smolt stocking and removal of adults for broodstock occurred in all 50 years (i.e., ten generations). Sensitivities were run with the hatchery turned off for the whole time period, the hatchery turned on for the first 25 years and off for the second 25 years, and the hatchery turned off for the first 25 years and on for the second 25 years.

6.9.1 Base Case

Adult abundance and distribution differed between Base Case scenarios. The median number of 2SW females decreased to zero when the hatchery was turned off but maintained a low level of abundance when the hatchery was turned on (Table 6.9.1.1; Figure 6.9.1.1). The proportion of iterations when at least one 2SW female was located in each Penobscot River watershed area equaled one in generation 1 for all three areas (Table 6.9.1.2; Figure 6.9.1.2). The proportion of iterations decreased to zero or near zero in all three areas when the hatchery was turned off. Adult abundance and distribution increased in generation 6 in the scenario with the hatchery turned on for the first 25 years, whereas adult abundance and distribution remained nearly the same in generation 6 in the scenario with the hatchery turned on for the whole time period. The increase in generation 6 mentioned above was caused by leaving 150 2SW females in the river to spawn when the hatchery was turned off instead of removing them above Veazie Dam for broodstock.

The number and proportion of smolts killed differed between Base Case scenarios. The median total number and total proportion of smolts killed during emigration due to direct and indirect cumulative mortality associated with dam passage decreased to zero when the hatchery was turned off (Table 6.9.1.3 and Table 6.9.1.4; Figure 6.9.1.3). The number and proportion of smolts killed remained stable when the hatchery was on.

6.9.2 Recovery

Adult abundance and distribution differed between Recovery scenarios. The median number of 2SW females moved toward a lower or higher equilibrium point when the hatchery was turned off or on, respectively (Table 6.9.2.1; Figure 6.9.2.1). By generation 10, adult abundance in the scenario with the hatchery turned on for the first 25 years approached adult abundance in the scenario with the hatchery turned off for the whole time series, and adult abundance in the scenario with the hatchery turned on in the second 25 years approached adult abundance in the scenario with the hatchery turned on for the whole time series. In the scenarios with the hatchery turned on for the first 25 years and the hatchery turned on for the second 25 years, the median number of 2SW females in generation 6 differed from the scenarios with the hatchery turned on and off, respectively. The difference in generation 6 was caused by leaving in or removing 150 2SW females above Veazie Dam. The proportion of iterations when at least one 2SW female was located in each Penobscot River watershed area equaled or was near one in all generations, areas, and scenarios except in the scenario with the hatchery turned off for the first 25 years and on for the second 25 years (Table 6.9.2.2; Figure 6.9.2.2). In the latter scenario, the proportion of iterations dropped in generation 6 (when the hatchery was turned on) in the Upper Penobscot and the Piscataquis and rebounded in subsequent generations. This was caused by the removal of 150 2SW females above Veazie Dam for broodstock.

The number and proportion of smolts killed differed between Recovery scenarios. The median total number of smolts killed during emigration due to direct and indirect cumulative mortality associated with dam passage decreased overall when the hatchery was turned off and increased when the hatchery was turned on in the Recovery scenarios (Table 6.9.2.3; Figure 6.9.2.3). The median total proportion of smolts killed during emigration due to direct and indirect cumulative mortality associated with dam passage decreased when the hatchery was turned off but remained constant at 0.11 when the hatchery was turned on (Table 6.9.2.4; Figure 6.9.2.3).

6.10 Hatchery Discount

The hatchery discount was applied to hatchery-origin smolts to convert the number of wild-equivalents before they migrated out to sea and was drawn from a log logistic distribution, with γ = 1, β = 1.4271, α = 1.9922, and maximum = 12 (Figure 3.8.1). This distribution was used as the base input, and sensitivities were run at values of 0.25, 0.5, 2, and 4 times the base.

6.10.1 Base Case

Adult abundance and distribution differed between Base Case scenarios. The median number of 2SW females increased in the scenario with 0.25 times the base hatchery discount rate and decreased overall in the scenarios with 0.5, 1, 2, and 4 times the base (Table 6.10.1.1; Figure 6.10.1.1). The proportion of iterations when at least one 2SW female was located in each Penobscot River watershed area equaled one in generation 1 for all three areas and remained at one for the Lower Penobscot in generations 2-10 (Table 6.10.1.2; Figure 6.10.1.2). The proportion of iterations for the Upper Penobscot and Piscataquis were similar in each scenario and decreased more as the hatchery discount increased.

The number of smolts killed decreased in all five Base Case scenarios, whereas the proportion of smolts killed remained constant. The median total number of smolts killed during emigration due to direct and indirect cumulative mortality associated with dam passage was the highest in generation 1, declined from generation 1 to generation 2, and varied without trend in generations 2-10 (Table 6.10.1.3; Figure 6.10.1.3). The median total proportion of smolts killed during emigration due to direct and indirect cumulative mortality associated with dam passage equaled 0.10 or 0.11 in all generations (Table 6.10.1.4; Figure 6.10.1.3).

6.10.2 Recovery

Adult abundance increased in all five Recovery scenarios, whereas adult distribution differed between scenarios. The median number of 2SW females increased in all scenarios and was the highest in the scenario with the lowest hatchery discount (i.e., 0.25 times the base) (Table 6.10.2.1; Figure 6.10.2.1). The proportion of iterations when at least one 2SW female was located in each Penobscot River watershed area equaled one in generation 1 for all three areas and remained at one for the Lower Penobscot in generations 2-10 (Table 6.10.2.2; Figure 6.10.2.2). The proportion of iterations for the Upper Penobscot and Piscataquis were similar in each scenario, with the proportion decreasing as the hatchery discount increased.

The number of smolts killed differed between Recovery scenarios, whereas the proportion of smolts killed remained constant. The median total number of smolts killed during emigration due to direct and indirect cumulative mortality associated with dam passage increased overall in the scenarios with 0.25, 0.5, and 1 times the base hatchery discount rate and decreased overall in the scenarios with 2 and 4 times the base (Table 6.10.2.3; Figure 6.10.2.3). The median total proportion of smolts killed during emigration due to direct and indirect cumulative mortality associated with dam passage equaled 0.10 or 0.11 for all generations (Table 6.10.2.4; Figure 6.10.2.3).

6.11 Number of Smolts Stocked

In the base hatchery input, 550,000 smolts were stocked annually. This value was used as the base input for the number of smolts stocked, and sensitivities were run at values of 0.25, 0.5, 2, and 4 times the base value (i.e., 137,500, 275,000, 1,100,000, and 2,200,000, respectively).

6.11.1 Base Case

Adult abundance and distribution differed between Base Case scenarios. The median number of 2SW females decreased overall in the scenarios with 0.25, 0.5, 1, and 2 times the base number of smolts stocked and increased in the scenario with 4 times the base (Table 6.11.1.1; Figure 6.11.1.1). The proportion of iterations when at least one 2SW female was located in each Penobscot River watershed area equaled one in generation 1 for all three areas and remained at one for the Lower Penobscot in generations 2-10 in all scenarios (Table 6.11.1.2; Figure 6.11.1.2). The proportion of iterations for the Upper Penobscot and Piscataquis were the lowest in the scenario with 0.25 times the base number of smolts stocked and increased as the number of smolts stocked increased.

The number of smolts killed decreased in all five Base Case scenarios, whereas the proportion of smolts killed remained constant. The median total number of smolts killed during emigration due to direct and indirect cumulative mortality associated with dam passage was the highest in generation 1, declined from generation 1 to generation 2, and varied without trend in generations 2-10 (Table 6.11.1.3; Figure 6.11.1.3). The median total proportion of smolts killed during emigration due to direct and indirect cumulative mortality associated with dam passage equaled 0.11 in all generations and scenarios (Table 6.11.1.4; Figure 6.11.1.3).

6.11.2 Recovery

Adult abundance increased overall in all five Recovery scenarios, whereas adult distribution differed between scenarios. The median number of 2SW females increased in all scenarios and was the highest in the scenario with 4 times the base number of smolts stocked (Table 6.11.2.1; Figure 6.11.2.1). The proportion of iterations when at least one 2SW female was located in each Penobscot River watershed area equaled one in generation 1 for all three areas and remained at one for the Lower Penobscot in generations 2-10 in all scenarios (Table 6.11.2.2; Figure 6.11.2.2). In the scenarios with 0.25, 0.5, 1, and 2 times the base number of smolts stocked, the proportion of iterations for the Upper Penobscot and Piscataquis declined from generation 1 to generation 2 and increased in subsequent generations. In the scenario with 4 times the base number of smolts stocked, the proportion of iterations for the Upper Penobscot and Piscataquis equaled one in all generations.

The number and proportion of smolts killed differed between Recovery scenarios. The median total number of smolts killed during emigration due to direct and indirect cumulative mortality associated with dam passage decreased from generation 1 to generation 2 and then increased in subsequent generations in the scenarios with 0.25, 0.5, and 1 times the base number of smolts stocked (Table 6.11.2.3; Figure 6.11.2.3). The median total number of 2SW females increased in the scenarios with 2 and 4 times the base. The median total proportion of smolts killed during emigration due to direct and indirect cumulative mortality associated with dam passage decreased from 0.11 to 0.10 in the scenarios with 0.25 and 0.5 times the base number of smolts stocked and remained constant at 0.11 for all generations in the scenarios with 1, 2, and 4 times the base (Table 6.11.2.4; Figure 6.11.2.3).

6.12 Stocking Distribution

Smolts were distributed throughout the Penobscot River watershed according to the mean proportion stocked in each PU during 2003-2012 (Table 3.4.1). This distribution of stocked smolts was used as the base input, and sensitivities were run with all smolts stocked in the Piscataquis River, all smolts stocked in PU 2, smolts stocked equally among PUs, and all smolts stocked below Veazie Dam. In the scenario with all smolts stocked in the Piscataquis River, smolts were stocked in PUs 4, 5, and 6 according to the proportion of habitat units in each PU (i.e., 66%, 0.4%, and 33.6%, respectively). In the scenario with all smolts stocked equally among PUs, no smolts were stocked in PU 1. In the scenario with all smolts stocked below Veazie Dam, smolts were stocked in PUs 13 and 14 according to the proportion of habitat units in each PU (i.e., 21.3% and 78.7%, respectively).

6.12.1 Base Case

Adult abundance decreased overall in all five Base Case scenarios, whereas adult distribution differed between scenarios. The median number of 2SW females decreased overall in all scenarios, but the number of 2SW females was highest in the scenario with all smolts stocked below Veazie dam (Table 6.12.1.1; Figure 6.12.1.1). The proportion of iterations when at least one 2SW female was located in each Penobscot River watershed area equaled one in generation 1 for all three areas and remained at one for the Lower Penobscot in generations 2-10 in all scenarios (Table 6.12.1.2; Figure 6.12.1.2). The proportion of iterations for the Upper Penobscot and Piscataquis decreased from generation 1 to generation 3 and varied without trend in subsequent generations in all scenarios. The proportion of iterations in the Piscataquis was higher than the proportion of iterations in the Upper Penobscot in the scenario where all smolts were stocked in the Piscataquis. The proportion of iterations in the Upper Penobscot was higher than the proportion of iterations in the Piscataquis in the scenario where all smolts were stocked in PU 2. The proportion of iterations in the Upper Penobscot and Piscataquis were similar to each other in the other three scenarios and were highest in the base scenario.

The number of smolts killed decreased in all five Base Case scenarios, whereas the proportion of smolts killed differed between scenarios. The median total number of smolts killed during emigration due to direct and indirect cumulative mortality associated with dam passage was the highest in generation 1, declined from generation 1 to generation 2, and varied without trend in generations 2-10 in all scenarios (Table 6.12.1.3; Figure 6.12.1.3). The number of smolts killed was lowest in the scenario where all smolts were stocked below Veazie Dam and was highest in the scenario where all smolts were stocked in PU 2. The median total proportion of smolts killed during emigration due to direct and indirect cumulative mortality associated with dam passage was highest (0.13) in the scenario where all smolts were stocked in PU 2, lowest (0.06 in generations 2-10) in the scenario where all smolts were stocked below Veazie Dam, and equaled 0.10 or 0.11 in all generations for the other three scenarios (Table 6.12.1.4; Figure 6.12.1.3).

6.12.2 Recovery

Adult abundance increased overall in all five Recovery scenarios, whereas adult distribution differed between scenarios. The median number of 2SW females increased in all scenarios and was the highest in the scenario where all smolts were stocked below Veazie Dam (Table 6.12.2.1; Figure 6.12.2.1). The proportion of iterations when at least one 2SW female was located in each Penobscot River watershed area equaled one in generation 1 for all three areas and remained at one for the Lower Penobscot in generations 2-10 in all scenarios (Table 6.12.2.2; Figure 6.12.2.2). The proportion of iterations for the Upper Penobscot and Piscataquis declined from generation 1 to generation 2 and increased in subsequent generations. The proportion of iterations was closest to one in the base scenario.

The number and proportion of smolts killed differed between Recovery scenarios. The median total number of smolts killed during emigration due to direct and indirect cumulative mortality associated with dam passage decreased from generation 1 to generation 2 and then increased in subsequent generations in all scenarios (Table 6.12.2.3; Figure 6.12.2.3). The number of smolts killed increased overall in the scenario where all smolts were stocked in the Piscataquis and in the base scenario and decreased overall in the other three scenarios. The number of smolts killed was the lowest in the scenario with all smolts stocked below Veazie Dam. The median total proportion of smolts killed during emigration due to direct and indirect cumulative mortality associated with dam passage was highest (0.13) in the scenario where all smolts were stocked in PU 2, lowest (0.09 in generations 2-10) in the scenario where all smolts were stocked below Veazie Dam, and equaled 0.10 or 0.11 in all generations for the other three scenarios (Table 6.12.2.4; Figure 6.12.2.3).

6.13 Straying

A set of logical rules was developed to assist with estimating PU-specific homing rates and straying patterns by using specific study results combined with the Expert Panel opinions and local knowledge (Table 3.10.1). These rules were used as the base input, and alternate sets of rules were developed to run sensitivities. The first alternate set of rules (RulesX1) was developed using study results, fishway trap data, and Expert Panel opinions, but local knowledge was excluded (Table 6.13.1). The second alternate set of rules (RulesX2) was the RulesX1 table applied to itself to further distribute straying fish (Table 6.13.2). In the third alternate set of rules (100% home), all adults returned to their natal PU (Table 6.13.3). In the fourth alternate set of rules (=straying), all returning adults strayed to other PUs equally (Table 6.13.4).

6.13.1 Base Case

Adult abundance and distribution decreased in all five Base Case scenarios and were not sensitive to straying in the Base Case scenarios. The median number of 2SW females declined from generation 1 to generation 2 and varied without trend in subsequent generations (Table 6.13.1.1; Figure 6.13.1.1). The proportion of iterations when at least one 2SW female was located in each Penobscot River watershed area equaled one in generation 1 for all three areas and remained at one for the Lower Penobscot in generations 2-10 (Table 6.13.1.2; Figure 6.13.1.2). The proportion of iterations for the Upper Penobscot and Piscataquis were similar, declining from generation 1 to 3 and varying without trend in subsequent generations for all scenarios.

The number of smolts killed decreased in all five Base Case scenarios, whereas the proportion of smolts killed remained constant. The number and proportion of smolts killed were not sensitive to straying in the Base Case scenarios. The median total number of smolts killed during emigration due to direct and indirect cumulative mortality associated with dam passage was the highest in generation 1, declined from generation 1 to generation 2, and varied without trend in generations 2-10 (Table 6.13.1.3; Figure 6.13.1.3). The median total proportion of smolts killed during emigration due to direct and indirect cumulative mortality associated with dam passage remained constant at 0.11 for all generations and scenarios (Table 6.13.1.4; Figure 6.13.1.3).

6.13.2 Recovery

Adult abundance increased and adult distribution remained near one in all five Recovery scenarios. The median number of 2SW females increased from generation 1 and reached a plateau by generation 10 (Table 6.13.2.1; Figure 6.13.2.1). Adult abundance was highest in the base scenario and lowest in the scenario with 100% homing. The proportion of iterations when at least one 2SW female was located in each Penobscot River watershed area equaled one or close to one in all generations, areas, and scenarios (Table 6.13.2.2; Figure 6.13.2.2).

The number of smolts killed differed between the Recovery scenarios, whereas the proportion of smolts killed remained constant between the scenarios. The median total number of smolts killed during emigration due to direct and indirect cumulative mortality associated with dam passage declined from generation 1 to generation 2 and increased in subsequent generations in all straying scenarios except 100% home, where the number of smolts killed varied without trend in generations 2-10 (Table 6.13.2.3; Figure 6.13.2.3). The number of smolts killed decreased overall in the RulesX1 and RulesX2 scenarios, whereas the number of smolts killed increased overall in the base and =straying scenarios. The median total proportion of smolts killed during emigration due to direct and indirect cumulative mortality associated with dam passage equaled 0.10 or 0.11 for all generations and scenarios (Table 6.13.2.4; Figure 6.13.2.3).

6.14 Proportion Dying

The fate of adult Atlantic salmon that were unsuccessful at passing an individual dam was determined by the Expert Panel, and one fate was that a proportion of the fish die (Table 3.12.1). The Expert Panel's decision on the proportion of fish that die was used as the base input, and sensitivities were run at alternate values (i.e., 0, 0.012, 0.024, and 0.048). The alternate proportions of fish dying were applied only to dams where the base proportion dying input was greater than zero (Table 3.12.1). The proportion of fish remaining downstream was adjusted accordingly so that the proportion dying, the proportion returning to sea, and the proportion remaining downstream still summed to one.

6.14.1 Base Case

Adult abundance and distribution decreased in all five Base Case scenarios, and these results were not sensitive to the proportion of fish dying in the Base Case scenarios. The median number of 2SW females declined to approximately the same value from generation 1 to generation 2 and varied without trend in subsequent generations in all scenarios (Table 6.14.1.1; Figure 6.14.1.1). The proportion of iterations when at least one 2SW female was located in each Penobscot River watershed area equaled one in generation 1 for all three areas and remained at one for the Lower Penobscot in generations 2-10 (Table 6.14.1.2; Figure 6.14.1.2). The proportion of iterations for the Upper Penobscot and Piscataquis were similar, declining from generation 1 to 3 and varying without trend in subsequent generations in all scenarios.

The number of smolts killed decreased in all five Base Case scenarios, whereas the proportion of smolts killed remained constant. The number and proportion of smolts killed were not sensitive to the proportion of fish dying in the Base Case scenarios. The median total number of smolts killed during emigration due to direct and indirect cumulative mortality associated with dam passage was highest in generation 1, declined from generation 1 to generation 2, and varied without trend in generations 2-10 (Table 6.14.1.3; Figure 6.14.1.3). The numbers of smolts killed were similar among scenarios. The median total proportion of smolts killed during emigration due to direct and indirect cumulative mortality associated with dam passage remained constant at 0.11 for all generations and scenarios (Table 6.14.1.4; Figure 6.14.1.3).

6.14.2 Recovery

Adult abundance increased and adult distribution remained near one in all five Recovery scenarios. Adult abundance and distribution were not sensitive to the proportion of fish dying in the Recovery scenarios. The median number of 2SW females increased from generation 1 and reached a plateau by generation 10 at approximately the same abundance in all scenarios (Table 6.14.2.1; Figure 6.14.2.1). The proportion of iterations when at least one 2SW female was located in each Penobscot River watershed area equaled one or close to one in all generations, areas, and scenarios (Table 6.14.2.2; Figure 6.14.2.2).

The number of smolts killed increased overall in all five Recovery scenarios, whereas the proportion of smolts killed remained constant. The number and proportion of smolts killed were not sensitive to the proportion of fish dying in the Recovery scenarios. The median total number of smolts killed during emigration due to direct and indirect cumulative mortality associated with dam passage was the highest in generation 1, declined from generation 1 to generation 2, and increased in subsequent generations to approximately the same level in all scenarios by generation 10 (Table 6.14.2.3; Figure 6.14.2.3). The median total proportion of smolts killed during emigration due to direct and indirect cumulative mortality associated with dam passage was equal to 0.10 or 0.11 for all generations and scenarios (Table 6.14.2.4; Figure 6.14.2.3).

6.15 Proportion Returning to Sea

Another fate of adult Atlantic salmon that were unsuccessful at passing an individual dam was that a proportion of the fish return to the ocean un-spawned. The proportion of fish returning to sea at each dam was determined by the Expert Panel, and this decision was used as the base input (Table 3.12.1). Sensitivities were run at values of 0, 0.5, 2, and 4 times the base value. The proportion of fish remaining downstream was adjusted accordingly so that the proportion dying, the proportion returning to sea, and the proportion remaining downstream still summed to one.

6.15.1 Base Case

Adult abundance and distribution decreased in all five Base Case scenarios. The median number of 2SW females declined from generation 1 to generation 2 and varied without trend in subsequent generations in all scenarios (Table 6.15.1.1; Figure 6.15.1.1). Abundance was the lowest in the scenario with 4 times the base. The proportion of iterations when at least one 2SW female was located in each Penobscot River watershed area equaled one in generation 1 for all three areas and remained at one for the Lower Penobscot in generations 2-10 (Table 6.15.1.2; Figure 6.15.1.2). The proportion of iterations for the Upper Penobscot and Piscataquis were similar, declining from generation 1 to 3 and varying without trend in subsequent generations in all scenarios.

The number of smolts killed decreased in all five Base Case scenarios, whereas the proportion of smolts killed remained constant. The median total number of smolts killed during emigration due to direct and indirect cumulative mortality associated with dam passage was highest in generation 1, declined from generation 1 to generation 2, and varied without trend in generations 2-10 (Table 6.15.1.3; Figure 6.15.1.3). The median total proportion of smolts killed during emigration due to direct and indirect cumulative mortality associated with dam passage remained constant at 0.11 for all generations and scenarios (Table 6.15.1.4; Figure 6.15.1.3).

6.15.2 Recovery

Adult abundance increased and adult distribution remained near one in all five Recovery scenarios. The median number of 2SW females increased from generation 1 and reached a plateau by generation 10 in all scenarios (Table 6.15.2.1; Figure 6.15.2.1). Abundance was highest in the scenario with 0 times the base proportion returning to sea and lowest in the scenario with 4 times the base. The proportion of iterations when at least one 2SW female was located in each Penobscot River watershed area equaled one or close to one in all generations, areas, and scenarios (Table 6.15.2.2; Figure 6.15.2.2).

The number of smolts killed increased overall in all five Recovery scenarios, whereas the proportion of smolts killed remained constant. The median total number of smolts killed during emigration due to direct and indirect cumulative mortality associated with dam passage was the highest in generation 1, declined from generation 1 to generation 2, and increased in subsequent generations (Table 6.15.2.3; Figure 6.15.2.3). The median total proportion of smolts killed during emigration due to direct and indirect cumulative mortality associated with dam passage was equal to 0.10 or 0.11 for all generations and scenarios (Table 6.15.2.4; Figure 6.15.2.3).

6.16 Proportion Remaining Downstream

The third, and final, possible fate of adult Atlantic salmon that were unsuccessful at passing an individual dam in the DIA Model was that a proportion of the fish go elsewhere, specifically to a downstream PU, to spawn. The proportion of fish spawning in a downstream PU was determined by the Expert Panel, and this decision was used as the base input (Table 3.12.1). Sensitivities were run with all adults that were unsuccessful at passing an individual dam spawning in the PU immediately below that dam (Table 6.16.1) and with the adults evenly distributed between all PUs below the dam that was not passed (Table 6.16.2).

6.16.1 Base Case

Adult abundance and distribution decreased in all three Base Case scenarios, and these results were not sensitive to the proportion of fish remaining downstream in the Base Case scenarios. The median number of 2SW females declined to approximately the same value from generation 1 to generation 2 and varied without trend in subsequent generations in all scenarios (Table 6.16.1.1; Figure 6.16.1.1). The proportion of iterations when at least one 2SW female was located in each Penobscot River watershed area equaled one in generation 1 for all three areas and remained at one for the Lower Penobscot in generations 2-10 (Table 6.16.1.2; Figure 6.16.1.2). The proportion of iterations for the Upper Penobscot and Piscataquis were similar, declining from generation 1 to 3 and varying without trend in subsequent generations in all scenarios.

The number of smolts killed decreased in all three Base Case scenarios, whereas the proportion of smolts killed remained constant. The number and proportion of smolts killed were not sensitive to the proportion of fish remaining downstream in the Base Case scenarios. The median total number of smolts killed during emigration due to direct and indirect cumulative mortality associated with dam passage was highest in generation 1, declined from generation 1 to generation 2, and varied without trend in generations 2-10 (Table 6.16.1.3; Figure 6.16.1.3). The numbers of smolts killed were similar among scenarios. The median total proportion of smolts killed during emigration due to direct and indirect cumulative mortality associated with dam passage remained constant at 0.11 for all generations and scenarios (Table 6.16.1.4; Figure 6.16.1.3).

6.16.2 Recovery

Adult abundance increased and adult distribution remained near one in all three Recovery scenarios. Adult abundance and distribution were not sensitive to the proportion of fish remaining downstream in the Recovery scenarios. The median number of 2SW females increased from generation 1 and reached a plateau by generation 10 at approximately the same abundance in all scenarios (Table 6.16.2.1; Figure 6.16.2.1). The proportion of iterations when at least one 2SW female was located in each Penobscot River watershed area equaled one or close to one in all generations, areas, and scenarios (Table 6.16.2.2; Figure 6.16.2.2).

The number of smolts killed increased overall in all three Recovery scenarios, whereas the proportion of smolts killed remained constant. The number and proportion of smolts killed were not sensitive to the proportion of fish dying in the Recovery scenarios. The median total number of smolts killed during emigration due to direct and indirect cumulative mortality associated with dam passage was the highest in generation 1, declined from generation 1 to generation 2, and increased in subsequent generations (Table 6.16.2.3; Figure 6.16.2.3). The median total proportion of smolts killed during emigration due to direct and indirect cumulative mortality associated with dam passage was equal to 0.10 or 0.11 for all generations and scenarios (Table 6.16.2.4; Figure 6.16.2.3).

6.17 Downstream Dam Passage Survival Rates

Downstream dam passage survival rates of smolts were estimated by Alden (Amaral et al. 2012). These rates were used as the base input values, except for Upper Dover Dam, which equaled 92.15% (see Section 3.6.1). Sensitivities were run at -10%, -5%, +5%, and +10% of the base survival rates, with survival capped at one. These data adjustments were applied to each dam, except Dover Upper Dam.

6.17.1 Base Case

Adult abundance and distribution decreased in all five Base Case scenarios. The median number of 2SW females declined from generation 1 to generation 2 and varied without trend in subsequent generations (Table 6.17.1.1; Figure 6.17.1.1). Abundance was lowest in the scenario with downstream dam passage survival rates decreased by 10% and highest in the scenario with these rates increased by 10%. The proportion of iterations when at least one 2SW female was located in each Penobscot River watershed area equaled one in generation 1 for all three areas and remained at one for the Lower Penobscot in generations 2-10 (Table 6.17.1.2; Figure 6.17.1.2). The proportion of iterations for the Upper Penobscot and Piscataquis declined from generation 1 to 3 and varied without trend in subsequent generations. The proportion of iterations in the Upper Penobscot and Piscataquis were lowest in the scenario with downstream dam passage survival rates decreased by 10% and highest in the scenario with these rates increased by 10%.

The number of smolts killed decreased in all five Base Case scenarios, whereas the proportion of smolts killed remained constant in each of the Base Case scenarios. The median total number of smolts killed during emigration due to direct and indirect cumulative mortality associated with dam passage was highest in generation 1, declined from generation 1 to generation 2, and varied without trend in generations 2-10 (Table 6.17.1.3; Figure 6.17.1.3). The median total proportion of smolts killed during emigration due to direct and indirect cumulative mortality associated with dam passage remained constant for all generations in each scenario but varied among scenarios (Table 6.17.1.4; Figure 6.17.1.3). The number and proportions of smolts killed were highest in the scenario with downstream dam passage survival rates decreased by 10% and lowest in the scenario with these rates increased by 10%.

6.17.2 Recovery

Adult abundance increased and adult distribution remained near one in all five Recovery scenarios. The median number of 2SW females increased from generation 1 and reached a plateau by generation 10 in all scenarios (Table 6.17.2.1; Figure 6.17.2.1). The proportion of iterations when at least one 2SW female was located in each Penobscot River watershed area equaled one or close to one in all generations, areas, and scenarios (Table 6.17.2.2; Figure 6.17.2.2). The adult abundance and distribution were lowest in the scenario with downstream dam passage survival rates decreased by 10% and highest in the scenario with these rates increased by 10%.

The number of smolts killed differed between Recovery scenarios, whereas the proportion of smolts killed remained constant in each of the Recovery scenarios. The median total number of smolts killed during emigration due to direct and indirect cumulative mortality associated with dam passage decreased overall in the scenario with survival rates decreased by 10% but increased overall in the other four Recovery scenarios (Table 6.17.2.3; Figure 6.17.2.3). The median total proportion of smolts killed during emigration due to direct and indirect cumulative mortality associated with dam passage remained constant for all generations in each scenario but varied among scenarios (Table 6.17.2.4; Figure 6.17.2.3). The number and proportions of smolts killed were highest in the scenario with downstream dam passage survival rates decreased by 10% and lowest in the scenario with these rates increased by 10%.

6.18 Upstream Dam Passage Survival Rates

Upstream dam passage survival rate distributions of adults were estimated using telemetry studies or previous model estimates (Table 3.11.1.1). Sensitivities were run at -10%, -5%, +5%, and +10% of the base survival rates, with survival capped at one. Four dams (i.e., Medway, Milo, Sebec, and Orono) do not have any upstream passage, and so upstream passage values at these dams were set to zero in both the base input and the sensitivity runs.

6.18.1 Base Case

Adult abundance and distribution decreased in all five Base Case scenarios, and these results were not sensitive to upstream dam passage survival rates in the Base Case scenarios. The median number of 2SW females declined from generation 1 to generation 2 and varied without trend in subsequent generations (Table 6.18.1.1; Figure 6.18.1.1). The proportion of iterations when at least one 2SW female was located in each Penobscot River watershed area equaled one in generation 1 for all three areas and remained at one for the Lower Penobscot in generations 2-10 (Table 6.18.1.2; Figure 6.18.1.2). The proportion of iterations for the Upper Penobscot and Piscataquis declined from generation 1 to 3 and varied without trend in subsequent generations. The proportion of iterations in the Upper Penobscot and Piscataquis were lowest in the scenario with upstream dam passage survival rates decreased by 10% and highest in the scenario with these rates increased by 10%.

The number of smolts killed decreased in all five Base Case scenarios, whereas the proportion of smolts killed remained constant. The number and proportion of smolts killed were not sensitive to the upstream dam passage survival rates in the Base Case scenarios. The median total number of smolts killed during emigration due to direct and indirect cumulative mortality associated with dam passage was highest in generation 1, declined from generation 1 to generation 2, and varied without trend in generations 2-10 (Table 6.18.1.3; Figure 6.18.1.3). The median total proportion of smolts killed during emigration due to direct and indirect cumulative mortality associated with dam passage remained constant at 0.11 for all generations and scenarios (Table 6.18.1.4; Figure 6.18.1.3).

6.18.2 Recovery

Adult abundance increased and adult distribution remained near one in all five Recovery scenarios. Adult abundance and distribution were not sensitive to upstream dam passage survival rates in the Recovery scenarios. The median number of 2SW females increased from generation 1 and reached a plateau by generation 10 in all scenarios (Table 6.18.2.1; Figure 6.18.2.1). The proportion of iterations when at least one 2SW female was located in each Penobscot River watershed area equaled one or close to one in all generations, areas, and scenarios (Table 6.18.2.2; Figure 6.18.2.2). The adult abundance and distribution were lowest in the scenario with upstream dam passage survival rates decreased by 10% and highest in the scenario with these rates increased by 10%.

The number of smolts killed differed between Recovery scenarios, whereas the proportion of smolts killed remained constant. The number and proportion of smolts killed were not sensitive to the upstream dam passage survival rates in the Recovery scenarios. The median total number of smolts killed during emigration due to direct and indirect cumulative mortality associated with dam passage decreased overall in the scenarios with survival rates decreased by 5 and 10% but increased overall in the other three Recovery scenarios (Table 6.18.2.3; Figure 6.18.2.3). The number of smolts killed was lowest in the scenario with downstream dam passage survival rates decreased by 10% and highest in the scenario with these rates increased by 10%. The median total proportion of smolts killed during emigration due to direct and indirect cumulative mortality associated with dam passage equaled 0.10 or 0.11 for all generations and scenarios (Table 6.18.2.4; Figure 6.18.2.3).

6.19 Indirect Latent Mortality

An indirect latent mortality rate of 10% per dam was applied, and this rate was used as the base input. Sensitivities were run at values of 2.5%, 5%, 20%, and 40% indirect latent mortality per dam.

6.19.1 Base Case

Adult abundance and distribution decreased in all five Base Case scenarios. The median number of 2SW females declined from generation 1 to generation 2 and varied without trend in subsequent generations (Table 6.19.1.1; Figure 6.19.1.1). The proportion of iterations when at least one 2SW female was located in each Penobscot River watershed area equaled one in generation 1 for all three areas (Table 6.19.1.2; Figure 6.19.1.2). The proportion of iterations remained at one for the Lower Penobscot in generations 2-10 for the scenarios with 2.5%, 5%, 10%, and 20% indirect latent mortality per dam but declined to below 0.10 in the scenario with 40% indirect latent mortality per dam. The proportion of iterations for the Upper Penobscot and Piscataquis declined from generation 1 to 3 and varied without trend in subsequent generations. Adult abundance and distribution were highest in the scenario with 2.5% indirect latent mortality per dam and lowest (zero or close to zero in generations 2-10) in the scenario with 40% indirect latent mortality per dam.

The number of smolts killed decreased in all five Base Case scenarios, whereas the proportion of smolts killed remained constant. The median total number of smolts killed during emigration due to direct and indirect cumulative mortality associated with dam passage was highest in generation 1, declined from generation 1 to generation 2, and varied without trend in generations 2-10 (Table 6.19.1.3; Figure 6.19.1.3). The median total proportion of smolts killed during emigration due to direct and indirect cumulative mortality associated with dam passage remained constant at 0.11 for all generations and scenarios (Table 6.19.1.4; Figure 6.19.1.3).

6.19.2 Recovery

Adult abundance and distribution differed between the Recovery scenarios. The median number of 2SW females increased overall in the scenarios with 2.5%, 5%, 10% and 20% indirect latent mortality per dam but decreased to zero in the scenario with 40% indirect latent mortality per dam (Table 6.19.2.1; Figure 6.19.2.1). The proportion of iterations when at least one 2SW female was located in each Penobscot River watershed area equaled one or close to one in all generations and areas in the scenarios with 2.5%, 5%, and 10% indirect latent mortality per dam, decreased from generation 1 to generation 2 and increased in subsequent generations in the scenario with 20% indirect latent mortality per dam, and decreased to zero or close to zero in the scenario with 40% indirect latent mortality per dam (Table 6.19.2.2; Figure 6.19.2.2).

The number of smolts killed differed between Recovery scenarios, whereas the proportion of smolts killed remained constant. The median total number of smolts killed during emigration due to direct and indirect cumulative mortality associated with dam passage increased overall in the scenarios with 2.5%, 5%, and 10% indirect latent morality per dam but decreased overall in the scenarios with 20% and 40% indirect latent mortality per dam (Table 6.19.2.3; Figure 6.19.2.3). The median total proportion of smolts killed during emigration due to direct and indirect cumulative mortality associated with dam passage equaled 0.10 or 0.11 for all generations and scenarios (Table 6.19.2.4; Figure 6.19.2.3).

6.20 Downstream Path Choice

Smolts originating upriver of the Stillwater Branch have the option of migrating to the ocean via the Stillwater Branch or the mainstem. Telemetry data were used to develop a distribution for smolt use of the Stillwater Branch (Figure 3.6.3.2), and this was used as the base input. Sensitivities were run at 0.25, 0.5, 2, and 4 times the base. Downstream path choice for the Stillwater Branch was capped at one.

6.20.1 Base Case

Adult abundance and distribution decreased in all five Base Case scenarios, and these results were not sensitive to downstream path choice in the Base Case scenarios. The median number of 2SW females declined to approximately the same value from generation 1 to generation 2 and varied without trend in subsequent generations in all scenarios (Table 6.20.1.1; Figure 6.20.1.1). The proportion of iterations when at least one 2SW female was located in each Penobscot River watershed area equaled one in generation 1 for all three areas and remained at one for the Lower Penobscot in generations 2-10 (Table 6.20.1.2; Figure 6.20.1.2). The proportion of iterations for the Upper Penobscot and Piscataquis were similar, declining from generation 1 to 3 and varying without trend in subsequent generations in all scenarios.

The number of smolts killed decreased in all five Base Case scenarios, whereas the proportion of smolts killed remained constant in each scenario. The median total number of smolts killed during emigration due to direct and indirect cumulative mortality associated with dam passage was highest in generation 1, declined from generation 1 to generation 2, and varied without trend in generations 2-10 (Table 6.20.1.3; Figure 6.20.1.3). The number of smolts killed was highest in the scenario with 0.25 times the base Stillwater Branch use rate and lowest in the scenario with 4 times the base Stillwater use rate. The median total proportion of smolts killed during emigration due to direct and indirect cumulative mortality associated with dam passage remained constant for all generations at either 0.10 (in the scenarios with 2 and 4 times the base) or 0.11 (in the scenarios with 0.25, 0.5, and 1 times the base) (Table 6.20.1.4; Figure 6.20.1.3).

6.20.2 Recovery

Adult abundance increased and adult distribution remained near one in all five Recovery scenarios, and these results were not sensitive to downstream path choice in the Recovery scenarios. The median number of 2SW females increased from generation 1 and reached a plateau by generation 10 at approximately the same abundance in all scenarios (Table 6.20.2.1; Figure 6.20.2.1). The proportion of iterations when at least one 2SW female was located in each Penobscot River watershed area equaled one or close to one in all generations, areas, and scenarios (Table 6.20.2.2; Figure 6.20.2.2).

The number of smolts killed increased overall in all five Recovery scenarios, whereas the proportion of smolts killed remained constant in each scenario. The median total number of smolts killed during emigration due to direct and indirect cumulative mortality associated with dam passage was the highest in generation 1, declined from generation 1 to generation 2, and increased in subsequent generations (Table 6.20.2.3; Figure 6.20.2.3). The number of smolts killed was highest in the scenario with 0.25 times the base Stillwater Branch use rate and lowest in the scenario with 4 times the Stillwater Branch use rate. The median total proportion of smolts killed during emigration due to direct and indirect cumulative mortality associated with dam passage remained constant for all generations at either 0.10 (in the scenarios with 2 and 4 times the base) or 0.11 (in the scenarios with 0.25, 0.5, and 1 times the base) (Table 6.20.2.4; Figure 6.20.2.3).

6.21 Freshwater and Marine Survival Rates With the Hatchery Turned On or Off

A series of sensitivities were run with a range of freshwater and marine survival rates. Scenarios with five freshwater survival rates (0.25, 0.5, 1, 2, and 4 times the base case egg to smolt survival) were run with five marine survival rates (0.25, 0.5, 1, 2, and 4 times the base case). Each freshwater and marine survival rate combination was run with stocking of hatchery-reared smolts turned on and turned off in the DIA Model.

6.21.1 Hatchery On

Adult abundance and distribution increased as freshwater and marine survival increased with the hatchery component of the model turned on. The median number of 2SW females decreased from generation 1 to generation 2 and varied without trend in subsequent generations when the marine survival rate was low (Tables 6.21.1.1 - 6.21.1.3; Figures 6.21.1.1 - 6.21.1.3). The number of 2SW females increased above the starting abundance of 587 fish by generation 10 in the scenarios with the marine survival increased by 2 times the base and freshwater survival increased by 2 and 4 times the base and in all scenarios with marine survival increased by 4 times the base (Table 6.21.1.4 and Table 6.21.1.5; Figure 6.21.1.4 and Figure 6.21.1.5). The proportion of iterations when at least one 2SW female was located in each Penobscot River watershed area equaled one in generation 1 for all three areas and remained at or near one for the Lower Penobscot in generations 2-10 for all combinations of freshwater and marine survival rates (Tables 6.21.1.6 - 6.21.1.10; Figures 6.21.1.6 - 6.21.1.10). The proportion of iterations in the Upper Penobscot and Piscataquis declined from generation 1 to generation 3 and varied without trend in subsequent generations when the marine survival rate was low, but the proportion of iterations in these areas was close to or equal to one in all generations in scenarios when the freshwater survival rate was 1, 2, or 4 times the base and marine survival was 4 times the base.

The trend in the number of smolts killed differed based on the freshwater and marine survival rates with the hatchery turned on, but the proportion of smolts killed remained constant. The median total number of smolts killed during emigration due to direct and indirect cumulative mortality associated with dam passage generally increased as freshwater and marine survival increased (Tables 6.21.1.11 - 6.21.1.15; Figures 6.21.1.11 - 6.21.1.15). The number of smolts killed decreased overall in scenarios when marine and freshwater survival rates were low and increased overall only in scenarios with 4 times the base marine survival rate and 1, 2, and 4 times the freshwater survival rate. The median total proportion of smolts killed during emigration due to direct and indirect cumulative mortality associated with dam passage remained constant for all generations (0.10 and 0.11 in the scenario with 2 times the marine survival base case rate and 4 times the freshwater survival base case rate and the scenarios with 4 times the marine survival base case rate and 1, 2, and 4 times the freshwater base case rate; 0.11 for all other combinations of marine and freshwater survival rates) (Tables 6.21.1.16 - 6.21.1.20; Figures 6.21.1.11 - 6.21.1.15).

6.21.2 Hatchery Off

Adult abundance and distribution increased as freshwater and marine survival increased with the hatchery component of the model turned off. The median number of 2SW females decreased to zero or near zero when the marine or freshwater survival rates were low (Tables 6.21.2.1 - 6.21.2.5; Figures 6.21.2.1 - 6.21.2.5). The number of 2SW females increased above the starting abundance of 587 fish by generation 10 in the scenarios with the marine survival increased by 2 times the base and freshwater survival increased by 4 times the base and with marine survival increased by 4 times the base and freshwater survival increased by 2 and 4 times the base. The proportion of iterations when at least one 2SW female was located in each Penobscot River watershed area equaled one in generation 1 for all three areas for all combinations of freshwater and marine survival rates (Tables 6.21.2.6 - 6.21.2.10; Figures 6.21.2.6 - 6.21.2.10). The proportion of iterations in all three areas decreased to zero or near zero by generation 10 in scenarios with low marine or freshwater survival rates but equaled one or close to one in all generations in scenarios with 2 times the base marine survival and 4 times the base freshwater survival rates and 4 times the base marine survival rate and 2 and 4 times the base freshwater survival rates.

The number and proportion of smolts killed decreased overall in all scenarios with a range of freshwater and marine survival rates and the hatchery turned off. The median total number of smolts killed during emigration due to direct and indirect cumulative mortality associated with dam passage decreased to zero or near zero in scenarios with low marine or freshwater survival rates (Tables 6.21.2.11 - 6.21.2.15; Figures 6.21.2.11 - 6.21.2.15). The number of smolts killed began to increase by generation 10 in scenarios with 2 times the base marine survival and 4 times the base freshwater survival rates and 4 times the base marine survival rate and 2 and 4 times the base freshwater survival rates. The median total proportion of smolts killed during emigration due to direct and indirect cumulative mortality associated with dam passage decreased to zero in scenarios with low marine or freshwater survival rates (Tables 6.21.2.16 - 6.21.2.20; Figures 6.21.2.11 - 6.21.2.15). The decrease in the proportion of smolts killed was less in scenarios with increased marine and freshwater survival rates.

6.22 Median and Mean Marine Survival Rates for Different Time Series

Estimates of the median 2SW female marine survival rates during 1971-2010 were fitted to an inverse gaussian distribution with μ = 0.006265, shape parameter λ = 0.0068723, and a shift of 0.00000813424 (Figure 3.9.4). This distribution was used as the base input, and five alternate distributions were used to run sensitivities. The distributions for the base and sensitivities were developed using the same process, but different data were used in each distribution (i.e., mean or median, different time series). The first alternate distribution was median estimates of 2SW female marine survival rates during 1971-1990 fitted to an inverse gaussian distribution with μ = 0.012056, shape parameter λ = 0.080705, and a shift of -0.0020676. The second alternate distribution was median estimates of 2SW female marine survival rates during 1991-2010 fitted to an inverse gaussian distribution with μ = 0.003916, shape parameter λ = 0.0480996, and a shift of -0.0013579. The third alternate distribution was mean estimates of 2SW female marine survival rates during 1971-2010 fitted to an inverse gaussian distribution with μ = 0.0070894, shape parameter λ = 0.0077758, and a shift of 0.00000954211. The fourth alternate distribution was mean estimates of 2SW female marine survival rates during 1971-1990 fitted to an inverse gaussian distribution with μ = 0.013649, shape parameter λ = 0.091492, and a shift of -0.0023463. The fifth alternate distribution was mean estimates of 2SW female marine survival rates during 1991-2010 fitted to an inverse gaussian distribution with μ = 0.0044284, shape parameter λ = 0.0543302, and a shift of -0.0015336. A Base Case scenario was run with each of the above distributions and the base case freshwater survival rate input, and a Recovery scenario was run with each of the above distributions increased by four times and the freshwater survival base case increased by two times.

6.22.1 Base Case

Adult abundance and distribution differed between the Base Case scenarios. The median number of 2SW females declined from generation 1 to generation 2 in all median and mean scenarios, varied without trend in generations 2-10 in the median and mean scenarios using 1991-2010 and 1971-2010 data, and increased in generations 2-10 in the median and mean scenarios using 1971-1990 data (Tables 6.22.1.1 and 6.22.1.2; Figures 6.22.1.1 and 6.22.1.2). The number of 2SW females was highest in the scenario using the mean 2SW female marine survival rates from 1971 to 1990 and lowest in the scenario using the median 2SW female marine survival rates from 1991 to 2010, although the number of 2SW females killed was similar between median and mean scenarios that used marine survival rates from the same time period. The proportion of iterations when at least one 2SW female was located in each Penobscot River watershed area equaled one in generation 1 for all three areas and remained at one for the Lower Penobscot in generations 2-10 for all scenarios (Tables 6.22.1.3 and 6.22.1.4; Figures 6.22.1.3 and 6.22.1.4). The proportion of iterations for the Upper Penobscot and Piscataquis were similar, declining from generation 1 to 3 and varying without trend in subsequent generations for all scenarios. The proportion of iterations for the Upper Penobscot and Piscataquis were closest to one in the scenario using the mean 2SW female marine survival rates from 1971 to 1990 and closest to zero in the scenario using the median 2SW female marine survival rates from 1991 to 2010, although the proportion of iterations was similar between median and mean scenarios that used marine survival rates from the same time period.

The number of smolts killed decreased in all of the median and mean Base Case scenarios, whereas the proportion of smolts killed remained constant. The median total number of smolts killed during emigration due to direct and indirect cumulative mortality associated with dam passage was the highest in generation 1, declined from generation 1 to generation 2 or 3, and varied without trend in subsequent generations (Tables 6.22.1.5 and 6.22.1.6; Figures 6.22.1.5 and 6.22.1.6). The number of smolts killed was highest in the scenario using the mean 2SW female marine survival rates from 1971 to 1990 and lowest in the scenario using the median 2SW female marine survival rates from 1991 to 2010, although the number of smolts killed was similar between median and mean scenarios that used marine survival rates from the same time period. The median total proportion of smolts killed during emigration due to direct and indirect cumulative mortality associated with dam passage equaled 0.11 for all generations and scenarios (Tables 6.22.1.7 and 6.22.1.8; Figures 6.22.1.5 and 6.22.1.6).

6.22.2 Recovery

Adult abundance increased overall and adult distribution remained near one in all of the median and mean Recovery scenarios. The median number of 2SW females increased after generation 1 or 2 and reached a plateau by generation 10 (Tables 6.22.2.1 and 6.22.2.2; Figures 6.22.2.1 and 6.22.2.2). The number of 2SW females was highest in the scenario using the mean 2SW female marine survival rates from 1971 to 1990 and lowest in the scenario using the median 2SW female marine survival rates from 1991 to 2010, although the number of 2SW females was similar between median and mean scenarios that used marine survival rates from the same time period. The proportion of iterations when at least one 2SW female was located in each Penobscot River watershed area equaled one or close to one in all generations, areas, and scenarios (Tables 6.22.2.3 and 6.22.2.4; Figures 6.22.2.3 and 6.22.2.4).

The number of smolts killed differed between the Recovery scenarios, whereas the proportion of smolts killed remained constant. The median total number of smolts killed during emigration due to direct and indirect cumulative mortality associated with dam passage increased overall in the median and mean scenarios using the mean 2SW female marine survival rates from 1971 to 1990 and 1971 to 2010 but decreased overall in the median and mean scenarios using the mean 2SW female marine survival rates from 1991 to 2010 (Tables 6.22.2.5 and 6.22.2.6; Figures 6.22.2.5 and 6.22.2.6). The number of smolts killed was highest in the scenario using the mean 2SW female marine survival rates from 1971 to 1990 and lowest in the scenario using the median 2SW female marine survival rates from 1991 to 2010, although the number of smolts killed was similar between median and mean scenarios that used marine survival rates from the same time period. The median total proportion of smolts killed during emigration due to direct and indirect cumulative mortality associated with dam passage equaled 0.10 or 0.11 for all generations and scenarios (Tables 6.22.2.7 and 6.22.2.8; Figures 6.22.2.5 and 6.22.2.6).

6.23 Summary

Model diagnostics were used to decide the appropriate number of iterations and whether or not the results of model runs using that number of iterations were stable. Five thousand iterations was determined to be the appropriate number of model iterations, and the variation in the results of DIA Model runs using 5,000 iterations was considered acceptable.

Sensitivity of model inputs were examined using the percent difference of 2SW female abundance in generation 10 (Table 6.23.1). The percent difference was calculated as the difference in the median number of 2SW females in generation 10 when the input was changed, compared to the median number of 2SW females in generation 10 when all inputs were set at their base estimates. Highly sensitive model inputs caused the median number of 2SW females to deviate by more than the percent change from the base (e.g., if the base input was multiplied by two, a highly sensitive input would cause more than a 100% difference) (Essington 2003). Linearly sensitive model inputs caused the median number of 2SW females to deviate by the percent change from the base (e.g., if the base input was multiplied by two, a linearly sensitive input would cause a 100% difference). Insensitive model inputs caused the median number of 2SW females to deviate by less than the percent change from the base (e.g., if the base input was multiplied by two, an insensitive input would cause less than a 100% difference). Several sensitivities (hatchery stocking, stocking distribution, straying, proportion dying, proportion remaining downstream, and marine survival using mean and median-based estimates with different time series) could not be classified using this system because the base model was changed in a manner that did not allow for direct quantitative comparison.

A few of the model inputs were highly sensitive, but the majority of the inputs were insensitive. Marine survival and downstream dam survival were highly sensitive in all Base Case and Recovery scenarios, the number of smolts stocked was highly sensitive in all Base Case scenarios, and the hatchery discount was highly sensitive when this input was decreased from the base value in Base Case and Recovery scenarios. The sensitivity of the model to these inputs means a demographic response could reasonably be expected if one of these dynamics changed due to management intervention (e.g., downstream dam passage survival rate) or a shift in the natural range (e.g., marine survival). Other model inputs that could be compared quantitatively were insensitive.

The sensitivities that could not be classified as highly sensitive, linearly sensitive, or insensitive were compared to each other (Table 6.23.1). The percent differences in the sensitivities for hatchery stocking, stocking distribution, and marine survival using mean and median-based estimates with different time were relatively high compared to percent differences of straying, the proportion dying, and the proportion remaining downstream.

7 Conclusions

Marine survival and downstream dam passage survival rates had the greatest impact of all DIA Model inputs. These findings are consistent with expectations expressed by the NRC (2004), who reviewed the status of Atlantic salmon in Maine. The model was highly sensitive to the estimates of marine survival and downstream dam passage survival in all Base Case and Recovery scenarios. Model results showed that the population of Atlantic salmon in the Penobscot River watershed declined when marine survival was low and all 15 hydroelectric dams were present in the watershed. Persistence of the modeled Atlantic salmon population at low levels was sustained by the stocking of hatchery-reared smolts. When marine survival and freshwater survival rates were increased, the abundance and distribution of adult Atlantic salmon abundance increased even when no smolts were stocked.

The locations of dams also had a large affect on the modeled population of Atlantic salmon. Mainstem dams on the Penobscot River were more detrimental to Atlantic salmon than dams in tributaries. Adult abundance and distribution throughout the watershed were higher in scenarios with mainstem dams turned off and dams in tributaries turned on than in scenarios with mainstem dams turned on and dams in tributaries turned off, even though the former scenario has fewer dams turned off (i.e., 100% survival during fish passage) in the model. This result likely occurred because mainstem dams impacted access to the rest of the watershed.

Several model inputs were difficult to estimate due to insufficient data from the Penobscot River. Even though Atlantic salmon are well-studied, especially Penobscot River Atlantic salmon, more data from this population would be beneficial for input value estimates. Marine survival, the hatchery discount, and indirect latent mortality were three such inputs, and the possibility exists that the estimates of mortality incorporated by these three inputs overlap. Marine survival estimates for 2SW females may include a significant portion of the mortality accounted for in the discount of hatchery smolt survival and the indirect latent mortality of fish that passed multiple dams. One benefit of the DIA Model is the flexibility to change input parameters. The hatchery discount and indirect latent mortality rates could be easily lowered if future studies suggest that this is appropriate.

Model inputs that could be improved include egg to smolt survival rate, marine survival distribution, and flow distribution for downstream dam passage. Egg to smolt survival rate was estimated using values from the literature, but a distribution developed from a time series of Penobscot River monitoring data would be preferable. A monitoring program of egg deposition by reach could be initiated to provide data for this model input. The marine survival distribution was corrected to attempt to represent a 'true' marine survival distribution for the Penobscot River, but a time series of post-smolt abundance at Verona and number of adult returns at Verona would be better. The flow allocation provided by Alden (Amaral et al. 2012) was based on their best estimate of the probability of flow at certain levels, but the flow patterns in real life may or may not mirror the Alden estimates. A long, continuous time series of flow data from gauges at each dam would be ideal for this data input. The model inputs are considered the best available information for each input. To continuously improve model performance and realism, inputs should be updated as new information becomes available.

One peculiar model result that may have been caused by model input values was the large decrease in adult abundance, adult distribution, and the number of smolts killed from generation 1 to generation 2 in the Base Case model runs. A possible reason for this decrease could be because the numbers of adults in generations 1 and 2 were not counted at the same point in their life stage. The starting population size for generation 1 in all model runs was the mean annual number of 2SW female spawners at Veazie Dam during 2002-2011. The number of adults in generation 2 (and subsequent generations) was the total number of adults throughout the Penobscot River watershed that reached a PU and spawned after accounting for homing, straying, and upstream dam passage dynamics. These dynamics could have resulted in increased mortality of adult spawners in generation 2 and subsequent smolts, which was not reflected in the number of adults used to seed the model in generation 1. The seeding locations in the model in generation 1 also resulted in a more widely dispersed population through the drainage than normally occurs, resulting in longer migrations because a proportion of the adults, and subsequently smolts, were located farther from the mouth of the Penobscot River and, therefore, had to pass more dams to reach the ocean and home to their natal PU. Migrating longer distances and over multiple dams increased the mortality to which fish were subjected. Although the decrease from generation 1 to generation 2 may not be realistic, the DIA Model is not meant to be a predictive tool but should instead be used to evaluate the relative changes in the Penobscot River population of Atlantic salmon as input values are modified.

The DIA Model can be updated and developed further as new information becomes available. To date, the DIA Model has been used to evaluate the impacts of several hydroelectric dams on the Atlantic salmon population in the Penobscot River (NMFS 2012), and other analyses are possible. For instance, the DIA Model can be used to estimate which of the 15 modeled hydroelectric dams has the most impact on Atlantic salmon productivity, to describe the relative impacts of dams versus marine survival on the Atlantic salmon population at the southern end of the species range, and to address a variety of ecologically-related hypotheses associated with Atlantic salmon, dams, and fresh- and saltwater survival.

8 Acknowledgements

This project could not have been completed without the contributions of many people from various Federal and State Agencies, Universities, and the private sector. Collaborators can be grouped according to the following: Alden Research Laboratory, Inc., Black Bear Hydro (Scott Hall), Brookfield Power (Kevin Bernier), Connecticut Department of Energy and Environmental Protection (Steve Gephard), Department of Fisheries and Ocean Canada (Ross Jones), Maine Department of Marine Resources (Oliver Cox, Norm Dube, Randy Spencer, and Joan Trial), NOAA Fisheries Service (Al Blott, Mary Colligan, Kiersten Curti, Kim Damon-Randall, Jon Deroba, Don Dow, Dan Kircheis, John Kocik, Chris Legault, Christine Lipsky, Molly McCarthy, Alicia Miller, Paul Music, Paul Rago, Mark Renkawitz, Dan Tierney, Max Tritt, and Danielle Watson), NOAA Restoration Center (Matt Collins), U.S. Fish and Wildlife Service (Mike Bailey, Denise Buckley, Dimitry Gorsky, Fred Seavey, and Steve Shepard), U.S. Geological Survey (Alex Haro and Chris Holbrook), and U.S. Geological Survey Maine Cooperative Fish and Wildlife Research Unit (Doug Sigourney, Daniel Stitch, and Joe Zydlewski). We thank them for their willingness to help improve this model and manuscript.

9 References Cited

Amaral S, Fay C, Hecker G, Perkins N. 2012. Atlantic salmon survival estimates at mainstem hydroelectric projects on the Penobscot River. Holden (MA): Alden Research Laboratory, Inc. Phase 3 Final Report.

Aprahamian MW, Barnard S, Farooqi MA. 2004. Survival of stocked Atlantic salmon and coarse fish and an evaluation of costs. Fisheries Manag Ecol 11:153-163.

Baum ET. 1983. The Penobscot River: an Atlantic salmon river management report. Bangor (ME): State of Maine Atlantic Sea Run Salmon Commission.

Baum ET. 1997. Maine Atlantic salmon: a national treasure. 1st ed. Hermon (ME): Atlantic Salmon Unlimited.

Beissinger SR. 2002. Population viability analysis: past, present and future. In: Beissinger SR, McCullough DR, editors. Population viability analysis. Chicago (IL): The University of Chicago Press.

Budy P, Thiede GP, Bouwes N, Petrosky CE, Schaller H. 2002. Evidence linking delayed mortality of Snake River salmon to their earlier hydrosystem experience. N Am J Fish Manage 22:35-51.

Chaput G, Legault CM, Reddin DG, Caron F, Amiro PG. 2005. Provision of catch advice taking account of non-stationarity in productivity of Atlantic salmon (Salmo salar L.) in the Northwest Atlantic. ICES J Mar Sci 62:131-143.

Coghlan SM, Ringler NH. 2004. A comparison of Atlantic salmon embryo and fry stocking in the Salmon River, New York. N Am J Fish Manage 24: 1385-1397.

Coghlan SM, Connerton MJ, Ringler NH, Stewart DJ, Mead JV. 2007. Survival and growth responses of juvenile salmonines stocked in eastern Lake Ontario tributaries. Trans Am Fish Soc 136:56-71.

Cross PC, Beissinger SR. 2001. Using logistic regression to analyze the sensitivity of PVA models: a comparison of methods based on African wild dog models. Conserv Biol 15(5):1335-1346.

Crozier WW, Kennedy GJA. 1993. Marine survival of wild and hatchery-reared Atlantic salmon (Salmo salar L.) from River Bush, Northern Ireland. In: Mills D, editor. Salmon in the sea and new enhancement strategies. Oxford: Fishing News Books. p. 139-162.

Cunjak RA, Prowse TD, Parrish DL. 1998. Atlantic salmon (Salmo salar) in winter: "the season of parr discontent"? Can J Fish Aquat Sci 55(Suppl. 1):161-180.

De Leaniz CG, Verspoor E, Hawkins AD. 1989. Genetic determination of the contribution of stocked and wild Atlantic salmon, Salmo salar L., to the angling fisheries in two Spanish rivers. J Fish Biol 35(Suppl. A):261-270.

Dennis B, Munholland PL, Scott JM. 1991. Estimation of growth and extinction parameters for endangered species. Ecol Monogr 61(2):115-143.

Deriso R, Marmorek D, Parnell I. 1996. Retrospective analysis of passage mortality of spring chinook of the Columbia River. In: Marmorek DR and 21 coauthors, compilers, and editors. Plan for analyzing and testing hypotheses (PATH): final report of retrospective analysis for fiscal year 1996. Vancouver (BC): ESSA Technologies, Ltd.

Dumas J, Marty S. 2006. A new method to evaluate egg-to-fry survival in salmonids, trials with Atlantic salmon. J Fish Biol 68:284-304.

Egglishaw HJ, Shackley PE. 1973. An experiment on faster growth of salmon (Salmo salar L.) in a Scottish stream. J Fish Biol 5:197-204.

Egglishaw HJ, Shackley PE. 1980. Survival and growth of salmon, Salmo salar (L.), planted in a Scottish stream. J Fish Biol 16:565-584.

Einum S, Fleming IA. 2001. Implications of stocking: ecological interactions between wild and released salmonids. Nord J Freshwater Res: 75:56-70.

Essington TE. 2003. Development and sensitivity analysis of bioenergetics models for skipjack tuna and albacore: a comparison of alternative life histories. Trans Am Fish Soc 132:759-770.

Fay C, Bartron M, Craig S, Hecht A, Pruden J, Saunders R, Sheehan T, Trial J. 2006. Status review for anadromous Atlantic salmon (Salmo salar) in the United States. Report to the National Marine Fisheries Service and U.S. Fish and Wildlife Service. Silver Spring (MD): National Oceanic and Atmospheric Administration, Office of Protected Resources.

Flanagan JJ, Jones RA, Wissink R, Caissie A. 2008. An evaluation of the survival of captive-reared Atlantic salmon (Salmo salar) eggs in incubation baskets in two inner Bay of Fundy rivers. Fisheries and Oceans Canada. Can Tech Rep Fish Aquat Sci 2775.

Fleming IA. 1996. Reproductive strategies of Atlantic salmon: ecology and evolution. Rev Fish Biol Fisher 6(4):379-416.

Fleming IA, Lamberg A, Jonsson B. 1997. Effects of early experience on the reproductive performance of Atlantic salmon. Behav Ecol 8(5):470-480.

Gardiner R, Shackley P. 1991. Stock and recruitment and inversely density-dependent growth of salmon, Salmo salar L., in a Scottish stream. J Fish Biol 38:691-696.

Gorsky D. 2005. Site fidelity and the influence of environmental variables on migratory movements of adult Atlantic salmon (Salmo salar L.) in the Penobscot River Basin, Maine [Master's thesis]. [Orono (ME)]: University of Maine.

Gorsky D, Trial J, Zydlewski J, McCleave J. 2009. The effects of smolt stocking strategies on migratory path selection of adult Atlantic salmon in the Penobscot River, Maine. N Am J Fish Manage 29(4):949-957.

Haeseker SL, McCann JA, Tuomikoski J, Chockley B. 2012. Assessing freshwater and marine environmental influences on life-stage-specific survival rates of Snake River spring-summer Chinook salmon and steelhead. Trans Am Fish Soc 141:121-138.

Heggberget TG, Hansen LP, Naesje TP. 1988. Within-river spawning migration of Atlantic salmon (Salmo salar). Can J Fish Aquat Sci 45:1691-1698.

Hilborn R, Walters CJ. 1992. Quantitative fisheries stock assessment: choice, dynamics and uncertainty. New York (NY): Chapman and Hall, Inc.

Holbrook CM, Kinnison MT, Zydlewski J. 2011. Survival of migrating Atlantic salmon smolts through the Penobscot River, Maine: a prerestoration assessment. Trans Am Fish Soc 140:1255-1268.

Holbrook CM, Zydlewski J, Gorsky D, Shepard SL, Kinnison MT. 2009. Movements of prespawn adult Atlantic salmon near hydroelectric dams in the lower Penobscot River, Maine. N Am J Fish Manage 29:495-505.

Jokikokko E, Kallio-Nyberg I, Saloniemi I, Jutila E. 2006. The survival of semi-wild, wild and hatchery-reared Atlantic salmon smolts of the Simojoki River in the Baltic Sea. J Fish Biol 68:430-442.

Jonsson B. 1997. A review of ecological and behavioural interactions between cultured and wild Atlantic salmon. ICES J Mar Sci 54:1031-1039.

Jonsson B, Fleming IA. 1993. Enhancement of wild salmonids populations. In: Sundnes G, editor. Human impact on self-recruiting populations. An international symposium; 1993 Jun 7-11; Kongsvoll, Norway.Tapir, Trondheim, Norway. P. 209-242.

Jonsson B, Jonsson N, Hansen LP. 1991. Differences in life history and migratory behaviour between wild and hatchery-reared Atlantic salmon in nature. Aquaculture 98:69-78.

Jonsson N, Jonsson B, Hansen LP. 2003. The marine survival and growth of wild and hatchery-reared Atlantic salmon. J Appl Ecol 40:900-911.

Jordan RM, Beland KF. 1981. Atlantic salmon spawning and evaluation of natural spawning success. Augusta (ME): State of Maine Atlantic Sea Run Salmon Commission. Final Project Report AFS-20-R.

Jutila E, Jokikokko E, Kallio-Nyberg I, Saloniemi I, Pasanen P. 2003. Differences in sea migration between wild and reared Atlantic salmon (Salmo salar L.) in the Baltic Sea.Fish Res 60:333-343.

Kallio-Nyberg I, Jutila E, Saloniemi I, Jokikokko E. 2004. Association between environmental factors, smolt size and the survival of wild and reared Atlantic salmon from the Simojoki River in the Baltic Sea. J Fish Biol 65:122-134.

Kallio-Nyberg I, Saloniemi I, Jutila E, Jokikokko E. 2011. Effect of hatchery rearing and environmental factors on the survival, growth and migration of growth and migration of Atlantic salmon in the Baltic Sea. Fish Res 109:285-294.

Kennedy GJA, Strange CD. 1980. Population changes after two years of salmon (Salmo salar L.) stocking in upland trout (Salmo trutta L.) streams. J Fish Biol 17:577-586.

Kennedy GJA, Strange CD. 1986. The effects of intra- and inter-specific competition on the distribution of stocked juvenile Atlantic salmon, Salmo salar L., in relation to depth and gradient in an upland trout, Salmo trutta L., stream. J Fish Biol 29:199-214.

Kocik JF, Ferreri CP. 1998. Juvenile production variation in salmonids: population dynamics, habitat, and the role of spatial relationships. Can J Fish Aquat Sci 55(Suppl. 1):191−200.

Legault CM. 2004. Salmon PVA: a population viability analysis model for Atlantic salmon in the Maine Distinct Population Segment. U.S. Department of Commerce Northeast Fisheries Science Center Reference Document 04-02. Available from: National Marine Fisheries Service, 166 Water St., Woods Hole, MA 02543-1026.

Legault CM. 2005. Population viability analysis of Atlantic salmon in Maine, USA. Trans Am Fish Soc 134:549-562.

Letcher BH, Gries G, Juanes F. 2002. Survival of stream-dwelling Atlantic salmon: effects of life history variation, season, and age. Trans Am Fish Soc 131:838-854.

McCarthy MA, Burgman MA, Ferson S. 1996. Logistic sensitivity and bounds for extinction risks. Ecol Model 86:297-303.

McGowan CP, Ryan MR. 2009. A quantitative framework to evaluate incidental take and endangered species population viability. Biol Conserv 142:3128-3136.

McGowan CP, Ryan MR. 2010. Arguments for using population models in incidental take assessments for endangered species. J Fish Wildlife Manage 1(2):183-188.

McMenemy JR. 1995. Survival of Atlantic salmon fry stocked at low density in the West River, Vermont. N Am J Fish Manage 15:366-374.

Meister AL. 1962. Atlantic salmon production in Cove Brook, Maine. Trans Am Fish Soc 91(2):208-212.

Millard CJ. 2005. An evaluation of stocking Atlantic salmon (Salmo salar) fry in tributaries of the Oswego River, New York [Master's thesis]. [Syracuse (NY)]: State University of New York.

Morris WF, Doak DF. 2002. Quantitative conservation biology: theory and practice of population viability analysis. Sunderland (MA): Sinauer Associates, Inc.

Myers RA. 1984. Demographic consequences of precocious maturation of Atlantic salmon (Salmo salar). Can J Fish Aquat Sci 41:1349-1353.

NMFS (National Marine Fisheries Service). 2012. Biological Opinion regarding effects to the endangered GOM DPS for Atlantic salmon (Salmo salar), endangered shortnose sturgeon (Acipenser brevirostrum), and the threatened and endangered DPSs of Atlantic sturgeon (Acipenser oxyrhynchus oxyrhynchus) from proposed FERC license amendments for Black Bear Hydro Partners' Hydroelectric Projects on the Penobscot River in Maine. Gloucester [MA]: NMFS Northeast Regional Office; August 31, 2012. Available from: http://www.nero.noaa.gov/prot_res/section7/FERC-signedBOs/Black-BearHydroBO.pdf.

NRC (National Research Council). 1995. Science and the Endangered Species Act. Washington (DC): National Academy Press.

NRC (National Research Council). 2004. Atlantic Salmon in Maine. Washington (DC): National Academies Press.

Orciari RD, Leonard GH, Mysling DJ, Schluntz EC. 1994. Survival, growth, and smolt production of Atlantic salmon stocked as fry in a southern New England stream. N Am J Fish Manage 14:588-606.

Peyronnet A, Friedland KD, O'Maoileidigh N. 2008. Different ocean and climate factors control the marine survival of wild and hatchery Atlantic salmon Salmo salar in the north-east Atlantic Ocean. J Fish Biol 73:945-962.

Power JH, McCleave JD. 1980. Riverine movements of hatchery-reared Atlantic salmon (Salmo salar) upon return as adults. Environ Biol Fish 5(1):3-13.

Raffenberg MJ, Parrish DL. 2003. Interactions of Atlantic salmon (Salmo salar) and trout (Salvelinus fontinalis and Oncorhynchus mykiss) in Vermont tributaries of the Connecticut River. Can J Fish Aquat Sci 60:279-285.

Ruggles CP. 1980. A review of the downstream migration of Atlantic salmon. Canadian Technical Report of Fisheries and Aquatic Sciences 952.

Salminen M, Alapassi T, Ikonen E. 2007. The importance of stocking age in the enhancement of River Kymijoki salmon (Salmo salar). J Appl Ichthyol 23:46-52.

Saloniemi I, Jokikokko E, Kallio-Nyberg I, Jutila E, Pasanen P. 2004. Survival of reared and wild Atlantic salmon smolts: size matters more in bad years. ICES J Mar Sci 61:782-787.

Samson FB. 2002. Population viability analysis, management and conservation planning at large scales. In: Beissinger SR, McCullough DR, editors. Population viability analysis. Chicago (IL): The University of Chicago Press.

Saunders RL. 1967. Seasonal pattern of return of Atlantic salmon in the Northwest Miramichi River, New Brunswick. J Fish Res Board Can 24(1):21-32.

Schaller HA, Petrosky CE. 2007. Assessing hydrosystem influence on delayed mortality of Snake River stream-type Chinook salmon. N Am J Fish Manage 27:810-824.

Shepard SL. 1995. Atlantic salmon spawning migrations in the Penobscot River, Maine: fishways, flows and high temperatures [Master's thesis]. [Orono (ME)]: University of Maine.

Trinko Lake TR, Ravana KR, Saunders R. 2012. Evaluating changes in diadromous species distributions and habitat accessibility following the Penobscot River Restoration Project. Mar Coast Fish 4:284-293.

USASAC (Unites States Atlantic Salmon Assessment Committee). 2011. Annual Report of the U.S. Atlantic Salmon Assessment Committee. Portland (ME): 2011 Mar 2-5. Report No. 23 - 2010 Activities.

USFWS (United States Fish and Wildlife Service). 1988. Projected impacts of the proposed Basin Mills Dam on the Penobscot River Atlantic salmon restoration program. East Orland (ME): ASAL Working Group.

White GC, Burnham KP. 1999. Program MARK: survival estimation from populations of marked animals. Bird Study 46(Suppl.):S120-S139.

Wright J, Sweka J, Abbott A, Trinko T. 2008. GIS-based Atlantic salmon habitat model. In: National Marine Fisheries Service. 2009. Biological valuation of Atlantic salmon habitat within the Gulf of Maine Distinct Population Segment. Gloucester (MA).

www.nefsc.noaa.gov
NMFS Search
Link Disclaimer
webMASTER
Privacy Policy
(File Modified Jul. 01 2016)

This page has had 4 visits today, 4 visits this week, 137 visits this month, 1,445 visits this year