|
the film forum library tutorial contact |
|
The Density Dilemma: Limitations on
by Annika W. Walters, Timothy Copeland, David A. Venditti
|
(All Figures & Tables are available at original PDF)
bluefish excerpts some pertinent highlights from Discussion:
HABITAT QUALITY
In our study, the populations with the greatest degree of density dependence (Lemhi River, East Fork, and Crooked Fork) also experience a higher degree of anthropogenic disturbance; therefore, habitat restoration may be effective in some situations. However, many of the populations studied are in relatively undisturbed areas suggesting that habitat quality is not a universal limiting factor.MARINE-DERIVED NUTRIENTS
The idea that there are insufficient resources due to nutrient limitation also has proponents (Achord et al. 2003). In Atlantic salmon Salmo salar, variation in prey biomass has been shown to be one of the most important factors driving growth (Ward et al. 2009). In the Pacific Northwest, it is estimated that returning spawners supply only 6-7% of the historic load of marine-derived nutrients (Gresh et al. 2000) resulting in nutrient-limited streams (Sanderson et al. 2009). In addition, studies of nutrient additions in the form of salmon carcasses have often resulted in increased salmon growth (Wipfli et al. 2003; Rinella et al. 2009). Lowered resource availability could also be interacting with abiotic factors; Crozier et al. (2010) found that density-dependent growth was more prevalent at higher temperatures, which they attributed to increased resource requirements at higher temperatures.CLUSTERING & HATCHERY SUPPLEMENTATION
The importance of spatial clustering, either due to clustered redd distribution and limited movement or due to predation risk, is less well explored. Two popular management techniques, population supplementation with hatchery fish and restoration of spawning habitat, will be ineffective if juvenile clustering is not addressed. Increased numbers of juveniles due to stocking can actually lead to increased severity of density-dependent growth and survival if juveniles do not disperse (Walters & Juanes 1993; Buhle et al. 2009). Renovation of spawning habitat is also ineffective unless adults colonise restored areas, and there is sufficient rearing habitat available.REFUGIA - MAINSTEM HABITAT - WINTER MORTALITY
Conservation actions should focus on life-history stages that are most susceptible to density depen- dence. Our analyses suggest that winter mortality is important for these populations. Use of winter refugia can be density dependent (Armstrong & Griffiths 2001); however, winter habitats have not been identified for parr. If these are delineated, managers could consider how to increase refugia quantity. Given these areas are likely in larger main stem habitat, this problem needs careful and creative planning to make the issue tractable. Bjornn (1971) found that fewer fish emigrated in the fall from experimental channels provided with rubble substrate rather than gravel, showing the importance of appropriate habitat to fish preparing for winter. NUTRIENT Another approach to reduce winter mortality is to increase juvenile quality going into that period. Several authors (Gresh et al. 2000; Wipfli et al. 2003; Kohler et al. 2012) have proposed augmenting nutrient levels to increase growth. To our knowledge, none have demonstrated a population-level benefit to date. Nutrient augmentation would likely be most successful if complemented by an increase in refugia in summer rearing reaches, which may allow juveniles to safely access resources currently too risky to use (Walters & Juanes 1993).HABITAT RESTORATION OF QUESTIONABLE VALUE
The scale of the actions necessary to address density dependence across the study area is likely to be considerable and varies between streams. While some streams would clearly benefit from local habitat restoration efforts, many of these populations inhabit minimally impacted areas in remote settings. If density dependence is a natural feature of these populations, then it will be hard to generate many more smolts from the currently occupied areas. Indeed, Petrosky et al. (2001) found that the productivity rate (smolts per spawner) of the aggregated Snake River spring/summer Chinook salmon populations did not change significantly between the 1962 and 1999 brood years, indicating that the quality of currently occupied habitats has not changed greatly in the last few decades. Hilborn (1985) posited that stressed populations lose the less productive subunits first; surviving subunits are more productive and respond quickly but might not fill former areas.
SMOLT-TO-ADUL SURVIVAL RATES
Therefore, recovery to levels capable of sustainable harvest may depend on an increase in smolt-to-adult survival rates, so there are sufficient adults to recolonise non-core areas.
The basic concept that density dependence exists in fish populations is not new and has been institutionalised in the use of stock-recruit functions to analyse fisheries worldwide (Hilborn & Walters 1992; Quinn & Deriso 1999; Myers 2002). Density dependence theoretically allows populations to be resilient to stressors such as human exploitation; however, if stressors also reduce capacity, populations may become trapped in a lower productivity state (Peterman 1987). This is alarming because small populations are susceptible to extinction by chance catastrophes. Currently, there is substantial uncertainty about the strength of density dependence in threatened populations (Achord et al. 2003).
Density dependence has repeatedly been shown to have an important role in salmonid population dynamics (Milner et al. 2003; Einum et al. 2006). Density dependence has been attributed to competition for habitat (interference competition) or competition for available resources (exploitation competition) (Grant & Imre 2005). These competitive processes can interact with density-independent factors, such as temperature (Utz & Hartman 2009; Crozier et al. 2010) or flow (Vollestad & Olsen 2008; Teichert et al. 2010).
Density dependence can present itself in a number of ways, including density-dependent fecundity, age at maturity, growth, mortality and dispersal. The strength and relative importance of these processes can vary. Density-dependent mortality is often thought to be strongest at early life stages, especially if juveniles have limited mobility. In an experimental study, Einum et al. (2006) found density-dependentmortality for early juveniles and density-dependent dispersal for older juveniles. Density-dependent growth has been found to be most important at lower juvenile densities (Grant & Imre 2005; Lobon-Cervia 2007). The processes of density dependence are not independent; many studies report evidence of density-dependent growth (reviewed by Vincenzi et al. 2012), but if mortality is size-selective, density-dependent growth will have delayed demographic consequences. The processes of density dependence and the life stages affected have important implications for habitat restoration (Greene & Beechie 2004; Einum et al. 2008) and population supplementation (Lorenzen 2005).
In this study, we demonstrate that density dependence is ubiquitous in juvenile production for spring/summer Chinook salmon Oncorhynchus tshawytscha that spawn in central Idaho, USA. We examine patterns in relevant population parameters (movement, growth, survival) to understand the processes that produce the overall pattern. Insight into the role of density dependence can help guide and prioritise management strategies for these threatened populations (Einum et al. 2008).
The study populations are considerably below historical capacity for adult spawners. It is estimated there were greater than 1.5 million adult spawners in late 1800s, but this dropped to approximately 100,000 in the 1950s and to less than 10,000 during the 1980s (Matthews & Waples 1991). All populations are part of the Snake River spring/summer run Chinook salmon Evolutionarily Significant Unit (ESU), which was listed as threatened under the U.S. Endangered Species Act in 1992. At the time of listing, the adult population was estimated to be 0.5% of the historical abundance (Matthews & Waples 1991). Note that this period of extreme low abundance coincides with the beginning of our study. In terms of juvenile capacity, yearly smolt production for the ESU reaches an asymptote of approximately 1.6 million since 1990 (Copeland et al. 2009) compared with an average of 2.5 million during 1964-1969 (Raymond 1979).
Snake River spring/summer run Chinook salmon are considered to have a stream-type life history (Good et al. 2005); that is, they have an extended freshwater rearing phase and enter the ocean as yearlings. In one of our study populations (Pahsimeroi River), a sizeable fraction of the juveniles emigrate to the ocean as subyearlings (Copeland & Venditti 2009), but this group has very low adult return rates and was excluded from this study. In general, juvenile Chinook salmon emigrants display two distinct migratory phenotypes: leaving the spawning grounds as subyearlings during June through November (parr) and emigrating one full year after emergence during their second spring (age-1 smolts). Parr emigrants spend the winter in main stem reaches and pass Lower Granite Dam (Fig. 1, the first of eight dams that juveniles must pass to reach the Pacific Ocean) the following spring. Age-1 smolts leave the streams between March and June and travel quickly to pass Lower Granite Dam that spring.
Estimates of juvenile production, survival, growth and migration were all based on monitoring data. In all populations, emigrating juveniles were collected using rotary screw traps. The fish caught at the traps were counted and measured (fork length), and a subsample was PIT-tagged. Traps on the Lemhi River, Pahsimeroi River, upper Salmon River, Marsh Creek, South Fork Salmon, Red River and Crooked Fork Creek were operated by the Idaho Department of Fish and Game. The Secesh River trap was operated by the Nez Perce Tribe Department of Fisheries Resources, and the East Fork Salmon River trap was operated by the Shoshone-Bannock Tribes Fisheries Department.
Total juvenile production is a combination of abundance and survival estimates. More specifically, we estimated the total abundance of each life stage (i.e., parr and age-1 smolt) at the rotary screw traps, estimated survival for each life stage to Lower Granite Dam using a subsample of PIT-tagged fish and applied these survival rates to the screw trap abundance estimates.
We calculated life-stage abundance estimates within the brood year from rotary screw trap operations with Bailey's modification of the Lincoln-Petersen estimator (Steinhorst et al. 2004). To detect changes in trap efficiency, we conducted efficiency trials using marked fish released upstream of the traps. We divided each trap season into temporal strata corresponding to parr and smolt movement timing, and these strata were further subdivided into shorter substrata in response to changes in environmental conditions (e.g., flow and temperature). To maintain robustness for analysis, we set a lower limit of seven mark recaptures for any substrata (Steinhorst et al. 2004). If a substratum did not contain a sufficient number of recaptures, it was included with the previous or subsequent strata depending on stream and trap conditions. Young-of-the-year Chinook salmon fry (identified by readily observable differences in size, coloration and lateral body markings) were not included in smolt estimates for the spring season, and precocial Chinook salmon (large juveniles that freely expressed milt when handled) were not included in parr abundance.
Juvenile survival from emigration from the traps to Lower Granite Dam (calculated separately for parr and age-1 smolts) was estimated from the detection of PIT-tagged individuals in the lower Snake and Columbia rivers. Daily detection records were obtained by querying the PTAGIS database (www. ptagis.org) for all observations of fish tagged at each trap by calendar year. We estimated survival to Lower Granite Dam by emigrant type within each cohort using a Cormack-Jolly-Seber model implemented by SURPH software (Lady et al. 2010). Model inputs were records of the PIT tags released at each trap and their subsequent detection at downstream sites. Model outputs were the probability of being detected at Lower Granite Dam (based on detections there and downstream) and the probability of survival to Lower Granite Dam. The number of each emigrant type surviving to Lower Granite Dam was computed by multiplying abundance estimates at the trap by survival probability. Juvenile production was estimated as the total number of juveniles (includes both parr and age-1 smolt emigrants) that pass Lower Granite Dam.
We used length at time of PIT tagging as our growth estimate. We queried PTAGIS for each screw trap and used average length in October, November and December as our index of growth for parr and average length in February, March and April as our index of growth for smolts (Copeland & Venditti 2009). Obvious data entry mistakes (e.g., lengths greater than 250 mm) were deleted.
For movement, we focused on emigration from the spawning grounds as estimated by the date of capture in the screw trap. For parr, we looked at the start of emigration or the day that 10% of the population had passed the trap, as we hypothesised that at high densities, parr would start emigrating earlier. In contrast, for smolts, we used the end of emigration or the day that 90% of the population had passed the trap. We hypothesised that the start of smolt emigration in the spring would be controlled by environmental factors, but in high-density years, emigration would continue for a longer period of time. We also examined the ratio of juveniles (calculated at the screw trap) that emigrate as parr relative to smolts. We analysed the ratio of parr abundance to smolt abundance and hypothesised that in high-density years, a higher proportion would emigrate as parr.
Hierarchical models have become widely used in fisheries and ecology over the last decade (Myers et al. 2001; Bolker 2009). In hierarchical models, parameters for each grouping factor (individual, population, etc.) are assumed to be distributed around global or shared parameters (Gelman & Hill 2007). The strength of hierarchical models or meta-analyses is that by combining data from multiple sources, the precision of estimated global parameters can be greatly improved (Liermann & Hilborn 1997). It also provides better estimates for populations with limited or variable data, because it can draw on the data from the other populations.
We considered four potential models: a linear model forced through the origin, a linear model with estimated intercept, the Ricker model and the Beverton-Holt model (Table 1). For the linear models, we treated the slope and intercept as hierarchical parameters, and for the Ricker and Beverton-Holt models, the 'a' and 'b' parameters were treated as hierarchical. We used deviance information criterion (DIC), a Bayesian model selection technique, to see which of the four models the data supported (Spiegelhalter et al. 2002). The posterior parameter estimates from the best model provided a global model and estimates of parameters for each individual population. Analyses were run using JAGS (Just Another Gibbs Sampler) in R utilising the runjags and R2 jags packages. All analyses were run in R (version 2.13.1; R Development Core Team 2011).
The population-level effect of density dependence occurs via processes affecting individuals. Therefore, we also considered the effect of density on growth, survival and movement for the two emigrant life-history strategies (parr and age-1 smolts). We plotted each metric versus the redd count for the corresponding year and found the negative power function (y = ax&b) provided a good fit to the data (Grant & Imre 2005; Vincenzi et al. 2010). We loge-transformed the data to linearise the relationship (ln (y) = ln(a) & b*ln(x)) and used a linear model to fit the data using the same hierarchical modelling approach for parameter estimation as described above. For the linear models, the slope (b) provides an indicator of the degree of density dependence. A negative slope suggests density dependence, with steeper slopes indicative of greater density dependence, a slope of zero suggests no density dependence, and a positive slope suggests inverse density dependence. We considered the slope to be significantly different from zero if the 95% confidence intervals for b did not overlap zero.
There was strong support for density dependence in juvenile production. The Beverton-Holt model best fit the data, and the Ricker model was the next best fit (Table 1). In the Beverton-Holt model, juvenile production is a function of the initial cohort size (redd count), intrinsic productivity (a) and the per capita strength of density-dependent mortality (b). The global model had parameter estimates for the Beverton-Holt model of a = 275.1 and b = 0.0107 (Table 1). In other words, on average 275 juveniles per redd would survive to migrate past Lower Granite Dam in the absence of density dependence. Together a and b parameters provide an estimate of average carrying capacity (a/b = 275.1/0.0107 = 25,710 juveniles). The individual populations showed substantial variation in both a (228-358) and b (0.0037-0.0409) parameters (Table 2). In terms of juvenile production, the Lemhi River, East Fork Salmon River and Crooked Fork Creek populations showed the strongest degree of density dependence (high b values). For those populations, the asymptotic production was lower due to density dependence causing a greater curvature in the stock-recruit relationship. In contrast, the upper Salmon River and the Secesh River populations had low b values and high asymptotic production and did not approach asymptotic production as quickly (Fig. 2, Table 2).
We also found evidence of density-dependent effects on growth and survival, but not on migration. The average length of parr and smolts for all populations showed a significant decrease with increasing redd count suggesting strong density-dependent growth (Fig. 3, Table 3). Both parr and smolt survival decreased with increasing redd count, but the slopes were steeper for parr survival (Fig. 4, Table 3). The end of smolt migration occurred later at higher redd counts, as hypothesised, for all populations, but was not significant for any, suggesting the timing of smolt migration is not highly density dependent (Fig. 5, Table 3). The beginning of parr migration occurred significantly earlier in two populations suggesting the potential for density-dependent movement in these populations, but there were negligible shifts in timing for theremaining populations (Fig. 5, Table 3). The ratio of parr to smolt migrants showed no significant trends in relation to redd count, but the intercept of the regression of the ratio on redds was significantly greater than one for six of nine populations (95% confidence interval of the a parameter did not overlap one). Thus, there were consistently more parr than smolt emigrants.
Previous studies have noted density dependence for Chinook salmon at a couple of our study sites. Sekulich (1980) saw density-dependent growth and dispersal in Marsh Creek, and Bjornn (1978) found evidence for density-dependent growth in the Lemhi River. Density dependence has also been seen in Snake River steelhead Oncorhynchus mykiss populations (e.g., Bowersox et al. 2011; T. Copeland, Idaho Fish and Game, unpublished data). Density dependence is not unexpected and can even promote recovery after disturbance events (Vincenzi et al. 2008), but its continued presence despite relatively undisturbed locations and historically low population abundances should be addressed by managers in planning conservation and recovery actions.
We also found strong evidence for density-dependent growth, which supports previous research that found density-dependent growth was prevalent for populations at low abundances (Grant & Imre 2005; Lobon-Cervia 2007). Early density-dependent growth may be the process driving density-dependent juvenile survival because winter mortality is often size-selective (Hurst 2007). Growth integrates fish performance over the first spring/summer rearing stage, and reduced growth can have important delayed effects on the performance of individuals and cohorts (Beckerman et al. 2002; Stamps 2006). In salmonids, density-dependent juvenile growth has been shown to affect reproductive output (Vincenzi et al. 2010), lifetime body length trajectory (Vincenzi et al. 2008) and survival (Connor & Tiffan 2012).
We did not find convincing evidence of density-dependent movement at the scale of our investigation. There were two populations that did show significantly earlier parr migration in high-density years, as one would expect, but the majority did not show this pattern, and there was no indication of an extended smolt emigration at high densities as we hypothesised. Evidence suggests that young Chinook have limited movements during their first summer (Richards & Cernera 1989; Peery & Bjornn 2000). Any density-dependent movement during the summer may have been at small scales within spawning reaches. Chinook salmon in Idaho typically make a major habitat shift in the fall in preparation for winter. This is likely tied to photoperiod and cued by weather (i.e., temperature and flow) instead of density (Bjornn 1971). It is unclear whether larger fish tend to stay in spawning reaches for the winter or whether they preferentially move downstream. Zabel & Achord (2004) found larger fish emigrate sooner, but their work focused on the spring emigration much farther downstream in the main stem Snake River.
The idea that there are insufficient resources due to nutrient limitation also has proponents (Achord et al. 2003). In Atlantic salmon Salmo salar, variation in prey biomass has been shown to be one of the most important factors driving growth (Ward et al. 2009). In the Pacific Northwest, it is estimated that returning spawners supply only 6-7% of the historic load of marine-derived nutrients (Gresh et al. 2000) resulting in nutrient-limited streams (Sanderson et al. 2009). In addition, studies of nutrient additions in the form of salmon carcasses have often resulted in increased salmon growth (Wipfli et al. 2003; Rinella et al. 2009). Lowered resource availability could also be interacting with abiotic factors; Crozier et al. (2010) found that density-dependent growth was more prevalent at higher temperatures, which they attributed to increased resource requirements at higher temperatures.
The importance of spatial clustering, either due to clustered redd distribution and limited movement or due to predation risk, is less well explored. Two popular management techniques, population supplementation with hatchery fish and restoration of spawning habitat, will be ineffective if juvenile clustering is not addressed. Increased numbers of juveniles due to stocking can actually lead to increased severity of density-dependent growth and survival if juveniles do not disperse (Walters & Juanes 1993; Buhle et al. 2009). Renovation of spawning habitat is also ineffective unless adults colonise restored areas, and there is sufficient rearing habitat available.
Conservation actions should focus on life-history stages that are most susceptible to density dependence. Our analyses suggest that winter mortality is important for these populations. Use of winter refugia can be density dependent (Armstrong & Griffiths 2001); however, winter habitats have not been identified for parr. If these are delineated, managers could consider how to increase refugia quantity. Given these areas are likely in larger main stem habitat, this problem needs careful and creative planning to make the issue tractable. Bjornn (1971) found that fewer fish emigrated in the fall from experimental channels provided with rubble substrate rather than gravel, showing the importance of appropriate habitat to fish preparing for winter. Another approach to reduce winter mortality is to increase juvenile quality going into that period. Several authors (Gresh et al. 2000; Wipfli et al. 2003; Kohler et al. 2012) have proposed augmenting nutrient levels to increase growth. To our knowledge, none have demonstrated a population-level benefit to date. Nutrient augmentation would likely be most successful if complemented by an increase in refugia in summer rearing reaches, which may allow juveniles to safely access resources currently too risky to use (Walters & Juanes 1993).
The scale of the actions necessary to address density dependence across the study area is likely to be considerable and varies between streams. While some streams would clearly benefit from local habitat restoration efforts, many of these populations inhabit minimally impacted areas in remote settings. If density dependence is a natural feature of these populations, then it will be hard to generate many more smolts from the currently occupied areas. Indeed, Petrosky et al. (2001) found that the productivity rate (smolts per spawner) of the aggregated Snake River spring/summer Chinook salmon populations did not change significantly between the 1962 and 1999 brood years, indicating that the quality of currently occupied habitats has not changed greatly in the last few decades. Hilborn (1985) posited that stressed populations lose the less productive subunits first; surviving subunits are more productive and respond quickly but might not fill former areas. Therefore, recovery to levels capable of sustainable harvest may depend on an increase in smolt-to-adult survival rates, so there are sufficient adults to recolonise non-core areas.
In summary, effective conservation and management of these populations will require a thorough consideration of density dependence. We found strong evidence for density-dependent growth and survival in multiple populations across a wide range of spawner and juvenile abundances and environmental conditions. Density in spawning reaches affects growth of all juveniles, which in turn affects survival of parr emigrants downstream and overwinter survival of smolt emigrants before they start their movements in spring. There are several reasons why density dependence could be occurring. Of these, habitat loss and degradation are being addressed, while further research is needed into the role of resource availability, spatial clustering and life-history trade-offs due to predation risk. Density dependence at the population level is common in anadromous salmonids with substantial freshwater residence time, even if the population has experienced serious declines, and must be considered in demographic analysis and management.
learn more on topics covered in the film
see the video
read the script
learn the songs
discussion forum
