Strength of evidence for density dependence in abundance time series of 1198 species

Population limitation is a fundamental tenet of ecology, but the relative roles of exogenous and endogenous mechanisms remain unquantified for most species. Here we used multi-model inference (MMI), a form of model averaging, based on information theory (Akaike's Information Criterion) to evaluate the relative strength of evidence for density-dependent and density-independent population dynamical models in long-term abundance time series of 1198 species. We also compared the MMI results to more classic methods for detecting density dependence: Neyman-Pearson hypothesis-testing and best-model selection using the Bayesian Information Criterion or cross-validation. Using MMI on our large database, we show that density dependence is a pervasive feature of population dynamics (median MMI support for density dependence = 74.7-92.2%), and that this holds across widely different taxa. The weight of evidence for density dependence varied among species but increased consistently, with the number of generations monitored. Best-model selection methods yielded similar results to MMI (a density-dependent model was favored in 66.2-93.9% of species time series), while the hypothesis-testing methods detected density dependence less frequently (32.6-49.8%). There were no obvious differences in the prevalence of density dependence across major taxonomic groups under any of the statistical methods used. These results underscore the value of using multiple modes of analysis to quantify the relative empirical support for a set of working hypotheses that encompass a range of realistic population dynamical behaviors.     

Fitting complex population models by combining particle filters with Markov chain Monte Carlo

We show how a recent framework combining Markov chain Monte Carlo (MCMC) with particle filters (PFMCMC) may be used to estimate population state-space models. With the purpose of utilizing the strengths of each method, PFMCMC explores hidden states by particle filters, while process and observation parameters are estimated using an MCMC algorithm. PFMCMC is exemplified by analyzing time series data on a red kangaroo (Macropus rufus) population in New South Wales, Australia, using MCMC over model parameters based on an adaptive Metropolis-Hastings algorithm. We fit three population models to these data; a density-dependent logistic diffusion model with environmental variance, an unregulated stochastic exponential growth model, and a random-walk model. Bayes factors and posterior model probabilities show that there is little support for density dependence and that the random-walk model is the most parsimonious model. The particle filter Metropolis-Hastings algorithm is a brute-force method that may be used to fit a range of complex population models. Implementation is straightforward and less involved than standard MCMC for many models, and marginal densities for model selection can be obtained with little additional effort. The cost is mainly computational, resulting in long running times that may be improved by parallelizing the algorithm.     

Irruptive population dynamics in Yellowstone pronghorn.

Irruptive population dynamics appear to be widespread in large herbivore populations, but there are few empirical examples from long time series with small measurement error and minimal harvests. We analyzed an 89-year time series of counts and known removals for pronghorn (Antilocapra americana) in Yellowstone National Park of the western United States during 1918-2006 using a suite of density-dependent, density-independent, and irruptive models to determine if the population exhibited irruptive dynamics. Information-theoretic model comparison techniques strongly supported irruptive population dynamics (Leopold model) and density dependence during 1918-1946, with the growth rate slowing after counts exceeded 600 animals. Concerns about sagebrush (Artemisia spp.) degradation led to removals of >1100 pronghorn during 1947-1966, and counts decreased from approximately 700 to 150. The best models for this period (Gompertz, Ricker) suggested that culls replaced intrinsic density-dependent mechanisms. Contrary to expectations, the population did not exhibit enhanced demographic vigor soon after the termination of the harvest program, with counts remaining between 100 and 190 animals during 1967 1981. However, the population irrupted (Caughley model with a one-year lag) to a peak abundance of approximately 600 pronghorn during 1982-1991, with a slowing in growth rate as counts exceeded 500. Numbers crashed to 235 pronghorn during 1992-1995, perhaps because important food resources (e.g., sagebrush) on the winter range were severely diminished by high densities of browsing elk, mule deer, and pronghorn. Pronghorn numbers remained relatively constant during 1996-2006, at a level (196-235) lower than peak abundance, but higher than numbers following the release from culling. The dynamics of this population supported the paradigm that irruption is a fundamental pattern of growth in many populations of large herbivores with high fecundity and delayed density-dependent effects on recruitment when forage and weather conditions become favorable after range expansion or release from harvesting. Incorporating known removals into population models that can describe a wide range of dynamics can greatly improve our interpretation of observed dynamics in intensively managed populations.     

Model selection analysis of temporal variation in benefit for an ant-tended treehopper

Recent studies of mutualism have emphasized both that the net benefit to participants depends on the ecological context and that the density-dependent pattern of benefit is key to understanding the population dynamics of mutualism. Indeed, changes in the ecological context are likely to drive changes in both the magnitude of benefit and the density-dependent pattern of benefit. Despite the close linkage between these two areas of research, however, few studies have addressed the factors underlying variation in the density-dependent pattern of benefit. Here I use model selection to evaluate how variation in the benefits of a mutualism drives temporal variation in the density-dependent pattern of net benefit for the ant-tended treehopper Publilia concava. In the interaction between ants and treehoppers in the genus Publilia, ants collect the sugary excretions of treehoppers as a food resource, and treehoppers benefit both directly (e.g., by feeding facilitation) and indirectly (e.g., by predator protection). Results presented here show that temporal changes in the relative magnitude of direct and indirect benefit components of ant tending, especially the effectiveness of predator protection by ants, qualitatively change the overall pattern of density-dependent benefit between years with maximum benefit shifting from treehoppers in small to large aggregations. These results emphasize the need for empirical studies that evaluate the long-term dynamics of mutualism and theoretical studies that consider the population dynamics consequences of variation in the density-dependent pattern of benefit.     

Ontogenetic and sex-specific differences in density-dependent habitat selection of a marine fish population

The spatial dynamics of species are the result of complex interactions between density-independent and density-dependent sources of variability. Disentangling these two sources of variability has challenged ecologists working in both terrestrial and aquatic ecosystems. Using a novel spatially explicit statistical model, we tested for the presence of density-independent and density-dependent habitat selection in yellowfin sole (Limanda aspera) in the eastern Bering Sea. We found specificities in the density-dependent processes operating across ontogeny and particularly with gender. Density-dependent habitat expansion occurred primarily in females, and to a lesser degree in males. These patterns were especially evident in adult stages, while juvenile stages of both sexes exhibited a mix of different dynamics. Association of yellowfin sole with substrate type also varied by sex and to a lesser degree with size, with large females distributed over a wider range of substrates than males. Moreover, yellowfin sole expanded northward as cold subsurface waters retracted in summer, suggesting high sensitivity to arctic warming. Our findings illustrate how marginal habitats can play an important role in buffering density-dependent habitat expansion, with direct implications for resource management. Our spatially explicit modeling approach is effective in evaluating density-dependent spatial dynamics, and can easily be used to test similar hypotheses from a variety of aquatic and terrestrial ecosystems.     

Population dynamics and control of invasive Spartina alterniflora: inference and forecasting under uncertainty

Managing invaded ecosystems entails making decisions about control strategies in the face of scientific uncertainty and ecological stochasticity. Statistical tools such as model selection and Bayesian decision analysis can guide decision-making by estimating probabilities of outcomes under alternative management scenarios, but these tools have seldom been applied in invasion ecology. We illustrate the use of model selection and Bayesian methods in a case study of smooth cordgrass (Spartina alterniflora) invading Willapa Bay, Washington. To address uncertainty in model structure, we quantified the weight of evidence for two previously proposed hypotheses, that S. alterniflora recruitment varies with climatic conditions (represented by sea surface temperature) and that recruitment is subject to an Allee effect due to pollen limitation. By fitting models to time series data, we found strong support for climate effects, with higher per capita seedling production in warmer years, but no evidence for an Allee effect based on either the total area invaded or the mean distance between neighboring clones. We used the best-supported model to compare alternative control strategies, incorporating uncertainty in parameter estimates and population dynamics. For a fixed annual removal effort, the probability of eradication in 10 years was highest, and final invaded area lowest, if removals targeted the smallest clones rather than the largest or randomly selected clones. The relationship between removal effort and probability of eradication was highly nonlinear, with a sharp threshold separating -0% and -100% probability of success, and this threshold was 95% lower in simulations beginning early rather than late in the invasion. This advantage of a rapid response strategy is due to density-dependent population growth, which produces alternative stable equilibria depending on the initial invasion size when control begins. Our approach could be applied to a wide range of invasive species management problems where appropriate data are available.     

Are there reasons against open-ended research into solar radiation management? A model of intergenerational decision-making under uncertainty

Solar radiation management (SRM) has been proposed as a means of last resort against dangerous climate change. We propose a stylized model of intergenerational decision making on SRM research, greenhouse-gas abatement and SRM deployment, under uncertainties about (a) the extent of future climate damage and (b) effectiveness and potential harmful side-effects of SRM. Open-ended research may reveal either that SRM effectively reduces climate damage, or that it would cause more harm than benefits. We find that SRM research increases the likelihood of deployment (“slippery slope”), and derive conditions that it decreases abatement effort in expectation (“moral hazard”). Neither of these provides a rationale against SRM research, though. The rational decision is to perform SRM research, unless (i) discounting is hyperbolic and (ii) the absolute prudence of expected climate damage is smaller than absolute risk aversion. These results generalize to the case where SRM research also provides information on climate sensitivity.     

Estimating the functional form for the density dependence from life history data

Two contrasting approaches to the analysis of population dynamics are currently popular: demographic approaches where the associations between demographic rates and statistics summarizing the population dynamics are identified; and time series approaches where the associations between population dynamics, population density, and environmental covariates are investigated. In this paper, we develop an approach to combine these methods and apply it to detailed data from Soay sheep (Ovis aries). We examine how density dependence and climate contribute to fluctuations in population size via age- and sex-specific demographic rates, and how fluctuations in demographic structure influence population dynamics. Density dependence contributes most, followed by climatic variation, age structure fluctuations and interactions between density and climate. We then simplify the density-dependent, stochastic, age-structured demographic model and derive a new phenomenological time series which captures the dynamics better than previously selected functions. The simple method we develop has potential to provide substantial insight into the relative contributions of population and individual-level processes to the dynamics of populations in stochastic environments.     

Terrestrial habitat selection and strong density-dependent mortality in recently metamorphosed amphibians

To predict the effects of terrestrial habitat change on amphibian populations, we need to know how amphibians respond to habitat heterogeneity, and whether habitat choice remains consistent throughout the life-history cycle. We conducted four experiments to evaluate how the spatial distribution of juvenile wood frogs, Rana sylvatica (including both overall abundance and localized density), was influenced by habitat choice and habitat structure, and how this relationship changed with spatial scale and behavioral phase. The four experiments included (1) habitat manipulation on replicated 10-ha landscapes surrounding breeding pools; (2) short-term experiments with individual frogs emigrating through a manipulated landscape of 1 m wide hexagonal patches; and habitat manipulations in (3) small (4-m2); and (4) large (100-m2) enclosures with multiple individuals to compare behavior both during and following emigration. The spatial distribution of juvenile wood frogs following emigration resulted from differences in the scale at which juvenile amphibians responded to habitat heterogeneity during active vs. settled behavioral phases. During emigration, juvenile wood frogs responded to coarse-scale variation in habitat (selection between 2.2-ha forest treatments) but not to fine-scale variation. After settling, however, animals showed habitat selection at much smaller scales (2-4 m2). This resulted in high densities of animals in small patches of suitable habitat where they experienced rapid mortality. No evidence of density-dependent habitat selection was seen, with juveniles typically choosing to remain at extremely high densities in high-quality habitat, rather than occupying low-quality habitat. These experiments demonstrate how prediction of the terrestrial distribution of juvenile amphibians requires understanding of the complex behavioral responses to habitat heterogeneity. Understanding these patterns is important, given that human alterations to amphibian habitats may generate extremely high densities of animals, resulting in high density-dependent mortality.     

Spatial and temporal variability modify density dependence in populations of large herbivores

A central challenge in ecology is to understand the interplay of internal and external controls on the growth of populations. We examined the effects of temporal variation in weather and spatial variation in vegetation on the strength of density dependence in populations of large herbivores. We fit three subsets of the model ln(Nt) = a + (1 + b) x ln(N(t-1)) + c x ln(N(t-2)) to five time series of estimates (Nt) of abundance of ungulates in the Rocky Mountains, USA. The strength of density dependence was estimated by the magnitude of the coefficient b. We regressed the estimates of b on indices of temporal heterogeneity in weather and spatial heterogeneity in resources. The 95% posterior intervals of the slopes of these regressions showed that temporal heterogeneity strengthened density-dependent feedbacks to population growth, whereas spatial heterogeneity weakened them. This finding offers the first empirical evidence that density dependence responds in different ways to spatial heterogeneity and temporal heterogeneity.     

A model for behavioral regulation of metapopulation dynamics

Habitat fragmentation can decrease local population persistence by reducing connectivity, which is a function of dispersal of individuals among habitat fragments. Dispersal is often treated as diffusion in population models, even though for many species it is a result of a series of behavioral decisions. We developed a metapopulation model to explore the potential importance of dispersal behaviors in driving metapopulation dynamics. We incorporated types of behavior that affect dispersal—colonization inhibiting, colonization enhancing, extinction inhibiting, extinction enhancing, rescue enhancing, rescue inhibiting—into Levins’ (1969) metapopulation model and projected occupancy rates for a variety of parameter values. Examples from the literature of behaviors associated with each of these parameters are provided. Our model simplifies into previously published metapopulation models that incorporate only a single behavior, and we present a density-dependent rescue function that leads to multiple non-zero equilibria. We found a variety of behavioral effects on metapopulations. Rescue enhancement fills patches faster than does colonization enhancement or extinction inhibition, and declines in patch occupancy are moderate with extinction enhancement, but colonization inhibition causes metapopulation extinction. We also found that with colonization and extinction inhibitions, equilibrium patch occupancy is inversely related to patch turnover rate. With density-dependent rescue, persistence depends not only on the strength of the strong rescue effect, but also on having a sufficient initial fraction of patches occupied; the stronger the rescue effect, the lower this fraction can be. This study suggests that dispersal behavior can have strong influences on metapopulation dynamics. It confirms the importance of understanding the relationship between landscape structure and dispersal behavior in understanding population persistence.     

A dynamic host selection model for mountain pine beetle,Dendroctonus ponderosae Hopkins

The link between individual habitat selection decisions (i.e., mechanism) and the resulting population distributions of dispersing organisms (i.e., outcome) has been little-studied in behavioural ecology. Here we consider density-dependent habitat (i.e., host) selection for an energy- and time-limited forager: the mountain pine beetle (Dendroctonus ponderosae Hopkins). We present a dynamic state variable model of individual beetle host selection behaviour, based on an individual’s energy state. Field data are incorporated into model parameterization which allows us to determine the effects of host availability (with respect to host size, quality, and vigour) on individuals’ decisions. Beetles choose larger trees with thicker phloem across a larger proportion of the state-space than smaller trees with thinner phloem, but accept lower quality trees more readily at low energy- and time-states. In addition, beetles make habitat selection decisions based on host availability, conspecific attack densities, and beetle distributions within a forest stand. This model provides a framework for the development of a spatial game model to examine the implications of these results for attack dynamics of beetle populations.     

Modelling landscape effects on density-contact rate relationships of deer in eastern Alberta: Implications for chronic wasting disease

Managing wildlife diseases requires an understanding of disease transmission, which may be strongly affected by host population density and landscape features. Transmission models are typically fit from time-series disease prevalence data and modelled based on how the contact rate among hosts is affected by density, which is often assumed to be a linear (density-dependent transmission) or constant (frequency-dependent transmission) relationship. However, long-term time-series data is unavailable for emerging diseases, and this approach cannot account for independent effects of landscape. We developed a mechanistic model based on ecological data to empirically derive the contact rate-density relationship in white-tailed and mule deer in an enzootic region of chronic wasting disease (CWD) in Alberta, Canada and to determine whether it was affected by landscape. Using data collected from aerial surveys and GPS-telemetry, we developed empirical relationships predicting deer group size, home range size, and habitat selection to iteratively simulate deer distributions across a range of densities and landscapes. We calculated a relative measure of total per-capita contact rate, which is proportional to the number of other deer contacted per individual per unit time, for each distribution as the sum of pairwise contact rates between a target deer and all other individuals. Each pairwise contact rate was estimated from an empirical relationship developed from GPS-telemetry data predicting pairwise contact rates as a function of home range overlap and landscape structure. Total per-capita contact rates increased as a saturating function of density, supporting a transmission model intermediate between density- and frequency-dependent transmission. This pattern resulted from group sizes that reached an asymptote with increasing deer density, although this relationship was mediated by tree and shrub coverage in the landscape, such that in heavily wooded areas, the contact rate saturated at much lower densities. These results suggest that CWD management based on herd reductions, which require a density-dependent contact rate to be effective, may have variable effects on disease across a single management region. The novel mechanistic approach we employed for estimating effects of density and landscape on transmission is a powerful complement to typical data-fitting approaches for modelling disease transmission.     

Density dependence and density independence during the early life stages of four marine fish stocks

Recruitment variability caused by density-dependent and density-independent processes is an important area within the study of fish dynamics. These processes can exhibit nonlinearities and nonadditive properties that may have profound dynamic effects. We investigate the importance of population density (i.e., density dependence) and environmental forcing (i.e., density independence) on the age-0 and age-1 abundance of capelin (Mallotus villosus), northeast Arctic cod (Gadus morhua), northeast Arctic haddock (Melanogrammus aeglefinus), and Norwegian spring spawning herring (Clupea harengus) in the Barents Sea. We use statistical methods that explicitly account for nonlinearities and nonadditive interactions between internal and external variables in the abundance of these two pre-recruitment stages. Our results indicate that, during their first five months of life, cod, haddock, and herring experience higher density-dependent survival than capelin. The abundance of age-0 cod depends on the mean age and biomass of the spawning stock, a result which has implications for the management of the entire cod stock. Temperature is another important factor influencing the abundance at age-0 and age-1 of all four species, except herring at age-1. Between age-0 and age-1, there is an attenuation of density-dependent survival for cod and herring, while haddock and capelin experience density dependence at high and low temperatures, respectively. Predation by subadult cod is important for both capelin and cod at age-1. We found strong indications for interactions among the studied species, pointing to the importance of viewing the problem of species recruitment variability as a community, rather than as a population phenomenon.     

Nonlinear effects of climate and density in the dynamics of a fluctuating population of reindeer

Nonlinear and irregular population dynamics may arise as a result of phase dependence and coexistence of multiple attractors. Here we explore effects of climate and density in the dynamics of a highly fluctuating population of wild reindeer (Rangifer tarandus platyrhynchus) on Svalbard observed over a period of 29 years. Time series analyses revealed that density dependence and the effects of local climate (measured as the degree of ablation [melting] of snow during winter) on numbers were both highly nonlinear: direct negative density dependence was found when the population was growing (Rt > 0) and during phases of the North Atlantic Oscillation (NAO) characterized by winters with generally high (1979-1995) and low (1996-2007) indices, respectively. A growth-phase-dependent model explained the dynamics of the population best and revealed the influence of density-independent processes on numbers that a linear autoregressive model missed altogether. In particular, the abundance of reindeer was enhanced by ablation during phases of growth (Rt > 0), an observation that contrasts with the view that periods of mild weather in winter are normally deleterious for reindeer owing to icing of the snowpack. Analyses of vital rates corroborated the nonlinearity described in the population time series and showed that both starvation mortality in winter and fecundity were nonlinearly related to fluctuations in density and the level of ablation. The erratic pattern of growth of the population of reindeer in Adventdalen seems, therefore, to result from a combination of the effects of nonlinear density dependence, strong density-dependent mortality, and variable density independence related to ablation in winter.     

Non-renewable resource prices: Deterministic or stochastic trends?

In this paper, we examine temporal properties of 11 natural resource real price series from 1870 to 1990. Recent studies by Ahrens and Sharma [Trends in natural resource commodity prices: deterministic or stochastic? J. Environ. Econom. Manage. 33(1997)59–74], Berck and Roberts [Natural resource prices: will they ever turn up? J. Environ. Econom. Manage. 31(1996)65–78], and Slade [Grade selection under uncertainty: least cost last and other anomalies, J. Environ. Econom. Manage. 15(1988)189–205], among others, find that many non-renewable resource prices have a stochastic trend. We revisit this issue by employing a Lagrangian multiplier unit root test that allows for two endogenously determined structural breaks with and without a quadratic trend. Contrary to previous research, we find evidence against the unit root hypothesis for all price series. Our findings support characterizing natural resource prices as stationary around deterministic trends with structural breaks. We additionally show that both pre-testing for unit roots with breaks and allowing for breaks in the forecast model can improve forecast accuracy. Overall, the results in this paper are important in both a positive and normative sense; without an appropriate understanding of the dynamics of a time series, empirical verification of theories, forecasting, and proper inference are potentially fruitless.     

Directional selection by fisheries and the timing of sockeye salmon (Oncorhynchus nerka) migrations.

The timing of migration from feeding to breeding areas is a critical link between the growth and survival of adult animals, their reproduction, and the fitness of their progeny. Commercial fisheries often catch a large fraction of the migrants (e.g., salmon), and exploitation rates can vary systematically over the fishing season. We examined daily records of sockeye salmon (Oncorhynchus nerka) in the Egegik and Ugashik management districts in Bristol Bay, Alaska (USA), for evidence of such temporally selective fishing. In recent years, the early migrants have experienced lower fishing rates than later migrants, especially in the Egegik district, and the median migration date of the fish escaping the fisheries has been getting progressively earlier in both districts. Moreover, the overall runs (catch and escapement) in the Egegik district and, to a lesser extent the Ugashik district, have been getting earlier, as predicted in response to the selection on timing. The trends in timing were not correlated with sea surface temperature in the region of the North Pacific Ocean where the salmon tend to concentrate, but the trends in the two districts were correlated with each other, indicating that there may be some common environmental influence in addition to the effect of selection. Despite the selection, both groups of salmon have remained productive. We hypothesize that this resilience may result from representation of all component populations among the early and late migrants, so that the fisheries have not eliminated entire populations, and from density-dependent processes that may have helped maintain the productivity of these salmon populations.     

Information processing in models for density-dependent emigration: A comparison

Density-dependent emigration has been recognized as a fitness enhancing strategy. Yet, especially in the modelling literature there is no consensus about how density-dependent emigration should quantitatively be incorporated into metapopulation models. In this paper we compare the performance of five different dispersal strategies (defined by the functional link between density and emigration probability). Four of these strategies are based on published functional relationships between local population density and emigration probability, one assumes density-independent dispersal. We use individual-based simulations of time-discrete metapopulation dynamics and conduct evolution experiments for a broad range of values for dispersal mortality and environmental stochasticity. For each set of these conditions we analyze the evolution of emigration rates in ‘monoculture experiments’ (with only one type of dispersal strategy used by all individuals in the metapopulation) as well as in selection experiments that allow a pair-wise comparison of the performance of each functional type. We find that a single-parameter ‘asymptotic threshold’ strategy - derived from the marginal value theorem - with a decelerating increase of emigration rate with increasing population density, out-competes any other strategy, i.e. density-independent emigration, a ‘linear threshold’ strategy and a flexible three-parameter strategy. Only when environmental conditions select for extremely high emigration probabilities (close to one), strategies may perform approximately equally. A simple threshold strategy derived for the case of continuous population growth performs even worse than the density-independent strategy. As the functional type of the dispersal function implemented in metapopulation models may severely affect predictions concerning the survival of populations, range expansion, or community changes we clearly recommend to carefully select adequate functions to model density-dependent dispersal.     

A Habitat-Based Metapopulation Model of the California Gnatcatcher

We present an analysis of the metapopulation dynamics of the federally threatened coastal California Gnatcatcher (Polioptila c. californica) for an approximately 850 km² region of Orange County, California. We developed and validated a habitat suitability model for this species using data on topography, vegetation, and locations of gnatcatcher pair observations. Using this habitat model, we calculated the spatial structure of the metapopulation, including size and location of habitat patches and the distances among them. We used data based on field studies to estimate parameters such as survival, fecundity, dispersal, and catastrophes, and combined these parameters with the spatial structure to build a stage-structured, stochastic, spatially-explicit metapopulation model. The model predicted a fast decline and high risk of population extinction with most combinations of parameters. Results were most sensitive to density-dependent effects, the probability of weather-related catastrophes, adult survival, and adult fecundity. Based on data used in the model, the greatest difference in results was given when the simulation's time horizon was only a few decades, suggesting that modeling based on longer or shorter time horizons may underestimate the effects of alternative management actions.     

Habitat complexity modifies post-settlement mortality and recruitment dynamics of a marine fish

For species that have an open population structure, local population size may be strongly influenced by a combination of propagule supply and post-settlement survival. While it is widely recognized that supply of larvae (or recruits) is variable and that variable recruitment may affect the relative contribution of pre- and post-settlement factors, less effort has been made to quantify how variation in the strength of post-settlement mortality (particularly density-dependent mortality) will affect the importance of processes that determine population size. In this study, I examined the effects of habitat complexity on mortality of blue rockfish (Sebastes mystinus) within nearshore reefs off central California. I first tested whether variation in habitat complexity (measured as three-dimensional complexity of rocky substrate) affected the magnitude of both density-independent and density-dependent mortality. I then used limitation analysis to quantify how variation in habitat complexity alters the relative influence of recruitment, density-independent mortality, and density-dependent mortality in determining local population size. Increased habitat complexity was associated with a reduction in both density-independent and density-dependent mortality. At low levels of habitat complexity, limitation analysis revealed that mortality was strong and recruitment had relatively little influence on population size. However, as habitat complexity increased, recruitment became more important. At the highest levels of habitat complexity, limitation by recruitment was substantial, although density-dependent mortality was ultimately the largest constraint on population size. In high-complexity habitats, population dynamics may strongly reflect variation in recruitment even though fluctuations may be dampened by density-dependent mortality. By affecting both density-independent and density-dependent mortality, variation in habitat complexity may result in qualitative changes in the dynamics of populations. These findings suggest that the relative importance of pre- vs. post-settlement factors may be determined by quantifiable habitat features, rather than ambient recruitment level alone. Because the magnitude of recruitment fluctuations can affect species coexistence and the persistence of populations, habitat-driven changes in population dynamics may have important consequences for both community structure and population viability.