Information-theoretic Model Selection Research Articles

Psittacine reintroductions: Common denominators of success

With 28% of the 350 species of parrots considered threatened, numerous conservation efforts have been initiated for these species. Among these, the restoration or establishment of new populations has increasingly relied on reintroductions as a conservation strategy, often with mixed or uncertain results. We reviewed the results and methodologies of 47 distinct releases and reintroductions of psittacines in nine different countries worldwide over the past 25years to identify common denominators of successful efforts. To do so, we established a uniform and objective definition of reintroduction success (first-year survival>0.50 and released birds breeding with conspecifics, either captive-reared or wild), and applied generalized linear models and information-theoretic model selection to multiple datasets to identify important predictor variables. We identified several likely predictors of successful psittacine reintroductions, relating to predation mitigation, habitat quality, and post-release supplementation that may provide guidance for future efforts. We also advocate SWOT analysis for objectively evaluating the suitability of potential reintroduction sites.

Analysis and modelling of tree succession on a recent rockslide deposit

Dendroecological methods were combined with vegetation and soil mapping to study recolonisation of European larch (Larix decidua), silver birch (Betula pendula) and Norway spruce (Picea abies) on a recently formed rockslide cone (deposit of 30 × 106 m3) in the Valais Alps (Switzerland). Tree density and tree height were predicted with regression models that we derived using an information-theoretic model selection approach. Results demonstrate that the deposits of the 1991 rockslide have been colonised relatively rapidly with larch (ecesis time 2 years), birch (5 years) and spruce (2 years). Most of the colonisation occurred 5–11 years following the rockslide. Clast size was the primary factor driving tree colonisation with the highest tree densities found on plots with mainly smaller (<30 cm) clast sizes. Tree height was affected by a combination of different influences, with tree age and tree density showing the most obvious effects. This study demonstrates how dendroecological methods allow reconstruction of spatio-temporal patterns of tree succession on rockslides, which may ultimately facilitate a more accurate dating of similar landforms of unknown age.

Dependence of Bayesian Model Selection Criteria and Fisher Information Matrix on Sample Size

Geostatistical analyses require an estimation of the covariance structure of a random field and its parameters jointly from noisy data. Whereas in some cases (as in that of a Matern variogram) a range of structural models can be captured with one or a few parameters, in many other cases it is necessary to consider a discrete set of structural model alternatives, such as drifts and variograms. Ranking these alternatives and identifying the best among them has traditionally been done with the aid of information theoretic or Bayesian model selection criteria. There is an ongoing debate in the literature about the relative merits of these various criteria. We contribute to this discussion by using synthetic data to compare the abilities of two common Bayesian criteria, BIC and KIC, to discriminate between alternative models of drift as a function of sample size when drift and variogram parameters are unknown. Adopting the results of Markov Chain Monte Carlo simulations as reference we confirm that KIC reduces asymptotically to BIC and provides consistently more reliable indications of model quality than does BIC for samples of all sizes. Practical considerations often cause analysts to replace the observed Fisher information matrix entering into KIC with its expected value. Our results show that this causes the performance of KIC to deteriorate with diminishing sample size. These results are equally valid for one and multiple realizations of uncertain data entering into our analysis. Bayesian theory indicates that, in the case of statistically independent and identically distributed data, posterior model probabilities become asymptotically insensitive to prior probabilities as sample size increases. We do not find this to be the case when working with samples taken from an autocorrelated random field.

Using a genetic network to parameterize a landscape resistance surface for fishers, Martes pennanti

Knowledge of dispersal-related gene flow is important for addressing many basic and applied questions in ecology and evolution. We used landscape genetics to understand the recovery of a recently expanded population of fishers (Martes pennanti) in Ontario, Canada. An important focus of landscape genetics is modelling the effects of landscape features on gene flow. Most often resistance surfaces in landscape genetic studies are built a priori based upon nongenetic field data or expert opinion. The resistance surface that best fits genetic data is then selected and interpreted. Given inherent biases in using expert opinion or movement data to model gene flow, we sought an alternative approach. We used estimates of conditional genetic distance derived from a network of genetic connectivity to parameterize landscape resistance and build a final resistance surface based upon information-theoretic model selection and multi-model averaging. We sampled 657 fishers from 31 landscapes, genotyped them at 16 microsatellite loci, and modelled the effects of snow depth, road density, river density, and coniferous forest on gene flow. Our final model suggested that road density, river density, and snow depth impeded gene flow during the fisher population expansion demonstrating that both human impacts and seasonal habitat variation affect gene flow for fishers. Our approach to building landscape genetic resistance surfaces mitigates many of the problems and caveats associated with using either nongenetic field data or expert opinion to derive resistance surfaces.

Habitat size determine algae biomass in tank-bromeliads

The primary goal of this study was to evaluate the relative importance of habitat physical properties, bottom-up and top-down factors, and their interaction on algae biomass in tank-bromeliads. We sampled algae biomass (chlorophyll-a concentration), micro-metazoan density, mosquito abundance, and several environmental variables, including nutrient concentrations and characteristics of the habitat physical structure, in a survey of 64 tank-bromeliads of four different species (Aechmea nudicaulis, Aechmea lingulata, Neoregelia cruenta, and Vriesea neoglutinosa). We analyzed the complete and individual bromeliad species datasets using an information-theoretic model selection approach (Akaike’s information criterion). Bromeliad species, maximum water volume, and bromeliad diameter comprised the best model for determining chlorophyll-a concentrations for the complete dataset. The maximum water volume also comprised the best model to explain chlorophyll-a concentrations in three of four bromeliad species datasets. Interactions between consumers and nutrient concentration were included in the subsequent models, but they were not statistically significant. Taken together, these results demonstrate that the impact of habitat size on the associated autotrophic biomass occurs possibly via changes in community susceptibility to disturbances, particularly drought. We can conclude that habitat size is more important than resource availability or herbivory on phytotelm autotrophic biomass regulation in these natural microcosms.

Seasonal pulses of migratory prey and annual variation in small mammal abundance affect abundance and reproduction by arctic foxes

We examined how large seasonal influxes of migratory prey influenced population dynamics of arctic foxes and how this varied with fluctuations in small mammal (lemming and vole) abundance—the main prey of arctic foxes throughout most of their range. Specifically, we compared how arctic fox abundance, breeding density and litter size varied inside and outside a large goose colony and in relation to annual variation in small mammal abundance. Information-theoretic model selection showed that (1) breeding density and fox abundance were 2–3 times higher inside the colony than they were outside the colony and (2) litter size, breeding density and annual variation in fox abundance in the colony tracked fluctuations in lemming abundance. The influence of lemming abundance on reproduction and abundance of arctic foxes outside the colony was inconclusive, largely because fox densities outside the colony were low, which made it difficult to detect such relationships. Lemming abundance was, thus, the main factor governing reproduction and abundance of arctic foxes in the colony, whereas seasonal influxes of geese and their eggs provided foxes with external subsidies that elevated breeding density and fox abundance above that which lemmings could support. This study highlights (1) the relative importance of migratory prey and other foods on the abundance and reproduction by local consumers and (2) how migratory animals function as vectors of nutrient transfer between distant ecosystems such as Arctic environments and wintering areas by geese thousands of kilometres to the south.

What do simple models reveal about the population dynamics of a cooperatively breeding species?

Research on cooperatively breeding species has shown that their population dynamics differ from those of conventional breeders. Populations of cooperators are structured into groups, and group‐level Allee effects are likely common. We assess the ability of phenomenological models, lacking explicit group structure, to describe population dynamics in cooperative meerkats Suricata suricatta, and we assess potential Allee effects at the population level. Using maximum likelihood model fitting and information theoretic model selection, applied to time series data from a wild meerkat population, we find simple models that incorporate rainfall and conventional density dependence to be the most parsimonious of the models considered. Detecting no population‐level Allee effect, we conclude that explicit consideration of population structure will be key to understanding the mechanisms behind population dynamics in cooperatively breeding species.

Variation in plasma levels of insulin-like growth factor-I (IGF-I) in lingcod: relationships among season, size, and gonadal steroids

The physiological response of a fish to its environment is mediated through the endocrine axis controlling growth. Therefore, growth-regulating hormone levels can serve as ecologically relevant indicators of fish growth rate. We quantified variation in plasma insulin-like growth factor-I (IGF-I) to evaluate its potential as an indicator of growth in lingcod, an economically and ecologically important bottomfish species in the northeast Pacific. An information-theoretic model selection approach was used to test the hypothesis that variation in lingcod IGF-I is related to season, body size, and gonadal steroid concentration. Season and a length × season interaction were the most important predictors of plasma IGF-I among the variables we evaluated, suggesting that season and body size should be explicitly accounted for when interpreting endocrine patterns in wild fish populations. This is among the few studies that have measured and interpreted patterns of IGF-I in wild fish and the first to describe seasonal endocrine profiles in lingcod.

Deconstructing the mammal species richness pattern in Europe - towards an understanding of the relative importance of climate, biogeographic history, habitat heterogeneity and humans

Aim We deconstructed the mammal species richness pattern in Europe to assess the importance of large-scale gradients in current macroclimate relative to biogeographic history, habitat heterogeneity and human influence (HHH variables) as richness determinants for total species, and for widespread and endemic species separately. Location Europe, west of 30° E. Methods We deconstructed total species richness (50-km resolution) into its widespread and endemic species richness components. We used simultaneous autoregressive modelling (SAR) with information-theoretic model selection and variation partitioning to assess the importance of macroclimate and HHH variables. The HHH variables included two historical factors, estimated by novel methodologies: (1) ice-age-driven dynamics, represented by accessibility to recolonization from hindcasting-estimated glacial refugia, and (2) biogeographic peninsular dynamics, represented by distance to the entry region for the main European faunal source in western Asia. Results A large fraction of explained variation was shared between macroclimate and HHH in the SAR models. For total species richness, more variation could be uniquely attributed to macroclimate than to HHH, whereas for the deconstructed patterns (widespread and endemic species) the opposite was the case. Considering the individual factors, there was a strong peninsula effect on both widespread and endemic species richness but not on total richness. Main conclusions Both macroclimate and HHH variables (history, habitat heterogeneity and human influence) proved important predictors of species richness, but also difficult to disentangle. Notably, biogeographic history, in particular peninsular dynamics, is an important determinant of widespread and endemic species richness.

Distribution and nest site selection of red-shouldered hawks ( Buteo lineatus) in forests of northeastern Wisconsin (USA)

Red-shouldered hawks ( Buteo lineatus) are threatened in Wisconsin and when nest sites are found during the cruising or marking stage of timber harvesting, the harvest is altered to accommodate the hawks. If nest site locations are known before initiation of timber harvest, foresters can employ a proactive approach to manage red-shouldered hawks while maintaining timber production. We searched for red-shouldered hawks nest sites on Marinette County Forest (MCF) which encompasses 94,000 ha in northeastern Wisconsin and is the second largest county forest in the state. We used a comparative modeling approach to evaluate distribution and habitat relationships of red-shouldered hawk nest sites in relation to a suite of environmental variables in MCF. Models were used to develop forest management recommendations for red-shouldered hawks in Wisconsin. During the spring of 2006 and 2007, we broadcasted conspecific calls to survey 1121 calling stations along forest roads and trails. We located 20 and 25 active nesting territories in 2006 and 2007, respectively (11 of which were active in both years). To understand nest site selection, we measured 22 habitat variables within 0.04-ha plots at active nest sites ( n = 34) and at stratified random sites ( n = 61). Logistic regression with information-theoretic model selection identified a model including greater tree species richness and closeness to forested wetland as the best-approximating model. Variable selection with Discriminant Function Analysis (DFA) indicated that nest selection was best explained by greater number of tree species, closer distance to forested wetlands, greater volume of downed woody debris, fewer small sawlogs, and increased proximity to streams. Univariate comparisons identified four of the five aforementioned variables in the DFA model as significant. Red-shouldered hawks are likely more common in Wisconsin than their state status suggests. Forest management for red-shouldered hawk nest sites should focus on increasing tree species richness, increasing down woody debris volume, and protecting forested wetlands. These recommendations may assist property managers to locate and plan for continued persistence of this species on MCF.

Model selection and model averaging in behavioural ecology: the utility of the IT-AIC framework

Behavioural ecologists often study complex systems in which multiple hypotheses could be proposed to explain observed phenomena. For some systems, simple controlled experiments can be employed to reveal part of the complexity; often, however, observational studies that incorporate a multitude of causal factors may be the only (or preferred) avenue of study. We assess the value of recently advocated approaches to inference in both contexts. Specifically, we examine the use of information theoretic (IT) model selection using Akaike’s information criterion (AIC). We find that, for simple analyses, the advantages of switching to an IT-AIC approach are likely to be slight, especially given recent emphasis on biological rather than statistical significance. By contrast, the model selection approach embodied by IT approaches offers significant advantages when applied to problems of more complex causality. Model averaging is an intuitively appealing extension to model selection. However, we were unable to demonstrate consistent improvements in prediction accuracy when using model averaging with IT-AIC; our equivocal results suggest that more research is needed on its utility. We illustrate our arguments with worked examples from behavioural experiments.

Taking the chaos out of genetic patchiness: seascape genetics reveals ecological and oceanographic drivers of genetic patterns in three temperate reef species

Marine species frequently show weak and/or complex genetic structuring that is commonly dismissed as 'chaotic' genetic patchiness and ecologically uninformative. Here, using three datasets that individually feature weak chaotic patchiness, we demonstrate that combining inferences across species and incorporating environmental data can greatly improve the predictive value of marine population genetics studies on small spatial scales. Significant correlations in genetic patterns of microsatellite markers among three species, kelp bass Paralabrax clathratus, Kellet's whelk Kelletia kelletii and California spiny lobster Panulirus interruptus, in the Southern California Bight suggest that slight differences in diversity and pairwise differentiation across sampling sites are not simply noise or chaotic patchiness, but are ecologically meaningful. To test whether interspecies correlations potentially result from shared environmental drivers of genetic patterns, we assembled data on kelp bed size, sea surface temperature and estimates of site-to-site migration probability derived from a high resolution multi-year ocean circulation model. These data served as predictor variables in linear models of genetic diversity and linear mixed models of genetic differentiation that were assessed with information-theoretic model selection. Kelp was the most informative predictor of genetics for all three species, but ocean circulation also played a minor role for kelp bass. The shared patterns suggest a single spatial marine management strategy may effectively protect genetic diversity of multiple species. This study demonstrates the power of environmental and ecological data to shed light on weak genetic patterns and highlights the need for future focus on a mechanistic understanding of the links between oceanography, ecology and genetic structure.

Evidence of source–sink metapopulations in a vulnerable native galaxiid fish driven by introduced trout

Introduced predators with patchy distributions can create demographic sinks within native prey populations. Such invasions may give rise to source-sink metapopulations if there are still sources of native species colonists in the landscape. In New Zealand, introduced brown trout (Salmo trutta) and rainbow trout (Oncorhychus mykiss) are linked with declines in native non-diadromous galaxiids but co-occur with these galaxiids in some locations. We investigated whether trout create sinks in Galaxias vulgaris populations, and whether trout-free reaches could act as sources, allowing persistence in the sink habitat. We conducted quantitative seasonal monitoring of G. vulgaris population structure across two subcatchments of the Waimakariri River, South Island. Two trout-free and seven trout-invaded sites in the Porter River catchment and two trout-free and five trout-invaded sites in the Broken River catchment were monitored over two winters and the adjoining summer. Spatially continuous monitoring of young-of-the-year (YOY) galaxiid distributions and apparent survival across the Broken River catchment was also undertaken. Galaxias vulgaris YOY recruitment was high in trout-free reaches, indicating positive population growth. Galaxias vulgaris was absent from three trout-invaded sites, and the remaining invaded sites had significantly depleted juvenile recruitment. Information-theoretic model selection indicated that trout, rather than habitat, drove recruitment failure. Trout-invaded sites could be divided into "sinks" that retained no YOY galaxiids, indicating no local recruitment, and "pseudosinks," which had very few recruits. Absence of small G. vulgaris at sink sites suggested population maintenance through immigration of adults from sources, whereas pseudosink sites appear capable of self-recruitment at low carrying capacities. Trout-free reaches appear to act as sources in a river network but are susceptible to future invasions by trout. Thus, not only may invasive species cause source-sink metapopulations in native species, but also the potential of refugia for natives (sources) to become future sinks highlights the vulnerability of these metapopulations when invasive predators are the principal demographic driver.

Are the declining trends in forest grouse populations due to changes in the forest age structure? A case study of Capercaillie in Finland

In Finland, Capercaillie ( Tetrao urogallus) populations have a history of serious decrease starting from the mid-20th century. The decline is temporally in line with the expansion of modern forestry practices that created major changes in the landscape. We used tetraonid route-censuses from 18 forestry board districts and Finnish forest inventories (data on forest stand structure) to analyze the decline in 1965–1988. We used information theoretical model selection to evaluate a set of log-linear second order autoregressive models, allowing for spatially correlated process errors. The average trend throughout the country corresponded to an annual decline of 4.01% (mean of local trends) ± 0.24% (SEM), parallel to a half-life of 17 years. The decline was surprisingly uniform throughout the country (SD = 1.01%) and most parsimoniously explained by a geographically constant log-linear trend. At the large scale of observation applied here, population trends could not be explained by the proportional increase of younger forest age classes (<40 years old and <80 years old, respectively). Our analysis does not support the hypothesis that the decline in Capercaillie numbers is due to changes in the forest age structure, but we cannot exclude the possibility that other factors behind the decline may have interacted with forestry in general. From a conservation point of view, we caution against over-emphasizing the role of forest age especially at large spatial scales, but leaning also on other research, we recommend that more management efforts would go into the preservation of the overall forest cover and the original physiognomy in single forest patches.

Potential Component Allee Effects and Their Impact on Wetland Management in the Conservation of Endangered Anurans

Effective management of wetland quantity and quality is crucial for effective conservation of declining amphibian populations. In particular, frogs and toads that employ aggregative breeding strategies may suffer negative population impacts in response to changes in availability of aquatic breeding habitat, including overabundance of suitable habitat, if density of conspecifics attending aggregations is positively correlated with reproductive success. Here we document such a positive relationship, potentially the first example of a component Allee effect in an anuran, in the critically endangered Houston toad (Bufo houstonensis). We assessed the relationship between mean yearly chorus size and reproductive success of males at the pond level using an information theoretic model selection approach and a two-sample t-test. The chosen model contained the single variable of mean yearly chorus size to predict probability of reproduction, as selected using the Akaike Information Criterion corrected for small sample size and Akaike weight. Mean chorus sizes were significantly higher among ponds exhibiting evidence of reproduction than in those that showed no evidence of reproduction. Our results suggest that chorusing alone is a poor proxy for inference of population stability and highlight a need for reassessment of widely-used amphibian monitoring protocols. Further, amphibian conservation efforts should account for potential Allee effects in order to optimize benefits and avoid underestimating critical population thresholds, particularly in species exhibiting rapid population declines.

Response of arthropod species richness and functional groups to urban habitat structure and management

Urban areas are a particular landscape matrix characterized by a fine-grained spatial arrangement of very diverse habitats (urban mosaic). We investigated arthropods to analyse biodiversity- habitat associations along five environmental gradients (age, impervious area, management, con- figuration, composition) in three Swiss cities (96 study sites). We considered total species richness and species richness within different functional groups (zoophagous, phytophagous, pollinator, low mobility, and high mobility species). Information theoretical model selection procedures were applied and predic- tions were calculated based on weighted models. Urban areas yielded on average 284 arthropod species (range: 169-361), with species richness correlating mostly with heterogeneity indices (configuration and composition). Species richness also increased with age of urban settlement, while enlarged proportions of impervious area and intensified habitat manage- ment was negatively correlated. Functional groups showed contrasted, specific responses to environmen- tal variables. Overall, we found surprisingly little variation in species richness along the gradients, which is possibly due to the fine-grained spatial interlinkage of good (heterogeneous) and bad (sealed) habitats. The highly fragmented nature of urban areas may not represent a major obstacle for the arthropods currently existing in cities because they have prob- ably been selected for tolerance to fragmentation and for high colonisation potential. Given that built areas are becoming denser, increasing spatial heterogeneity of the urban green offers potential for counteracting the detrimental effects of densification upon urban biodiversity. By quantifying the expected effects along environmental gradients, this study provides guidance for managers to set priorities when enhanc- ing urban arthropod species richness.

Modeling disease vector occurrence when detection is imperfect: infestation of Amazonian palm trees by triatomine bugs at three spatial scales.

BackgroundFailure to detect a disease agent or vector where it actually occurs constitutes a serious drawback in epidemiology. In the pervasive situation where no sampling technique is perfect, the explicit analytical treatment of detection failure becomes a key step in the estimation of epidemiological parameters. We illustrate this approach with a study of Attalea palm tree infestation by Rhodnius spp. (Triatominae), the most important vectors of Chagas disease (CD) in northern South America.Methodology/Principal FindingsThe probability of detecting triatomines in infested palms is estimated by repeatedly sampling each palm. This knowledge is used to derive an unbiased estimate of the biologically relevant probability of palm infestation. We combine maximum-likelihood analysis and information-theoretic model selection to test the relationships between environmental covariates and infestation of 298 Amazonian palm trees over three spatial scales: region within Amazonia, landscape, and individual palm. Palm infestation estimates are high (40–60%) across regions, and well above the observed infestation rate (24%). Detection probability is higher (∼0.55 on average) in the richest-soil region than elsewhere (∼0.08). Infestation estimates are similar in forest and rural areas, but lower in urban landscapes. Finally, individual palm covariates (accumulated organic matter and stem height) explain most of infestation rate variation.Conclusions/SignificanceIndividual palm attributes appear as key drivers of infestation, suggesting that CD surveillance must incorporate local-scale knowledge and that peridomestic palm tree management might help lower transmission risk. Vector populations are probably denser in rich-soil sub-regions, where CD prevalence tends to be higher; this suggests a target for research on broad-scale risk mapping. Landscape-scale effects indicate that palm triatomine populations can endure deforestation in rural areas, but become rarer in heavily disturbed urban settings. Our methodological approach has wide application in infectious disease research; by improving eco-epidemiological parameter estimation, it can also significantly strengthen vector surveillance-control strategies.

Model Selection in Seismic Hazard Analysis: An Information-Theoretic Perspective

Abstract Although the methodological framework of probabilistic seismic hazard analysis is well established, the selection of models to predict the ground motion at the sites of interest remains a major challenge. Information theory provides a powerful theoretical framework that can guide this selection process in a consistent way. From an information-theoretic perspective, the appropriateness of models can be expressed in terms of their relative information loss (Kullback–Leibler distance) and hence in physically meaningful units (bits). In contrast to hypothesis testing, information-theoretic model selection does not require ad hoc decisions regarding significance levels nor does it require the models to be mutually exclusive and collectively exhaustive. The key ingredient, the Kullback–Leibler distance, can be estimated from the statistical expectation of log-likelihoods of observations for the models under consideration. In the present study, data-driven ground-motion model selection based on Kullback–Leibler-distance differences is illustrated for a set of simulated observations of response spectra and macroseismic intensities. Information theory allows for a unified treatment of both quantities. The application of Kullback–Leibler-distance based model selection to real data using the model generating data set for the Abrahamson and Silva (1997) ground-motion model demonstrates the superior performance of the information-theoretic perspective in comparison to earlier attempts at data-driven model selection (e.g., Scherbaum et al. , 2004).

Using a model selection criterion to identify appropriate complexity in aquatic biogeochemical models

Aquatic biogeochemical models are widely used as tools for understanding aquatic ecosystems and predicting their response to various stimuli (e.g., nutrient loading, toxic substances, climate change). Due to the complexity of these systems, such models are often elaborate and include a large number of estimated parameters. However, correspondingly large data sets are rarely available for calibration purposes, leading to models that may be overfit and possess reduced predictive capabilities. We apply, for the first time, information-theoretic model-selection techniques to a set of spatially explicit (1D) algal dynamics models of varying parameter dimension. We demonstrate that increases in complexity tend to produce a better model fit to calibration data, but beyond a certain degree of complexity the benefits of adding parameters are diminished (the risk of overfitting becomes greater). The particular approach taken here is computationally expensive, but several suggestions are made as to how multimodel methods may practically be extended to more sophisticated models.

Marbled Murrelet nest predation risk in managed forest landscapes: dynamic fragmentation effects at multiple scales

The effects of forest fragmentation on bird populations have been studied primarily as static phenomena. Yet when forests are allowed to regenerate, local edge contrast and landscape matrix composition change with time, and we would expect fragmentation effects to change accordingly. Describing this process is critical for the conservation of avian species sensitive to forest fragmentation, including the Marbled Murrelet (Brachyramphus marmoratus), a seabird threatened by ongoing harvest of old-growth forest nesting habitat. We experimentally assessed potential murrelet nest predation probability in four regions of southwestern British Columbia, Canada. We compared the fates of 448 simulated murrelet nests at paired edge and interior treatments, at sites with "hard" edges (recent clearcuts), "soft" edges (regenerating forest), and natural edges (i.e., riparian areas). Motion-sensitive digital nest cameras enabled us to focus on known predators of real nests, and patterns of nest fates did not differ between real and simulated nests. Using information-theoretic model selection (AIC) with the combined data set (116 sites), we assessed effects at patch (approximately 13 ha), landscape (approximately 1700 ha), and regional (approximately 96000 ha) scales. Nest disturbance probability at hard edges was 2.5 times that of interior sites, but soft edges had less than half the disturbance probability of interiors. There was no edge effect at natural edges. At the landscape scale, overall avian disturbance risk declined by as much as 50% with increasing amounts of regenerating forest in the surrounding matrix. These results indicate that initially negative fragmentation effects decrease as forests regenerate, at both patch and landscape scales. There was no evidence that these patterns differed between regions. Predator surveys suggested that Steller's Jays drive patterns of nest predation risk at the regional scale. Assuming that corvids are the most important predators, larger reserves of habitat will lessen negative hard-edge effects. Smaller reserves should be embedded in a protective matrix of regenerating forest to reduce predation risk at both patch and landscape scales. Our results suggest that dynamic fragmentation effects are generalizable across widespread regions and can be broadly applied to both murrelet management and the conservation of old-growth forest-breeding birds in general.