No full-text available
To read the full-text of this research,
you can request a copy directly from the authors.
Range wide analysis of northern spotted owl habitat relations: Reply to comments
Abstract
Bell et al. (2015) and Dunk et al. (2015) comment on our appraisal (Loehle et al., 2015) of biological insights from the US Fish and Wildlife Service models for northern spotted owl critical habitat. We here respond to those comments. We argue that while the low predictability of vegetation plot data by the gradient nearest neighbor (GNN) models may average out at very large scales and thus be useful in that context, errors at the site-specific scale may confound the modeling used to develop critical habitat designations. We further found that GNN errors violate statistical assumptions and are not propagated through the modeling exercise. We found multiple lines of evidence for habitat model instability, which may result from GNN uncertainty. We believe our evidence for lack of demographic predictability from the MaxEnt RHS values remains relevant to judicious use of these models for conservation. We similarly respond to other particular concerns with our analysis and conclude with suggestions.
A synthesis of northern spotted owl science with focus on published literature from 2006-2016.
The use of habitat suitability index (HSI) models is often criticized because of unreliable model performance; however, no consistent framework to validate these models exists. We offer a framework to evaluate the thoroughness of HSI model validation studies. We evaluated 17 studies that tested the reliability of 58 HSI models according to 7 criteria. The criteria included model components evaluated, input data variability, validity of comparative test(s) used, scale, range of HSIs, population index, and duration of population data collection. Guidelines that indexed the adequacy of study design were established for each criterion based on published information and past experience. All studies were deficient (maximum score was 4.05 out of 7.00 possible) according to adequacy guidelines. Most common deficiencies included inadequate consideration of input parameter variability, application of the models to inappropriate spatial scales, sampling too narrow a range of HSI values, and population data that were collected over too short a time frame.
Recently the US Fish and Wildlife Service, as part of a critical habitat analysis for the northern spotted owl (Strix occidentalis caurina), developed habitat suitability models based on thousands of owl nest sites distributed across 11 regions using the MaxEnt tool. Because these models formed the basis for critical habitat designations on millions of hectares of land, we undertook an independent evaluation of the FWS effort. We evaluated the accuracy of vegetation data used as input to develop the models, conducted out of sample analyses, correlated model output with owl reproductive success in two study areas, and developed alternate models using two different statistical methods. Vegetation data appeared accurate for only a few variables, and accuracy varied among model regions. Out of sample testing gave a high rate of classification errors and owl productivity was not correlated with MaxEnt model output in two study areas. Alternate statistical methods produced reasonable models with fewer variables. Critically, neither the models compared across regions nor the regions analyzed with different tools led to comparable use of variables. Thus biological interpretation of owl habitat selection models seemed ambiguous. In addition, for MaxEnt and one of the other tools, a highly significant trend by regression was found showing decreasing model accuracy as number of training nest sites increased. Together, these two results suggest that the generated models may be spurious to some unknown degree, perhaps because the underlying vegetation data, also derived from a model, are not sufficiently accurate to support the analysis and/or because the owls themselves affect habitat suitability by consuming their prey base. We suggest that the USFWS exercise caution in using MaxEnt models as a basis for regulatory purposes such as consultation, estimating likelihood of occupancy by owls, or evaluation of site-specific recovery actions.
Hutchinson defined species' realized niche as the set of environmental conditions in which populations can persist in the presence of competitors. In terms of demography, the realized niche corresponds to the environments where the intrinsic growth rate ( r ) of populations is positive. Observed species occurrences should reflect the realized niche when additional processes like dispersal and local extinction lags do not have overwhelming effects. Despite the foundational nature of these ideas, quantitative assessments of the relationship between range‐wide demographic performance and occurrence probability have not been made. This assessment is needed both to improve our conceptual understanding of species' niches and ranges and to develop reliable mechanistic models of species geographic distributions that incorporate demography and species interactions.
The objective of this study is to analyse how demographic parameters (intrinsic growth rate r and carrying capacity K ) and population density ( N ) relate to occurrence probability ( P occ ). We hypothesized that these relationships vary with species' competitive ability. Demographic parameters, density, and occurrence probability were estimated for 108 tree species from four temperate forest inventory surveys (Québec, western USA, France and Switzerland). We used published information of shade tolerance as indicators of light competition strategy, assuming that high tolerance denotes high competitive capacity in stable forest environments.
Interestingly, relationships between demographic parameters and occurrence probability did not vary substantially across degrees of shade tolerance and regions. Although they were influenced by the uncertainty in the estimation of the demographic parameters, we found that r was generally negatively correlated with P occ , while N, and for most regions K, was generally positively correlated with P occ . Thus, in temperate forest trees the regions of highest occurrence probability are those with high densities but slow intrinsic population growth rates. The uncertain relationships between demography and occurrence probability suggests caution when linking species distribution and demographic models.
The federally threatened northern spotted owl (Strix occidentalis caurina) is the focus of intensive conservation efforts that have led to much forested land being reserved as habitat for the owl and associated wildlife species throughout the Pacific Northwest of the United States. Recently, however, a relatively new threat to spotted owls has emerged in the form of an invasive competitor: the congeneric barred owl (S. varia). As barred owls have rapidly expanded their populations into the entire range of the northern spotted owl, mounting evidence indicates that they are displacing, hybridizing with, and even killing spotted owls. The range expansion by barred owls into western North America has made an already complex conservation issue even more contentious, and a lack of information on the ecological relationships between the 2 species has hampered recovery efforts for northern spotted owls. We investigated spatial relationships, habitat use, diets, survival, and reproduction of sympatric spotted owls and barred owls in western Oregon, USA, during 2007-2009. Our overall objective was to determine the potential for and possible consequences of competition for space, habitat, and food between these previously allopatric owl species. Our study included 29 spotted owls and 28 barred owls that were radio-marked in 36 neighboring territories and monitored over a 24-month period. Based on repeated surveys of both species, the number of territories occupied by pairs of barred owls in the 745-km2 study area (82) greatly outnumbered those occupied by pairs of spotted owls (15). Estimates of mean size of home ranges and core-use areas of spotted owls (1,843 ha and 305 ha, respectively) were 2-4 times larger than those of barred owls (581 ha and 188 ha, respectively). Individual spotted and barred owls in adjacent territories often had overlapping home ranges, but interspecific space sharing was largely restricted to broader foraging areas in the home range with minimal spatial overlap among core-use areas. We used an information-theoretic approach to rank discrete-choice models representing alternative hypotheses about the influence of forest conditions, topography, and interspecific interactions on species-specific patterns of nighttime resource selection. Spotted owls spent a disproportionate amount of time foraging on steep slopes in ravines dominated by old (>120 yr) conifer trees. Barred owls used available forest types more evenly than spotted owls, and were most strongly associated with patches of large hardwood and conifer trees that occupied relatively flat areas along streams. Spotted and barred owls differed in the relative use of old conifer forest (greater for spotted owls) and slope conditions (steeper slopes for spotted owls), but we found no evidence that the 2 species differed in their use of young, mature, and riparian-hardwood forest types. Mean overlap in proportional use of different forest types between individual spotted owls and barred owls in adjacent territories was 81% (range = 30-99%). The best model of habitat use for spotted owls indicated that the relative probability of a location being used was substantially reduced if the location was within or in close proximity to a core-use area of a barred owl. We used pellet analysis and measures of food-niche overlap to determine the potential for dietary competition between spatially associated pairs of spotted owls and barred owls. We identified 1,223 prey items from 15 territories occupied by spotted owls and 4,299 prey items from 24 territories occupied by barred owls. Diets of both species were dominated by nocturnal mammals, but diets of barred owls included many terrestrial, aquatic, and diurnal prey species that were rare or absent in diets of spotted owls. Northern flying squirrels (Glaucomys sabrinus), woodrats (Neotoma fuscipes, N. cinerea), and lagomorphs (Lepus americanus, Sylvilagus bachmani) were primary prey for both owl species, accounting for 81% and 49% of total dietary biomass for spotted owls and barred owls, respectively. Mean dietary overlap between pairs of spotted and barred owls in adjacent territories was moderate (42%; range = 28-70%). Barred owls displayed demographic superiority over spotted owls; annual survival probability of spotted owls from known-fate analyses (0.81, SE = 0.05) was lower than that of barred owls (0.92, SE = 0.04), and pairs of barred owls produced an average of 4.4 times more young than pairs of spotted owls over a 3-year period. We found a strong, positive relationship between seasonal (6-month) survival probabilities of both species and the proportion of old (>120 yr) conifer forest within individual home ranges, which suggested that availability of old forest was a potential limiting factor in the competitive relationship between these 2 species. The annual number of young produced by spotted owls increased linearly with increasing distance from a territory center of a pair of barred owls, and all spotted owls that attempted to nest within 1.5 km of a nest used by barred owls failed to successfully produce young. We identified strong associations between the presence of barred owls and the behavior and fitness potential of spotted owls, as shown by changes in movements, habitat use, and reproductive output of spotted owls exposed to different levels of spatial overlap with territorial barred owls. When viewed collectively, our results support the hypothesis that interference competition with barred owls for territorial space can constrain the availability of critical resources required for successful recruitment and reproduction of spotted owls. Availability of old forests and associated prey species appeared to be the most strongly limiting factors in the competitive relationship between these species, indicating that further loss of these conditions can lead to increases in competitive pressure. Our findings have broad implications for the conservation of spotted owls, as they suggest that spatial heterogeneity in vital rates may not arise solely because of differences among territories in the quality or abundance of forest habitat, but also because of the spatial distribution of a newly established competitor. Experimental removal of barred owls could be used to test this hypothesis and determine whether localized control of barred owl numbers is an ecologically practical and socio-politically acceptable management tool to consider in conservation strategies for spotted owls.
A new approach to habitat distribution modeling is presented and tested with data on North American plants. The relative frequency function (RFF) algorithm compares the relative frequencies of a species’ sample points to that of random points or absence points on the landscape to compute a frequency ratio. The relative frequency ratio r is smoothed across the range of values using moving, overlapping windows. The ratio of frequencies at a sample point for each variable is used to compute the geometric mean score for all data with non-missing values. Variables are added using a forward stepwise method. Confidence intervals are computed with bootstrap resampling. The method was tested with artificial and species habitat and geographic range data. The RFF method in all cases gave results comparable to other methods tested. For the species with good geographic range maps, the results were consistent with known biogeography.
The RFF method is particularly well-suited to irregularly shaped distributions and can classify sample points even when the data contain missing values. The method is extremely simple to use and comes with a free software tool, does not require a large sample size, does not require absence data, and is more interpretable and portable than certain other methods.
We examined territory selection, occupancy and reproduction among northern spotted owls (Strix occidentalis caurina) relative to environmental factors influencing wildfire regimes in the eastern Cascade Mountains, Washington, USA. Territory selection was influenced positively by the amount of (a) forest with 30-70% overstory canopy cover; (b) grand fir (Abies grandis) and (c) riparian habitat, and negatively by (a) late-successional forest (LSF); (b) trees 13-19 cm diameter; and (c) elevation. Owl pairs in late-successional reserves (LSRs) were not more productive than those in intervening forests. Owl reproductive rates were lower in territories with more pole-sized trees and with greater annual precipitation at higher, wetter elevations and glaciated landtypes where LSF was abundant. As a result, LSF was not correlated to owl reproduction. The most productive owl pairs occurred in dry forests that were at highest risk to uncharacteristic wildfires. Short-term successional advances toward shade-tolerant, pole-sized trees may have led to abandonment of 45 owl territories in mesic forests. Sustainable conservation in this area should account for physical environmental factors that influence owl population performance and reduce continuity and density of fuels.
We studied prey populations and the use and composition of home ranges of 47 Northern Spotted Owls (Strix occidentalis caurina) over 12 mo in five landscapes in two forest types in southwestern Oregon. We measured I-yr home ranges of 23 owl pairs, 2-yr home ranges of 13 pairs, and 3-yr home ranges of 3 pairs. The landscapes differed in the degree to which old forest had been fragmented by wildfire and logging. Prey populations were measured at 47 sites in southwestern Oregon. Further data on prey populations were gathered on 14 sites on the Olympic Peninsula in northern Washington. where owls use larger ranges than in Oregon. Owls in Washington used ≈ 1700 ha of old forest annually and primarily one prey species; available prey biomass was 61 g/ha. Owls in Oregon Douglas-fir (Pseudotsuga menziesii) forests used 813 ± 133 ha (0 ± SE) of old forest annually and concentrated on two prey species that had a combined biomass of 244 g/ha. Owls in Oregon mixed-conifer forest used 454 - 84 ha of old forest annually and three primary prey whose availability averaged 338 g/ha. The amount of old forest used by owls studied for 2 yr was 40% greater in the 2nd yr than that used in the 1st yr. No increase in use of old forest was seen in the 3rd yr in Douglas-fir forest; 50% more old forest was used in 3 yr than in the 1st yr in mixed-conifer forest. The most common prey in Washington and Oregon was the northern flying squirrel (Glaucomys sabrinus). In areas where the flying squirrel was the primary prey and where predation was intense (as judged by telemetry), flying squirrel populations were depressed. The addition of medium-sized mammal species. especially woodrats (.veotoma spp.), to the prey base appeared to reduce markedly the amount of old forest used for foraging. Owls traversed 85% more Douglas-fir forest and 3 times more mixed-conifer forest in the heavily fragmented areas than in the lightly fragmented areas. Overlap among pairs and separation of birds within pairs in space increased with fragmentation. In the most heavily fragmented landscape, social structure appeared to be abnormal, as judged by the proportion of adult-subadult pairs, instances of adult nomadism, and overlap among the home ranges of pairs. The pattern of fragmentation affected the ability of owls to find concentrations of old forest in the landscapes. Even so, almost all the owls consistently selected old forests for foraging and roosting; only one owl selected a younger type as part of its foraging range. Selection of old forest was significant at three levels: landscape, annual home ranges of pairs, and foraging and roosting sites of individuals. The most important prey species, the northern flying squirrel, was twice as abundant in old forest as in young forest in all areas. Landscape indices (dominance, contagion, variance in density of old forest) had less predictive ability than indices based on owl home ranges because owls selected areas of concentrated old forest and because patterning was complex, reflecting four processes, each operating at a different scale: physiography, human land ownership (259-ha scale), history of catastrophic fires, and history of small-scale fires and timber harvesting.
Home ranges of 14 pairs of spotted owls (Strix occidentalis) were studied over four to six seasons in three southwest Oregon landscapes that differed in prey base and degree of forest fragmentation. The general theory of space use and the biology of spotted owls was used to construct a scheme for spatiotemporal analyses that provided insights beyond those possible with polygon home-range methods. Landscape units, precisely defined by habitat type and aspect, were classified as not used, used ≤expected based on area, and used > expected (selected). Selected units were categorized further by consistency of use: frequent, intermittent, and sporadic. Ratios of use categories and costs of use based on distances to units from activity centers revealed that spotted owls adopted different tactics depending on prey base and degree and type of fragmentation. Strategies were in accordance with predictions for central-place foragers exploiting patchy environments. Owls generally concentrated their foraging in old forests, but selectively used particular young forest units, especially when dusky-footed woodrats (Neotoma fuscipes) were present. Costs of fragmentation were high, equivalent to the zoogeographic gain of adding the dusky-footed woodrat to the prey base. Selective use of young forests suggests that manipulation of young forests to provide habitat for prey and foraging by the owl would be of value in conservation efforts in this region.
The Northwest Forest Plan (NWFP), which aims to conserve late-successional and old-growth forests (older forests) and associated species, established new policies on federal lands in the Pacific Northwest USA. As part of monitoring for the NWFP, we tested nearest-neighbor imputation for mapping change in older forest, defined by threshold values for forest attributes that vary with forest succession. We mapped forest conditions on >19millionha of forest for the beginning (Time 1) and end (Time 2) of a 13-year period using gradient nearest neighbor (GNN) imputation. Reference data were basal area by species and size class from 17,000 forest inventory plots measured from 1993 to 2008. Spatial predictors were from Landsat time-series and GIS data on climate, topography, parent material, and location. The Landsat data were temporally normalized at the pixel level using LandTrendr algorithms, which minimized year-to-year spectral variability and provided seamless multi-scene mosaics. We mapped older forest change by spatially differencing the Time 1 and Time 2 GNN maps for average tree size (MNDBH) and for old-growth structure index (OGSI), a composite index of stand age, large live trees and snags, down wood, and diversity of tree sizes. Forests with higher values of MNDBH and OGSI occurred disproportionately on federal lands. Estimates of older forest area and change varied with definition. About 10% of forest at Time 2 had OGSI>50, with a net loss of about 4% over the period. Considered spatially, gross gain and gross loss of older forest were much greater than net change. As definition threshold value increased, absolute area of mapped change decreased, but increased as a percentage of older forest at Time 1. Pixel-level change was noisy, but change summarized to larger spatial units compared reasonably to known changes. Geographic patterns of older forest loss coincided with areas mapped as disturbed by LandTrendr, including large wildfires on federal lands and timber harvests on nonfederal lands. The GNN distribution of older forest attributes closely represented the range of variation observed from a systematic plot sample. Validation using expert image interpretation of an independent plot sample in TimeSync corroborated forest changes from GNN. An advantage of imputed maps is their flexibility for post-classification, summary, and rescaling to address a range of objectives. Our methods for characterizing forest conditions and dynamics over large regions, and for describing the reliability of the information, should help inform the debate over conservation and management of older forest.
Spatially explicit information on the species composition and structure of forest vegetation is needed at broad spatial scales for natural resource policy analysis and ecological research. We present a method for predictive vegetation mapping that applies direct gradient analysis and nearest-neighbor imputation to ascribe detailed ground at - tributes of vegetation to each pixel in a digital landscape map. The gradient nearest neighbor method integrates vegeta - tion measurements from regional grids of field plots, mapped environmental data, and Landsat Thematic Mapper (TM) imagery. In the Oregon coastal province, species gradients were most strongly associated with regional climate and geographic location, whereas variation in forest structure was best explained by Landsat TM variables. At the regional level, mapped predictions represented the range of variability in the sample data, and predicted area by vegetation type closely matched sample-based estimates. At the site level, mapped predictions maintained the covariance structure among multiple response variables. Prediction accuracy for tree species occurrence and several measures of vegetation structure and composition was good to moderate. Vegetation maps produced with the gradient nearest neighbor method are appropriately used for regional-level planning, policy analysis, and research, not to guide local management deci - sions.
A controversy exists in the Pacific Northwest of the United States between logging of old-growth coniferous forests and conservation of Northern Spotted Owl (Strix occidentalis caurina) populations. This species has a strong association with old-growth forests that also have economic value as timber. Research questions relevant to conservation of this species include how temporal trends in Northern Spotted Owl populations are influenced and how spatial configuration of old-growth forests affects these populations. To address these questions, we studied a population of marked Northern Spotted Owls on 95 territories in northwestern California from 1985 through 1994. We examined the magnitude of temporal and spatial variation in life history traits (survival, reproductive output, and recruitment), the effects of climate and landscape characteristics on temporal and spatial variation in these traits, respectively, and how this variation affected aspects of population dynamics. We used a components-of-variation analysis to partition sampling from process variation, and a model selection approach to estimate life history traits using capturerecapture and random-effects models. Climate explained most of the temporal variation in life history traits. Annual survival varied the least over time, whereas recruitment rate varied the most, suggesting a 'bet-hedging' life history strategy for the owl. A forecast of annual rates of population change (λ), estimated from life history traits, suggested that Northern Spotted Owl populations may change solely due to climate influences, even with unchanging habitat conditions. In terms of spatial variation, annual survival on territories was positively associated both with amounts of interior old-growth forest and with length of edge between those forests and other vegetation types. Reproductive output was negatively associated with interior forest, but positively associated with edge between mature and old-growth conifer forest and other vegetation types. A gradient existed in territory-specific estimates of fitness derived from these life history estimates. This gradient suggested that a mosaic of older forest interspersed with other vegetation types promoted high fitness in Northern Spotted Owls. Habitat quality, as defined by fitness, appeared to buffer variation in annual survival but did not buffer reproductive output. We postulated that the magnitude of λ was determined by habitat quality, whereas variation of λ was influenced by recruitment and reproductive output. As habitat quality declines, variation in λ should become more pronounced.
A common sampling design in resource selection studies involves measuring resource attributes at sample units used by an animal and at sample units considered available for use. Few models can estimate the absolute probability of using a sample unit from such data, but such approaches are generally preferred over statistical methods that estimate a relative probability of use.
The case–control model that allows for contaminated controls, proposed by Lancaster & Imbens (1996) and Lele (2009), can estimate the absolute probability of using a sample unit from use‐availability data. However, numerous misconceptions have likely prevented the widespread application of this model to resource selection studies. We address common misconceptions regarding the case–control model with contaminated controls and demonstrate its ability to estimate the absolute probability of use, prevalence and parameters associated with categorical covariates from use‐availability data.
We fit the case–control model with contaminated controls to simulated data with varying prevalence (defined as the average probability of use across all sample units) and sample sizes ( n 1 = 500 used and n a = 500 available samples; n 1 = 1000 used and n a = 1000 available samples). We then applied this model to estimate the probability Ozark hellbenders ( Cryptobranchus alleganiensis bishopi ) would use a location within a stream as a function of covariates.
The case–control model with contaminated controls provided unbiased estimates of all parameters at N = 2000 sample size simulation scenarios, particularly at low prevalence. However, this model produced increasingly variable maximum likelihood estimates of parameters as prevalence increased, particularly at N = 1000 sample size scenarios. We thus recommend at least 500–1000 used samples when fitting the case–control model with contaminated controls to use‐availability data. Our application to hellbender data revealed selection for locations with coarse substrate that are close to potential sources of cover.
This study unites a disparate literature, addresses and clarifies many commonly held misconceptions and demonstrates that the case–control model with contaminated controls is a viable alternative for estimating the absolute probability of use from use‐availability data.
Forest communities, defined by the size and configuration of cover types and stand ages, have commonly been used as proxies for the abundance or viability of wildlife populations. However, for community types to succeed as proxies for species abundance, several assumptions must be met. We tested these assumptions for birds in an Oregon forest environment. Measured habitat was a weak proxy for species abundance and vegetation cover type was a weak proxy for habitat, explaining only 4% of the variance in species abundance. The adequacy of forest community types as habitat proxies was highly dependent on classification rules and the spatial scales at which communities were defined. Habitat was perceived differently by species guilds and a single, generalized characterization of habitat is therefore unlikely to provide a reliable basis for multi-species conservation efforts. Given the weak relations between forest vegetation and species abundance, evaluation of landscape pattern is unlikely to be an effective replacement for the direct monitoring of species population size and distribution.
We are assessing the potential for current and alternative policies in the Oregon Coast Range to affect habitat capability for a suite of forest resources. We provide an example of a spatially explicit habitat capability model for northern spotted owls (Strix occidentalis caurina) to illustrate the approach we are taking to assess potential changes in habitat capability for vertebrates across the Coast Range. The model was based on vegetation structure at five spatial scales: the potential nest tree, a 0.5 ha potential nest patch, 28 ha around a potential nest patch, 212 ha around a potential nest patch, and a 1,810 ha home range area around a potential nest patch. Sensitivity analyses indicated that the proportion of the 28 ha patch in large trees around a potential nest patch, and the number of potential nest trees per ha in the nest patch, had the greatest influence on habitat capability estimates. The model was verified using georeferenced locations of spotted owl nests from systematically surveyed areas. Logistic regression analysis indicated that habitat capability scores were significantly associated with the probability of a site having a nest. Alternative model structures were tested during verification to test assumptions associated with four variables. The final model allowed development of a map of habitat capability for spotted owl nesting. The model will be linked to a model of forest dynamics to project changes in habitat capability under alternative land management policies. FOR. SCI. 48(2):203216.
To assess spotted owl use of young forests, we studied home-range sizes and habitat-use patterns of 24 adult northern spotted owls (Strix occidentalis caurina) on 2 sites in the Oregon Coast Range: the Elliott State Forest (ESF) and state forest lands in the Northern Coast Range (NCR). Conifer forests at ESF were characterized by a mixture of old, mature, and pole-sized conifer, similar to other areas occupied by spotted owls in western Oregon, USA. In contrast, conifer forests at NCR were younger than most other sites occupied by spotted owls in western Oregon and consisted primarily of conifers <80 years old. Broadleaf forest also was abundant (approx 22%) at both ESF and NCR. We used an information–theoretic approach and Akaike's Information Criterion (AIC) to evaluate a priori hypotheses about spotted owl home-range sizes and habitat-use patterns on our study areas. Considering previous knowledge about habitat requirements of the species, we predicted that owls occupying sites with fewer old conifer stands would have larger home ranges and that owls would select the oldest and most structurally diverse forest available for foraging and roosting. Our top model for evaluating home-range sizes indicated that the proportion of older conifer forest within the home range best explained the variability in home-range sizes. Although we found considerable variation in home-range size among owls, home-range sizes at ESF generally were smaller than home-range sizes at NCR, and home ranges at both sites were smaller than those reported for other study areas in western Oregon. Habitat-use patterns also varied widely among owls both within and between sites. Models containing distance to the nest tree, proximity to nearest forest edge, and proximity to nearest broadleaf-forest edge were the most parsimonious models for distinguishing owl locations from random points. On average, owl locations at both study areas were closer to ecotones between broadleaf forest and other cover types and farther from forest–nonforest ecotones than random points. Overall, we did not observe strong selection or avoidance of any cover type, although owls at ESF showed greatest use of older conifer forest while owls at NCR showed greatest use of broadleaf forest. Use of these habitat configurations and cover types by spotted owls had not been well documented prior to our study. The predictive power of our models was not great, however, indicating that factors in addition to those we included in our analysis may have influenced owl habitat-use patterns at our study areas. Based on our results, we recommend that managers at these sites maintain existing old and mature conifer forest, broadleaf forest, broadleaf-forest edges, and forested riparian areas as owl habitat; avoid timber harvest in core use areas; and plan the size of areas managed for spotted owls to reflect actual home-range and core-area sizes for owls in those forests.
We examined associations between annual reproduction and climate for 6 populations of individually marked northern spotted owls (Strix occidentalis caurina) in Washington and Oregon. We used an information-theoretical approach and mixed models to evaluate statistical models representing a priori hypotheses about the effects of weather and climate on reproduction. Reproduction was higher for adult than subadult owls and declined as the proportion of spotted owl territories with barred owl (Strix varia) detections increased. Similar to other spotted owl studies, we found that reproduction was negatively associated with cold, wet winters and nesting seasons at 3 of 6 study areas. In addition, we identified new relationships between reproduction, annual precipitation, storms, and regional climate cycles. For 3 of 6 areas, we found a quadratic relation between precipitation (rain and snow) and reproduction, with the number of young fledged per pair per year declining as precipitation in the previous year deviated from average levels. A meta-analysis conducted across all 6 areas indicated that reproduction at the regional level had a quadratic association with total winter snowfall in the preceding winter and was positively related to temperatures during the previous summer and fall. The amount of annual variation in reproduction accounted for by weather and climate varied widely across the 6 areas (4–79%), whereas variation in weather and climate across owl territories accounted for little of the spatial variation in reproduction (0–4%). Our results suggest that across the range of the species climate factors affecting prey abundance may have a greater effect on reproduction than direct effects of weather on nestlings. © 2011 The Wildlife Society.
Two problems, spurious regression bias and naive data mining, conspire to mislead analysts about predictive models for stock returns. This article demonstrates the two problems, how they interact, and makes suggestions for what to do about it.
Prediction of species’ distributions is central to diverse applications in ecology, evolution and conservation science. There is increasing electronic access to vast sets of occurrence records in museums and herbaria, yet little effective guidance on how best to use this information in the context of numerous approaches for modelling distributions. To meet this need, we compared 16 modelling methods over 226 species from 6 regions of the world, creating the most comprehensive set of model comparisons to date. We used presence-only data to fit models, and independent presence-absence data to evaluate the predictions. Along with well-established modelling methods such as generalised additive models and GARP and BIOCLIM, we explored methods that either have been developed recently or have rarely been applied to modelling species’ distributions. These include machine-learning methods and community models, both of which have features that may make them particularly well suited to noisy or sparse information, as is typical of species’ occurrence data. Presence-only data were effective for modelling species’ distributions for many species and regions. The novel methods consistently outperformed more established methods. The results of our analysis are promising for the use of data from museums and herbaria, especially as methods suited to the noise inherent in such data improve.
Satellite sensors are well suited to monitoring changes on the Earth's surface through provision of consistent and repeatable measurements at a spatial scale appropriate for many processes causing change on the land surface. Here, we describe and test a new conceptual approach to change detection of forests using a dense temporal stack of Landsat Thematic Mapper (TM) imagery. The central premise of the method is the recognition that many phenomena associated with changes in land cover have distinctive temporal progressions both before and after the change event, and that these lead to characteristic temporal signatures in spectral space. Rather than search for single change events between two dates of imagery, we instead search for these idealized signatures in the entire temporal trajectory of spectral values. This trajectory-based change detection is automated, requires no screening of non-forest area, and requires no metric-specific threshold development. Moreover, the method simultaneously provides estimates of discontinuous phenomena (disturbance date and intensity) as well as continuous phenomena (post-disturbance regeneration). We applied the method to a stack of 18 Landsat TM images for the 20-year period from 1984 to 2004. When compared with direct interpreter delineation of disturbance events, the automated method accurately labeled year of disturbance with 90% overall accuracy in clear-cuts and with 77% accuracy in partial-cuts (thinnings). The primary source of error in the method was misregistration of images in the stack, suggesting that higher accuracies are possible with better registration.
Accuracy assessments of remote sensing products are necessary for identifying map strengths and weaknesses in scientific and management applications. However, not all accuracy assessments are created equal. Motivated by a recent study published in Forest Ecology and Management (Volume 342, pages 8–20), we explored the potential limitations of accuracy assessments related to characteristics of the field data: sampling bias and spatial resolution. The authors of the previous paper used data from variable radius plots near northern spotted owl nest sites to assess the predictive accuracy of gradient nearest neighbor (GNN) maps in portions of Oregon and Washington, USA. The field plots used for accuracy assessment (1) potentially biased the accuracy assessment toward older forests and (2) examined accuracy at finer scales than the imputation map predictions under consideration. To examine both the impacts of bias and scale in accuracy assessment, we assessed the predictive accuracy of GNN maps in western and southern Oregon. We found correlation coefficients between predicted (900 m2) and observed forest attributes for small plots (506 m2) were consistently lower than accuracy assessments using larger plots (4048 m2). Similarly, correlation coefficients based only on field plots near nest sites were lower than correlations based on all field plots. These results imply that sampling bias and small plot areas result in accuracy assessments that underestimate map predictive performance. In particular, assessing accuracy at spatial scales below the resolution of the map products are overly pessimistic (i.e., low correlation coefficients). While accuracy assessment is important, care needs to be taken to ensure that the sampling design for field data does not limit inference on map accuracy.
We describe the structure of forests at 105 nest sites of northern spotted owls (Strix occidentalis caurina) in the Klamath, Coast, and Cascade provinces of western Oregon and the Olympic province of Washington. This information is critical for management and recovery of this threatened species. We compared forest stand data at nest sites with data from 105 random sites, using logistic regression for 1:1 matched pairs. All random sites were located in older forests (overstory trees >50 cm diameter at breast height [dbh]) within owl home ranges. Most nests in Oregon were in Douglas-fir (Pseudotsuga menziesii) trees (88%), whereas nests in the Olympic province were equally divided among Douglas-fir, western hemlock (Tsuga heterophylla), and western redcedar (Thuja plicata). In all 4 provinces, nests were most often located in live trees (73-97%) with broken tops (60-93%), most of which were fire scarred (77-83%). Mean diameter of all nest trees (n = 105) was 139.4 ± 5.2 cm (x̄ ± SE). Most nests (83%) were in cavities; of the 17% nests that were in platforms, most (67%) were in the Klamath province. The majority of nest sites (95%) were found from the middle to the bottom of slopes. Mean aspects at nest sites were southerly in Oregon and northwesterly in the Olympics. Elevations at nest sites were lower than at their paired random sites, and evidence of fire was present at 86% of nest sites. Univariate analyses indicted nest sites were associated with structurally diverse older forests exhibiting characteristics typical of old-growth forests in the Pacific Northwest. Mean diameter of large trees (>100 cm dbh) was greater at nest than random sites (130.6 ± 1.8 cm vs. 123.1 ± 1.3 cm; P 53.3 cm dbh with 1 or more secondary crowns) were also greater at nest sites than at random sites in all physiographic provinces (P
We compared principal components derived from sets of real data with dimensions of 120 x 7, 120 x 4, 150 x 11, 150 x 8, 150 x 5, 454 x 12, 454 x 8, and 454 x 5, to those from sets of randomly generated data of corresponding size. Principal components from subsets of 25, 50, 75, and 100 observations from the 120- and 150-observation data sets and those from subsets of 25, 50, 75, 100, 150, 200, 300, and 400 observations from the 454-observation data sets were compared. Percent variance association with components from real data was relatively constant over all sample sizes; percent variance decreased with larger samples of random data. A bootstrap method was used to develop standard error estimates on percent variance and percent of remaining variance associated with components from real data. Percent of remaining variance associated with the first four components from real data was significantly higher than analogous components from random data.
Traditionally overlooked by foresters as unproductive and ecologists as disorganized, naturally regenerating forests in the Pacific Northwest (PNW) are perhaps the least understood forest condition in the region. More recently, concerns over the rarity of this forest condition have sparked interest in identifying ecological characteristics unique to forested sites after a canopy-opening disturbance and before the re-establishment of a closed conifer canopy. Here we review the literature to identify the plant and animal associates of early-seral pre-forests in the PNW in order to provide baseline information pertaining to the recognition and conservation role of early-seral pre-forest ecosystems. We describe a number of bird, mammal, insect, amphibian and reptile species associated with PNW early-seral pre-forests either by empirical observation or inferred through life-history characteristics in an attempt to formally identify unique species indicators of naturally regenerating pre-forest communities. For Washington, Oregon, and northern California, we also review the state lists of endangered, threatened, monitored or otherwise conservation-listed species (664 unique species or subspecies for the combined region) to assess the proportion of protected species that rely on the structural attributes of early-seral pre-forests. Here, we found that these proportions are comparable to the proportions reliant on mature or late-seral forest. In addition, greater than 50% of all listed species for each of the three regions were partial or facultative users of early-seral pre-forest ecosystems. This assessment suggests that naturally structured early-seral pre-forests in the PNW provide key habitat for many species, including obligates and near-obligates, and that future research should seek to refine our understanding of the specific structural and compositional attributes that form the basis of these associations.
Wildlife habitat mapping has evolved at a rapid pace over the last few decades. Beginning with simple, often subjective, hand-drawn maps, habitat mapping now involves complex species distribution models (SDMs) using mapped predictor variables derived from remotely sensed data. For species that inhabit large geographic areas, remote sensing technology is often essential for producing range wide maps. Habitat monitoring for northern spotted owls (Strix occidentalis caurina), whose geographic covers about 23 million ha, is based on SDMs that use Landsat Thematic Mapper imagery to create forest vegetation data layers using gradient nearest neighbor (GNN) methods. Vegetation data layers derived from GNN are modeled relationships between forest inventory plot data, climate and topographic data, and the spectral signatures acquired by the satellite. When used as predictor variables for SDMs, there is some transference of the GNN modeling error to the final habitat map.
Recent increases in the use of light detection and ranging (lidar) data, coupled with the need to produce spatially accurate and detailed forest vegetation maps have spurred interest in its use for SDMs and habitat mapping. Instead of modeling predictor variables from remotely sensed spectral data, lidar provides direct measurements of vegetation height for use in SDMs. We expect a SDM habitat map produced from directly measured predictor variables to be more accurate than one produced from modeled predictors.
We used maximum entropy (Maxent) SDM modeling software to compare predictive performance and estimates of habitat area between Landsat-based and lidar-based northern spotted owl SDMs and habitat maps. We explored the differences and similarities between these maps, and to a pre-existing aerial photo-interpreted habitat map produced by local wildlife biologists. The lidar-based map had the highest predictive performance based on 10 bootstrapped replicate models (AUC = 0.809 ± 0.011), but the performance of the Landsat-based map was within acceptable limits (AUC = 0.717 ± 0.021). As is common with photo-interpreted maps, there was no accuracy assessment available for comparison. The photo-interpreted map produced the highest and lowest estimates of habitat area, depending on which habitat classes were included (nesting, roosting, and foraging habitat = 9962 ha, nesting habitat only = 6036 ha). The Landsat-based map produced an estimate of habitat area that was within this range (95% CI: 6679–9592 ha), while the lidar-based map produced an area estimate similar to what was interpreted by local wildlife biologists as nesting (i.e., high quality) habitat using aerial imagery (95% CI: 5453–7216). Confidence intervals of habitat area estimates from the SDMs based on Landsat and lidar overlapped.
We concluded that both Landsat- and lidar-based SDMs produced reasonable maps and area estimates for northern spotted owl habitat within the study area. The lidar-based map was more precise and spatially similar to what local wildlife biologists considered spotted owl nesting habitat. The Landsat-based map provided a less precise spatial representation of habitat within the relatively small geographic confines of the study area, but habitat area estimates were similar to both the photo-interpreted and lidar-based maps.
Photo-interpreted maps are time consuming to produce, subjective in nature, and difficult to replicate. SDMs provide a framework for efficiently producing habitat maps that can be replicated as habitat conditions change over time, provided that comparable remotely sensed data are available. When the SDM uses predictor variables extracted from lidar data, it can produce a habitat map that is both accurate and useful at large and small spatial scales. In comparison, SDMs using Landsat-based data are more appropriate for large scale analyses of amounts and general spatial patterns of habitat at regional scales.
Species distribution models (SDMs) are widely used to explain and predict species ranges and environmental niches. They are most commonly constructed by inferring species' occurrence–environment relationships using statistical and machine-learning methods. The variety of methods that can be used to construct SDMs (e.g. generalized linear/additive models, tree-based models, maximum entropy, etc.), and the variety of ways that such models can be implemented, permits substantial flexibility in SDM complexity. Building models with an appropriate amount of complexity for the study objectives is critical for robust inference. We characterize complexity as the shape of the inferred occurrence–environment relationships and the number of parameters used to describe them, and search for insights into whether additional complexity is informative or superfluous. By building ‘under fit’ models, having insufficient flexibility to describe observed occurrence–environment relationships, we risk misunderstanding the factors shaping species distributions. By building ‘over fit’ models, with excessive flexibility, we risk inadvertently ascribing pattern to noise or building opaque models. However, model selection can be challenging, especially when comparing models constructed under different modeling approaches. Here we argue for a more pragmatic approach: researchers should constrain the complexity of their models based on study objective, attributes of the data, and an understanding of how these interact with the underlying biological processes. We discuss guidelines for balancing under fitting with over fitting and consequently how complexity affects decisions made during model building. Although some generalities are possible, our discussion reflects differences in opinions that favor simpler versus more complex models. We conclude that combining insights from both simple and complex SDM building approaches best advances our knowledge of current and future species ranges.
Species distribution models (SDMs) are important tools for forecasting the potential impacts of future environmental changes but debate remains over the most robust modelling approaches for making projections.
Suggested improvements in SDMs vary from algorithmic development through to more mechanistic modelling approaches. Here, we focus on the improvements that can be gained by conditioning SDMs on more detailed data. Specifically, we use breeding bird data from across Europe to compare the relative performances of SDMs trained on presence–absence data and those trained on abundance data.
Species distribution models trained on presence–absence data, with a poor to slight fit according to Cohen's kappa, show an average improvement in model performance of 0·32 (SE ± 0·12) when trained on abundance data. Even those species for which models trained on presence–absence data are classified as good to excellent show a mean improvement in Cohen's kappa score of 0·05 (SE ± 0·01) when corresponding SDMs are trained on abundance data. This improved explanatory power is most pronounced for species of high prevalence.
Our results illustrate that even using coarse scale abundance data, large improvements in our ability to predict species distributions can be achieved. Furthermore, predictions from abundance models provide a greater depth of information with regard to population dynamics than their presence–absence model counterparts. Currently, despite the existence of a wide variety of abundance data sets, species distribution modellers continue to rely almost exclusively on presence–absence data to train and test SDMs. Given our findings, we advocate that, where available, abundance data rather than presence–absence data can be used to more accurately predict the ecological consequences of environmental change. Additionally, our findings highlight the importance of informative baseline data sets. We therefore recommend the move towards increased collection of abundance data, even if only coarse numerical scales of recording are possible.
Aim To test the prediction that environmental suitability derived from species distribution modelling (SDM) could be a surrogate for jaguar local population density estimates.
Location Americas.
Methods We used 1409 occurrence records of jaguars to model the distribution of the species using 11 SDM methods. We tested whether models’ suitability is linearly correlated with jaguar population densities estimated from 37 different locations. We evaluated whether the relationship between density and suitability forms a constraint envelope, in which higher densities are found mainly in regions with high suitability, whereas low densities can occur in regions with variable suitability. We tested this using heteroscedasticity test and quantile regressions.
Results A positive linear relationship between suitability and jaguar density was found only for four methods [bioclimatic envelope (BIOCLIM), genetic algorithm for rule set production (GARP), maximum entropy (Maxent) and generalized boosting models (GBM)], but with weak explanatory power. BIOCLIM showed the strongest relationship. Variance of suitability for lower densities values was larger than for higher values for many of the SDM models used, but the quantile regression was significantly positive only for BIOCLIM and random forests (RF). RF and GBM provided the most accurate models when measured with the standard SDM evaluation metrics, but possess poor relationship with local density estimates.
Main conclusions Results indicate that the relationship between density and suitability could be better described as a triangular constraint envelope than by a straight positive relationship, and some of the SDM methods tested here were able to discriminate regions with high or low local population densities. Low jaguar densities can occur in areas with low or high suitability, whereas high values are restricted to areas where the suitability is greater. In high suitability areas but with low jaguar density estimates, we discuss how extrinsic factors driving abundance could act at local scales and then prevent higher densities that would be expected by the favourable regional environmental conditions.
With the rise of new powerful statistical techniques and GIS tools, the development of predictive habitat distribution models has rapidly increased in ecology. Such models are static and probabilistic in nature, since they statistically relate the geographical distribution of species or communities to their present environment. A wide array of models has been developed to cover aspects as diverse as biogeography, conservation biology, climate change research, and habitat or species management. In this paper, we present a review of predictive habitat distribution modeling. The variety of statistical techniques used is growing. Ordinary multiple regression and its generalized form (GLM) are very popular and are often used for modeling species distributions. Other methods include neural networks, ordination and classification methods, Bayesian models, locally weighted approaches (e.g. GAM), environmental envelopes or even combinations of these models. The selection of an appropriate method should not depend solely on statistical considerations. Some models are better suited to reflect theoretical findings on the shape and nature of the species’ response (or realized niche). Conceptual considerations include e.g. the trade-off between optimizing accuracy versus optimizing generality. In the field of static distribution modeling, the latter is mostly related to selecting appropriate predictor variables and to designing an appropriate procedure for model selection. New methods, including threshold-independent measures (e.g. receiver operating characteristic (ROC)-plots) and resampling techniques (e.g. bootstrap, cross-validation) have been introduced in ecology for testing the accuracy of predictive models. The choice of an evaluation measure should be driven primarily by the goals of the study. This may possibly lead to the attribution of different weights to the various types of prediction errors (e.g. omission, commission or confusion). Testing the model in a wider range of situations (in space and time) will permit one to define the range of applications for which the model predictions are suitable. In turn, the qualification of the model depends primarily on the goals of the study that define the qualification criteria and on the usability of the model, rather than on statistics alone.
A fundamental ecological modeling task is to estimate the probability that a species is present in (or uses) a site, conditional on environmental variables. For many species, available data consist of “presence” data (locations where the species [or evidence of it] has been observed), together with “background” data, a random sample of available environmental conditions.
Recently published papers disagree on whether probability of presence is identifiable from such presence–background data alone. This paper aims to resolve the disagreement, demonstrating that additional information is required.
We defined seven simulated species representing various simple shapes of response to environmental variables (constant, linear, convex, unimodal, S‐shaped) and ran five logistic model‐fitting methods using 1000 presence samples and 10 000 background samples; the simulations were repeated 100 times. The experiment revealed a stark contrast between two groups of methods: those based on a strong assumption that species' true probability of presence exactly matches a given parametric form had highly variable predictions and much larger RMS error than methods that take population prevalence (the fraction of sites in which the species is present) as an additional parameter. For six species, the former group grossly under‐ or overestimated probability of presence. The cause was not model structure or choice of link function, because all methods were logistic with linear and, where necessary, quadratic terms. Rather, the experiment demonstrates that an estimate of prevalence is not just helpful, but is necessary (except in special cases) for identifying probability of presence. We therefore advise against use of methods that rely on the strong assumption, due to Lele and Keim (recently advocated by Royle et al.) and Lancaster and Imbens. The methods are fragile, and their strong assumption is unlikely to be true in practice. We emphasize, however, that we are not arguing against standard statistical methods such as logistic regression, generalized linear models, and so forth, none of which requires the strong assumption.
If probability of presence is required for a given application, there is no panacea for lack of data. Presence–background data must be augmented with an additional datum, e.g., species' prevalence, to reliably estimate absolute (rather than relative) probability of presence.
In order to test the veracity of currently accepted ideas about Northern Spotted Owl (Strix occidentalis caurina). habitat associations in the Klamath Province of northern California (USA) we compared different habitat descriptions using predictive habitat-association models. The current description used by federal agencies and new descriptions based on research results and field biologists' best estimates of owl nesting/roosting habitat and foraging habitat were evaluated. For each habitat description, three habitat metrics and three forms of the relationship between owl occupancy and quantities of these habitat metrics were evaluated, each at three spatial scales. Our refined descriptions of owl nesting and roosting, and foraging habitat, were better at predicting owl occupancy than the habitat description currently used by federal land managers. The best-fitting model for predicting owl occupancy was at the 200-ha scale and exhibited a pseudo-threshold relationship to nesting and roosting habitat and a quadratic relationship to foraging habitat. This model correctly classified owl-occupied sites 94% of the time for the developmental data set and between 85% and 92% of the time on four independent test data sets. The current description of owl habitat in northern California ranked among the worst in the collection of models we examined. The testing of multiple models on the four independent data sets was very important for determining the goodness-of-fit and predictive capabilities of the best models. We explored the use of the best-fitting model to predict number of owls on several independent study areas and found a strong correlation between predicted and observed number of owls. The results of this study are beginning to be used to make land-management decisions regarding harvesting and prescribed-burning activities on federal forestlands and were specifically designed to be amenable to adaptive resource management.
Consider developing a regression model in a context where substantive theory is weak. To focus on an extreme case, suppose that in fact there is no relationship between the dependent variable and the explanatory variables. Even so, if there are many explanatory variables, the R 2 will be high. If explanatory variables with small t statistics are dropped and the equation refitted, the R 2 will stay high and the overall F will become highly significant. This is demonstrated by simulation and by asymptotic calculation.
We compared vegetative structures in 4-16-ha patches in forest stands used by 12 pairs of Northern Spotted Owls (Strix ocddentalis caurina) for nesting (N = 44) and foraging (N = 38) with habitat structures in 50 stands located randomly throughout annual home ranges in a young and mid- successional forest landscape (25-79 yr-old stands) in the foothills of the western Cascades in Oregon. Forest stand structures influenced selection for stands used for foraging and nesting by Spotted Owls, and abundance of these structures varied with successional development as represented by five age classes. Conifer saplings (10-19 cm in diameter at breast height (dbh)) and trees 50-79 cm dbh were more abundant in foraging areas than nest sites or random sites. Large snags (>40 cm dbh) tended to be more abundant, down woody debris was more abundant, and cover of herbs and low-growing shrubs (
We describe 83 nest sites (0.2-ha areas) of northern spotted owls (Strix ocidentalis caurina) in mixed conifer forests on the eastern slope of the Cascade Mountains, Washington. Approximately 74% of the nest sites were in forests in intermediate stages of succession, and 27% were in old-growth forests (median = 122 yr, range 54-700 yr). Most sites were naturally regenerated after fire, but 23% of the nest sites had been partially harvested greater-than-or-equal-to 40 years ago. We tested the hypothesis that habitat structure does not influence nest site selection within forested stands, because such knowledge would aid conservation strategies that may include silvicultural prescriptions for creating future habitat. We compared habitat characteristics at 62 nest sites with those at 62 random sites within the same forest stands. Compared with random sites, spotted owl nest sites had canopies of dominant and/or codominant and intermediate trees that were farther aboveground (P = 0.02 and 0.07, respectively), more 35-60-cm-dbh (diam at breast height) Douglas-fir (Pseudotsuga menziesii) trees (P = 0.03), greater basal area of Douglas-fir trees (P = 0.02), more 61-84-cm-dbh ponderosa pine (Pinus ponderosa) trees (P = 0.03), greater live tree basal area (P = 0.09), greater basal area of Class IV snags (broken snags with no branches and little bark; P < 0.001), less basal area of a group of relatively uncommon conifer species (P = 0.02), fewer 10-34-cm-dbh uncommon conifer species (P = 0.08), and less basal area of Class I and II snags (intact or nearly intact snags with branches and most bark remaining; P = 0.08 and 0.095, respectively). Volume of coarse woody debris (P > 0.13 in all decay classes) and percent canopy closure (P = 0.45) did not differ between nest and random sites. Data support the hypothesis that nest sites are selected as part of an antipredator strategy.
We used data from Northern Spotted Owl (Strix occidentalis caurina) territories to model the effects of habitat (particularly intermediate-aged forest stand types), climate, and nonhabitat covariates (i.e., age, sex) on owl reproductive rate and apparent survival in southwestern Oregon. Our best model for reproductive rate included an interaction between a cyclic, annual time trend and male breeding experience, with higher reproductive rates in even years compared to odd, particularly for males with previous breeding experience. Reproductive rate was also negatively related to the amount of winter precipitation and positively related to the proportion of old-growth forest near the owl territory center. Apparent survival was not associated with age, sex, climate or any of the intermediate-aged forest types, but was positively associated with the proportion of older forest near the territory center in a pseudothreshold pattern. The quadratic structure of the proportion of nonhabitat farther from the nest or primary roost site was also part of our best survival model. Survival decreased dramatically when the amount of nonhabitat exceeded similar to 50%. Habitat fitness potential estimates (lambda(h)) for 97 owl territories ranged from 0.29-1.09, with a mean of 0.86 +/- 0.02. Owl territories with habitat fitness potentials < 1.0 were generally characterized by < 40%-50% old forest habitat near the territory center. Our results indicate that both apparent survival and reproductive rate are positively associated with older forest types close to the nest or primary roost site. We found no support for either a positive or negative direct effect of intermediate-aged forests on either survival or reproductive rate.
The Northern Spotted Owl (Strix occidentalis caurina) population is declining throughout its range in the United States and adjacent Canada and is facing increasing pressure from the invading Barred Owl (Strix varia). In this study, we characterize Spotted Owl habitat associations and develop 2 new habitat selection models for the eastern Washington Cascade Range. Topographic and habitat data were compiled at 2 scales (0.25 and 1.0 mi) around 224 Spotted Owl activity centers, or sites, and at 160 random locations in the same geographic region, and used to develop models for predicting owl distributions. Univariate analysis found that owl sites occurred below 5000-ft elevation and were more likely to occur as area in the >71% crown-cover class increased. Owl sites were found to be more likely to occur closer to streams and to be rare in the Subalpine Fir (Abies lasiocarpa) vegetation type. The 9–25″ tree size-class was a significant predictor of the distribution of owl sites. Habitat models were constructed that were moderately successful at predicting owl-site distribution. Models from the largest scale tested (1.0-mi radius) were the most predictive, at 80% accuracy. Top-ranked models included overstory canopy cover, tree size, elevation, precipitation, distance to stream, and tree species as predictors. The resulting models can be used to help identify likely sites for surveys and to inform conservation and landscape management activities associated with forest-health restoration.
Northern spotted owls (Strix occidentalis caurina) are known to be associated with late-successional forests in the Pacific Northwest of the United States, but the effects of habitat on their demographic performance are relatively unknown. We developed statistical models relating owl survival and productivity to forest cover types within the Roseburg Study Area in the Oregon Coast Range of Oregon, USA. We further combined these demographic parameters using a Leslie-type matrix to obtain an estimate of habitat fitness potential for each owl territory (n = 94). We used mark–recapture methods to develop models for survival and linear mixed models for productivity. We measured forest composition and landscape patterns at 3 landscape scales centered on nest and activity sites within owl territories using an aerial photo-based map and a Geographic Information System (GIS). We also considered additional covariates such as age, sex, and presence of barred owls (Strix varia), and seasonal climate variables (temperature and precipitation) in our models. We used Akaike's Information Criterion (AIC) to rank and compare models. Survival had a quadratic relationship with the amount of late- and mid-seral forests within 1,500 m of nesting centers. Survival also was influenced by the amount of precipitation during the nesting season. Only 16% of the variability in survival was accounted for by our best model, but 85% of this was due to the habitat variable. Reproductive rates fluctuated biennially and were positively related to the amount of edge between late- and mid-seral forests and other habitat classes. Reproductive rates also were influenced by parent age, amount of precipitation during nesting season, and presence of barred owls. Our best model accounted for 84% of the variability in productivity, but only 3% of that was due to the habitat variable. Estimates of habitat fitness potential (which may range from 0 to infinity) for the 94 territories ranged from 0.74 to 1.15 (x̄ = 1.05, SE = 0.07). All but 1 territory had 95% confidence intervals overlapping 1.0, indicating a potentially stable population based on habitat pattern. Our results seem to indicate that while mid- and late-seral forests are important to owls, a mixture of these forest types with younger forest and nonforest may be best for owl survival and reproduction. Our results are consistent with those of researchers in northern California, USA, who used similar methods in their analyses. However, we believe that given the low variability in survival and productivity attributed to habitat, further study is needed to confirm our conclusions before they can be used to guide forest management actions for spotted owls.
Determine the type of habitat that owls like.
ABSTRACT Weighted distributions can be used to fit various forms of resource selection probability functions (RSPF) under the use-versus-available study design (Lele and Keim 2006). Although valid, the numerical maximization procedure used by Lele and Keim (2006) is unstable because of the inherent roughness of the Monte Carlo likelihood function. We used a combination of the methods of partial likelihood and data cloning to obtain maximum likelihood estimators of the RSPF in a numerically stable fashion. We demonstrated the methodology using simulated data sets generated under the log—log RSPF model and a reanalysis of telemetry data presented in Lele and Keim (2006) using the logistic RSPF model. The new method for estimation of RSPF can be used to understand differential selection of resources by animals, an essential component of studies in conservation biology, wildlife management, and applied ecology.
Conservation planning for the federally threatened northern spotted owl (Strix occidentalis caurina) requires an ability to predict their responses to existing and future habitat conditions. To inform such planning we modeled habitat selection by northern spotted owls based upon fine-scale (approx. 1.0 ha) characteristics within stands comprised primarily of mixed-aged, mixed coniferous forests of southwestern Oregon and north-central California. We sampled nocturnal (i.e., primarily foraging) habitat use by 71 radio-tagged spotted owls over 5 yr in 3 study areas and sampled vegetative and physical environmental conditions at inventory plots within 95% utilization distributions of each bird. We compared conditions at available forest patches, represented by the inventory plots, with those at patches used by owls using discrete-choice regressions, the coefficients from which were used to construct exponential resource selection functions (RSFs) for each study area and for all 3 areas combined. Cross-validation testing indicated that the combined RSF was reasonably robust to local variation in habitat availability. The relative probability that a fine-scale patch was selected decreased nonlinearly with distances from nests and streams; varied unimodally with increasing average diameter of coniferous trees and also with increasing basal area of Douglas-fir (Pseudotsuga menziesii) trees; increased linearly with increasing basal areas of sugar pine (Pinus lambertiana) and hardwood trees and with increasing density of understory shrubs. Large-diameter trees (>66 cm) appeared important <400 m from nest sites. The RSF can support comparative risk assessments of the short- versus long-term effects of silvicultural alternatives designed to integrate forest ecosystem restoration and habitat improvement for northern spotted owls. Results suggest fine-scale factors may influence population fitness among spotted owls. © 2011 The Wildlife Society.
1] The aim of this paper is to foster the development of an end-to-end uncertainty analysis framework that can quantify satellite-based precipitation estimation error characteristics and to assess the influence of the error propagation into hydrological simulation. First, the error associated with the satellite-based precipitation estimates is assumed as a nonlinear function of rainfall space-time integration scale, rain intensity, and sampling frequency. Parameters of this function are determined by using high-resolution satellite-based precipitation estimates and gauge-corrected radar rainfall data over the southwestern United States. Parameter sensitivity analysis at 16 selected 5° Â 5° latitude-longitude grids shows about 12–16% of variance of each parameter with respect to its mean value. Afterward, the influence of precipitation estimation error on the uncertainty of hydrological response is further examined with Monte Carlo simulation. By this approach, 100 ensemble members of precipitation data are generated, as forcing input to a conceptual rainfall-runoff hydrologic model, and the resulting uncertainty in the streamflow prediction is quantified. Case studies are demonstrated over the Leaf River basin in Mississippi. Compared with conventional procedure, i.e., precipitation estimation error as fixed ratio of rain rates, the proposed framework provides more realistic quantification of precipitation estimation error and offers improved uncertainty assessment of the error propagation into hydrologic simulation. Further study shows that the radar rainfall-generated streamflow sequences are consistently contained by the uncertainty bound of satellite rainfall generated streamflow at the 95% confidence interval.
Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed page of such transmission. The JSTOR Archive is a trusted digital repository providing for long-term preservation and access to leading academic journals and scholarly literature from around the world. The Archive is supported by libraries, scholarly societies, publishers, and foundations. It is an initiative of JSTOR, a not-for-profit organization with a mission to help the scholarly community take advantage of advances in technology. For more information regarding JSTOR, please contact support@jstor.org.
Accurate modeling of geographic distributions of species is crucial to various applications in ecology and conservation. The best performing techniques often require some parameter tuning, which may be prohibitively time-consuming to do separately for each species, or unreliable for small or biased datasets. Additionally, even with the abundance of good quality data, users interested in the application of species models need not have the statistical knowledge required for detailed tuning. In such cases, it is desirable to use ‘‘default settings’’, tuned and validated on diverse datasets. Maxent is a recently introduced modeling technique, achieving high predictive accuracy and enjoying several additional attractive properties. The performance of Maxent is influenced by a moderate number of parameters. The first contribution of this paper is the empirical tuning of these parameters. Since many datasets lack information about species absence, we present a tuning method that uses presence-only data. We evaluate our method on independently collected high-quality presenceabsence data. In addition to tuning, we introduce several concepts that improve the predictive accuracy and running time of Maxent. We introduce ‘‘hinge features’ ’ that model more complex relationships in the training data; we describe a new logistic output format that gives an estimate of probability of presence; finally we explore ‘‘background sampling’’ strategies that cope with sample selection bias and decrease model-building time. Our evaluation, based on a diverse dataset of 226 species from 6 regions, shows: 1) default settings tuned on presence-only data achieve performance which is almost as good as if they had been tuned on the evaluation data itself; 2) hinge features substantially improve model
1. Understanding the factors affecting species occurrence is a pre‐eminent focus of applied ecological research. However, direct information about species occurrence is lacking for many species. Instead, researchers sometimes have to rely on so‐called presence‐only data (i.e. when no direct information about absences is available), which often results from opportunistic, unstructured sampling. maxent is a widely used software program designed to model and map species distribution using presence‐only data.
2. We provide a critical review of maxent as applied to species distribution modelling and discuss how it can lead to inferential errors. A chief concern is that maxent produces a number of poorly defined indices that are not directly related to the actual parameter of interest – the probability of occurrence ( ψ ). This focus on an index was motivated by the belief that it is not possible to estimate ψ from presence‐only data; however, we demonstrate that ψ is identifiable using conventional likelihood methods under the assumptions of random sampling and constant probability of species detection.
3. The model is implemented in a convenient r package which we use to apply the model to simulated data and data from the North American Breeding Bird Survey. We demonstrate that maxent produces extreme under‐predictions when compared to estimates produced by logistic regression which uses the full (presence/absence) data set. We note that maxent predictions are extremely sensitive to specification of the background prevalence, which is not objectively estimated using the maxent method.
4. As with maxent , formal model‐based inference requires a random sample of presence locations. Many presence‐only data sets, such as those based on museum records and herbarium collections, may not satisfy this assumption. However, when sampling is random, we believe that inference should be based on formal methods that facilitate inference about interpretable ecological quantities instead of vaguely defined indices.
The case of the Northern spotted owl (Strix occidentalis caurina) has now become a classic case study in conservation biology, characterized by a harsh social battle but also by the quantity
and quality of the research performed. Based on this example, I study the way the research-management interface was organized.
The main lessons I have learned were: 1. laws that involve science in management are crucial but should be more precise; 2.
scientific ad-hoc groups are useful reviewers of management plans and interpreters of the best scientific data available,
even if more transparent scientific argumentation is needed on some points; 3. in such applied cases, even science that has
not been strongly integrated with management can produce results that are useful for management; 4. stronger links between
science and management appear necessary, but difficult to implement. This last point makes me wonder whether environmental
laws should not more frequently target the incorporation of science into the management process itself rather than “only”
basing management on the best scientific data available. On a more ecological level, perhaps the habitat issue has been underrated
during the last few years compared to other emerging threats such as the invasion of the spotted owl range by barred owls.
This article presents the results of a simulation study of variable selection in a multiple regression context that evaluates the frequency of selecting noise variables and the bias of the adjusted R2 of the selected variables when some of the candidate variables are authentic. It is demonstrated that for most samples a large percentage of the selected variables is noise, particularly when the number of candidate variables is large relative to the number of observations. The adjusted R2 of the selected variables is highly inflated.
We introduce and test LandTrendr (Landsat-based detection of Trends in Disturbance and Recovery), a new approach to extract spectral trajectories of land surface change from yearly Landsat time-series stacks (LTS). The method brings together two themes in time-series analysis of LTS: capture of short-duration events and smoothing of long-term trends. Our strategy is founded on the recognition that change is not simply a contrast between conditions at two points in time, but rather a continual process operating at both fast and slow rates on landscapes. This concept requires both new algorithms to extract change and new interpretation tools to validate those algorithms. The challenge is to resolve salient features of the time series while eliminating noise introduced by ephemeral changes in illumination, phenology, atmospheric condition, and geometric registration. In the LandTrendr approach, we use relative radiometric normalization and simple cloud screening rules to create on-the-fly mosaics of multiple images per year, and extract temporal trajectories of spectral data on a pixel-by-pixel basis. We then apply temporal segmentation strategies with both regression-based and point-to-point fitting of spectral indices as a function of time, allowing capture of both slowly-evolving processes, such as regrowth, and abrupt events, such as forest harvest. Because any temporal trajectory pattern is allowable, we use control parameters and threshold-based filtering to reduce the role of false positive detections. No suitable reference data are available to assess the role of these control parameters or to test overall algorithm performance. Therefore, we also developed a companion interpretation approach founded on the same conceptual framework of capturing both long and short-duration processes, and developed a software tool to apply this concept to expert interpretation and segmentation of spectral trajectories (TimeSync, described in a companion paper by Cohen et al., 2010). These data were used as a truth set against which to evaluate the behavior of the LandTrendr algorithms applied to three spectral indices. We applied the LandTrendr algorithms to several hundred points across western Oregon and Washington (U.S.A.). Because of the diversity of potential outputs from the LTS data, we evaluated algorithm performance against summary metrics for disturbance, recovery, and stability, both for capture of events and longer-duration processes. Despite the apparent complexity of parameters, our results suggest a simple grouping of parameters along a single axis that balances the detection of abrupt events with capture of long-duration trends. Overall algorithm performance was good, capturing a wide range of disturbance and recovery phenomena, even when evaluated against a truth set that contained new targets (recovery and stability) with much subtler thresholds of change than available from prior validation datasets. Temporal segmentation of the archive appears to be a feasible and robust means of increasing information extraction from the Landsat archive.
Broad-scale maps of forest characteristics are needed throughout the United States for a wide variety of forest land management applications. Inexpensive maps can be produced by modelling forest class and structure variables collected in nationwide forest inventories as functions of satellite-based information. But little work has been directed at comparing modelling techniques to determine which tools are best suited to mapping tasks given multiple objectives and logistical constraints. Consequently, five modelling techniques were compared for mapping forest characteristics in the Interior Western United States. The modelling techniques included linear models (LMs), generalized additive models (GAMs), classification and regression trees (CARTs), multivariate adaptive regression splines (MARS), and artificial neural networks (ANNs). Models were built for two discrete and four continuous forest response variables using a variety of satellite-based predictor variables within each of five ecologically different regions. All techniques proved themselves workable in an automated environment. When their potential mapping ability was explored through simulations, tremendous advantages were seen in use of MARS and ANN for prediction over LMs, GAMs, and CART. However, much smaller differences were seen when using real data. In some instances, a simple linear approach worked virtually as well as the more complex models, while small gains were seen using more complex models in other instances. In real data runs, MARS and GAMS performed (marginally) best for prediction of forest characteristics.
Habitat requirements of northern spotted owls Strix occidentalis caurina have become the focus of a major controversy over how much old-growth forest in the western United States should be preserved. Analysis of three large data sets showed that the subspecies was rare or absent in areas with little older (i.e. >80-year-old) forest but with extensive stands nearing harvest age. The owls were also rare in areas with the small amounts of old-growth typically left after harvest operations. Old-growth stands in Wilderness Areas supported sparse populations of northern spotted owls, and their reproductive success was only about half that of owls outside Wilderness Areas. The results indicate that timber harvest operations, as currently practiced, lead to declines in numbers of northern spotted owls, and that currently protected old-growth stands are unlikely to provide enough high-quality habitat to maintain self-supporting populations of northern spotted owls.