Censoring Trace-Level Environmental Data: Statistical Analysis Considerations to Limit Bias.

Trace-level environmental data typically include values near or below detection and quantitation thresholds where health effects may result from low-concentration exposures to one chemical over time or to multiple chemicals. In a cook stove case study, bias in dibenzo[a,h]anthracene concentration means and standard deviations (SDs) was assessed following censoring at thresholds for selected analysis approaches: substituting threshold/2, maximum likelihood estimation, robust regression on order statistics, Kaplan-Meier, and omitting censored observations. Means and SDs for gas chromatography-mass spectrometry-determined concentrations were calculated after censoring at detection and calibration thresholds, 17% and 55% of the data, respectively. Threshold/2 substitution was the least biased. Measurement values were subsequently simulated from two log-normal distributions at two sample sizes. Means and SDs were calculated for 30%, 50%, and 80% censoring levels and compared to known distribution counterparts. Simulation results illustrated (1) threshold/2 substitution to be inferior to modern after-censoring statistical approaches and (2) all after-censoring approaches to be inferior to including all measurement data in analysis. Additionally, differences in stove-specific group means were tested for uncensored samples and after censoring. Group differences of means tests varied depending on censoring and distributional decisions. Investigators should guard against censoring-related bias from (explicit or implicit) distributional and analysis approach decisions.

The basis function approach for modeling autocorrelation in ecological data.

Analyzing ecological data often requires modeling the autocorrelation created by spatial and temporal processes. Many seemingly disparate statistical methods used to account for autocorrelation can be expressed as regression models that include basis functions. Basis functions also enable ecologists to modify a wide range of existing ecological models in order to account for autocorrelation, which can improve inference and predictive accuracy. Furthermore, understanding the properties of basis functions is essential for evaluating the fit of spatial or time-series models, detecting a hidden form of collinearity, and analyzing large data sets. We present important concepts and properties related to basis functions and illustrate several tools and techniques ecologists can use when modeling autocorrelation in ecological data.

Model selection and assessment for multi-species occupancy models.

While multi-species occupancy models (MSOMs) are emerging as a popular method for analyzing biodiversity data, formal checking and validation approaches for this class of models have lagged behind. Concurrent with the rise in application of MSOMs among ecologists, a quiet regime shift is occurring in Bayesian statistics where predictive model comparison approaches are experiencing a resurgence. Unlike single-species occupancy models that use integrated likelihoods, MSOMs are usually couched in a Bayesian framework and contain multiple levels. Standard model checking and selection methods are often unreliable in this setting and there is only limited guidance in the ecological literature for this class of models. We examined several different contemporary Bayesian hierarchical approaches for checking and validating MSOMs and applied these methods to a freshwater aquatic study system in Colorado, USA, to better understand the diversity and distributions of plains fishes. Our findings indicated distinct differences among model selection approaches, with cross-validation techniques performing the best in terms of prediction.

Dynamic occupancy models for explicit colonization processes.

The dynamic, multi-season occupancy model framework has become a popular tool for modeling open populations with occupancies that change over time through local colonizations and extinctions. However, few versions of the model relate these probabilities to the occupancies of neighboring sites or patches. We present a modeling framework that incorporates this information and is capable of describing a wide variety of spatiotemporal colonization and extinction processes. A key feature of the model is that it is based on a simple set of small-scale rules describing how the process evolves. The result is a dynamic process that can account for complicated large-scale features. In our model, a site is more likely to be colonized if more of its neighbors were previously occupied and if it provides more appealing environmental characteristics than its neighboring sites. Additionally, a site without occupied neighbors may also become colonized through the inclusion of a long-distance dispersal process. Although similar model specifications have been developed for epidemiological applications, ours formally accounts for detectability using the well-known occupancy modeling framework. After demonstrating the viability and potential of this new form of dynamic occupancy model in a simulation study, we use it to obtain inference for the ongoing Common Myna (Acridotheres tristis) invasion in South Africa. Our results suggest that the Common Myna continues to enlarge its distribution and its spread via short distance movement, rather than long-distance dispersal. Overall, this new modeling framework provides a powerful tool for managers examining the drivers of colonization including short- vs. long-distance dispersal, habitat quality, and distance from source populations.

Spatial segregation in eastern North Pacific skate assemblages.

Skates (Rajiformes: Rajoidei) are common mesopredators in marine benthic communities. The spatial associations of individual species and the structure of assemblages are of considerable importance for effective monitoring and management of exploited skate populations. This study investigated the spatial associations of eastern North Pacific (ENP) skates in continental shelf and upper continental slope waters of two regions: central California and the western Gulf of Alaska. Long-term survey data were analyzed using GIS/spatial analysis techniques and regression models to determine distribution (by depth, temperature, and latitude/longitude) and relative abundance of the dominant species in each region. Submersible video data were incorporated for California to facilitate habitat association analysis. We addressed three main questions: 1) Are there regions of differential importance to skates?, 2) Are ENP skate assemblages spatially segregated?, and 3) When skates co-occur, do they differ in size? Skate populations were highly clustered in both regions, on scales of 10s of kilometers; however, high-density regions (i.e., hot spots) were segregated among species. Skate densities and frequencies of occurrence were substantially lower in Alaska as compared to California. Although skates are generally found on soft sediment habitats, Raja rhina exhibited the strongest association with mixed substrates, and R. stellulata catches were greatest on rocky reefs. Size segregation was evident in regions where species overlapped substantially in geographic and depth distribution (e.g., R. rhina and Bathyraja kincaidii off California; B. aleutica and B. interrupta in the Gulf of Alaska). Spatial niche differentiation in skates appears to be more pronounced than previously reported.

Spatial occupancy models applied to atlas data show Southern Ground Hornbills strongly depend on protected areas.

Determining the range of a species and exploring species--habitat associations are central questions in ecology and can be answered by analyzing presence--absence data. Often, both the sampling of sites and the desired area of inference involve neighboring sites; thus, positive spatial autocorrelation between these sites is expected. Using survey data for the Southern Ground Hornbill (Bucorvus leadbeateri) from the Southern African Bird Atlas Project, we compared advantages and disadvantages of three increasingly complex models for species occupancy: an occupancy model that accounted for nondetection but assumed all sites were independent, and two spatial occupancy models that accounted for both nondetection and spatial autocorrelation. We modeled the spatial autocorrelation with an intrinsic conditional autoregressive (ICAR) model and with a restricted spatial regression (RSR) model. Both spatial models can readily be applied to any other gridded, presence--absence data set using a newly introduced R package. The RSR model provided the best inference and was able to capture small-scale variation that the other models did not. It showed that ground hornbills are strongly dependent on protected areas in the north of their South African range, but less so further south. The ICAR models did not capture any spatial autocorrelation in the data, and they took an order, of magnitude longer than the RSR models to run. Thus, the RSR occupancy model appears to be an attractive choice for modeling occurrences at large spatial domains, while accounting for imperfect detection and spatial autocorrelation.

Using detection dogs to conduct simultaneous surveys of northern spotted (Strix occidentalis caurina) and barred owls (Strix varia).

State and federal actions to conserve northern spotted owl (Strix occidentalis caurina) habitat are largely initiated by establishing habitat occupancy. Northern spotted owl occupancy is typically assessed by eliciting their response to simulated conspecific vocalizations. However, proximity of barred owls (Strix varia)-a significant threat to northern spotted owls-can suppress northern spotted owl responsiveness to vocalization surveys and hence their probability of detection. We developed a survey method to simultaneously detect both species that does not require vocalization. Detection dogs (Canis familiaris) located owl pellets accumulated under roost sites, within search areas selected using habitat association maps. We compared success of detection dog surveys to vocalization surveys slightly modified from the U.S. Fish and Wildlife Service's Draft 2010 Survey Protocol. Seventeen 2 km × 2 km polygons were each surveyed multiple times in an area where northern spotted owls were known to nest prior to 1997 and barred owl density was thought to be low. Mitochondrial DNA was used to confirm species from pellets detected by dogs. Spotted owl and barred owl detection probabilities were significantly higher for dog than vocalization surveys. For spotted owls, this difference increased with number of site visits. Cumulative detection probabilities of northern spotted owls were 29% after session 1, 62% after session 2, and 87% after session 3 for dog surveys, compared to 25% after session 1, increasing to 59% by session 6 for vocalization surveys. Mean detection probability for barred owls was 20.1% for dog surveys and 7.3% for vocal surveys. Results suggest that detection dog surveys can complement vocalization surveys by providing a reliable method for establishing occupancy of both northern spotted and barred owl without requiring owl vocalization. This helps meet objectives of Recovery Actions 24 and 25 of the Revised Recovery Plan for the Northern Spotted Owl.

Novel methods reveal shifts in migration phenology of barn swallows in South Africa.

Many migratory bird species, including the barn swallow (Hirundo rustica), have advanced their arrival date at Northern Hemisphere breeding grounds, showing a clear biotic response to recent climate change. Earlier arrival helps maintain their synchrony with earlier springs, but little is known about the associated changes in phenology at their non-breeding grounds. Here, we examine the phenology of barn swallows in South Africa, where a large proportion of the northern European breeding population spends its non-breeding season. Using novel analytical methods based on bird atlas data, we show that swallows first arrive in the northern parts of the country and gradually appear further south. On their north-bound journey, they leave South Africa rapidly, resulting in mean stopover durations of 140 days in the south and 180 days in the north. We found that swallows are now leaving northern parts of South Africa 8 days earlier than they did 20 years ago, and so shortened their stay in areas where they previously stayed the longest. By contrast, they did not shorten their stopover in other parts of South Africa, leading to a more synchronized departure across the country. Departure was related to environmental variability, measured through the Southern Oscillation Index. Our results suggest that these birds gain their extended breeding season in Europe partly by leaving South Africa earlier, and thus add to scarce evidence for phenology shifts in the Southern Hemisphere.