1. Camera traps deployed in grids or stratified random designs are a well-established survey tool for wildlife but there has been little evaluation of study design parameters.2. We used an empirical subsampling approach involving 2,225 camera deployments run at 41 study areas around the world to evaluate three aspects of camera trap study design (number of sites, duration and season of sampling) and their influence on the estimation of three ecological metrics (species richness, occupancy and detection rate) for mammals.3. We found that 25-35 camera sites were needed for precise estimates of species richness, depending on scale of the study. The precision of species-level estimates of occupancy (ψ) was highly sensitive to occupancy level, with <20 camera sites needed for precise estimates of common (ψ > 0.75) species, but more than 150 camera sites likely needed for rare (ψ < 0.25) species. Species detection rates were more difficult to estimate precisely at the grid level due to spatial heterogeneity, | 701Methods in Ecology and Evoluঞon KAYS et Al.

Advances in species distribution modeling continue to be driven by a need to predict species responses to environmental change coupled with increasing data availability. Recent work has focused on development of methods that integrate multiple streams of data to model species distributions. Combining sources of information increases spatial coverage and can improve accuracy in estimates of species distributions. However, when fusing multiple streams of data, the temporal and spatial resolutions of data sources may be mismatched. This occurs when data sources have fluctuating geographic coverage, varying spatial scales and resolutions, and differing sources of bias and sparsity. It is well documented in the spatial statistics literature that ignoring the misalignment of different data sources will result in bias in both the point estimates and uncertainty. This will ultimately lead to inaccurate predictions of species distributions. Here, we examine the issue of misaligned data as it relates specifically to integrated species distribution models. We then provide a general solution that builds off work in the statistical literature for the change‐of‐support problem. Specifically, we leverage spatial correlation and repeat observations at multiple scales to make statistically valid predictions at the ecologically relevant scale of inference. An added feature of the approach is that addressing differences in spatial resolution between data sets can allow for the evaluation and calibration of lesser‐quality sources in many instances. Using both simulations and data examples, we highlight the utility of this modeling approach and the consequences of not reconciling misaligned spatial data. We conclude with a brief discussion of the upcoming challenges and obstacles for species distribution modeling via data fusion.

The use of camera traps as a tool for studying wildlife populations is commonplace. However, few have considered how the number of detections of wildlife differ depending upon the number of camera traps placed at cameras-sites, and how this impacts estimates of occupancy and community composition. During December 2015–February 2016, we deployed four camera traps per camera-site, separated into treatment groups of one, two, and four camera traps, in southern Illinois to compare whether estimates of wildlife community metrics and occupancy probabilities differed among survey methods. The overall number of species detected per camera-site was greatest with the four-camera survey method (P<0.0184). The four-camera survey method detected 1.25 additional species per camera-site than the one-camera survey method, and was the only survey method to completely detect the ground-dwelling silvicolous community. The four-camera survey method recorded individual species at 3.57 additional camera-sites (P = 0.003) and nearly doubled the number of camera-sites where white-tailed deer (Odocoileus virginianus) were detected compared to one- and two-camera survey methods. We also compared occupancy rates estimated by survey methods; as the number of cameras deployed per camera-site increased, occupancy estimates were closer to naïve estimates, detection probabilities increased, and standard errors of detection probabilities decreased. Additionally, each survey method resulted in differing top-ranked, species-specific occupancy models when habitat covariates were included. Underestimates of occurrence and misrepresented community metrics can have significant impacts on species of conservation concern, particularly in areas where habitat manipulation is likely. Having multiple camera traps per site revealed significant shortcomings with the common one-camera trap survey method. While we realize survey design is often constrained logistically, we suggest increasing effort to at least two camera traps facing opposite directions per camera-site in habitat association studies, and to utilize camera-trap arrays when restricted by equipment availability.

Citizen science projects that use sensors (such as camera traps) to collect data can collect large-scale data without compromising information quality. However, project management challenges are increased when data collection is scaled up. Here, we provide an overview of our efforts to conduct a large-scale citizen science project using camera traps-North Carolina's Candid Critters. We worked with 63 public libraries to distribute camera traps to volunteers in all 100 counties in North Carolina, USA. Candid Critters engaged 580 volunteers to deploy cameras at 4,295 locations across private and public lands, collecting 120,671 wildlife records and 2.2 million photographs. We provide eight key suggestions for overcoming challenges in study design, volunteer recruitment and management, equipment distribution, outreach, training, and data management. We found that citizen science was a successful and economical method for collecting large-scale wildlife records, and the use of sensors allowed for inspectable quality and streamlined acquisition. In three years, we collected roughly five times the number of verified mammal records than were previously available in North Carolina, and completed the work for less than the typical cost of collecting data with field assistants. The project also yielded many positive outcomes for adult and youth volunteers. Although citizen science presents many challenges, we hope that sharing our experiences will provide useful insight for those hoping to use sensors for citizen science over large scales.

Broad-scale ecological research on species distributions commonly presumes that the correlative relationships discovered are stationary over space. This is an assumption of most species distribution models (SDMs) that combine observations of species occurrence with environmental characteristics to understand current ecological correlates and to predict distributions based on those relationships. However, ecological relationships may vary spatially because of changes in the environment (i.e., resource availability) or the organism itself (i.e., local adaptation). Discovering this within-species variation typically requires dense datasets over large geographic areas, which are now being provided by the recent proliferation of open-access biodiversity occurrence records. Using nearly 4000 sampling locations from an open-access, state-wide camera-trapping project, we explore the space-varying effects of covariates on the distribution of four mammal species at two scales: region-specific and fine resolution, with the latter estimated using spatially varying coefficients (SVC) models, to understand the scale of spatial variation in ecological relationships. Among the four species tested, the ecological relationships for two were best explained with the regional models, equivocal results for one species, while the SVC model had superior fit and predictive performance for the final species (white-tailed deer, Odocoileus virginianus). Spatial nonstationarity was more common in relationships with landscape composition characteristics, such as housing density, than in landscape configuration metrics, such as patch richness density. One of the most appealing results of an SVC approach is not only the improved predictions across large landscapes but also understanding how animals are responding to the environment differently at the management unit level. For example, we found that deer's spatially varying relationship with forest cover was best explained by an interactive effect of deer management units (i.e., differences in deer populations) and predator pressure. These findings lead to a new hypothesis about how deer may be differentially using forested environments across space and could be a promising area of future research. Given sufficient data, accounting for nonstationarity in SDMs can show large-scale ecological patterns