We investigated the demography of a common aquatic turtle (Chrysemys picta) along a gradient of urbanization in southeastern New Hampshire. Using a geographic information system and live trapping of turtles, we compared the proportion of males, the proportion of adults, and the relative abundance of turtles in 37 ponds. We used satellite images, aerial photographs, and field visits to describe pond‐specific features and habitat composition up to 2000 m from the perimeter of each pond. The proportion of males was positively associated with the percentage of forest cover within 500 m, greater road density within 100 m, and an index of predator activity at a pond. The proportion of adults in a population was associated with road density within 250 m of the pond and the percentage of the pond perimeter that was forested. Abnormal population structures associated with greater road densities did not necessarily result in lower abundances of turtles in ponds. Turtle abundance increased as the distance to neighboring wetlands decreased and the amount of nesting habitat near pond edges increased. Pond‐specific features also affected turtle abundance where populations were larger in ponds with organic substrates and abundant coverage by shoreline vegetation than in ponds lacking these features. Few turtles were encountered in ponds with an abundance of herbaceous emergent vegetation, and fewer turtles were captured during a summer with abundant precipitation. Suburban and urban developments, with dense road networks and abundant populations of generalist predators (especially the raccoon, Procyon lotor), can alter the structure of aquatic turtle populations. Although these alterations may not result in immediate changes in turtle abundance within a specific population, we suspect that the reduction in recruitment caused by habitat alterations will eventually reduce or eliminate local populations. Even though there are life‐history differences among species of turtles, our results may provide new insight into the causes of recent declines of other turtles.

No abstract

Unlike other regions of North America, forested habitats in New England bave increased substantially in the past 100 years. The proportion of land in New Hampshire covered by forests was 47% in 1880 and 87% in 1980. This increase was largely the result of a region‐wide abandonment of farms and the subsequent colonization of these lands by second‐growth forests. I examined the sequence of farm abandonment, forest colonization and forest maturation that occurred in New Hampshire in relation to changes in the abundance and distribution of a group of forest mammals and birds that have undergone substantial declines. A modeled pattern of secondary succession resulted in the availability of approximately 195,000 ha of early seral habitats (10–25 years after abandonment) from 1905 to 1940. These habitats then mutured into closed‐canopy forests by about 1960. Concurrent to the loss of early successional habitats, populations of New England cottontails (Sylvilagus transitionalis) decreased from an apparent continuous distribution throughout 60% of New Hampshire to fragmented populations that occupy less than 20% of the state Bobacts (Felis rufus) responded functionally (S. transitionalis in diet: 1951–1954 = 43%, 1961–1964 = 10%) and numerically (mean annual harvest of bobcats: 1951–1954 = 350, 1965–1969 = 36) to changes in cottontail abundance. Eighteen of 26 species of migratory passerines that nest in the forests of northern New England also declined during the period their populations were monitored (1934–1987). Eight (44%) of the species that declined are associated with early successional habitats, and these species consistently exhibited population declines during the 1950s. The reduction of early successional species may be extended in space and time by current land uses that fragment and isolate patches of habitat. Ownership patterns of forest lands in New England (excluding Maine) reveal 88% private ownership with an average holding of 10 ha. This suggests that large tracts of early successional habitats will be restricted to industrial and state/national forests. Although even‐aged management of a portion of these forests may be prerceived as incompatible with area‐sensitive and interior species, clustering of clearcuts and maintaining large tracts of mature habitats could sustain diverse populations of forest vertebrates.

Increased predation has been suggested as a proximate factor causing the decline of vertebrate diversity in many human-altered landscapes. Previous studies on this topic have provided conflicting results, perhaps as a consequence of the limited spatial scale used in these investigations. We incorporated a multiscaled approach (using site, plot (1.44 km2), and landscape (54 km2)) to investigate the distribution of activity of medium-sized carnivores relative to habitat edges and the numeric responses of these predators to habitat diversity. Among the taxa surveyed, raccoons (Procyon lotor) did not show an affinity for habitat edges at any spatial scale. However, raccoons were more abundant in landscapes characterized by a diversity of cover types. Free-ranging domestic dogs (Canis familiaris) and cats (Felis domesticus) did not respond to the proximity of habitat edges in summer but showed a strong affinity for edge habitats (especially those associated with human dwellings) during winter. Wild canids (Vulpes vulpes and Canis latrans) also selected sites in close proximity to edges in winter and were more abundant in diverse landscapes. Although human-dominated habitats (agricultural areas, grass–brushland, and developed sites) represented only 7–27% of the three landscapes studied, populations of generalist predators (raccoons and wild canids) increased as landscapes became more diverse. As a result, even moderate levels of habitat fragmentation may elevate predation rates and subsequently alter the composition of prey communities.

ABSTRACT/Predictive models of wildlife-habitat relationships often have been developed without being tested The apparent classification accuracy of such models can be optimistically biased and misleading. Data rasampling methods exist that yield a more realistic estimate of model classification accuracy These methods are simple and require no new sample data. We illustrate these methods (cross-validation. jackknife resampling, and bootstrap resampling) with computer simulation to demonstrate the increase in precision of the estimate. The bootstrap method is then applied to field data as a technique for model comparison We recommend that biologists use some resampling procedure to evaluate wildlife habitat models prior to field evaluation.The increased involvement of wildlife biologists in habitat inventory, impact assessment, and land-use planning has generated a need for accurate models of wildlife-habitat relationships (Berry 1986). As a result, several modeling techniques have been developed, including the US Fish and Wildlife Service's habitat suitability index models (Fish and Wildlife Service 1981) and more rigorous statistical models (Capen 1981).During the development of a statistical model, a biologist measures many variables that are potentially important to the target species and then applies some procedure (e.g., discriminant analysis or logistic regression) to build a model that predicts the presence or absence of the target species. This method of model construction has been widely applied by wildlife biologists (Verner and others 1986). However, the problems of misleading statistical models also have been recognized (Verbyla 1986, Rexstad and others 1988). For example, if many predictor variables are measured and utilized in the model, spurious sample relationships may occur, especially if the sample size is small (e.g., Magnusson 1983, Verbyla 1986. Because of this, multivariate statistical models may predict well when applied to the data that were used in developing the model, but predict poorly if they are tested with an independent data set. This does not mean that wildlife biologists should abandon multivariate statistics. However, these models should be used cautiously and tested thoroughly.The most rigorous test of a statistical model is to In some instances, a resampling procedure may be the only practical method available to evaluate a model, such as in exploring species-habitat relationships during unusual climate conditions. The objectives of this article are to: (1) describe several resampling methods that can be applied to habitat models, (2) use a computer simulation to illustrate how these methods can be used to evaluate classification accuracy, and (3) demonstrate the application of these methods to field data. Resampling MethodsSuppose a wildlife biologist wants to predict the presence or absence of a rare species from habitat measurements. Access to areas where the species occurs is difficult, and therefore only 30 sample sites are established within the study area. Ten habitat v...

No abstract

Among the most conspicuous environmental effects of roads are vehicle-related mortalities of wildlife. Research to understand the factors that contribute to wildlife-vehicle collisions can be partitioned into several major themes, including (i) characteristics associated with roadkill hot spots, (ii) identification of road-density thresholds that limit wildlife populations, and (iii) species-specific models of vehicle collision rates that incorporate information on roads (e.g., proximity, width, and traffic volume) and animal movements. We suggest that collision models offer substantial opportunities to understand the effects of roads on a diverse suite of species. We conducted simulations using collision models and information on Blanding's turtles (Emydoidea blandingii), bobcats (Lynx rufus), and moose (Alces alces), species endemic to the northeastern United States that are of particular concern relative to collisions with vehicles. Results revealed important species-specific differences, with traffic volume and rate of movement by candidate species having the greatest influence on collision rates. We recommend that future efforts to reduce wildlife-vehicle collisions be more proactive and suggest the following protocol. For species that pose hazards to drivers (e.g., ungulates), identify collision hot spots and implement suitable mitigation to redirect animal movements (e.g., underpasses, fencing, and habitat modification), reduce populations of problematic game species via hunting, or modify driver behavior (e.g., dynamic signage that warns drivers when animals are near roads). Next, identify those species that are likely to experience additive (as opposed to compensatory) mortality from vehicle collisions and rank them according to vulnerability to extirpation. Then combine information on the distribution of at-risk species with information on existing road networks to identify areas where immediate actions are warranted.

No abstract