Wildlife Responses to Recreational Disturbance in Mountain Habitats

Knowledge Graph (51 nodes, 235 connections)

Research Primer

Background

Outdoor recreation in mountain landscapes — hiking, mountain biking, dog walking, and motorized travel — is one of the fastest-growing pressures on wildlife in places like the Gunnison Basin. The trails surrounding the Rocky Mountain Biological Laboratory (RMBL) draw thousands of visitors each summer, generating revenue for local communities while passing through habitats used by deer, foxes, coyotes, ground squirrels, marmots, chipmunks, and mice. Recreational trail impacts refer to the ways these human activities alter wildlife behavior, distribution, and physiology, even when no animals are directly harmed. Understanding these impacts matters because subtle changes — an animal feeding less, shifting active hours, or avoiding a meadow — can cascade through food webs and shape which species persist near popular trails.

Researchers studying these questions rely heavily on non-invasive sampling methods that record wildlife without capturing or handling animals. Two approaches dominate the work summarized here. Camera trapping uses motion-triggered cameras placed along trails or inside baited buckets to photograph animals as they pass; the resulting images can be tallied to estimate small mammal abundance (an index of how many animals are using an area) and to test for changes in activity timing. Hair tube traps, a complementary non-invasive method, collect hair samples that animals leave behind as they enter a tube. Acoustic localization, a third major approach, uses networks of microphones — sensor networks deployed across a meadow — to detect animal calls (such as marmot alarm whistles) and pinpoint where the caller is located. When multiple microphones share information, an approach called collaborative sensing, the system can overcome spatial aliasing (an ambiguity in locating high-pitched sounds when microphones are spaced too far apart) and track individual animals across the landscape.

Two other ideas recur throughout this work. A distance gradient is a sampling design that places monitoring stations at progressively greater distances from a trail or road, allowing researchers to ask how far recreational effects extend into the surrounding habitat. Detection probability acknowledges that wildlife surveys rarely detect every animal present; statistical models separate true absence from simple non-detection so that comparisons across trails or seasons are fair. Together, these tools and concepts let researchers ask not just whether wildlife are present near trails, but how their behavior, community composition, and use of habitat shift in response to people.

Foundational work

The technological foundation for acoustic monitoring at RMBL was built in a sequence of engineering papers in the late 2000s. (Ali et al., 2007) demonstrated that an Approximated Maximum Likelihood algorithm could localize yellow-bellied marmot alarm calls within a few meters in noisy field conditions, and showed that distributing microphones across multiple sub-arrays solved the spatial aliasing problem that limits single large arrays. (Allen et al., 2008) introduced VoxNet, a weather-resistant, rapidly deployable platform that detected and localized marmot calls in RMBL meadows with a 99.3% success rate. (Ali et al., 2009) later confirmed that the platform could self-localize its own microphones to within 5 cm in open terrain, a prerequisite for accurate animal tracking. These advances were synthesized in a widely cited review by (Blumstein et al., 2011), which laid out how microphone arrays enable remote, non-invasive monitoring of acoustically active animals across spatial and temporal scales relevant to behavior and conservation.

A parallel foundation for small mammal work emerged from validation studies of non-invasive sampling. (Sandoval, 2018) compared hair tube traps to traditional Longworth live traps and found that hair-tube abundance was strongly correlated with mark-recapture estimates (R² = 0.965), establishing that hair tubes could replace live trapping for abundance work — though species identification from hair alone proved difficult.

Key findings

Findings on how recreation affects mammals at RMBL are nuanced and species-specific. Across multiple summers of camera-trap work along trails of varying use intensity, several student-led studies found that overall small mammal species richness and abundance did not decline with heavy human use. (Bermudez et al., 2020) reported no significant negative impacts of high trail use on rodent richness or abundance, and even detected a marginally positive relationship between small mammal activity and human activity, with deer mice (Peromyscus maniculatus) increasing with pedestrian traffic but declining slightly with biker traffic. (Macias, 2019) similarly found that diurnal ground squirrels and chipmunks showed no negative trail effect, and (Rodriguez, 2021) confirmed that hiking, biking, and dogs had no significant negative impact on rodents or predators overall.

Beneath these aggregate patterns, however, finer-grained analyses reveal behavioral shifts. (Novoa, 2021) used a machine-learning image classifier to separate diurnal rodents and showed that, while total presence was unchanged, the daytime activity of chipmunks and ground squirrels near trails decreased significantly as pedestrian and biker traffic rose — suggesting animals were still using the area but adjusting when they moved. Larger mammals show even clearer differential responses. (Uetrecht et al., 2023) used camera traps along a gradient of recreation intensity and found that mule deer were more likely to use areas with more vehicles and bikers and farther from trailheads, while coyotes tracked their prey with no detectable response to recreation, and red foxes responded mainly to vegetation cover. The picture that emerges is that human use of trails reshuffles which species are active, when, and where, rather than simply emptying the landscape of wildlife.

This work also confirmed the practical value of the methodological foundations. Bucket camera traps captured both nocturnal and diurnal species that live trapping missed (Rodriguez, 2021), and hair tubes detected biologically meaningful disturbances such as the negative effect of avalanches on small mammal abundance at three of four sites studied (Sandoval, 2019). Companion work using geographic information systems showed that distance from cover and meadow size continue to structure where chipmunks and ground squirrels coexist (Cohen, 2023), providing a habitat-template baseline against which recreational effects can be compared.

Current frontier

Early work in the 2000s focused on building the sensor and algorithm infrastructure needed to monitor wildlife remotely ((Ali et al., 2007); (Trifa et al., 2007)). Studies since 2019 have shifted decisively toward applying that infrastructure — together with camera traps and hair tubes — to recreation questions in the Upper East River Valley. The most recent publications combine multiple methods and emphasize species-specific responses along disturbance gradients, exemplified by the multi-species camera-trap design of (Uetrecht et al., 2023) and the machine-learning classification pipeline of (Novoa, 2021). The COVID-era surge in trail use provided an unplanned natural experiment, and several of the post-2020 studies leverage that intensity to test effects that would have been hard to detect in quieter years.

Research is heading toward integrating automated image and sound classification with formal occupancy models that account for detection probability, allowing weaker but ecologically important effects to be teased apart from background variation. The combination of acoustic localization for vocal species like marmots, camera traps for visual species, and hair tubes for cryptic small mammals positions RMBL to build a multi-sensor picture of how the entire mammal community responds as recreation grows.

Open questions

Several important questions remain. How do behavioral shifts — such as the reduced daytime activity of chipmunks and ground squirrels near busy trails — translate into long-term consequences for survival, reproduction, and population persistence? At what distance from a trail do effects fade, and does that distance depend on the type of recreation (quiet hikers versus mountain bikers versus dogs)? Why do some species (mule deer) appear to tolerate or even associate with human activity while others avoid it, and what role do predators play in mediating these patterns? Finally, can the acoustic and camera networks already deployed at RMBL be linked into a continuous, basin-wide monitoring system that detects change before it becomes irreversible? Answering these questions over the next decade will require longer time series, coordinated multi-species designs, and tighter integration between the engineering and ecological communities that have grown up around RMBL.

References

Ali, A. M., et al. (2007). Acoustic source localization using the acoustic ENSBox. →

Ali, A. M., et al. (2009). An empirical study of collaborative acoustic source localization. Journal of Signal Processing Systems. →

Ali, A. M., Yao, K., Collier, T. C., Taylor, C. E., Blumstein, D. T., & Girod, L. (2007). An empirical study of acoustic source localization. →

Allen, M., Girod, L., Newton, R., Madden, S., Blumstein, D. T., & Estrin, D. (2008). VoxNet: an interactive, rapidly-deployable acoustic monitoring platform. ISPN 2008: Information Processing in Sensor Networks. →

Bermudez, et al. (2020). The effects of recreational trail use on small mammal species richness and abundance. →

Blumstein, D. T., et al. (2011). Acoustic monitoring in terrestrial environments using microphone arrays: applications, technological considerations and prospectus. Journal of Applied Ecology. →

Cohen (2023). Using GIS techniques to test a model of the coexistence of the golden-mantled ground squirrel and the least chipmunk. →

Macias (2019). Effects of recreational trails on small mammal communities in north-central Gunnison County, Colorado. →

Novoa, D. (2021). Development of Keras image classification model for use with a study on the effects of recreational trail use on small mammal species richness and activity. →

Rodriguez (2021). The influence of human recreational trail use has on rodent and predator activity using motion triggered cameras. →

Sandoval, G. (2018). Comparing Longworth live traps to hair tubes for describing small mammal communities. →

Sandoval, G. (2019). Effects of avalanches on rodent populations. →

Trifa, V., et al. (2007). Automated wildlife monitoring using self-configuring sensor networks deployed in natural habitats. Proceedings of the Twelfth International Symposium on Artificial Life and Robotics. →

Uetrecht, J., et al. (2023). Differential response of three large mammal species to human recreation in the Rocky Mountains of Colorado, USA. Frontiers in Conservation Science. →

Concept (11) →

Protocol (7) →

Author (12) →

Publication (14) →