Lessons Learned from Parkwide Visitor Use Modeling

• Ashley D’Antonio, Associate Professor of Nature-based Recreation Management, Department of Forest Ecosystems and Society, Oregon State University
• Evan M. Bredeweg, Postdoctoral Fellow, Department of Forest Ecosystems and Society, Oregon State University
• David Pettebone, Applied Research Coordinator, Social Science Program, National Park Service
• Scott Esser, Director, Continental Divide Research Learning Center, Rocky Mountain National Park, National Park Service
• Lauren Miller, Social Scientist, Yellowstone National Park, National Park Service (currently Pacific Southwest Social Scientist, U.S. Fish and Wildlife Service)

Abstract

Many of the most complex national park management challenges relate to understanding the relationships between visitor use of the park and associated ecological impacts of those visitors on the park. Most visitor use studies in national parks occur at relatively small spatial scales, such as a single trail system or a feature of interest (e.g., a lake, mountain summit, etc.). In contrast, ecological disturbances can occur at larger spatial and temporal scales. The required management decisions that apply to this social and ecological system must span these very different scales. A recent research project supported by the National Park Foundation Social Science Fellowship aimed to evaluate how visitor use of roads and trails throughout two national parks might impact the fragmentation of wildlife habitats and connectivity between habitat patches. This analysis required estimating visitor use on park roads and trails at the scale of a whole park. Parkwide visitor use monitoring data are rare, and these efforts are costly in terms of time and money. Since parkwide visitor use estimates were unavailable for our project, we leveraged visitor use estimation data to create system-wide models that predicted visitor use levels on unmonitored roads and trails. During our study, we realized that various limitations in the extent, resolution, and scale of the visitor use estimation datasets influenced the approaches we could apply to predict visitor use levels at a parkwide scale. This paper presents lessons learned from this recent research project for incorporating common visitor use estimation data into parkwide models of visitor use and recommends how visitor use data can be collected in systematic, accessible, and repeatable ways. We focus on suggested approaches for monitoring visitor use that may better allow for more systems-based understandings of visitor use that support modeling efforts, better integration with ecological datasets, and informing parkwide management strategies.

Visitors enjoy a hike on the Ute Trail in Rocky Mountain National Park.

NPS

Introduction

Integrated visitor use studies that combine social and ecological measures may better address the complex pressures facing national parks. There has been increasing recognition in the scientific literature that visitor use in national parks is a complex social-ecological system (Morse 2020). This means that both the social (i.e., the visitor use aspect) and ecological components of a national park are inherently linked. Additionally, systems thinking emphasizes that “the system as a whole cannot be understood by studying individual parts alone” (Morse et al. 2022: 2).

A recently funded National Park Foundation (NPF) Science Fellowship (formally the Conway Fellowship) aimed to use a more systems-focused approach to quantify the interaction of visitor use with the broader ecological landscape (NPS 2020). The Social Science Fellowship (SSF) project focused on understanding how park infrastructure and associated visitor use may contribute to wildlife habitat fragmentation and impact connectivity for wildlife between habitat patches in Rocky Mountain and Yellowstone national parks. A key project goal was to combine existing data on visitor use levels with data on park vegetation and habitats at the scale of the whole park. Rocky Mountain and Yellowstone national parks have both implemented multi-year, extensive visitor use monitoring programs across many park sites, making them excellent partners for the SSF project. However, an essential data component of the SSF project was estimates of visitor use levels across all park roads and trails to match the scale of the ecological data we’d be using in the project. While both partner parks have a wealth of visitor use estimation data, collecting visitor counts for every trail and road across entire park systems has not been feasible given the constraints of park resources and research capacity.

We used existing visitor use data to build models predicting visitor use levels on unmonitored roads and trails to address this challenge. During this process, we identified numerous visitor use monitoring practices used by Rocky Mountain and Yellowstone national parks that allow our modeling approaches to be successful. This paper summarizes these lessons learned by highlighting the importance of monitoring visitor use levels across space and time and makes recommendations for how national parks can approach visitor use monitoring more systematically. Systematic visitor use monitoring results in datasets that can be effectively integrated with ecological data or fed into parkwide predictive models. Both of which can help inform complex, parkwide management decisions.

Park entrance stations are important locations for estimating overall visitor use in a national park. Here cars are lined up at the north entrance to Yellowstone National Park.

NPS/Bob Greenburg

Field technician Anne Weiler installs a pneumatic tube vehicle counter along the side of a road in Grand Teton National Park. These automatic vehicle counters can help estimate visitor use along roads throughout national parks.

Oregon State University/Ashley D'Antonio

Visitor Use Levels

Visitor use levels are a vital measure of public use in national parks and are essential in understanding the social component of the system (Ziesler and Pettebone 2018). However, estimating visitor use levels can be incredibly time and labor intensive, and approaches for quantifying visitor use vary (Ziesler and Pettebone 2018). Common examples of techniques to estimate visitor use include counting visitors entering a trailhead or passing a trail section using automatic trail counters. Automatic trail counters record a count every time an infrared (or similar) beam is broken by a person passing the counter. The counts are then stored in the counter and can be summarized as total counts of people by hour or day. Visitor use estimates can also be based on the number of vehicles entering a park through an entrance station. Entrance station counts can be tallied by individuals working in the entrance station or counted automatically via an automatic vehicle counter. Automatic vehicle counters are often either pneumatic tubes or inductive loop counters. Pneumatic tubes are filled with air that is compressed every time a vehicle drives over the tube, the change in air pressure results in a count of a vehicle. Inductive loops can sense metal, counting a vehicle whenever the metal of a vehicle disrupts the magnetic field generated by the counter. Both types of vehicle counters can be used at park entrances or along park roads. Backcountry or camping permits are sometimes used to estimate visitor use in more remote settings, and even observational counts of individuals on trails, vehicles in parking lots, or visitors at one time at park destinations can be used to estimate use. Due to the wide variety of approaches for estimating visitor use, the collection of visitor use levels across parks is often context-specific or driven by research questions of interest and not necessarily repeatable across sites, systematic within a park, accessible for outside comparison, or collected at the scale of a whole park.

Ranger Jenni Burr installs an automatic trail counter in Yellowstone National Park.

NPS

Many visitor use management studies incorporate estimates of visitor use levels (i.e., D’Antonio et al. 2019). Visitor use estimates can help managers understand potential impacts on natural and cultural resources and provide opportunities to integrate visitor use with ecological data (i.e., vegetation maps, wildlife movement). Such integration can provide a start for understanding parks as a social-ecological system (Perry et al. 2020). However, visitor use studies are often designed around specific, small-scale management questions, such as visitor use levels at a single attraction site or trail and across only one or two visitor use seasons (Gutzweiller et al. 2017). This can result in the visitor use estimation data focusing on just certain park sites and seasons and being mismatched in both spatial and temporal terms with natural resource data that can range across landscapes and decades. These differences in spatial and temporal scale make integration across social and ecological data types challenging.

Estimating Visitor Use Across Space and Time

Estimates of visitor use can also vary in scale across space and time. The scale of visitor use estimates can be expressed in terms of extent (breadth of data) and granularity (smallest unit of observation of data) for both the spatial and temporal aspects of monitoring. For example, from a spatial perspective, visitor use can be reported for each parking lot across an entire national park (large spatial extent) or summarized for a single trailhead (small spatial extent). From a temporal perspective, visitor use can be tracked for an entire year (large temporal extent) or captured as a snapshot of use on a single day (small temporal extent). The same distinction can be compared to the grain of visitor use. For example, park visitor entrance counts could be tallied for a month (large grain), or vehicle counts can be tracked hourly along a road (small grain). The extent and grain of the visitor use estimation data define the scale of the observations and shape how the data can be analyzed or integrated with other data types.

The benefits of carefully considering the scale of visitor use level monitoring have become apparent for other aspects of park management beyond thinking about social-ecological systems. Recently, many parks have moved from site-based visitor use management strategies to more broad-scale and parkwide management strategies. This has been important as mechanisms that manage visitor experience have also moved to larger scales. For example, Yosemite National Park instituted permits for day hikes to Half Dome in 2010. The permit system was based on numerous small-scale, site-level studies that helped address acute visitor use issues that stemmed from broader, parkwide visitor use issues (Pettebone et al. 2013). Yosemite National Park recently implemented a temporary peak hours reservation system for the entire park to address crowding and congestion and was based on years of active visitor use monitoring throughout the park (NPS 2022).

Yosemite National Park was able to use many years of data to inform their permit system due to careful data management. Poor data management, archiving, or organization of visitor use data can also limit the ability to scale or combine visitor use estimates and other datasets. The data format, metadata, background information on monitoring protocols, and reporting can be inconsistent. This may not be a concern if visitor use data management approaches are consistent within an individual park. However, the data management aspect of visitor use estimation and creating repeatable workflows is essential for sharing visitor use estimates and allowing the data to be integrated across parks, with other datasets, or compared across time.

Visitor Use Estimates and Predictive Modeling

The ability to collect visitor use data at larger scales, particularly visitor use estimates, from site level to parkwide is an increasingly worthwhile effort. Understanding the potential unintended implications of parkwide visitor management actions requires monitoring systemwide. However, it is often not feasible given resource constraints (i.e., park size, personnel constraints, etc.) to monitor visitor use levels at large spatial extents and across varying temporal scales.

Thoughtful and efficient visitor use monitoring that considers grain, scale, and extent can feed into modeling efforts that can reliably predict visitor use levels at parkwide scales when resources may not be available for extensive on-the-ground monitoring. Predictive models can help provide park managers with an estimate of parkwide use levels and create a dataset that may better integrate with ecological datasets collected at large spatial extents (i.e., maps of habitat or species ranges).

For example, we generated qualitative (i.e., categorical) parkwide visitor use estimates (low, medium, high, and very high use) for Rocky Mountain National Park using existing qualitative data (based on park management experience and judgment) and a predictive model using a machine learning approach called random forest (NRSS 2022). This approach automatically looked for patterns between characteristics of the trails (e.g., slope, distance to water, parking lot size, etc.) with assigned qualitative categories and the qualitative categories themselves. These patterns were then used to assign qualitative use categories (low, medium, high, and very high use) to unmonitored trails throughout Rocky Mountain National Park (Table 1). Quantitative estimates of visitor use from automatic trail counters were used to validate our parkwide visitor use model, indicating the predictions were highly accurate.

Table 1: Summary of the visitor use estimation data used by the SSF team in creating parkwide models of visitor use in Rocky Mountain National Park. Importantly, data varied in spatial scale and temporal resolution.

Visitor Counts	#Trail Segments	%Trail Segments	Temporal Resolution
Trail Counters	22	5%	hourly
Categorical Use Levels Assigned	97	20%	none
Use Level - Low	14	14%	none
Use Level - Medium	32	33%	none
Use Level - High	24	25%	none
Use Level - Very high	27	28%	none
Total in Park Trail System	483	-	-
Total Segments Assigned	97	-	-
Entrance Stations (vehicle counts)	5	100%	monthly

In Yellowstone National Park, we used vehicle use data from a past visitor use management research project to infer the differential use of frontcountry trails. We used backcountry permit data in a network-based modeling approach to route visitors through backcountry trails following the camping itinerary described on their permit. We successfully predicted qualitative levels of backcountry use in Yellowstone National Park. Integrating these visitor use models with ecological data is ongoing, and results from that analysis are outside the scope of this paper. The forthcoming publications will detail the technical aspects of these predictive modeling approaches and their validation. Additionally, these publications will highlight how these parkwide visitation models were used to predict areas of potential fragmentation of habitat from different visitor use levels and where visitor use at different levels overlapped with corridors of connectivity between habitat patches.

A key takeaway from our project is that predictive modeling efforts are only successful if the visitor use estimation data used to create these models is of high quality and has variability in the types of visitor sites being monitored, spans spatial extents, and considers the granularity needed to inform management. Our two projects with Rocky Mountain and Yellowstone national parks have yielded some recommendations for public land agencies looking to collect visitor use estimation data that would allow for better quality inputs for parkwide predictive models of visitor use. Many of these practices are already being implemented in Rocky Mountain and Yellowstone national parks, which was directly related to satisfying the SFF project’s specific aims of examining visitor use and ecological disturbance at parkwide scales.

Visitor Use Monitoring Recommendations

These recommendations are for descriptive monitoring of visitor use focused on estimating use levels. We are not using the term “monitoring” in the same context as the Interagency Visitor Use Management Council, which focuses specifically on monitoring selected indicators (which may or may not be related to visitor use level) to compare to identified thresholds.

Spatial

Visitor use monitoring efforts should be extended throughout the whole park and apply systems-level thinking.

• Consider selecting monitoring locations that include low- and moderate-use areas and high-use locations and use a stratified, random sampling method.
  • For example, data collection at Yellowstone National Park has incorporated low- and medium-use trails in their monitoring network to complement monitoring of high-use destinations. Monitoring sites have also been geographically stratified so different regions of the park are rotated through, and all regions are monitored every few years.

• Include monitoring within the trail and road networks rather than exclusively at trailheads and entrance stations.
  • For example, trail counter monitoring at Rocky Mountain National Park has greatly expanded in recent years to include paired counters within single trail systems and has been used in conjunction with entrance station counts to inform parkwide management strategies.

• Consider including monitoring within central and satellite park areas. Central areas would include the need to go through entrance stations, while satellite areas are trailheads or areas that do not require a gate to enter.
  • For example, Rocky Mountain National Park has trails that enter the park from outside the park boundary. These locations could be considered satellite areas as visitors do not need to go through an entrance station to access these trails. Use on these trails is being monitored via automatic trail counters.

Temporal

Since data can only be analyzed at the smallest resolution they are collected, visitor use estimation data should be collected at the smallest resolution that may be relevant to current or future studies.

• Align various visitor use monitoring efforts (i.e., trail counters, tube counters, entrance station) with the same resolution (e.g., binned by the hour or time stamps of each individual).
  • For example, in the Bear Lake Road Corridor of Rocky Mountain National Park, vehicle counters on the Bear Lake Road and automatic trail counters in the trail system in the corridor can all summarize visitor count data by the hour.

• Coordinate monitoring efforts such that data are collected within the same time frame across all monitoring efforts (trail counters, tube counters, entrance station).
  • For example, if automatic trail counters are used to estimate shoulder season use (the time right before or after the busy season of the park), continue monitoring vehicle use at entrance stations or via tube counters into the shoulder seasons as well. This would allow for a system-level examination of use during the shoulder season.

Data Management and Accessibility

Better disciplined and standard conventions for managing visitor use estimation data could allow similar analytical approaches to be applied across parks.

• Look for examples of other data management and data sharing platforms that could be applied to visitor use estimation in national parks.
  • Examples include the NPS inventory and monitoring program or MoveBank (Wikelski et al. 2022 )

• Data repositories can simplify access when new or changing situations require visitor use data to inform management choices.
  • For example, a detailed and organized database of backcountry permits from Yellowstone National Park—that spanned many years—provided a robust and complete dataset for a network-based model of backcountry trail use. This model can be used to visualize the total number of both visitors and stock on backcountry trails across time and updated each year with new permit data.

We observed that each park's visitor use monitoring efforts were substantial and valuable in the context of the management issue they were designed to address. Still, they were not necessarily designed to be integrated with ecological data at a parkwide scale. However, using existing data that largely followed the above recommendations, the visitor use data available from each park varied in spatial scale and temporal resolution. This variability allowed us to use innovative modeling approaches to predict visitor use parkwide, and we will be able to integrate the visitor use data with ecological data to highlight areas of potential resource impact or concern. Through discussing and organizing available visitor use data for our project, we observed the opportunity that systematic and unified visitor monitoring data could provide in future research and park management.

Even parks like Yellowstone and Rocky Mountain that have a long history of visitor use research and monitoring do not have data available for all locations within the park. A modeling approach addresses this data gap and provides managers with estimates of use.

NPS/Neal Herbert

Conclusions

Historically, park managers tend to focus visitor use monitoring efforts on their busiest trailheads, most popular locations of interest, or where an increase in use is anticipated to address localized management issues with study designs without much consideration of broader uses for these data. Over time, many localized studies have been completed in parks, and subsequent monitoring efforts have become prevalent as a response to increasing levels of visitor use. Park managers and researchers, building on decades of previous visitor use research, are now asking new research questions at broader scales to address concerns about the impacts of visitor use.

Visitor use monitoring and research is time and staff-intensive and can be prohibitively expensive at large scales. Many parks have completed visitor use studies to address management concerns about visitor impacts to visitor experiences and park resources. Ideally, these data can be used to answer developing management issues and research questions. However, the primary challenge to this approach is an unbalanced level of data resolution, where there is extensive visitor use data associated with select locations and limited or anecdotal evidence for use levels in other locations. Certain spatial extents of a national park may be well understood regarding visitor use levels, but other locations may be less so (often lower use level locations). This “uneven” understanding of visitor use levels can limit how data can be integrated with ecological datasets or incorporated into modeling efforts.

The SSF project described in this paper sought to quantify the interaction of visitor use with the broader ecological landscape using available data. Although both study sites, Rocky Mountain National Park and Yellowstone National Park, have a long history of visitor use research and monitoring, visitor use data were not available at all locations throughout the parks. The modeling approach developed through this project addresses this data gap and provides researchers and managers with visitor use estimates that can be analyzed alongside broader ecological data. Adaptive management strategies rely on effectively monitoring system responses and using observed changes to inform decisions; creating data sets of uniform observations across the full extent of national parks can be essential for responsive management as novel challenges present themselves. Systems-focused monitoring incorporating best practices for data management and accessibility instead of problem-focused visitor use monitoring could help national parks better address system-level pressures.

References

D’Antonio, A., D. Taff, C. Monz, P. Newman, J. Baker, W. Rice, and Z. Miller. 2019.
Leigh and String Lakes visitor use study at Grand Teton National Park: 2018 data collection summary. Natural Resources Report.

Gutzwiller, K. J., A. L. D'Antonio, and C. A. Monz. 2017.
Wildland recreation disturbance: broad‐scale spatial analysis and management. Frontiers in Ecology and the Environment 15(9): 517-524.

Morse, W.C. 2020.
Recreation as a social-ecological complex adaptive system. Sustainability 12(3): 753.

Morse, W.C., S. Selin, L. K. Cerveny, and D. J. Blahna. 2022.
Introduction to sustainably managing outdoor recreation and nature-based tourism as social-ecological complex adaptive systems. Journal of Outdoor Recreation and Tourism 100519.

National Park Service (NPS). 2020.
National Park Foundation Science Fellowships. Available at: https://www.nps.gov/nature/conway.htm (accessed on June 20, 2022)

National Park Service (NPS). 2022.
Yosemite National Park launches temporary peak hours reservation system May 20. Available at: https://www.nps.gov/yose/learn/news/yosemite-national-park-launches-temporary-peak-hours-reservation-system-may-20.htm (accessed on June 20, 2022)

Natural Resource Stewardship and Science (NRSS). 2022.
NRSS monthly webinar series: How spatial interaction of visitor use & wildlife habitat can inform conservation, management, and planning. Available at: https://www.youtube.com/watch?v=WYQoRr50OH4 (accessed June 20, 2022)

Perry, E. E., J. M Thomsen, A. L. D’Antonio, W. C. Morse, N. P. Reigner, Y. F. Leung, J. Wimpey, and B. D. Taff. 2020.
Toward an integrated model of topical, spatial, and temporal scales of research inquiry in park visitor use management. Sustainability 12(15): 6183.

Pettebone, D., B. Meldrum, C. Leslie, S. R. Lawson, P. Newman, N. Reigner, and A. Gibson. 2013.
A visitor use monitoring approach on the Half Dome cables to reduce crowding and inform park planning decisions in Yosemite National Park. Landscape and Urban Planning 118:1-9.

Wikelski M., S. C. Davidson, and R. Kays. 2022.
Movebank: archive, analysis and sharing of animal movement data. Hosted by the Max Planck Institute of Animal Behavior. Available at: www.movebank.org (accessed on June 21, 2022)

Ziesler, P. S. and D. Pettebone. 2018.
Counting on visitors: A review of methods and applications for the National Park Service’s visitor use statistics program. Journal of Park and Recreation Administration 36(1): 39-55.

Cite this article as:

D'Antonio, A., E. M. Bredeweg, D. Pettebone, S. Esser, and L. Miller. 2025. Lessons learned from parkwide visitor use modeling. National Park Service, Intermountain Park Science 3(1), 2025.

The findings and conclusions in this article are those of the authors and do not necessarily represent the views of the United States Fish and Wildlife Service.