Leveraging digital mobility data to estimate visitation in National Wildlife Refuges

Abstract

The US Fish and Wildlife Service manages over 500 National Wildlife Refuges and dozens of National Fish Hatcheries across the United States. Accurately estimating visitor numbers to these areas is essential for understanding current recreation demand, planning for future use, and ensuring the ongoing protection of the ecosystems that refuges safeguard. However, accurately estimating visitation across the entire refuge system presents significant challenges. Building on previous research conducted on other federal lands, this study evaluates methods to overcome constraints in estimating visitation levels using statistical models and digital mobility data. We develop and test a visitation modeling approach using multiple linear regression, incorporating predictors from eight mobility data sources, including four social media platforms, one community science platform, and three mobile device location datasets from two commercial vendors. We find that the total number of observed visitors to refuges correlates with the volume of data from each mobility data source. However, neither social media nor mobile device location data alone provide reliable proxies for visitation due to inconsistent relationships with observed visitation; these relationships vary by data source, refuge, and time. Our results demonstrate that a visitation model integrating multiple mobility datasets accounts for this variability and outperforms models based on individual mobility datasets. We find that a refuge-level effect is the single most important predictor, suggesting that including site characteristics in future models will make them more generalizable. We conclude that statistical models which incorporate digital mobility data have the potential to improve the accuracy of visitor estimates, standardize data collection methods, and simplify the estimation process for agency staff.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 10 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.