Scientists analysed 11.8 million locations from 4,581 animals across the United States and found that wildlife responds not only to the roads and cities humans build, but to people themselves — more than 65% of the species studied changed how much space they used or which environments they occupied as human presence rose and fell.
Roads, farms and cities leave durable marks on wildlife habitat. People are less static. A park can be crowded one week and nearly empty the next, while its paths, car parks and buildings remain in place. That difference gave researchers a way to ask whether animals respond to human-built landscapes, to the people moving through them, or to both.
In a study published in Science, researchers analyzed about 11.8 million GPS locations from 4,581 animals representing 37 bird and mammal species across the United States. They paired those tracks with weekly estimates of human presence and a separate measure of landscape modification. More than 65 percent of the species changed either the amount of ground they used or the range of environmental conditions they occupied as human presence changed.
This is one study, not settled consensus. Its results are observational, the tagged animals were not a representative sample of all American wildlife, and the data cannot establish why a particular animal moved differently. What the study does unusually well is distinguish two forms of human pressure that are often bundled together.
A road stays put; people do not
The study covered January through August in both 2019 and 2020. COVID-19 disruptions altered where people spent time, helping loosen the usual connection between how developed a place was and how many people were present. The COVID-19 Bio-Logging Initiative had anticipated that this extraordinary shift could reveal how wildlife responds when human mobility changes abruptly.
Landscape modification was a relatively stable measure combining settlement, agriculture, transport, mining, energy and electrical infrastructure. Human presence was estimated daily from aggregated counts of mobile devices in US Census block groups, then aligned with the animals’ weekly movements. The researchers were not following named phone owners or recording direct encounters between a tagged animal and an identified person.
The distinction matters.
ScienceBlog’s earlier account of koala habitat fragmentation concerned durable changes to available habitat. This study asks a different question: what changes when the physical setting is much the same but the number of people present rises or falls?
How 11.8 million fixes became two weekly measures
The sample contained 22 bird species and 15 mammal species. For every animal and week, the researchers estimated two outcomes. The first was area size, or the amount of geographic space used. The second was environmental niche size, calculated from the range of environmental conditions represented in that animal’s locations.
That second term needs care. It did not describe a species’ entire ecological niche, nor did it show that an animal had permanently abandoned one habitat for another. It measured how varied the conditions occupied by an individual were during a particular week. The models also accounted for air temperature and vegetation productivity, and the niche analysis accounted for weekly area size.
The team then built species-specific statistical models relating those outcomes to human presence, landscape modification and the interaction between them. The pandemic supplied unusual contrast, but it did not turn the research into a controlled experiment. As the authors’ US Geological Survey record makes clear, the work identifies associations rather than randomized causal effects.
The 65 percent figure did not describe one response
Human presence was associated with a change in area size or environmental niche size for 67 percent of the mammal species and 68 percent of the bird species. Fifty-seven percent of all species responded to both human presence and landscape modification. Among the species that responded to human activity, 59 percent showed an interaction between the two pressures.
Those percentages do not mean that two-thirds of wildlife simply fled from people. A response could be a contraction or expansion in the area used, a narrowing or widening of environmental conditions occupied, or a pattern that changed with the level of landscape modification. The direction and size varied by species.
Among species with a statistically supported area response, mammals showed a median weekly change of about 11 percent less area per animal, while birds showed a median reduction of 1.8 percent. These are summaries across the responding species, not predictions for every individual. Large positive and negative changes could sit behind the same median.
Wolves, coyotes and ravens did different things
Most mammal species contracted the area they used in association with human presence or landscape modification. Human presence was associated with a smaller area for 41 percent of the bird species. But there was no universal avoidance pattern.
Gray wolves expanded their area as either human presence or landscape modification increased. One plausible interpretation is that avoidance made them travel farther, but the movement tracks cannot prove that mechanism. Across the modeled range of both human pressures, coyotes were estimated to use 11.3 fewer square kilometers per week per animal, while common ravens used 26 square kilometers more.
Environmental niche responses were just as varied. Ravens narrowed their occupied conditions by an estimated 46 percent, while cougars widened theirs by about 10 percent. White-tailed deer expanded their niche with greater landscape modification but contracted it with greater human presence; sandhill cranes showed the reverse. The University of California, Santa Barbara account emphasizes this species-by-species diversity rather than treating wildlife as a single behavioral category.
Less modified landscapes sometimes produced stronger responses
For 40 percent of mammal species that responded to human presence, the association was stronger in less modified habitats. Wild turkeys, for example, restricted the area they used as human presence rose, but that response was weaker in more modified settings.
Several explanations could fit. Animals living around cities might be more accustomed to people, or a developed landscape might leave them fewer options for changing course. Conversely, animals in a relatively intact landscape may have more room to shift their movements. The data do not choose among habituation, displacement, food subsidies or simple constraint.
This result also cautions against assuming that an intact-looking map is a complete measure of refuge. An animal can encounter changing levels of recreation, traffic or other activity without any new road appearing. A previous global camera-trap analysis of 163 mammal species likewise found that responses to changes in human activity during the pandemic depended on species traits and landscape context.
The same animals could change between years
Some individuals were tracked in both 2019 and 2020, allowing the researchers to compare an animal with itself. Year-to-year increases in human presence were generally associated with a smaller area used and a broader environmental niche. That pattern points to at least some behavioral flexibility within individuals, rather than every result coming from different animals occupying different regions.
Yet species varied, and the overall statistical interaction used to test the paired-year pattern was weak. The pandemic was a rare natural disruption, not a randomized intervention. Lockdowns, altered traffic, changes in work and recreation, weather and other regional events happened together. The analysis controlled for important environmental factors, but no observational model can remove every alternative explanation.
The Yale researchers involved in the study frame the result as evidence that animals notice short-term changes in human presence. I read that as the useful claim here. It is narrower, and stronger, than saying the pandemic showed nature “healing” whenever people stayed home.
What the tracks cannot tell us
A smaller weekly area is not automatically evidence of fear, stress or harm. It could reflect disturbance, access to a concentrated food source, a change in travel routes or something else. An expanded area might represent opportunity, or it might represent the extra distance required to avoid people and infrastructure. Without fitness data, the study cannot label each response a cost or a benefit.
The human-presence data also did not distinguish hiking from hunting, commuters from tourists or traffic from intentional feeding. Aggregated mobile-device counts are an imperfect proxy, particularly at fine spatial scales, and access to such commercial mobility data is limited. The period studied was unusual, and tagged animals are not a random census of US wildlife.
Those boundaries do not erase the finding.
They locate it. The study shows that static landscape maps can miss a dynamic part of the relationship between people and animals, while leaving the biological consequences for future work.
Conservation needs a clock as well as a map
The practical lesson is not that every reserve needs fewer visitors. It is that managers may need to ask which species respond, in what direction and at which times. Recreation can peak during migration, breeding or feeding periods. A trail that is quiet most of the year may have a different effect from one with constant low-level use, even if both look identical in a land-cover database.
That suggests testable options: seasonal trail closures, quiet corridors, adjusted visiting hours or monitoring before and after a change in access. Any intervention should be based on local evidence, because the same level of human presence did not produce the same response across wolves, deer, cranes, ravens and turkeys.
Static maps tell us where people have transformed the land. These 11.8 million locations show why knowing when people are actually there can add another layer.