Wild vertebrates usually avoid ground disturbed by humans with uncertain effects on their distribution and density. We report on a natural experiment that confirms and extends previous conclusions about how the local distribution of capercaillie is affected by track-based disturbance. We surveyed the birds’ droppings in two periods, before and after a predicted increase in disturbance adjacent to an expanding Scottish village, and model the probability of finding droppings in relation to period plus two disturbance gradients – distance to a much disturbed ‘entry zone’ with a high density of tracks lying adjacent to the village, plus distance to the nearest track. Throughout the study, no droppings were found within the entry zone and their density was depressed up to 400 m from it. Density of droppings was also depressed within 100–120 m of tracks but, compared to a notional scenario without tracks, was threefold greater on ground 120–250 m from tracks. Although the number of cocks counted at the lek in spring showed a small increase, there was an almost threefold decline in the probability of finding capercaillie droppings on and within about 40 m of tracks. We infer that, after the development, birds were less likely to use or cross tracks in pursuit of their daily living requirements, with unknown consequences for their population dynamics and vital rates. Our results have implications for refuge design. Birds on roughly half of a 50 ha refuge should be undisturbed by direct effects of track-based activities. However, birds shifting away from tracks may themselves cause local increases in density and associated, indirect disturbance such that a refuge would need to be over 300 ha  to keep half of it undisturbed.

Capercaille droppings were surveyed along fixed N-S transects 100m apart during 14 surveys in 2006-2021. These fell into two periods: Period 1 before the adjacent housing development and Period 2 after it.  Each observation was assigned to a 100 x 20m grid based on the transect lines. These observations were then used to prepare the heat map Figure 1. The distances from each grid point to the 'entry zone' dE and to the nearest track dT were measured via GIS. The probability of finding droppings Pf (response variable) was then modelled as (explanatory variables) a cubic polynomial in dE and a quadratic in dT plus the category period and its interactions with each of the polynomial terms via  the SAS (ver 9.1) Glimmix macro to enact a logistic GLMM with survey as a random effect. Temporal and spatial autocorrelations were accounted for as explained in the paper.