ABSTRACT
Bird collisions with windows are among the highest sources of human caused mortality to this group of animals. However, environmental correlates of spatial patterns in collision risk are poorly understood, thus making mitigation measures difficult to implement. We took advantage of Covid-19 lockdown in the spring of 2020, when people were obligated to stay mainly at home, and performed a memory-recall questionnaire survey concerning bird-window collisions in Poland. We received information on bird-window collisions with 1800 buildings across the whole country accompanied by characteristics of each building, its vicinity and resident's behavior (time spent home, window cleaning). We supplemented these data with landscape description and performed statistical models to estimate importance of 13 explanatory variables as predictors of number of bird-window collisions. Reported number of collisions increased with the share of forests and arable land within 2 km of the building, and with proximity to rivers. Number of collisions also increased when single trees were close to buildings. More collisions were reported for houses than for flats and for new buildings than for old ones. Reported number of collisions increased with window cleaning which might suggest that cleaning reduces glass visibility for birds. As bird-window collision risk is highly variable among buildings but can be reduced with several measures improving glass visibility for birds, we recommend to use predictive models to identify collision hotspots for applying these measures. New houses located near rivers, in forests or agricultural landscapes have highest collision risk, and trees near buildings, often planted to benefit birds, can additionally elevate collision rate, thus potentially creating ecological traps. In such collision hotspots, reduction of window cleaning frequency can be considered as a mitigation measure unless the visual markers improving glass visibility for birds are installed on the panels.