Grazing by wildlife on agricultural land is widespread across geographical regions, and can cause human-wildlife conflicts due to reduced crop yield when the grazing pressure is high. Growing red deer (Cervus elaphus) populations in Europe calls for an increased understanding of their grazing patterns to mitigate damages. We quantified how red deer grazing pressure (grazing presence and grazing level) on agricultural grass meadows (n = 60) in Norway varied across multiple spatial scales. We used a nested, hierarchical study design transcending from broad scale (meadows across the landscape) to intermediate (between nearby meadows) and local (within-meadow) scale, allowing us to identify at which scale the variation in grazing pressure was strongest. We estimated how grazing was determined by broad-scale factors influencing forage availability and quality through population density, distance to coastline and differences between the first versus second harvest, by intermediate-scale factors in terms of meadow management causing differences in botanical composition and quality, and by local-scale factors in terms of perceived predation risk and disturbance. At broad scale, higher population densities were associated with higher grazing pressure, and more grazing occurred before the first compared to the second harvest. Intermediate-scale factors explained the most variation of grazing pressure from red deer, with higher grazing pressure on newly renewed meadows compared to other nearby meadows. On local scale, more grazing occurred closer to the forest edge, providing cover, and further away from infrastructure, with increased risk and disturbance. Overall, our study highlights how drivers of grazing pressure on agricultural land vary across spatial scales. Population reductions on broader scale may have some effect in reducing the grazing pressure, but renewed meadows will nevertheless attract red deer causing higher grazing pressure compared to neighbouring meadows. This insight is crucial for determining effective mitigation strategies facing rising red deer populations across Europe.

The dataset was collected by observing the number of grazed grass blades on n=60 grass meadows on the Norwegian west coast during two field periods of 2021. On each meadow, two parallel transects with a random 30-70 m distance apart and minimum 10 m from the meadow edges were delineated (for each field period). For each transect, vegetational assessments were performed 20 times approximately 5 m apart, each time within a standard botanical metal frame (50 x 50 cm, divided into four sub-squares). 

Using a standard botanical frame (50x50 cm), the number of grazing incidents were counted. The number of grazing incidents in every sub-square of the botanical frame were categorised into four levels: 0 = no observed grazing; 1 = 1-3 grazing instances, 2 = 4-6 grazing instances, 3 = ≥7 grazing instances. One grazing incidence is counted for each grass blade that is grazed. The mean of this measure of grazing quantity was calculated for each sampling square (botanical frame), and re-scaled from count data to a distribution from 0-1 (0 being no grazing and 1 being highest quantity and degree of grazing, equal to 3 on the original scale). Vegetation cover assessments and plant height were averaged for each transect, and elevation and distance to coastline was averaged to meadow level. The variables timothy cover, and the distances to coast, buildings, forest edge, and roads, were log_e transformed. Total vegetation cover was inversed and square root transformed, and plant height was mean-centred.