Green treasures: Investigating the biodiversity potential of equine yards through the presence and quality of landscape features in the Netherlands

2 ha) and a total of 22 individual responses were removed due to (suspected) erroneous data (e.g. coverage of LFs exceeding the size of the property). All variables were tested for normality using the Kolmogorov-Smirnov Test which revealed non-normal distribution across all variables. A Mann-Whitney U test was conducted to deter- mine significant differences between commercial and private yards for Yard size, Number of horses, Number of LFs and PPC. univariable linear regression analysis was conducted initially to test the unique contribution of twenty-one (N = 21) predictor variables on PPC, including five (N = 5) “yard descriptors” (Type of ownership, Yard size, Number of horses, Soil type and Number of LFs and fifteen (N = 15) two dimensional LFs, namely Tree lanes, Tree rows, Fruit orchard, Wild hedges, Trimmed hedges, Woody strips, Forage walls, Flowering strips, Embankments, Buffer strips, Ponds, Permanent woodpiles, Cluttered corners, and Monocultural grassland and Herbaceous grassland. Even though the variables Monocultural grassland and Herbaceous grassland were not included in the calculation of PPC, they might be argued to be of importance to the total level of biodiversity, and were therefore included in the regression analysis. Variables were selected for inclusion in the hierarchical regression model based on their statistical significance in univariable regression analyses, with a significance level of p < 0.05. The hierarchical regression model was built in stages, with variables entered in two blocks

based on theoretical relevance: Yard descriptor variables were entered as block 1, as their impact on the spatial restrictions of the yard should be controlled for prior to investigating the impact of LFs on PPC. Variables encompassing two dimensional LFs were included in block 2. The fit of the model at each stage was compared using R-squared and adjusted R-squared, with a focus on the explanatory power of the model in relation to the number of variables. The model fit was evaluated by examining the residuals, estimating variance inflation factors (VIFs) to check for multicollinearity. The final model was chosen based on its ability to pro- vide a meaningful explanation of the predictive ability of two dimensional LFs on PPC, while controlling for yard descriptors. Semi-partial correlation coefficients squared were calculated to determine the percentage of variability of PPC uniquely accounted for by each predictor variable. Statistical significance was set at p<0.05. Results Commercial yards were significantly larger than private yards (p<0.001) and also had signifi- cantly more horses (p<0.001; Table 3). Commercial and private yards had a similar number of LFs but PPC was significantly higher on private yards (p<0.001; see Fig 1 and Table 3). All yards contained at least two features with only commercial yards featuring all 21 features (Fig 2). Most yards were situated on sandy soil, followed by clay soil. Most private and commercial yards contained LFs such as Birdhouses, Nesting sites and

Cluttered corners (Fig 3). The least common LF was forage wall followed by bat boxes and embankments (Fig 3). In general, the proportion of yards containing a particular feature was comparable between private and commercial yards, with the most noticeable differences being features such as tree lanes, trimmed hedges and tree rows (Fig 3). Detailed descriptives of LFs for all yards is provided in the S1–S3 Tables. Univariable linear regression results Fourteen of the 21 considered predictor variables in the univariate linear regression analyses explained significant variation in the percentage of a property covered by landscape features (Table 4). Yards that were larger and had more horses, had a lower PPC (Table 4). As can be PLOS ONE Biodiversity potential of equine yards through landscape features PLOS ONE | https://doi.org/10.1371/journal.pone.0301168 April 11, 2024 6 / 16