Introduction
Farmland bird populations have declined drastically in many countries due to agricultural intensification and land use changes that cause or exacerbate habitat loss, lack of food resources and high predation rates (Donald et al. 2001; Robinson and Sutherland 2002; Stanton et al. 2018). Ground-nesting farmland birds are particularly severely affected (Kamp et al. 2021).
Predation, mainly by mammalian predators, can be an important limiting factor for ground-nesting birds (Roos et al. 2018). Apart from high predator populations, habitat loss or fragmentation and the homogenization of agricultural landscapes can further increase predation risk in the remaining habitat patches (Whittingham and Evans 2004; Roos et al. 2018). Conservation measures often aim to improve habitat conditions for birds in arable farmland by providing nest sites and food resources (e.g., flower blocks and fallows), but high predation rates can remain a problem in some areas and can prevent bird population growth (Roos et al. 2018).
Advertisement
Lethal predator control is the most common method to reduce predation pressure, but reports on its effectiveness have been mixed (Tapper et al. 1996; Doherty and Ritchie 2017; Roos et al. 2018). Effective predator control can be particularly difficult to establish on a landscape level and may present ethical challenges (Doherty and Ritchie 2017; Roos et al. 2018; Kämmerle et al. 2019). Several authors have discussed landscape management as an alternative approach to mitigate predation pressure, e.g. by reducing the encounter probability between predator and prey or reducing predator activity in nesting habitats (Doherty and Ritchie 2017; Roos et al. 2018; Laidlaw et al. 2021). So far, however, there has been little research on the possibilities and effectiveness of landscape-based approaches, especially in agricultural landscapes (Doherty and Ritchie 2017; Roos et al. 2018; but see Eglington et al. 2009; Laidlaw et al. 2015, 2017).
Landscape configuration plays an important role in shaping predator activity, and predation risk in conservation measures likely depends on both the type of conservation measure and its location in the landscape (Chiavacci et al. 2018; Laux et al. 2022). On a large scale, landscape composition primarily affects the general abundance of predators in a given area (Thornton et al. 2011). At smaller scales, the location of conservation measures in relation to other landscape elements can affect their attractiveness and accessibility to predators. There could be two different ways in which neighbouring habitat types can affect predation risk in conservation measures: On the one hand, habitats that are particularly attractive to predators may act as focal points, drawing more predators into the vicinity. Consequently, this could lead to increased predator activity and heightened predation risk in neighbouring conservation measures (hypothesis A). On the other hand, attractive habitats for predators may act as a distraction and divert predator activity away from adjacent conservation measures. This, in turn, could reduce both predator activity and predation risk in these conservation measures (hypothesis B). Several studies found higher predator activity or higher predation rates at or near attractive structures such as hedges, field edges or patches of tall grass (e.g.: Morris and Gilroy 2008; Panek 2013; Laidlaw et al. 2015; Arbeiter and Franke 2018; Laux et al. 2022). However, the impact of those structures on nearby conservation measures and the extent to which attraction (hypothesis A) or distraction (hypothesis B) predominates is insufficiently studied and likely depends on the specific conservation measure and its surrounding environment.
Many conservation measures aim to provide nesting habitats for farmland birds and flower blocks, i.e., agri-environment schemes sown with a flowering seed mix, have been shown to be particularly effective (Schmidt et al. 2022). However, our knowledge of the predation risk associated with these conservation measures is limited. Previous studies have shown that predator activity and predation risk are strongly dependent on nesting habitat and the surrounding landscape (e.g., Valkama et al. 1999; Chalfoun et al. 2002; Morris and Gilroy 2008; Laidlaw et al. 2015; Chiavacci et al. 2018). For example, we found in earlier studies that predator activity and predation were lower in broad flower blocks and fallows than in narrow nesting habitats such as hedges (Laux et al. 2022, 2023). Nonetheless, there is little research on landscape effects on predation pressure in different types of conservation measures. If we want to implement conservation measures in a way that minimises predation risk, we need a better understanding of the relationship between landscape and predator activity, and how it depends on the conservation measure in question.
In this study, we focus on flower blocks targeted at Grey Partridges, i.e., agri-environment measures sown with a flowering seed mix, as an example for conservation measures. We use camera trap data to investigate how landscape structure and composition as well as the position of the flower block relative to other landscape elements (distance) affects mammalian predator activity in flower blocks in a Central European farmland. We focus only on mammalian predators, because mammals such as Red Foxes Vulpes vulpes are considered to be the main predators of Grey Partridges (Bro et al. 2000; Potts 2012; Gottschalk and Beeke 2014). Many mammals prey on both adult birds, including incubating hens, and eggs, whereas e.g., corvids mainly predate eggs and small chicks (Bro et al. 2000; Potts 2012; Gottschalk and Beeke 2014). Therefore, mammalian predators likely have a higher negative impact than avian predators. We compare landscape effects on predator activity in flower blocks at two different spatial scales, a 100 m radius as well as a 500 m radius around the camera trap. The 500 m radius has been shown to be well suited to describe landscape effects on predator activity (Laux et al. 2022) and is likely more important for the overall suitability of an area for predators, while the more local scale (100 m) probably has a greater impact on actual habitat use at the field level. With regard to different habitat types and landscape elements, we are further interested in whether attraction (hypothesis A) or distraction (hypothesis B) best explains the relationships between predator activity and different landscape elements such as hedgerows and woodlands, settlements, or edge structures. Grey Partridges are the focus of conservation efforts in this study, but we assume that the results also will be valuable for other ground-nesting birds in flower blocks that struggle with high predation rates.
Advertisement
Methods
Data collection
Study site
This study was conducted under the same framework as Laux et al. (2022). The study area was located near Göttingen, Germany, and consisted of two parts, “Diemarden” (35 km²) and “Eichsfeld” (131 km²), which were both characterised by hilly, semi-open agricultural landscapes dominated by arable land (Fig. 1). There were no targeted predator control efforts in the study area beyond general hunting activities, in which predators play only a minor role (Gräber et al. 2023).
Fig. 1
Map of the study area (NordNordWest 2008; CartoDB 2021; QGIS Development Team 2022) with all camera stations from 2019–2021, N = 132. Coordinates (centre, WGS84, EPSG 4326): Diemarden lat 51.55722, long 10.188629, Eichsfeld lat 51.47607, long 9.958103)
×
![]()
Flower blocks
All cameras were placed in “structurally rich flower blocks”, which were specifically designed for Grey Partridges (Niedersächsisches Ministerium für Ernährung, Landwirtschaft und Verbraucherschutz 2022). These flower blocks are sown with a mix of annual and perennial flowering plants and every year one half of each flower block is resown to create a mix of annual and biennial vegetation. Flower blocks varied from at least 6 m wide strips to extensive flowering fields of up to 6 ha (mean area 1.12 ha, min 0.09 ha - max. 6.94 ha; mean width at the camera site 55 m, min 6 m - max. 262 m).
Measuring predator activity
We used camera traps to measure mammalian predator activity (i.e., the number of predator captures) in flower blocks. Predator activity was used as a proxy for predation risk for ground-nesting birds, which is difficult to measure directly. We assumed that higher predator activity, i.e., more predator captures, corresponds with a higher predation risk. Even without a predation event, predator presence alone can cause disturbances and can have sublethal effects on the birds (Cresswell 2008; Cresswell and Quinn 2013).
Sampling design
The sampling design was an extension of the study described in Laux et al. (2022). We included the flower block data of Laux et al. (2022) where 24 flower blocks were surveyed 2019 and 2020, and expanded it. We surveyed 30 additional flower blocks in 2020 and repeated the survey with all 54 flower blocks in 2021. Overall, this resulted in 132 camera surveys in 54 flower blocks (24 in 2019, 54 in 2020, 54 in 2021).
Camera sites were selected randomly: we placed a 500 m x 500 m grid over each part of the study area (“Diemarden” and “Eichsfeld”) and randomly chose the required number of grid cells. The same grids were used in Laux et al. (2022) and in selecting the additional flower blocks. Only one camera was allowed per grid cell and we always selected the available flower block that was closest to the centre. Grid cells with more than 50% forest or settlement cover or less than 50% inside the study area were excluded. The number of camera stations in each part of the study area (“Diemarden” and “Eichsfeld”) was proportional to the available amount of flower blocks in that area. We obtained permission from all farmers and game tenants to install the cameras.
Data sampling took place each year between May and July to align with the breeding season of Grey Partridges. Camera sites remained the same each year, except when they had to be changed due to crop rotation. In these cases, the nearest suitable and available flower block was selected as a replacement.
For logistical reasons, camera traps were deployed in three successive time-blocks each year. We sampled 8 sites per time-block in 2019 and 18 sites per time-block in 2020 and 2021. Cameras were in operation for at least 20 full days (max. 27 days). If a camera had fewer than 15 consecutive sampling days, it was repeated once, either in the next time-block or in a fourth time-block at the end of the season. Only data collected during the longer collection period were included in the analysis.
Camera traps and picture analysis
We used Browning Strike Pro HD camera traps (HDPX-5, Browning Trail Cameras). They were mounted on wooden stakes approximately 40 cm above the ground and placed in the centre of the biennial part of the flower block (Grey Partridge nesting habitat). No bait was used, but cameras were placed along animal trails to ensure a similar field of view. The cameras were set to take two consecutive pictures after being triggered to facilitate species identification.
Pictures were sorted with digiKam 6.1.0 (digiKam developers team 2019), and all predators were identified to species or genus level. Consecutive detections of the same predator species were considered independent if they were at least 10 min apart, except when individuals could be identified. We chose this short timespan because we were interested in the number of predator visits at a particular camera site within the flower block and not in the flower block itself. We assumed 10 min to be sufficiently long to prevent multiple counts of predators that stayed next to the camera site, but short enough to capture predators that passed by the camera site multiple times while hunting in the flower block. Multiple animals in the same picture were counted individually. We summarised Stone Marten Martes foina and Pine Marten Martes martes as “marten” and Domestic Cats Felis catus and Wildcats Felis silvestris as “cats”, because identification to species level was not always possible. Wild Boars Sus scrofa frequently predate ground-nesting bird nests, therefore we included them as predators (Barrios-Garcia and Ballari 2012). The number of Domestic Dog Canis lupus familiaris captures is shown in Fig. 1, but we excluded dogs from all further analyses, because the number of their captures depends mostly on human behaviour, i.e., popular walking routes or proximity to car parks.
Environmental and site covariates
We generated detailed landscape composition and landscape structure metrics within buffers of 100 m and 500 m around each camera site and calculated the distances between each camera site and different landscape elements (Table 1, S6). 500 m was chosen because we have demonstrated in a previous study that 500 m is well suited to describe landscape effects on predator activity (Laux et al. 2022). 100 m was selected to study the effects of local habitat composition on predator activity in flower blocks.
Table 1
List of predictors considered in the analysis of predator activity
Predictor | Explanation | Unit | Source | |
|---|---|---|---|---|
Landscape composition, measured in 100 m and 500 m buffers | Arable_Areaa | Area of arable land | hectares | InVeKos |
Ext_Area | Area of permanent grassland, fallows, flower blocks and similar environmental schemes | hectares | InVeKos | |
Road_Densa | Area of roads and railways | hectares | B-DLM | |
Settl_Area | Area of settlements | hectares | B-DLM | |
Water_Area | Surface area of all running and standing water | hectares | B-DLM | |
Wood_Area | Hedges, small woods, and forests | hectares | B-DLM, own maps | |
Landscape structure | Border_Length | Length of field block borders | kilometre | InVeKos |
Hab_Div | Shannon-Index based on land cover types within a 100 m/ 500 m buffer: wood, water, settlement, field margin, crop type | Shannon-Index | B-DLM, InVeKos, own maps | |
Distances | Ext_Dist | Distance to the next extensive vegetation outside the flower block the camera is standing in | meter | InVeKos |
Road_Dist | Distance to next road, including railways | meter | B-DLM | |
Settl_Dist | Distance to next settlement | meter | B-DLM | |
Water_Dist | Distance to next running or standing water | meter | B-DLM | |
Wood_Dist | Distance to next wood, including hedges, small woods, and forests | meter | B-DLM, own maps | |
Edge_Dist | Distance to the edge of the flower block | meter | InVeKos, own maps | |
Site based | Year | 2019, 2020, or 2021 | factor | Empirical |
Block | Time-blocks 1-4 in each year | factor | Empirical | |
Runtime | Active camera time | minutes | Empirical |
All predictors were calculated in R 4.2.1 with the package sf (Pebesma et al. 2021; R Core Team 2022). We used the Digital Basic-Landscape Model (LGLN 2019; TLBG 2019) for settlements, streets, forests and water bodies and the 2019, 2020 and 2021 InVeKos data for Lower Saxony (SLA 2019a, b, SLA 2020, SLA 2021) for crop types and field borders. We developed our own maps for hedges and small woods, as well as for field margins, as there were no official maps available. Hedges, woods and field margins were mapped in QGIS 3.16.8 (QGIS Development Team 2022) based on Google Satellite imagery within a 500 m buffer area around each camera site, and later verified in the field.
Statistical analysis
All analyses were carried out with R 4.2.1 (R Core Team 2022). We analysed data from both parts of our study area together, because (a) both parts of the study area have similar landscape compositions and are very close together compared to their size, thus we do not expect predator responses to environmental parameters to vary between areas, (b) we were interested in the effects of environmental predictors on predator activity and these predictors should explain any differences between both areas, and (c) Moran’s I test showed no spatial autocorrelation in either the raw data or the model residuals (Table S1, Moran 1950; R-package ape, Paradis and Schliep 2019).
Comparison of predator capture rates
Using the combined data for 2019, 2020 and 2021, we compared standardised capture rates (i.e., the number of observations per species per 100 camera days per camera deployment (n = 132)) between species to identify the most frequent predators. All comparisons were calculated with a Kruskal-Wallis rank sum test (R Core Team 2022) followed by Dunn’s Post-Hoc test with Holm’s procedure to adjust p-values for multiple comparisons (R-package FSA, Ogle et al. 2021).
Modelling landscape effects
We used negative binomial generalized linear mixed models (GLMM) fitted with glmmTMB (Magnusson et al. 2021) to study landscape effects at the 100 m and 500 m scale on the number of mammalian predator captures in flower blocks. All predator species were grouped together to measure total predator activity. Separate models were fit for the 100 m scale and the 500 m scale.
Prior to modelling, we tested the environmental predictors at each scale for collinearity by calculating the Variance Inflation Factor (VIF) and sequentially dropped predictors with high VIF – values, if necessary, until all VIF < 3 (“HighstatLibV10.R”, Zuur et al. 2009, 2010). The area of arable land (Arable_Area) was excluded from both full models and road density (Road_Dens) was excluded from the 500 m full model, because they were closely related to the area of woodland (Wood_Area) resp. the distance to roads (Road_Dist). Road density was further excluded from the 100 m full model to keep parameterization similar across scales.
Full models at both scales included wood area (hedges, small woods, and forests), water surface area, settlement area, extensive vegetation area (permanent grassland, fallows, and flower blocks), distance to wood (hedges, small woods, and forests), distance to extensive vegetation (permanent grassland, fallows, and flower blocks), distance to water, distance to settlement, distance to road, length of field block borders (a field block consists of several contiguous fields without permanent inner boundaries) and habitat diversity (Shannon Index). See Table 1 for details on all parameters. Time block nested into year was added in all models as a random effect to account for temporal variation and the runtime of the cameras was used as an offset to correct for sampling periods of different length. Furthermore, both models include distance to the edge to account for differently sized flower blocks. All continuous predictors were scaled and centred prior to modelling. Based on the global models, we used backward selection based on the Akaike Information Criterion corrected for small sample sizes (AICc) to identify the most parsimonious model at each scale. We stopped when no further reduction in AICc occurred.
We used the DHARMa package (Hartig and Lohse 2021) to visually assess normality and homogeneity of variances. VIF was calculated with the performance package and was below 3 for all models (Lüdecke et al. 2021). Marginal and conditional R² for mixed models were calculated with the r.squared.GLMM function (Bartoń 2020).
Relative variable importance
We analysed the relative importance of variables for each model with a random permutation procedure following Thuiller et al. (2009). We randomised each variable individually and calculated the correlation between the predictions of the randomised and the original models. We repeated this procedure 100 times for each variable and averaged the correlation. Then, we calculated the importance of each variable as one minus the mean correlation and standardised the relative importance value to one (Thuiller et al. 2009).
Results
78.03% of all cameras recorded at least one mammalian predator (Table S2). Red Fox captures were significantly more frequent than other predator species (mean 7.09 captures/100 days, standard deviation (SD) 11.24), followed by those from Wild Boar, Racoon Procyon lotor, cats and European Badger Meles meles, and, more rarely, dogs, Stoat Mustela erminea, Mouse Weasel Mustela nivalis and martens (Fig. 2, Tables S3, S4). In the following, the term “predator” refers only to mammalian predators, unless explicitly stated otherwise.
Fig. 2
Mean capture rates (captures/100 days per camera station) for each predator. NCameras = 132; summers 2019, 2020, and 2021 together. Kruskal-Wallis χ² = 253.50, df = 8, p < 0.001, error bars show standard deviation (Table S4). Letters correspond to significant differences following a post-hoc Dunn’s Test (Table S4)
×
![]()
Landscape effects on predator captures
Full model results are given in the supporting information (Table S5). After backwards selection, the length of field block borders, distance to roads, and extensive vegetation area were retained as important predictors at both scales (Table 2). Additionally, distance to the next woody element (i.e., hedge, small wood, forest) was retained in the 500 m model (Table 2). Predator captures at both scales significantly decreased with increasing length of field block borders (Figs. 3 and 4; Table 2). Predator captures also decreased with increasing distance to roads and increased with increasing extensive vegetation area at both scales, although those effects were only significant at the 500 m scale and marginally significant at the 100 m scale (Figs. 3 and 4; Table 2). In the 500 m model predator captures also increased with greater distance to woody elements (Fig. 4; Table 2), although this effect was only marginally significant. Length of field block borders had the strongest effect: at the 500 m scale, predator captures decreased by 45% per 5 km of additional field block borders (β -0.341, 95% CI -0.571 - -0.110) and at the 100 m scale, predator captures decreased by 31% per 500 m of additional field block borders (β -0.239, 95% CI -0.421 - -0.056). Length of field block borders also had the highest explanatory power in both the 100 m and the 500 m model (48.63% resp. 37.57%), followed by extensive vegetation area (28.36% resp. 25.31%, Table 2).
Table 2
Model results of final models after backward selection
1) Final model 100 m Response variable: number of predator captures AICc = 568.003, conditional R2 = 0.256, marginal R2 = 0.082, dispersion parameter = 1.735 | |||||
|---|---|---|---|---|---|
Fixed effects | |||||
Predictors | Estimates | 95% CI | p-value | Relative importance | |
2.5% | 97.5% | ||||
Intercept | -9.338 | -9.663 | -9.012 | <0.001 | |
Ext_Area_100m | 0.177 | -0.003 | 0.356 | 0.054 | 28.36 |
Border_Length_100m | -0.239 | -0.421 | -0.056 | 0.01 | 48.627 |
Road_Dist | -0.205 | -0.415 | 0.004 | 0.054 | 23.013 |
Random effects | |||||
Variance | SD | Groups | N Observations | ||
Year: Block | 0.213 | 0.462 | 12 | 132 | |
2) Final model 500 m Response variable: number of predator captures AICc = 564.452, conditional R2 = 0.285, marginal R2 = 0.121, dispersion parameter = 1.816 | |||||
Fixed effects | |||||
Predictors | Estimates | 95% CI | p-value | Relative importance | |
2.5% | 97.5% | ||||
Intercept | -9.341 | -9.660 | -9.021 | <0.001 | |
Ext_Area_500m | 0.288 | 0.069 | 0.506 | 0.01 | 25.306 |
Border_Length_500m | -0.341 | -0.571 | -0.110 | 0.004 | 37.574 |
Road_Dist | -0.289 | -0.516 | -0.061 | 0.013 | 20.705 |
Wood_Dist | 0.17 | -0.011 | 0.350 | 0.065 | 16.415 |
Random effects | |||||
Variance | SD | Groups | N Observations | ||
Year: Block | 0.204 | 0.452 | 12 | 132 | |
Fig. 3
Estimated relationships between the number of predator captures and environmental parameters in a 100 m radius. Please note the different scales on the Y-axes, see Table 2 for effect sizes. NCameras = 132. Predictions are based on the 100 m final model (Table 2). Significant variables: field block borders 100 m (Table 2). For each panel, all other predictors were held at their mean
×
![]()
Fig. 4
Estimated relationships between the number of predator captures and environmental parameters in a 500 m radius. Please note the different scales on the Y-axes, see Table 2 for effect sizes. NCameras = 132. Predictions are based on the 500 m final model (Table 2). Significant variables: field block borders 500 m, extensive vegetation 500 m, distance to roads (Table 2). For each panel, all other predictors were held at their mean
×
![]()
Discussion
According to our study, landscape effects may play an important role in determining mammalian predator activity in flower blocks. Some landscape elements seemingly attract mammalian predators (hereafter: predators), thereby increasing predator activity and consequently predation risk in nearby flower blocks. Other landscape elements appear to be able to divert predators away from flower blocks.
Field block borders, which represent linear edge structures, were the most important factor and had a strong negative effect on predator activity at all scales, i.e., predator captures in flower blocks were significantly lower in highly structured landscapes. At the 500 m scale, predator activity was almost halved with each additional 5 km of field borders. In complex landscapes, predators may use a greater number of different vegetation and edge structures than in homogenous landscapes, which can lead to a “dilution effect” and reduce the likelihood of encounters between predators and birds at any given location (Whittingham and Evans 2004; Panek 2013). Furthermore, linear structures are frequently used as travelling lanes by predators (Šálek et al. 2009; Bischof et al. 2019) and could thus direct predators away from the interior of flower blocks. Field margins may also be preferred by predators for hunting because they are easily accessible and often harbour high populations of small mammals and other prey (Šálek et al. 2010). Thus, they could channel predators away from flower blocks, as proposed in hypothesis B. Besides reducing predation risk, landscape complexity has also been shown to benefit farmland birds and agricultural biodiversity in general (Guerrero et al. 2012; Ekroos et al. 2019; Tscharntke et al. 2021; Šálek et al. 2021). The amount of linear structures such as field block borders is often negatively correlated with farming intensity, because field sizes typically increase in intensively farmed landscapes. Besides preserving existing structures such as ditches and field margins and advocating for smaller field sizes and a reduction in farming intensity, e.g., biological farming, we need simple measures to preserve and increase structural richness and heterogeneity that can be applied on conventional farms. One possibility is to place flower blocks not along the edge of larger fields, as is usually done, but in the centre, so that the flower block effectively divides the field in two parts. An example of that can be found in the agri-environment schemes of Lower Saxony, where there is the possibility to divide large fields with a perennial flower block in the centre and receive an increased compensation (Niedersächsisches Ministerium für Ernährung, Landwirtschaft und Verbraucherschutz 2023).
Predator activity at both scales increased slightly with increasing area of extensive vegetation (i.e., permanent grasslands, fallows, and flower blocks), suggesting that, following hypothesis A, extensive vegetation attracted more predators into the area. Grassland and fallows can be attractive foraging areas for predators hunting small mammals and invertebrates, and these predators might then spill over into the surroundings (Aschwanden et al. 2007; Jacob et al. 2014; Laidlaw et al. 2015; Castañeda et al. 2022). The proportion of extensive vegetation can also be indicative of overall habitat quality, with higher quality habitats generally able to support a larger predator community (Thornton et al. 2011). However, our results indicated that landscape structure, i.e., the amount of borders between landscape elements, is more important for predator activity in flower blocks. Highly structured landscapes have been shown to reduce the probability of predator - bird encounters and may mitigate the positive effects of extensive vegetation on predator activity (Whittingham and Evans 2004; Panek 2013). Furthermore, in Laux et al. (2023) we have found a dilution effect of extensive vegetation area on Grey Partridge nest predation and, similarly, Bergin (2000) found a dilution effect of grassland area on predator activity. This suggests that larger areas of extensive vegetation increase the availability of potential nest sites and provide more opportunities for Grey Partridges to avoid predators, thus leading to lower predation losses, even though these areas are likewise attractive to predators. The discrepancy between these results and ours may be partly explained by the position of camera traps along small animal tracks in our study. If predators mainly stick to these tracks, increases in predator activity might not lead to a similar increase in predation risk throughout the flower block. However, this predicates that the flower blocks are large enough to allow Grey Partridges to move far enough away from tracks and edges (Laux et al. 2022). Consequently, we argue that the positive effects of large extensive vegetation areas outweigh the possible attraction effect on predators.
The negative effect of distance to roads on predator activity in flower blocks at both scales provided further evidence for hypothesis A, i.e., that landscape elements attract predators that then spill over into nearby flower blocks. Many predators use roads and similar infrastructure as travelling lanes because they are easy to navigate and may provide carrion and waste as food resources (Planillo et al. 2018; Bischof et al. 2019). Foxes, for example, also use roads as territory borders and tend to visit them frequently (Kolb 1984; Doncaster and Macdonald 1991; Meek and Saunders 2000).
An interesting observation in our study was the increase in predator activity in flower blocks the further away they were from woody structures (including hedges, small woods, and forests), although this effect was only marginally significant. In many previous studies, woody structures have been found to be highly attractive to predators and to increase predation in their vicinity (e.g., Kuehl and Clark 2002; Keuling et al. 2011; Douglas et al. 2014; Kaasiku et al. 2022; Laux et al. 2022, 2023). Our results, however, indicated that under certain circumstances woody structures might distract predators from nearby flower blocks, presumably because they are more attractive. It follows that flower blocks might become more attractive compared to their surroundings the further away they are from woody structures. However, this effect was only observed in the 500 m model and was only marginally significant. It should therefore only be considered as an indication of a possible interesting effect that warrants a more detailed investigation. In line with the existing literature, we continue to recommend that flower blocks and similar measures should not be placed near forests and large wooded areas in order to reduce predation pressure (Kuehl and Clark 2002; Keuling et al. 2011; Douglas et al. 2014; Laux et al. 2022, 2023).
The results at the 100 m and 500 m scale were very similar, indicating that similar processes shaped the landscape effects on predator activity at both local and larger scales. The 500 m model had better AICc and R² values, indicating that the 500 m scale was better suited to explain predator activity in flower blocks in our study. However, the scale-dependent parameters “area of extensive vegetation” and “length of field block borders” had larger effect sizes at the 100 m scale, suggesting that it is more important how many field block borders and areas with extensive vegetation are within 100 m than within 500 m. One possible conclusion from these seemingly contradictory results is that the landscape within 500 m is important because it determines how many predators are present in the general area, but that the small scale plays an important role in whether these predators actually use a particular flower block. Both scales should be taken into account in conservation planning by first selecting appropriate landscapes for conservation efforts and then ensuring that the immediate surroundings of a flowering strip are suitable.
However, these results apply only to flower blocks in our agricultural landscapes and might differ in other regions or vegetation types. Other studies, e.g., Chalfoun (2002) and Chiavacci (2018), found that landscape effects on predation risk strongly depend on spatial scale, predator species, habitat type and the larger landscape. Therefore, caution is needed when transferring our results to other contexts. While we studied total predation risk and pooled all predator observations to increase sampling size, different predators may react differently to environmental factors and more data is necessary to study predator-specific responses.
Interestingly, only 22% of all camera stations in flower blocks were without mammalian predator activity during the 20 days of the survey. However, Grey Partridges, for example, need 40 days to lay eggs and incubate them (Cramp 1980). In contrast to the low proportion of flower blocks without predator activity, we found a hatching success rate of 31.6% for Grey Partridge nests in flower blocks and fallows (excluding non-predation losses) in a previous study in the same area (Laux et al. 2023). One reason for this discrepancy is probably that we placed all cameras along small animal tracks to ensure a similar field of view, as these tracks are usually more frequented by predators than the surrounding higher vegetation (Laux et al. 2022). However, even if they follow set tracks, predators moving in flower strips pose a higher risk to ground-nesting birds than those outside, as they are potentially closer to nest sites and more likely to detect them. Therefore, these results suggest that microhabitat selection by Grey Partridges plays an important role in predator avoidance and the selection of safe nest sites within a habitat patch, as has been shown for other ground-nesting species (Benton et al. 2003; Coates and Delehanty 2010; Casas et al. 2022). Consequently, a high availability of potential nest sites, e.g., in broad nesting habitats or heterogeneous landscapes, likely facilitates microhabitat selection and predator avoidance, benefiting Grey Partridges and other ground-nesting farmland birds (Laux et al. 2023).
In summary, we found that landscape effects have an important impact on mammalian predator activity in flower blocks, demonstrating the potential of landscape-based approaches to reduce mammalian predator activity and thus predation risk. Landscape structure, represented by field block borders, was the most important parameter influencing predator activity in flower blocks. Landscapes rich in these edge structures were associated with lower predator activity in flower blocks, indicating that maintaining or enhancing small-scale landscapes can be beneficial for ground-nesting birds (Guerrero et al. 2012; Ekroos et al. 2019; Šálek et al. 2021). We found evidence that neighbouring habitat types can have both attracting and distracting effects on predators: Predator activity in flower blocks increased near roads, while linear edge structures seemed to divert predators away from flower blocks. Predator activity was also higher in areas with a high proportion of extensive vegetation (i.e., presumably in areas with a higher habitat quality), but this effect was mitigated by a large amount of edge structures. Consequently, adequate placement of flower blocks, e.g., in structurally rich areas, may reduce predation risk and increase nest success. This study focussed on flower blocks, but landscape effects on predation likely play a role in other contexts too, and further research on landscape-based approaches could also be valuable for other conservation measures aimed at reducing predation pressure (Langgemach and Bellebaum 2005; Laidlaw et al. 2015, 2017).
Landscape-based approaches to mitigate predation risk have the potential to reduce the need for controversial and difficult-to-implement predator control measures and can provide multiple benefits to agricultural biodiversity. However, landscape effects are highly context-specific, requiring careful consideration of the landscape and conservation measures in question before implementing a landscape-based approach.
Acknowledgements
We thank all farmers and game tenants who allowed us to install camera traps on their land. We thank L. Dumpe and W. Beeke for their help in contacting farmers and game tenants and for providing habitat maps. We thank N. Jaspert, L. Demel, L. Böttges and K. Mayer for help and advice during the field work.
Declarations
Competing interests
The authors declare no competing interests.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
Publisher’s note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
