Connecting habitats in European agricultural landscapes: Farmers' spatial preferences for linear wildlife corridors

Connecting habitats in European agricultural landscapes: Farmers’spatial preferences for linear wildlife corridors Fabian Klebl a,b,* , Jonathan R. Rhodes c , Kati H¨ afner a , Annette Piorr a a Leibniz Centre for Agricultural Landscape Research (ZALF), Eberswalder Straße 84, Müncheberg 15374, Germany b Agricultural and Food Policy Group, Albrecht Daniel Thaer Institute of Agricultural and Horticultural Sciences, Humboldt-Universit¨ at zu Berlin, Unter den Linden 6, Berlin 10099, Germany c School of Biology and Environmental Science, Queensland University of Technology, Brisbane, Queensland 4000, Australia HIGHLIGHTS •Farmers prefer to place measures to minimise disturbance and maximise productivity. •Landscape characteristics guide farmers’decisions. •Habitat connectivity is a relevant concept for farmers. •Local farming traditions and practices are crucial. •The study results can help to reduce transaction costs by targeting likely adopters. ARTICLE INFO Keywords: Biodiversity conservation Habitat fragmentation Landscape connectivity Ecological corridors Hedgerows Agri-environmental measures ABSTRACT Habitat fragmentation in agricultural landscapes threatens biodiversity. Enhancing landscape connectivity across cultivated areas requires a thorough understanding of farmers’spatial considerations and their willingness to create semi-natural habitats. We therefore conducted a spatial choice experiment with farmers from ten European countries to assess their preferences for placing linear wildlife habitats (hedgerows and wildflower strips) at the field scale under different scenarios, as well as the role of farm and personal factors. A total of 471 responses were analysed using multinomial logistic regression and generalised linear mixed models. The results indicate that landscape conditions, including field shape, slope, soil quality, and pre-existing landscape features, exert a significant influence on farmers’decisions, as do the size of machinery, cultural regions, attitudes towards biodiversity, and type of intervention. On the other hand, no statistical significance was found for other variables. In general, farmers’choices were driven by a desire to minimise disturbance to field work, optimise productivity, increase biodiversity, and address specific environmental challenges. The insights into farmers’ decision-making from this study can inform ecological network planning to reduce transaction costs by preselecting likely adopters, and to mitigate resistance and lower financial compensation by identifying best-fit options aligned with farmers’practices. Integrating these findings into geospatial models could improve predictions of the impact of spatially targeted biodiversity conservation strategies on landscape composition and future biodiversity trends in agricultural areas. 1. Introduction Habitat fragmentation poses an urgent threat to biodiversity in agricultural areas and is affected by how farmers manage their land (e. g., Haddad et al., 2015; Schlaepfer et al., 2018). The (re)introduction of wildlife corridors on farmland to connect existing habitats is seen as an important means to mitigate negative effects of habitat and species isolation. These corridors can take traditional forms, such as hedgerows, or more contemporary approaches, such as perennial wildflower strips. Yet, there is still limited understanding of farmers’perspectives on landscape connectivity and their decisions to contribute to it. Hedgerows in particular have demonstrated considerable potential for enhancing habitat connectivity. While the specific effects vary depending on the landscape and species under study, there is evidence that hedgerows in agricultural landscapes act as corridors for numerous species, enabling their movement and gene flow between small, isolated * Corresponding author at: Eberswalder Straße 84, Müncheberg 15374, Germany. E-mail address: [email protected] (F. Klebl). Contents lists available at ScienceDirect Landscape and Urban Planning journal homepage: www.elsevier.com/locate/landurbplan https://doi.org/10.1016/j.landurbplan.2025.105325 Received 12 August 2024; Received in revised form 17 December 2024; Accepted 13 February 2025 Landscape and Urban Planning 258 (2025) 105325 Available online 1 March 2025 0169-2046/© 2025 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY license ( http://creativecommons.org/licenses/by/4.0/ ). habitat patches (Dondina et al., 2016; Fischer et al., 2013; Gelling et al., 2007; Kratschmer et al., 2024; Liccari et al., 2022; Michel et al., 2006; Moorhouse et al., 2014; Pelletier-Guittier et al., 2020; Portela et al., 2020; Travers et al., 2021; Vasiliev &Greenwood, 2023; Wehling & Diekmann, 2009). Despite the risk that ecological corridors may facilitate the spread of diseases and invasive species (Montgomery et al., 2020), they remain a vital element in landscapes with small habitat patches where populations are unstable and habitat expansion is not feasible (Donaldson et al., 2017). However, establishing hedgerows requires substantial effort and a long-term commitment from farmers, often spanning multiple decades. As a trade-off, this study also considers perennial wildflower strips for the creation of linear connecting habitats. Although their overall impact on landscape connectivity is not as pronounced as that of hedgerows, they offer continuous resources and refuges for various species, thus supporting their dispersal and survival in otherwise inhospitable agricultural areas (Aviron et al., 2011; Szit´ ar et al., 2022). The strategic spatial placement of these strips by farmers is crucial for developing a comprehensive flowering network across entire landscapes, thereby maximising their ecological benefits (Buhk et al., 2018). To implement a network of connected habitats in agricultural areas, conservation strategies need to be based on an understanding of what motivates farmers to create these habitats and to connect them to the network. While much is known about farmers’general motivations for adopting biodiversity-friendly practices (Klebl, Feindt, &Piorr, 2024a), to the best of our knowledge, no study has yet assessed where farmers are most likely to place biodiversity measures at the field scale, what influences their underlying decisions, or whether habitat connectivity is a relevant consideration for them. This study investigates the factors influencing farmers’decisions about the placement of biodiversity measures. Specifically, it examines what drives farmers’choices in the spatial allocation of linear habitats such as hedgerows and flower strips, and how field and landscape characteristics impact their preferences. These questions were addressed through a spatial choice experiment in which farmers responded to hypothetical scenarios. The resulting data were statistically analysed to evaluate the influence of factors related to the farmers and their farms, as well as the hypothetical field and landscape settings. 2. Material and methods 2.1. Study areas and sample The present study relies on a survey conducted among farmers across several European countries (Fig. 1), representing a diverse range of landscapes, climate conditions, farming systems, and socio-cultural contexts. These countries were selected for their role as hosts of the ecological and socio-economic experimental sites within the SHOWCASE research project (https://showcase-project.eu/). To obtain a representative sample, farmers were randomly recruited through multiple channels, including farmers’associations, NGOs, and local networks. 2.2. Data collection Farmers were first asked to provide general personal information, including age and educational background, along with detailed farm characteristics, such as farm size, operational focus, average field size, field operations, machinery size, and landscape conditions. Likert scales were employed to evaluate farmers’attitudes towards biodiversity, assessing the importance they attribute to it and its conservation, both in general and in agricultural landscapes. The survey’s core component was a spatial choice experiment aimed at capturing farmers’preferences for allocating biodiversity measures at the field scale. Participants were randomly assigned to one of two groups: one focused on planting a wildflower strip as a temporary intervention (minimum width of four metres, lasting at least three years), while the other considered establishing a naturally grown hedgerow as a long-term landscape element (minimum height and width of two metres each). Respondents were then invited to envision a hypothetical field (or meadow or pasture, depending on their operational focus) that is accessible from all sides. In the baseline scenario, the field was described as being of average size as reported by the farmer, with a rectangular shape, flat surface, and uniform soil quality. In each subsequent scenario, one characteristic was altered while the others remained consistent with the baseline. These alterations included the introduction of a slope, a soil quality gradient, a different field shape, doubling the field size, a prevailing wind direction, as well as varying settings such as proximity to a forest or road, and the presence of existing hedges on neighbouring fields (Table 1). In each scenario, participants selected their first and second preferences from the available options by clicking on corresponding graphical illustrations and briefly explaining their choice. For example, in the baseline scenario, participants could choose from four options, as shown in Fig. 2. Although second preferences were recorded to gain additional insights, they were excluded from the final analysis as no sound approach was identified to weight or incorporate the second choice in a meaningful way. Further details on the survey can be found in the supplementary material. The survey was part of a broader project survey and was designed as an online application, accessible via web browsers and mobile devices. However, in some regions of Romania, where the online format was unsuitable, a printed version was used and completed during workshops organised by local project partners. Data collection occurred between January and October 2023, yielding 471 valid responses. 2.3. Data analysis 2.3.1. Data preparation The dataset contains relatively few missing values, as the choices and most survey questions were mandatory. Missing data, classified as Fig. 1. Overview of study areas: Estonia, France, Hungary, Portugal, Romania, Spain, Sweden, Switzerland, the Netherlands, and the UK. F. Klebl et al. Landscape and Urban Planning 258 (2025) 105325 2 ‘missing completely at random’, were addressed through the Multivariate Imputation by Chained Equations (MICE) method. Linear predictors were imputed via predictive mean matching, whereas categorical values were estimated using logistic regression. An iterative approach was employed to achieve a final imputation that closely matched the original dataset, utilising the {mice} package (van Buuren &GroothuisOudshoorn, 2011) in R version 4.4.1 (R Core Team, 2024). The data density before and after imputation exhibited an almost identical pattern (Fig. A1 in the Appendix A), indicating a high accuracy of the imputed data. Participants’textual explanations of their choices were translated and categorised. They were not analysed quantitatively due to frequent omissions or irrelevance. Nevertheless, an overview of the most commonly cited reasons is provided, which offers qualitative insights that help to contextualise the decision-making process. 2.3.2. Multinomial logistic regression model Since the choice options are categorical outcomes with multiple possible categories, we employed a multinomial logistic regression model fitted for each scenario to investigate the relationship between several predictor variables and the response variable (Agresti, 2007). The model can be described as follows: log( π j π J)=β0j+β1jX1+β2jX2⋯+βkjXk(1) for j=1,⋯,J−1, where π jis the probability of the outcome (i.e., the choice) being in category jand π Jin the reference category (J), β0jis the intercept for category j, and βij are the coefficients for the predictor variables Xi. As log-odds are unintuitive to interpret, we derived the odds ratio (OR) to estimate the change in the odds of being in a specific category of the outcome variable with a one-unit increase in the predictor variable, relative to the reference category (Hosmer et al., 2013). This is defined as  θij =e ˆ βij (2) In formulating the regression models, several variables were found not to significantly enhance the model’s explanatory power but to increase its complexity. These variables included farm size, the proportion of owned land to total cultivated land, the operational status of the farm (full-time or part-time), prior implementation of similar interventions (i. e., hedgerows or wildflower strips), farmers’general and agricultural education levels, and farmers’age. Consequently, these factors were excluded from the final models. The variables that provided an explanatory contribution and/or were an important component of the initial assumptions were retained as independent variables and are summarised in Table 2. Due to a high correlation between the countries and predictor variables, it was impractical to include country as an independent variable. However, regional influences were considered relevant beyond these factors. Therefore, a socio-cultural categorisation was conducted, emphasising major agricultural significance. The regions were defined as Mediterranean farming systems in southern Europe (ES, PT), European countries formerly aligned with the Soviet Union (EE, HU, RO), and western and northern European countries (CH, FR, NL, SE, UK), which served as the reference group. The regression models are founded on certain key assumptions. All observations in the sample are deemed to be independent and sufficient in number, with at least about ten observations per predictor variable for each outcome category. Categories that did not meet these conditions were excluded. 1 The independence of irrelevant alternatives (IIA) hypothesis was tested using the Hausman-McFadden Test via the mlogit:: hmftest() function and was rejected for each variable. The Box-Tidwell test (p=0.46) confirmed the assumption of linearity between Table 1 Overview of the scenarios in the spatial choice experiment. Scenario Sample sc00 baseline F/H sc01 slope (horizontal gradient) F/H sc02 soil quality (horizontal gradient) F/H sc03 soil quality (vertical gradient) F/H sc04 shape F sc05 increase in size F sc06 wind H sc07 forest F/H sc08 forest F/H sc09 road H sc10 existing hedges H sc11 existing hedges H sc12 existing hedges H The columns present the scenario code and name, along with a graphical illustration, and indicate the sample (wildflower strip (F) or hedgerow (H)) to which the scenario applies. Fig. 2. Choice options in the baseline scenario. The black bar indicates the biodiversity measure in the field. 1 i.e., [scenario code −choice option] sc00-D, sc04-D, sc04-B1, sc06-C/D, sc11-B2. F. Klebl et al. Landscape and Urban Planning 258 (2025) 105325 3 continuous predictors and the log-odds of outcome categories. 2 With a Cox’s distance value of 359.4 at 393 degrees of freedom, the model’s fit was decided to be satisfactory, suggesting a reasonable alignment between predicted probabilities and observed data. Multicollinearity was assessed using a variance inflation factor (VIF) tolerance of <5, and levels were generally very low. 3 Finally, the approximately symmetric distribution of residuals in each model confirmed the reliability and suitability of the model for analysis. The calculation of the log-odds for each choice option relative to the predictor variables was primarily based on the {nnet} package, combined with additional tests requiring the {mlogit} package, such as the Hausman-McFadden test. To determine the significance of individual regression coefficients, a Wald test was performed using the {lmtest} package, through which the zstatistics were obtained for each parameter. The overall relevance of each variable was assessed using a likelihood-ratio test, which is particularly reliable for small sample sizes (Agresti, 2007). 2.3.3. Generalised linear mixed model (GLMM) In order to evaluate the impact of different scenarios on the decisions, choices from each scenario were combined with those from the baseline, and a dummy variable (SCENARIO) was introduced to indicate whether a choice occurred under a specific scenario or the baseline conditions. As each respondent provided data for both the scenario and the baseline, which violates the assumption of independent observations, a GLMM was employed. This approach accounted for repeated measures by incorporating random intercepts for each respondent, thereby controlling for individual variability in the analysis (Agresti, 2007). The logit link function for the log-odds of the selected choice option in a given scenario relative to the baseline is: log(P(Yij =1) 1−P(Yij =1))=β0+β1Xij + μ 0i(3) where Yij is the binary response variable for the individual iand observation j,β0is the fixed intercept, β1is the fixed effect coefficient for the predictor Xij (i.e., SCENARIO), and μ 0iis the random effect for i, which follows a normal distribution μ 0iN(0, σ 2 μ ). The models were developed using the lme4::glmer() function for testing each choice option (n =37) against the variable SCENARIO as a fixed effect, with a random effect added for respondent ID. To ensure consistency with the symmetrical baseline conditions, sub-choices relevant only to specific scenarios were aggregated and standardised (e.g., merging the upper (B1) and lower (B2) edge options in the slope scenario into B). Odds ratios were calculated as described above. 2.4. Limitations and potential bias A challenge inherent to this study was the relatively limited sample size compared to the number of choice options, which led to some options being rarely selected and subsequently excluded from analysis. Moreover, there is a possibility of misinterpretation in the choice experiment, since some farmers assumed factors such as a north–south alignment or region-specific wind direction. When these assumptions were identified through written responses, the observations were removed from the analysis. However, it is likely that some respondents made implicit assumptions without stating them, which could have influenced the results. 3. Results 3.1. Data overview Of the 471 farmers who completed the survey, 55 % were primarily engaged in arable farming, 17 % in meadow management, and 28 % in grazing systems. Farm sizes ranged widely from 0.03 ha to 9200 ha, with a median of 70 ha. The majority of respondents (64 %) were full-time farmers who owned most of the land they worked on. The proportion of organic farms in this sample was disproportionately high at 42 %. A detailed overview of the sample and predictor variables is available in the supplementary material. Decisions on the placement of biodiversity measures showed distinct patterns. Under the baseline conditions (sc00), option A (lengthwise at the edge) was most frequently chosen due to minimal operational disturbance (Fig. 3). This is consistent with placing tramlines lengthwise to avoid obstructing machinery turning at the short edges. Option B (widthwise at the edge) was also popular, as it minimised both disturbance and area used for the measure. Option C (lengthwise in-field) was selected less often, mainly for its potential to promote biodiversity while allowing for lengthwise management. Option D (widthwise in-field) was rarely chosen, typically for dividing the field into separate plots or pasture parcels. The choices across all scenarios are summarised in Fig. A2 in the Appendix A. In the slope scenario (sc01), options B1 (widthwise at the Table 2 Description of the predictor variables included in the multinomial logistic regression models. Variable Description Values ATT_BIODIV attitudes towards biodiversity Likert scale (1–10) FIELDSIZE average field size [num] LU predominant land use of farmland arable land/grassland ORGANIC organic certification binomial REGION regions of cultural, historical, and geophysical similarities relevant for farming West and North Europe (1), former USSR-aligned countries (2), Mediterranean (3) SAMPLE sample allocation flower strip (F)/hedgerow (H) RELIEF topographic relief flat/hilly WORKWIDTH maximum work width of machinery [num] Fig. 3. Relative distribution of farmers’choices for the spatial allocation of hedgerows (H) and wildflower strips (F) in the baseline scenario. The y-axis labels denote the different choice options for each scenario, with graphics representing the biodiversity measure as a black bar in the field. The plot labels show the most frequently cited reasons for each choice. 2 The reference values are based on the baseline scenario but were also evaluated under each scenario. 3 The only exceptions are a moderate collinearity for FIELDSIZE in sc01 (VIF =6.3), and for WORKWIDTH (VIF =5.6) and REGION (VIF =6.9) in sc02. However, this did not substantially affect model performance. F. Klebl et al. Landscape and Urban Planning 258 (2025) 105325 4 higher edge) and B2 (widthwise at the lower edge) were most frequently selected due to erosion control concerns. However, the rationales differed: B1 was aimed at preventing run-off water from entering the field, while B2 was chosen to retain water and nutrients within the field. In the wind scenario (sc06), farmers prioritised placing hedgerows to mitigate wind effects, though some avoided wind barriers to ensure proper drying of hay or leaves, or to facilitate wind pollination. In the soil quality scenarios (sc02 and sc03), measures were generally placed in areas with the lowest soil quality. A similar pattern was observed in the forest scenarios (sc07 and sc08), where proximity to the forest was chosen in consideration of reduced productivity from shading. When integrating the hypothetical field into landscape constellations, many farmers sought to utilise biodiversity habitats as a means of preventing contamination from other fields (sc09-sc12), pollution and noise from roads (sc09), or wildlife entering from the forest (sc07 and sc08). While many farmers intended to increase biodiversity in the forest scenarios, perceptions of the best choice for increasing biodiversity varied. Some preferred options closer to the forest for habitat connectivity, whereas others favoured options farther from the forest, stating that forests already have a high level of biodiversity. In scenarios with existing hedges (sc10-sc12), the most frequently chosen options were those enhancing habitat connectivity and minimising disturbance of field work. It is worth noting that many farmers assumed scenario characteristics that were not provided, such as a north–south alignment. Consequently, several Estonian farmers placed the hedgerows to the north to avoid shading crops, while Spanish farmers positioned them to the south to shield crops from prevailing hot winds. Although these choices were excluded from the analysis, they offer insights into additional influencing factors. 3.2. The role of field and landscape characteristics The scenario characteristics had a significant impact on the farmers’ choices (Table 3). In sc01, the presence of a slope had a pronounced effect, with an exceptionally high odds ratio observed for widthwise options. In the scenarios related to soil quality, there was a notable shift in choices, as farmers preferred to place measures in areas with lower soil quality. An increase in field size was clearly associated with higher odds of subdividing the field. Although the results of the wind scenario should be interpreted with caution, there was a strong indication that the odds of implementing hedges widthwise, perpendicular to the wind direction, were substantially higher compared to the baseline scenario. The presence of forests and roads was found to increase the odds of creating habitats parallel to these features. For example, the odds of planting a hedgerow along the lengthwise edge were 82.3 times higher in the presence of a road. While an existing hedge along the shorter edge of the field (sc10) did not have an significant effect, the opportunity to connect two hedges on neighbouring fields (sc11) increased the odds of choosing the widthwise option by a factor of 3.7. The final scenario (sc12) differed from the others in that farmers were given the additional option of implementing a continuous hedgerow along both edges to connect existing habitats. In this scenario, the odds of selecting the option that deviated from the standard options A and B increased by a factor of 11.7, despite the higher effort required. 3.3. The impact of farmer and farm attributes In the baseline scenario (sc00), the region of the farm, farmers’attitudes towards biodiversity, the sample group (hedgerow or flower strip), and the maximum working width significantly influenced the choices (Table 4). Specifically, opting for alternatives to the favoured choice (lengthwise at the edge) was associated with higher odds for farmers from Eastern (Region 2) and Southern Europe (Region 3), highlighting regional differences in preferences. Farmers with strong biodiversity attitudes had lower odds of choosing option B (widthwise at the edge) but higher odds for in-field option C (lengthwise in-field). Region had the strongest overall effect and was significant across most scenarios (Table 5). Farmers in eastern and southern Europe were more inclined to split plots. For example, in Region 3, the odds of choosing option D (in-field widthwise) in the slope scenario sc01 were 7.7 times higher than in Region 1 (see Table S2 in the supplementary material). These farmers were also less keen on placing measures on low-quality land but had lower odds of selecting options for habitat connectivity. In particular, the odds of choosing the connectivity option E in sc12 were 89 % lower for Region 2 compared to Region 1 (Table S3). Instead, farmers in Regions 2 and 3 preferred the greatest possible distance from existing landscape elements, aiming to distribute the measures more evenly across the landscape. Furthermore, attitudes towards biodiversity had a significant influence on the choices in the majority of scenarios (Table 5). Farmers with strong attitudes had generally higher odds of choosing in-field options and considerably lower odds of selecting options adjacent to existing habitats (sc07, sc08, sc10, sc11). These farmers also demonstrated a greater willingness to dedicate land to biodiversity interventions to facilitate habitat connectivity (sc12). The responses to the open-ended questions revealed a number of distinct rationales for the value placed on the interventions. Flower strips were valued for soil fertility, wildlife support, and aesthetics, whereas hedgerows were appreciated for their effect on wind regimes and barrier functions. In the baseline scenario, the odds of placing a hedgerow within the field were 67 % lower compared to flower strips (Table 4). Additionally, hedgerows were most frequently selected to be at right angles to existing habitats such as forests. Moreover, the size of machinery, expressed as the maximum working width, exerted a considerable influence on the choices under the baseline conditions. In principle, there was a stronger preference for lengthwise options, even within the field, when larger machinery was involved. However, this influence was minimal in other scenarios. Contrary to our initial expectations, the type of land use (arable or grassland) had a negligible impact on the decisions. Nevertheless, the reasons cited in the open-text field suggest that hedges are often planted as fencing or to protect animals from the sun, pollution, or noise. Organic certification appeared to have only a minor effect, and no significant interaction with field size or topographical relief was observed for the regions in which the farmers worked. 4. Discussion By exploring farmers’spatial preferences for biodiversity measures, this study sheds light on how both operational and socio-cultural factors influence their choices. The findings emphasise the necessity of designing conservation strategies that are responsive to regional contexts and farmer priorities, thereby fostering practicability and acceptance among farmers. They also open avenues for future research to investigate how spatially informed approaches can enhance habitat connectivity while addressing multiple trade-offs. 4.1. Discussion of results Previous research has largely concentrated on the role of field, farm, and farmer characteristics on whether agri-environmental measures (AEMs) are adopted. The present study extends this discussion by examining where these measures are most likely to be placed, shifting the focus from the farm or regional scale to the plot level. Although the respective results are not directly comparable, common patterns F. Klebl et al. Landscape and Urban Planning 258 (2025) 105325 5 Table 3 Influence of field and landscape characteristics on farmers’preferences for allocating of linear wildlife habitats at the field scale (N=471). Scenario Choice OR Std. error zvalue Pr(>|z|) sc01 slope H L H L A 0.09 0.27 −8.76 <0.001*** B10.05 0.28 8.34 <0.001*** C0.00 0.99 −10.59 <0.001*** D13033.37 0.88 10.80 <0.001*** sc02 soil quality A  3.95 0.20 6.77 <0.001*** B †     C  0.22 0.41 −3.70 <0.001*** D  4.10 0.82 1.73 0.084 sc03 soil quality A  0.07 0.30 −9.03 <0.001*** B  16.27 0.31 9.12 <0.001*** C  0.00 0.91 −6.39 <0.001*** D †     sc04 shape A  3.69 0.31 4.17 <0.001*** B  0.16 0.63 −2.88 0.004** C  0.00 2.18 −6.96 <0.001*** D  6.36 2.18 0.85 0.395 sc05 size x2 A  0.32 0.31 −3.76 <0.001*** B  1.90 0.58 1.11 0.268 C  109.33 1.64 2.86 0.004** D  6813.31 1.77 4.98 <0.001*** sc06 wind A  0.10 0.33 −7.13 <0.001*** B ‡   11.59 0.23 10.44 <0.001*** C  0.00 4.26 −3.78 <0.001*** D  1.56 0.95 0.47 0.641 sc07 forest A  1.47 0.20 1.89 0.059 B  0.68 0.20 −1.89 0.059 sc08 forest A  0.15 0.22 −8.38 <0.001*** B  6.55 0.22 8.38 <0.001*** sc09 road A  82.27 1.09 4.03 <0.001*** B  0.01 1.10 −4.03 <0.001*** sc10 existing hedges A  1.57 0.28 1.62 0.105 B  0.64 0.28 −1.62 0.105 sc11 existing hedges A  0.27 0.27 −4.76 <0.001*** B  3.69 0.27 4.76 <0.001*** sc12 existing hedges A  0.18 0.29 −5.97 <0.001*** B  0.45 0.54 −1.49 0.135 E11.69 0.40 6.09 <0.001*** Signif. codes: 0 [***] 0.001 [**] 0.01 [*] 0.05 [.] 0.1 [ ] 1. The columns show the standardised choice options, odds ratios for selecting an option in the scenario compared to the baseline, and significance levels obtained from the GLMMs. Sc12 had the additional option to place a hedgerow at both the long and short edges (choice E) to connect existing hedgerows. † The models sc02_B and sc03_D were unidentifiable due to a lack of shared IDs, indicating that the data variability was insufficient to estimate the model parameters. ‡ The model sc06_B encountered boundary singular fits. Applying a log-gamma prior ( α =2, λ=0.001) to the random effects did not resolve the issue, which indicates insufficient data for reliable estimation. F. Klebl et al. Landscape and Urban Planning 258 (2025) 105325 6 emerge, indicating underlying similarities in the factors driving these decisions. The results imply that farmers’decisions regarding the placement of biodiversity measures are an expression of their efforts to balance productivity, operational efficiency, and the promotion of biodiversity through enhanced landscape connectivity. The specific landscape characteristics, machinery size, regional cultural factors, and individual attitudes towards biodiversity strongly influence these decisions. Key field parameters, such as slope, soil quality, field size, and proximity to existing landscape features, further guide the spatial allocation at the field scale. There is a notable tendency for farmers to favour lengthwise placement of biodiversity measures, especially when using large machinery. This preference is consistent with the typical arrangement of tractor tramlines in rectangular fields (Mederle &Bernhardt, 2017), where widthwise obstacles impede the turning of machinery. Lengthwise management is not only agronomically efficient but also offers ecological advantages, as it reduces headlands, where soil disturbance and compaction negatively affect invertebrate abundance (Carlesso et al., 2022). However, the preference for placing wildlife habitats lengthwise along tramlines appears to be counterbalanced by field-specific characteristics in other scenarios. Farmers consistently target areas of lower soil quality for AEM uptake, a pattern supported by earlier studies (Borsotto et al., 2008; Hynes &Garvey, 2009; Paulus et al., 2022; Rois-Díaz et al., 2018; Russi et al., 2016; Wool et al., 2023). Additionally, previous research has shown that concerns about soil erosion are a strong motivator for AEM adoption (Früh-Müller et al., 2019; van Herzele et al., 2013). The present study Table 4 Influence of selected variables on farmers’preferences for allocating linear wildlife habitats in the baseline scenario (n =417). sc00 [baseline] Variable Choice OR Std. error zvalue Pr(>|z|) Pr(>Chisq) (Intercept) B 2.90 0.65 1.65 0.100 C 0.01 1.02 −4.19 <0.001*** ATT_BIODIV B 0.76 0.07 −4.20 <0.001*** <0.001***   C 1.22 0.10 2.00 0.046* FIELDSIZE B 1.00 0.01 −0.26 0.795 0.946   C 1.00 0.01 −0.26 0.799 LU_GRASS B 1.26 0.28 0.81 0.415 0.711   C 1.00 0.01 −0.26 0.799 ORGANIC B 0.91 0.28 −0.32 0.749 0.215   C 1.64 0.31 1.59 0.111 REGION2 B 2.20 0.34 2.33 0.020* 0.003**   C 3.88 0.39 3.46 <0.001*** REGION3 B 1.30 0.33 0.81 0.419   C 2.66 0.42 2.35 0.019* RELIEF_HILL B 1.46 0.27 1.39 0.165 0.217   C 1.52 0.31 1.35 0.176 SAMPLE_H B 0.75 0.26 −1.13 0.257 0.014*   C 0.44 0.30 −2.83 0.005** WORKWIDTH B 0.98 0.02 −0.91 0.364 0.012**   C 1.06 0.02 2.68 0.007** Reference choice: long edge . Signif. codes: 0 [***] 0.001 [**] 0.01 [*] 0.05 [.] 0.1 [ ] 1, based on the Wald test (Pr(>|z|)) for each choice and the likelihood ratio test for the variables (Pr(>Chisq)). Table 5 Result of likelihood ratio tests estimating the influence of parameters in the scenario-specific models (N=471). ATT_BIODIV FIELDSIZE LU ORGANIC REGION RELIEF SAMPLE WORKWIDTH sc00 <0.001*** 0.946 0.711 0.215 0.003** 0.217 0.014* 0.012** sc01 <0.001*** 0.110 0.185 0.122 <0.001*** 0.450 0.022* 0.139 sc02 0.068 0.167 0.633 0.564 <0.001*** 0.717 0.044* 0.090 sc03 0.003** 0.231 0.361 0.286 <0.001*** 0.995 0.044* 0.061 sc04 <0.001*** 0.180 0.002** 0.022* 0.048* 0.214 0.445 sc05 0.002** 0.730 0.637 0.421 0.629 0.985 0.073 sc06 0.884 0.644 0.873 0.127 0.007** 0.206 0.199 sc07 <0.001*** 0.479 0.425 <0.001*** <0.001*** 0.832 <0.001*** 0.574 sc08 0.003** 0.173 0.963 0.007** 0.007** 0.677 0.018* 0.606 sc09 0.036* 0.289 0.958 0.413 0.011* 0.700 0.606 sc10 0.008** 0.328 0.995 0.827 0.063 0.186 0.899 sc11 0.130 0.794 0.696 0.660 0.003** 0.123 0.374 sc12 <0.001*** 0.759 0.239 0.243 0.018* 0.621 0.092 Pr(>Chisq). Signif. codes: 0 [***] 0.001 [**] 0.01 [*] 0.05 [.] 0.1 [ ] 1. F. Klebl et al. Landscape and Urban Planning 258 (2025) 105325 7 underscores the relevance of slopes and the associated risk of soil erosion, which are important factors in the spatial alignment of interventions at the plot level. At the field scale, farmers’hedgerow placement exemplifies how they weigh operational trade-offs against environmental priorities, particularly in response to specific landscape features. Farmers expressed a clear preference for locating interventions near forests, which is in line with Paulus et al. (2022) who observed a greater likelihood of AEM adoption near woody features. Furthermore, when given the opportunity to link hedgerows to existing semi-natural habitats, farmers adjusted their choices for the benefit of biodiversity. This suggests a sensitivity to habitat connectivity and the environmental impact of measures, with the prospect of continuing habitats outweighing the desire to minimise disturbance in management practices. Similarly, Früh-Müller et al. (2019) reported a higher uptake of AEMs in German regions where habitat fragmentation is a pressing concern. Alongside field and landscape attributes, farmers’personal characteristics are central to their decision-making. Numerous studies have demonstrated that positive attitudes towards biodiversity and nature in general increase farmers’willingness to implement biodiversity-friendly farming measures (see Klebl, Feindt, &Piorr, 2024a). This study adds that farmers with strong pro-biodiversity attitudes were more inclined to accept placement options that might complicate plot cultivation, prioritising conservation objectives over ease of operation. The impact of regional specificities as an overarching factor further emphasises the personal and contextual dimension of farmers’behaviour. The absence of collinearity between the predictor variable REGION and the characteristics of farms and farmers indicates that socio-cultural environments exert a considerable influence on farmers’decisions beyond farm structural factors. This may be attributed to regional habits and farming traditions, which significantly shape individual agricultural practices (Pavlis et al., 2016; Rois-Díaz et al., 2018), illustrating the importance of understanding the socio-cultural contexts behind these choices. While these results offer valuable insights into farmers’preferences, it is crucial to recognise the potential disparity between stated preferences and actual behaviour. Previous studies suggest that hypothetical scenarios, such as those applied in this study, can overestimate the participant’s willingness to act (e.g., Brownstone &Small, 2005; De Corte et al., 2021; Urama &Hodge, 2006). Nonetheless, the results reveal clear choice patterns and highlight trade-offs that are relevant in the design of targeted policies and interventions aimed at linking wildlife habitats. 4.2. Policy implications There is a particular need for spatial planning at the landscape scale to enhance habitat connectivity (Pe’er et al., 2020; Pe’er et al., 2022). Establishing a network of continuing hedgerows is considered a viable means of achieving this in European landscapes and should be a focus of policy efforts (Moorhouse et al., 2014; Staley et al., 2023). One key instrument of such policies is the provision of incentives to landowners (Cameron et al., 2022), which could be incorporated into the European Union’s Common Agricultural Policy (CAP). So far, evidence points to the conclusion that the CAP has not been effective in increasing landscape connectivity (Pardo et al., 2020). Despite the critical importance of spatially coordinating linear landscape features, the CAP Strategic Plans of the current legislation (2023–27) do not systematically address the spatial component in its instruments, i.e. the AEMs, now referred to as agri-environment-climate measures (AECMs), Good Agricultural and Environmental Conditions (GAEC), or eco-schemes (European Commission et al., 2023). This means that farmers are usually compensated for performing certain measures but, with some exceptions such as buffer strips, the specific placement within the field is not taken into account. There is potential for the widespread introduction of an agglomeration bonus that rewards additional payments for AECMs when linking a semi-natural habitat to an existing one (Parkhurst et al., 2002), or agglomeration/threshold payments when farmers cooperate to contribute to habitat connectivity (Drechsler et al., 2010; Nguyen et al., 2022; W¨ atzold &Drechsler, 2014). These instruments are relatively easy to implement and offer pragmatic options. Yet, we also advocate for more targeted approaches derived from ecological network planning, which may prove more impactful in connecting specific habitats to complement these broad schemes. The spatial coordination and implementation of planning-based strategies requires active collaboration between a range of stakeholders. Engaging farmers, government and public agencies, and conservation NGOs is central to developing a shared vision and effectively designing ecological networks (Keeley et al., 2018). Fostering cooperation among farmers and communities can further advance these efforts (McKenzie et al., 2013; Pe’er et al., 2022; Westerink et al., 2017), as collaborative initiatives have been shown to substantially enrich farmland biodiversity, notably in terms of butterfly and bird populations, which is largely attributed to an increase in landscape connectivity (Meier et al., 2024). Coordinated collaborative approaches are likely to deliver more significant environmental outcomes than individual actions, but they are associated with higher transaction costs. Although these types of schemes may prove even more cost-efficient at larger scales (Niemi et al., 2024), they must comply with international subsidy regulations set by the World Trade Organization (WTO). In response to this, it has been proposed to redirect financial resources from established management-based schemes to collaborative agri-environmental schemes (McKenzie et al., 2013). Moreover, policy frameworks such as the CAP have been criticised for failing to address structural disincentives to collaboration, including land tenure arrangements that often conflict with the duration of AECMs (Leventon et al., 2017). These misalignments are of particular relevance for permanent landscape features such as hedgerows, where both tenants and landowners need to be involved in conservation agreements. Beyond financial compensation, the creation of platforms for knowledge exchange and coordination between farmers, biodiversity advisors, and policy designers is crucial for evaluating regional sociocultural conditions and navigating farm operational and ecological objectives. Leventon et al. (2019) advocate for governance systems aligned with ecological scales that complement existing structures. This would entail the formation of a landscape-scale decision-making forum comprising diverse stakeholders to collaboratively develop conservation plans for a defined landscape. However, the authors acknowledge that these systems would require a fundamental reorganisation of current power structures and responsibilities. Such platforms can also play a critical role in amplifying attitudes towards biodiversity that relate to farmers’values. The results of this study indicate that environmental values and the concept of landscape connectivity resonate with many farmers across different regions, farm types, and farm sizes, as shown by their choices reflecting concerns about the broader ecological impact of plot-scale interventions. We therefore recommend greater investments in communicating the value of farmers’potential to contribute to ecosystem health and the natural environment in general. While conveying knowledge of ecosystem services derived from biodiversity interventions is important, previous research suggests that addressing farmers’connections to the land, for instance by emphasising the region-specific cultural meaning of hedgerows, can be more effective in securing long-term commitment to biodiversity-enhancing measures (Klebl et al., 2024b). F. Klebl et al. Landscape and Urban Planning 258 (2025) 105325 8 For strategies targeting ecological networks to be successful in terms of both biodiversity conservation and financial efficiency, Mossman et al. (2015) specify the need for sufficient information on the distribution of taxa. This is particularly pertinent in the context of ex-ante assessments designed to identify the most efficient way to connect habitats and to ensure that corridors contribute to the expansion of target species (Beier et al., 2011; Brodie et al., 2016). Although prioritising the shortest distance between fragmented habitats as the most cost-effective solution is a plausible strategy, we emphasise the consideration that the true costs of implementation include compensation payments to landowners or −managers, as also mentioned by Beier et al. (2011) and Mossman et al. (2015). Understanding landowners’motivations and preferences is therefore crucial for cost-effectiveness and practicality of conservation strategies. This stems from the assumption that the greater the deviation of a planned activity from farmers’preferences, the higher the financial compensation required to encourage participation. In this way, the insights of the study could help to estimate the costs of achieving connectivity. Linking woodland on a hill to a forest in a valley, for example, may be more costly than establishing corridors parallel to contour lines due to erosion concerns. Furthermore, corridors along roads are likely to be a low-cost solution, but potential trade-offs such as pollution, noise, and increased roadkill risks must be carefully weighed. The integration of predicted behavioural outcomes of stakeholders into biogeographical and spatial ecological models can be a powerful tool for identifying least-effort strategies and improving fine-grained regional connectivity maps (Beier et al., 2011). Such methods can enable practitioners to target stakeholders who are most likely to implement wildlife corridors on their land. However, Bergsten and Zetterberg (2013) highlighted the lack of a systematic approach in planning ecological networks. Our results could provide one component in developing these strategies. To this end, we propose to incorporate the parameters outlined in Table 6 into spatial network planning. 4.3. Research outlook The practical relevance of the study for ecological network planning reflects the need for research-driven analytical foundations to inform decision-making and implementation. For this purpose, approaches such as agent-based modelling have been recommended to simulate animal movement and behaviour at the human-environment interface, with interdisciplinary approaches being most effective (McLane et al., 2011). By conceptualising farmers and biodiversity as distinct agents, models can estimate the impact of different policy strategies on farmer behaviour and biodiversity outcome (e.g., Dai et al., 2020; Djenontin et al., 2022; Valbuena et al., 2010), and can be parameterised with the variables identified in this study. A potential key question to be addressed through such models is which spatial arrangements of biodiversity corridors enhance habitat connectivity at least cost while meeting farmers’operational needs. This could involve a multi-stage approach: discrete choice experiments to quantify the financial compensation required to implement different corridor scenarios and spatial modelling to simulate cost-effective corridor configurations within a sample region. Beyond evaluating cost-effective arrangements, estimating the ecological impact of different configurations is essential for understanding trade-offs between economic costs and ecological effectiveness, thus helping to determine the most viable options for habitat connectivity. The findings of this study can strengthen both newly developed and existing models of ecological network planning, such as the Paretobased approach proposed by Groot et al. (2010), which balances ecological coherence, landscape character, and implementation and maintenance costs. Integrating social factors with ecological indicators would improve the predictive power of these models for assessing tradeoffs and ecological outcomes. While hedgerows, perennial wildflower strips, and other wildlife corridors are widely recognised for promoting biodiversity, water retention, pollination, soil protection, and soil carbon sequestration (e. g., Holden et al., 2019; Kratschmer et al., 2024; Montgomery et al., 2020; Sutter et al., 2018; Van Vooren et al., 2017), they may act as vectors for the spread of diseases and invasive species (Montgomery et al., 2020), demanding careful planning and management (Wilkerson, 2014). However, in light of the ongoing decline of existing corridors and the growing fragmentation of habitats in Europe (Arnaiz-Schmitz et al., 2018; EEA et al., 2011; Van Den Berge et al., 2019), it remains questionable whether these risks outweigh the substantial benefits of (re) connecting fragmented habitats. Nevertheless, incorporating relevant risks into modelling approaches is recommended to ensure balanced and reliable conservation planning. Such refined models may serve as the basis for the development of practical tools to support decision-making by extension services, government agencies, and practitioners. To further extend their impact, future research could investigate which specific policy incentives deliver the greatest environmental benefits at the lowest cost. This has the potential to inform the design of policy initiatives that are effective and responsive to the needs of farmers. 5. Conclusions This study shows that farmers’decisions on the spatial distribution of biodiversity measures are shaped by their intrinsic motivations and practical considerations. While farmers seek to optimise productivity and minimise disturbance to field work, they also place importance on habitat connectivity. The significant impact of regional factors and local landscape characteristics on their decisions underscores the need for region-specific conservation plans. Understanding farmers’priorities and tailoring efforts to specific socio-cultural practices and regional conditions can increase their acceptance and effectiveness. There is an opportunity for policy schemes, such as those under the CAP, to better integrate interventions that contribute to habitat connectivity. Incentives, including the agglomeration bonus, can encourage farmers to link habitats, but broader, ecologically grounded, and locally adapted solutions are considered to be more fruitful. It is therefore recommended for policy designers to strategically address the spatial aspects of conservation measures to improve landscape connectivity. Effective practical conservation relies on active stakeholder engagement and collaboration among farmers. Collaborative initiatives would benefit from region-specific ecological guidance to enrich Table 6 Parameters proposed for consideration when simulating farmers’decisions to allocate linear wildlife habitats. Parameter Preference Motivation −lengthwise edge ease of management slope widthwise edge reduction of runoff and erosion soil quality low-quality land least impact on yield wind regime windward edge crop protection cardinal directions northern/southern edge climateand crop-specific shading landscape structure connection to habitats/ along roads, forests, fields biodiversity enhancement/ protection from pollution and wildlife regional habitats and traditions region-specific agricultural and aesthetic preferences F. Klebl et al. Landscape and Urban Planning 258 (2025) 105325 9