Multi-scale analysis of spatial drivers in wild boar space use at a biweekly scale across landscape patterns and step selection

Abstract

Background

Wild boar (Sus scrofa) populations have rapidly expanded across Europe, leading to increased agricultural damage and elevated risks of infectious disease transmission. Understanding how environmental conditions, population density and human activities shape movement across spatial scales is essential for effective management and disease control, particularly because the temporal windows when animals are infectious but asymptomatic align closely with the short-term movement dynamics relevant to transmission processes.

Methods

We analysed GPS tracking data from 99 adult wild boar (16 males, 83 females) in Sweden (2004–2021) to quantify movement at two complementary scales. Biweekly (14-day) space use was estimated using autocorrelated kernel density estimation (AKDE; 95% utilisation distributions) and modelled using a spatial Durbin model to account for both local and neighbouring landscape effects. To identify the behavioural mechanisms underlying these patterns, we applied integrated Step Selection Analysis (iSSA), comparing observed and available movement steps within a conditional logistic regression framework. Key predictors included land cover composition, wild boar density, hunting dog-allowed season, and sex.

Results

Biweekly space use decreased with increasing wild boar density and with higher proportions of forest and open land, indicating spatial contraction in resource-rich or high-density environments. Arable land had no clear local effect but reduced space use through neighbouring areas, highlighting the importance of landscape context. The Durbin model revealed significant spatial dependence, demonstrating that neighbouring conditions contribute to shaping local space use. At the movement scale, iSSA showed that higher density and hunting conditions were associated with reduced step lengths and altered movement behaviour, while individuals consistently selected forest cover over more open habitats. Males exhibited higher movement rates and larger space use than females. Importantly, the reduction in space use observed at the biweekly scale was consistent with density- and disturbance-driven changes in movement behaviour, indicating that spatial contraction emerges from local movement decisions.

Conclusions

By integrating the Durbin model and iSSA, this study links fine-scale movement behaviour to broader patterns of space use. Wild boar adjust their movement and space use in response to density, habitat conditions, and anthropogenic disturbance, with neighbouring landscape conditions contributing to these dynamics. The use of a biweekly temporal scale captures immediate behavioural adjustments within established ranges, providing a relevant framework for understanding temporal shifts in space use and optimising localised wildlife control. This multi-scale approach improves inference on wildlife movement processes and supports spatially targeted management strategies.

Supplementary information

The online version contains supplementary material available at 10.1186/s40462-026-00696-0.

Introduction

Wild boar (Sus scrofa) populations have increased rapidly across Europe in recent decades due to their high adaptability to diverse habitats and the increasing availability of anthropogenic food resources [1–3]. This expansion poses significant ecological and economic challenges, particularly through agricultural damage and the spread of infectious diseases [4–6]. As a result, understanding the spatial behaviour and movement ecology of wild boar has become increasingly important for effective population and landscape management, as well as for informing wildlife disease monitoring such as for African swine fever (ASF) [7, 8]. Consequently, interventions such as wildlife-population control, disease monitoring, and habitat management are essential to mitigate its spread [9–11]. Understanding how movement operates across spatial scales is therefore essential for effective management and disease control. Without linking these scales, the mechanisms through which local habitat conditions and neighbouring landscape contexts jointly shape wild boar spatial dynamics cannot be fully captured.

Wild boar movement is influenced by a combination of environmental and anthropogenic factors. Key drivers of spatial behaviour include land cover composition and configuration (e.g., forest, arable, and open areas) and resource availability [12, 13]. At the same time, human activities such as hunting, supplemental feeding, and changes in land-use can alter both activity and space-use patterns [14, 15]. Wild boar show notable behavioural flexibility, often adjusting their space use in response to shifting resource conditions and human disturbances [16, 17]. However, how environmental and anthropogenic factors jointly shape space use at intermediate temporal scales, and how conditions in surrounding areas influence local behaviour, remains insufficiently understood.

Wild boar frequently move across habitat and administrative boundaries, meaning that conditions in neighbouring areas can influence local space use [14, 18]. Accounting for such indirect spatial effects is therefore essential for ecological interpretation of wild boar space use. For instance, interactions among land cover composition, resource availability, human activities, and temporal dynamics may generate complex space-use patterns that cannot be fully captured by analyses restricted to local conditions or single drivers. Considering these interactions improves our ability to predict wild boar movement responses under diverse ecological and management scenarios, including hunting, supplemental feeding, and habitat change. However, most studies consider only local conditions, limiting inference about how landscape context shapes movement.

Most previous analyses have relied on coarse temporal aggregation, such as annual or seasonal scales, which can obscure important movement processes [19]. Although high-frequency GPS data enable fine-scale analyses, these are often applied to habitat selection at the location level within an individual’s range [20]. Despite extensive research on wild boar movement, studies have typically focused either on broad-scale space use, such as seasonal or annual home ranges [13, 21–23], or on fine-scale habitat selection at the level of individual locations or movement steps [24, 25]. While these approaches capture different aspects of spatial behaviour, they are rarely integrated. As a result, it remains unclear how conditions and processes operating in neighbouring areas shape local space use, and how underlying fine-scale movement decisions contribute to the formation of these landscape-scale spatial patterns. For highly infectious diseases such as ASF, focusing on solely fine-scale encounters may be less informative than understanding broader spatial dynamics, particularly when relocation frequency limits the ability to identify exact contact points [26, 27].

In this study, we focus on an intermediate temporal scale that captures spatial processes relevant to disease transmission. Since wild boar can be infectious several days before showing clinical signs [28–31], they are likely to transmit the virus across the full extent of their space use before mortality occurs. Consequently, the size and configuration of areas used over this period are expected to be more informative for transmission dynamics than fine-scale contact locations alone. This perspective aligns with landscape ecological concepts emphasising the importance of variation among neighbouring areas [32, 33].

To address these challenges, we adopted a multi-scale modelling framework that explicitly links patterns of space use to underlying movement behaviour. At the landscape-scale, which represent the landscape, we used a spatial Durbin model to quantify how environmental and anthropogenic factors within local areas, as well as in neighbouring areas, influence the extent of space use. This approach allowed us to capture indirect spatial effects that are not accounted for in conventional regression approaches or simpler spatial models [34, 35].

To analyse the fine-scale movements, we applied integrated Step Selection Analysis (iSSA) [36] to examine how individuals respond to local conditions through movement decisions. By jointly modelling movement and habitat selection, iSSA provided a mechanistic understanding of how animals move through the landscape. Together these approaches allowed us to evaluate whether large-scale patterns of space use emerge from changes in movement behaviour, thereby linking ecological processes across spatial scales.

In this study, we (i) quantified biweekly (14-day) space use of wild boar in southern and central Sweden, (ii) assessed how environmental and anthropogenic factors within focal and neighbouring areas influence space use using a Durbin model, and (iii) evaluated the movement processes underlying these patterns using iSSA. By linking movement behaviour to spatial extent, we provided a multi-scale perspective on wild boar spatial ecology with direct relevance for disease transmission and management.

Study area

We conducted the study in southern and central Sweden, between approximately 55°–60° N and 12°–18° E, where wild boar populations are established. The distribution range of wild boar in Sweden is primarily south of latitude 62° N, as hunting records indicate that the species is rarely observed north of this limit. Hunting bag statistics from the 2021–2022 season show spatial variation in wild boar density across the study area (Fig. 1). The landscape within the study area is heterogeneous, with boreal forests dominating the north and agricultural land in the south, interspersed with numerous lakes, wetlands, and forest-farmland mosaics. The area spans temperate to boreal vegetation zones and includes mixed and coniferous forests, grasslands, and wetlands [37], which provide varying resource availability and influence the space use of wild boar. The spatial distribution of land cover types is provided in Supplement Figure S1.

Fig. 1.

Spatial distribution of wild boar GPS locations and hunting bag statistics in Sweden. Red open circles represent wild boar GPS locations collected from 2004 to 2021. Background colours show hunting bag during the 2021–2022 season, indicating the number of wild boar shot per 1,000 hectares per year

Data collection and processing

All data cleaning, processing, analysis, spatial modelling, and visualisations were performed in R version 4.4.2 [38] using the packages sf [39, 40], spdep [39, 41], spatialreg [42], data.table [43], ctmm [44], amt [45], terra [46], and ggplot2 [47].

Wild boar tracking data

GPS location data were collected from 99 adult wild boar (16 males, 83 females) in southern and central Sweden between 2004 and 2021 [13, 48, 49]. Animals were captured using different protocols, including immobilization from a hide, capture from a vehicle, and live-trapping, and were fitted with GPS/GSM collars (Vertex Plus 2D/3D, Vertex Lite 2D, or GPS Pro Light 3D; Vectronic Aerospace GmbH, Berlin, Germany) recording locations at irregular intervals, typically every 30–60 minutes. To standardise the dataset and allow comparison across individuals, we retained only locations closest to regular 1-hour intervals. Tracking duration varied among individuals, ranging from 13 to 364 days.

Space use derivation

We estimated the biweekly areas used for each individual using autocorrelated kernel density estimation (AKDE) [50–52]. For every individual, all available GPS locations within each 14-day interval were used to generate a 95% AKDE contour representing the area used. Choosing a biweekly temporal resolution aligns with biologically and management-relevant timescales [53]. This temporal scale is wide enough to provide stable estimates of individual space use and maintain the statistical stationarity required for AKDE modelling, yet fine enough to capture immediate behavioural adjustments within established ranges rather than long-term range shifts [17, 23]. In total, 1,604 biweekly AKDE polygons were produced. To ensure adequate temporal representation and maintain the stationarity required for meaningful area estimation, we excluded observations where the data covered less than 10 days of GPS tracking data [51, 54]. To justify the use of AKDE over this temporal window, we assessed range residency and stationarity for each 14-day interval using continuous-time movement models fitted through variogram-based analysis. The results indicated that the vast majority of intervals (1,364 of 1,386; 98.4%) were best described by range-resident movement processes, predominantly Ornstein–Uhlenbeck Foraging (OUF) anisotropic models (n = 966), followed by OUF and OU anisotropic models and their variants. Only a small proportion of intervals (n = 22; 1.6%) were best described by the independent model, indicating a lack of autocorrelation and range residency. Consequently, 1,386 (86.4%) biweekly AKDE polygons were retained for the final analysis. The number of polygons per individual ranged from 1 to 26, with a mean of 14 (median = 13). Each polygon contained information on area (km²), sex, wild boar density, dog-allowed season, and the proportional composition of the three land cover classes within the area used.

Step selection

For the iSSA, we derived movement paths from the processed GPS tracking data. We identified movement bursts with at least three consecutive locations and calculated step lengths and turning angles from these bursts. We excluded steps with step lengths greater than 10,000 m to remove biologically unrealistic movements likely resulting from GPS errors or collar malfunction rather than true animal movement. For each observed step, we generated 10 random available steps to represent alternative movement options. We then combined observed and available steps into a single dataset to compare used and available movement choices within a step-selection framework. In total, 36,430 observed steps were retained for the analysis (mean = 383.5 steps per individual) and were paired with 364,300 available steps. This resulted in a total dataset of 400,730 observations organised into 36,430 step sets. Each step included information on step length, turning angle, sex, wild boar density, dog-allowed season, and land cover classes, enabling integrated analysis of movement and habitat selection.

Covariates

Wild boar density

We obtained national hunting bag statistics for the 2004–2022 seasons from the Swedish Association for Hunting and Wildlife Management’s Game Monitoring database. The dataset reports the number of wild boar harvested per 1,000 hectares per year within each hunting district. Hunting bag was used as a proxy for local wild boar density [9, 13, 55, 56].

Dog-allowed hunting season

We derived a categorical variable indicating whether hunting with dogs was permitted in each biweekly period (14 days) and district. This variable was included to account for potential effects of hunting practices on wild boar movement.

Interaction between density and dog-allowed hunting season

We additionally constructed an interaction term between wild boar density and the dog-allowed hunting season to capture potential effect modification by hunting practices.

Land cover

We collected land cover data from the Swedish Environmental Protection Agency [37]. The dataset provides nationwide coverage at a spatial resolution of 10 meters and originally included 55 land cover classes. We used the hierarchical structure of the land cover classes to aggregate them into six broader categories to facilitate analysis and reduce dimensionality: forest, wetland, arable land, open land, artificial surfaces, and water. For modelling, we further selected forest, arable land, and open land based on their ecological relevance to wild boar space use [13, 57, 58] to avoid multicollinearity among highly correlated land cover classes.

Sex

Sex (male/female) was included as an individual-level covariate to account for potential differences in movement and space use.

Spatial and movement modelling

To investigate the determinants of space use and movement dynamics, we employed a two-scale modelling approach: (1) the spatial Durbin model to determine drivers of biweekly range size at a landscape-scale, and (2) iSSA to elucidate the fine-scale movement decisions (i.e., underlying mechanisms) driving those spatial patterns.

Spatial Durbin model

We assessed spatial dependence using ordinary least squares (OLS) models and Lagrange Multiplier (LM) diagnostics [59] across alternative spatial-weight matrices (W). The diagnostics indicated the presence of both spatial lag and spatial error dependence, supporting the use of a spatial Durbin model.

The Durbin model incorporated spatial dependence in both the response and the covariates [60]:

where y is the vector of AKDE-derived space use areas, X is the matrix of explanatory variables, and W is the spatial weight matrix. The term Wy captures endogenous indirect spatial effects, while WX represents indirect effects arising from neighbouring covariates. The parameter ρ measures the strength of spatial dependence in the response variable, and θ captures the magnitude of indirect spatial effects associated with neighbouring covariates. Because the spatial-weight matrix explicitly excluded biweekly observations from the same individual, the spatial autoregressive process captured dependence among distinct individuals [61]. Under this specification, within-individual correlation was not propagated through the spatial lag structure, and therefore individual-level random effects were not included. Consequently, inference was conducted at the level of the biweekly space-use patch rather than the individual.

To identify the most appropriate spatial-weight specification, distance thresholds from 1 to 20 km were evaluated. For each candidate W matrix, we fitted the Durbin model and compared model fit using Akaike information criterion (AIC) values and Moran’s I statistics of model residuals [62]. We also examined the spatial lag coefficient P, to assess the magnitude and direction of indirect spatial effects. The optimal distance threshold and final model specification were selected based on a multi-criteria approach, prioritising models with the lowest AIC and non-significant Moran’s I of residuals, while also ensuring stability and ecological interpretability of estimated indirect spatial effects. For the selected Durbin model, we conducted Monte Carlo simulations with 1,000 replications of the Durbin model effect decomposition to estimate direct, indirect, and total effects [34]. We assessed statistical significance at the 5% level (p < 0.05). The full diagnostic workflow is detailed in Supplementary Information 1.

Integrated step selection analysis

The iSSA framework compared observed steps with randomly generated available steps within a conditional logistic regression framework, thereby accounting for both movement processes and habitat selection. The model included movement-related covariates, such as step length and turning angle, together with habitat and environmental predictors. Land cover was incorporated as a categorical variable, with forest as the reference category. To account for context-dependent movement, we included interaction terms among step length and sex, wild boar density, and hunting season. A three-way interaction between step length, wild boar density, and hunting season was also included to assess whether the effect of density on movement behaviour depends on hunting practices. By incorporating these terms, we evaluated how fine-scale movement behaviour and habitat selection vary across seasonal and density conditions, providing a behavioural perspective complementary to the landscape-scale spatial patterns estimated by the Durbin model. Statistical significance was assessed at the 5% level (p < 0.05).

Results

Distance threshold selection

The 7-km distance threshold was selected as the optimal spatial-weight specification (detailed statistics for all candidate thresholds are provided in Supplement Table S1). Among all candidates, the 7-km threshold provided the best model performance, yielding the lowest AIC values (OLS: 16,201.2; Durbin model: 16,135.5). In addition, this threshold maintained a well-connected spatial structure with 79.87% connectivity (mean neighbours = 1.5), ensuring sufficient interaction among spatial units while avoiding excessive network fragmentation. These results suggest that a 7-km scale represents an appropriate and ecologically meaningful scale for capturing spatial dependencies in wild boar biweekly space use.

Across all thresholds, OLS residuals exhibited significant spatial autocorrelation (p < 0.001), confirming the need for spatial modelling. LM diagnostics indicated strong spatial dependence in wild boar biweekly space use, with test statistics for both spatial error and spatial lag reaching their peak at the 7-km threshold (see Supplement Table S2 for detailed statistics). In contrast to the OLS models, which exhibited significant spatial autocorrelation in their residuals (e.g., at 7 km, Moran’s I = 0.218, p < 0.001), the Durbin model reduced spatial dependence in the residuals, which were no longer significant (e.g., at 7 km, Moran’s I = 0.031, p = 0.155). The spatial autoregressive coefficient (ρ) also reached its maximum at 7 km (ρ = 0.233), suggesting that indirect spatial effects on space-use area were strongest at this scale. At thresholds larger than 7 km, ρ declined or stabilized, and AIC values increased, indicating no further improvement in model fit. Comparative model performance and full residual statistics across thresholds are provided in Supplement Table S1.

Based on consistent spatial dependence, adequate network connectivity, the absence of residual spatial autocorrelation, competitive AIC values, and ecological relevance [22, 63, 64], we selected 7 km as the final distance threshold for subsequent analyses.

Estimated effects from the Durbin model

The overall mean biweekly space-use area was 20.31 km² (median = 3.16 km). Males exhibited substantially larger space-use areas (mean = 54.27 km², median = 6.90 km², n = 240) than females (mean = 14.07 km², median = 2.75 km², n = 1,307). Using the 7-km distance threshold, the Durbin model results indicated heterogeneous effects of environmental and anthropogenic covariates on biweekly wild boar space use (Table 1). The spatial autoregression coefficient (ρ) was 0.233 (p < 0.001, 95% CI: 0.19–0.28), confirming the presence of spatial dependence in space use area. Direct effects represent the local influence of a variable within a focal space-use area, while indirect effects reflect spatial influences from neighbouring space-use area. Total effects quantify the overall impact of a variable, combining both local and neighbouring influences.

Table 1.

Estimated direct, indirect, and total effects of explanatory variables on wild boar space-use area (km²) from the Durbin model, based on Monte Carlo simulations with 1,000 replications. Reported values are coefficients (in km²) with corresponding p-values in parentheses

Variable	Direct impact	Indirect impact	Total impact
Arable land	−0.32 (p = 0.348)	−0.76 (p = 0.017)	−1.07 (p = 0.043)
Forest	−0.64 (p = 0.042)	−0.42 (p = 0.002)	−1.06 (p = 0.010)
Open land	−1.63 (p = 0.011)	−0.17 (p = 0.807)	−1.80 (p = 0.098)
Wild boar density	−1.72 (p < 0.001)	0.38 (p = 0.467)	−1.34 (p = 0.001)
Dog-allowed season	0.30 (p = 0.637)	−0.35 (p = 0.656)	−0.05 (p = 0.935)
Wild boar density × Dog-allowed season	−10.04 (p = 0.367)	19.45 (p = 0.175)	9.41 (p = 0.356)
Sex - male	48.48 (p < 0.001)	18.68 (p = 0.075)	67.17 (p < 0.001)

The proportions of forest and open land within a biweekly area exhibited significant negative direct effects on space-use size. In contrast, the direct effect of arable land on biweekly area was not statistically significant. However, regarding indirect spatial effects, both forest and arable land proportions in neighbouring areas had significant negative indirect effects. Notably, the total effect of arable land remained significant, driven primarily by its indirect influence rather than a direct local effect. Open land showed a strong negative direct effect, but its indirect effect was negligible and non-significant.

Higher wild boar density within a biweekly area was associated with a significant negative direct effect on space-use size. While the indirect effect of density in neighbouring areas was positive, it was not statistically significant. The total effect remained significantly negative, indicating that increased density leads to a net reduction in individual space use. Regarding the dog-allowed hunting season, we observed no statistically significant direct, indirect, or total effects on space-use size. Furthermore, the interaction term between wild boar density and the dog-allowed season showed no significant influence.

We observed differences in movement patterns between sexes. Males showed a significant positive direct effect and a positive indirect effect from neighbouring males that approached statistical significance (p = 0.075), resulting in a significantly positive total effect. This indicates that males maintained larger biweekly space use than females, and the presence of males in surrounding areas was associated with a further expansion of male space-use size.

Movement dynamics and habitat selection

The iSSA indicated that both habitat composition and movement-related variables were associated with fine-scale movement paths and resource selection in wild boar (Table 2). Wild boars showed a consistent preference for forest habitats, as indicated by lower selection for arable land, open land, and other land cover types. Movement patterns were further structured by intrinsic behavioural processes, as indicated by the significant effects of step length and turning angle.

Table 2.

Estimated coefficients (β), 95% confidence intervals (CI), and p-values from the iSSA

Variable	Coefficient (β)	95% CI (β)	p-value
Arable land	−0.454	(−0.491, −0.417)	<0.001
Open land	−0.464	(−0.515, −0.413)	<0.001
Other land covers	−0.419	(−0.458, −0.380)	<0.001
Step length	−2.3 × 10−5	(−3.3 × 10−5, −1.3 × 10−5)	<0.001
log(Step length)	0.007	(0.003, 0.011)	0.002
cos(Turning angle)	−0.853	(−0.869, −0.837)	<0.001
Step length × Sex (Male)	6.4 × 10−5	(5.4 × 10−5, 7.4 × 10−5)	<0.001
Step length × Density	3.0 × 10−6	(2.0 × 10−6, 4.0 × 10−6)	<0.001
Step length × Dog-allowed	−2.6 × 10−5	(−4.6 × 10−5, −6.0 × 10−6)	0.009
Step length × Density × Dog-allowed	−4.0 × 10−6	(−6.0 × 10−6, −2.0 × 10−6)	<0.001

Movement velocity (step length) was significantly associated with individual and environmental contexts, although effect sizes were very small, with selection coefficients (β) close to zero. Males showed higher movement rates than females, as indicated by the positive interaction between step length and sex. Increased wild boar density was associated with reduced movement. The presence of hunting dogs also significantly influenced movement patterns, with reduced movement observed during dog-allowed periods. Furthermore, the negative effect of wild boar density on movement was slightly stronger under dog-allowed hunting conditions, although the effect size was very small.

Discussion

The drivers of biweekly wild boar space use operated across multiple spatial scales and were influenced by both landscape composition and population density. Ecological patterns of animal space use emerge from processes operating across different spatial and temporal scales, where fine-scale movement decisions can contribute to broader-scale spatial patterns [65, 66]. By combining a Durbin model with iSSA, we were able to examine how broader-scale patterns of space use emerged from fine-scale movement decisions. Overall, space use was primarily structured by local habitat conditions and density-dependent processes, while neighbouring landscape characteristics contributed additional, although more limited, influences on space-use dynamics.

Landscape effects on biweekly space use

The spatial dependence observed in the Durbin model (ρ = 0.233, p < 0.001) indicates that space use was influenced not only by characteristics within a focal area but also by features of surrounding areas. Although statistically significant, the relatively low magnitude of ρ suggests that indirect spatial effects from neighbouring areas were limited. This modest spatial dependence likely reflects the interaction between wild boar behavioural ecology and the 14-day temporal resolution, whereby space use is primarily structured by immediate landscape conditions, with neighbouring effects playing a secondary role. The iSSA results were broadly consistent with this pattern, although the estimated movement coefficients were small in magnitude. The negative effects of step length and turning angle were consistent with more localised movement behaviour, but these effects should be interpreted cautiously given their small magnitude [67–69].

The observed contraction of space use in forest-dominated areas suggests that forest habitats provide sufficient cover and foraging opportunities within relatively small and spatially restricted areas. Wild boar rely heavily on forested habitats for shelter, and these areas often contain both natural food resources and supplemental feeding sites [49, 70, 71]. When such resources are locally abundant, individuals may meet their energetic needs without expanding their space use [15, 49]. Thus, the observed contraction likely reflects repeated use of resource-rich areas within a limited spatial extent [13, 15, 72]. The presence of a significant neighbouring forest effect further suggests that forest habitats in surrounding areas may also contribute to reduced space-use size, potentially reflecting broader landscape contexts where forest resources are widely available.

Similarly, the space-use contraction associated with arable land likely reflects the high energetic value of concentrated crop resources [13, 73]. While the local direct effect was not statistically significant, the significant negative neighbouring effect suggests that the abundance of agricultural resources in the surrounding matrix is a primary driver of localised movement patterns. This aligns with the iSSA results, where arable land showed lower selection probability compared to forest, indicating that wild boar exploit these areas through selective and efficient foraging movements. When high-quality forage is widely available in adjacent areas, individuals may not need to extend their ranges, leading to more sedentary behaviour. This pattern is consistent with previous studies showing that wild boar exploit crops through short-distance foraging, typically at night, while avoiding extensive movement across exposed areas [13, 16, 74–76].

In contrast, open land exhibited the strongest local contraction effect among all land-cover types, likely reflecting a behavioural trade-off between resource acquisition and perceived risk. Open habitats generally provide less cover, increasing exposure to hunting and predation risk [16, 64, 77]. This suggests that increased perceived risk in open habitats may contribute to more spatially restricted space use rather than extensive use of exposed areas. The absence of a significant neighbouring effect for open land further suggests that the influence of open habitats is primarily associated with local landscape conditions rather than broader landscape context.

Density and hunting effects across spatial scales

Higher wild boar density within a biweekly area was associated with a significant reduction in space-use size. This pattern likely reflects the inverse relationship between population density and individual space requirements [13, 78], where individuals in high-density areas can satisfy their energetic needs within smaller spatial extents due to greater local resource availability [13, 55, 56]. However, fine-scale movement behaviour revealed a different response. The iSSA results showed that higher local density was associated with increased step lengths, suggesting that although overall space use becomes more spatially constrained, movement intensity within these restricted areas increases. This indicates that wild boar concentrate their activity within smaller resource-rich areas while exhibiting more active movement, potentially in response to local competition or resource exploitation.

This contrast between space-use and movement responses highlights the value of combining analyses across spatial scales. While density strongly influenced the extent of area used, movement behaviour within those areas remained dynamic and responsive to local conditions. Such patterns suggest that density-dependent behavioural adjustments can simultaneously constrain broader space-use patterns while increasing movement activity within established areas.

The dog-allowed hunting season did not have a statistically significant effect on biweekly space-use size. This suggests that, at the temporal scale considered here, space use is primarily shaped by habitat conditions and population density. In contrast, the iSSA results indicated that movement behaviour may still respond to hunting-related disturbance. The negative interaction between wild boar density and the dog-allowed season suggests reduced step lengths during periods when hunting dogs are present, indicating more constrained movement under disturbance [77]. However, the effect of hunting dogs cannot be disentangled from other hunting-related activities occurring during the same period. Collectively, these findings suggest that hunting-related disturbance primarily influences fine-scale movement decisions rather than broader patterns of space use.

Sex-specific movement strategies

Males maintained larger space use than females, consistent with previously reported sex differences in wild boar spatial behaviour [13, 79–81]. The positive direct effect observed for males reflects inherently larger space use, likely driven by mating strategies, territorial behaviour, and greater energetic requirements [21, 81]. The model also suggested a potential indirect spatial effect associated with neighbouring males, where the presence of males in surrounding areas was linked to a tendency for increased space use in focal males. This pattern suggests that, beyond the influence of local density, social or competitive interactions among males may contribute to expanded space use [69, 82], potentially reflecting a roaming mating strategy whereby males expand their movements to maximise mating opportunities and encounter multiple female groups in the presence of competitors. However, this association did not reach conventional levels of statistical significance and should therefore be interpreted with caution (p = 0.075).

Limitations and future directions

Some limitations related to the current investigation should be acknowledged. Hunting bag statistics were only available at an annual resolution, whereas wild boar movement was analysed at a biweekly temporal scale. Supplemental feeding was widespread but poorly documented, limiting the ability to quantify its effects. In addition, the GPS dataset included substantially fewer males (n = 16) than females (n = 83). Although sex was included as a covariate in the analyses, the unequal representation of sexes may have reduced the precision of sex-specific estimates. furthermore, the analysis focused on stationary adult individuals and did not include dispersing animals, meaning that the results primarily reflect space-use dynamics within established areas rather than long-distance dispersal movements. Future studies should aim to incorporate temporally finer-scale wild boar density data, more detailed information on supplemental feeding, and movement data covering dispersal phases. Such improvements would enhance behavioural inference and strengthen both ecological and applied management interpretations.

Conclusion

This study demonstrated that wild boar space use emerged from the interaction between movement behaviour and landscape-scale processes. By integrating a spatial Durbin model with iSSA, we showed how local and neighbouring landscape conditions shaped biweekly space-use patterns and linked these patterns to underlying movement processes. This multi-scale approach improved understanding of wild boar spatial ecology by connecting landscape-scale space use with fine-scale movement behaviour. From a disease ecology perspective, these findings highlighted the importance of incorporating spatial behaviour and landscape context into assessments of pathogen transmission dynamics and provided insights for targeted surveillance and spatial risk assessment.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Author contributions

H. Kim led the study and was responsible for the original draft preparation, investigation, conceptualization, data curation, methodology, analysis, visualization, funding acquisition, and project management. S. Widgren contributed to investigation, conceptualization, methodology, and funding acquisition. R. Johansson and T. Rosendal contributed to investigation, conceptualization, and methodology. P. Kjellander, J. Malmsten, J. Månsson, and H. Thurfjell contributed data. All authors participated in the review and editing of the manuscript and approved the final version.

Funding

This study was financial supported by FORMAS (2020-00733) and co-funded by the European Union’s Horizon Europe (101136346) EUPAHW. Views and opinions expressed are however those of the authors only and do not necessarily reflect those of the European Union or the European Research Executive Agency. Neither the European Union nor the granting authority can be held responsible for them. The wild boar GPS collar data were collected as part of previous projects funded by FORMAS (2016-000823), the Swedish Environmental Protection Agency (SEPA), the private foundations Marie-Claire Cronstedts Stiftelse and Karl Erik Önnesjös Stiftelse, and the Swedish Association for Hunting and Wildlife Management (Svenska Jägareförbundet).

Data availability

The data that support the findings of this study are available from the corresponding author upon reasonable request.

Declarations

Ethical approval

All wild boar captures and handling were approved by the Ethical Committee in Animal Research, Umeå, Sweden (permit A 18–04), and Uppsala, Sweden (permit C 80/9, C 77/10, C 5.2.18–2830/16, and 5.8.18–00845/2017), and in compliance with Swedish laws and regulations.

Competing interests

The authors declare no competing interests.