Introduction

As human population growth drives increasing demand for food, land-use expansion and intensification degrade landscape diversity and imperil species’ survival (Firbank et al. 2008; Rockström et al. 2009). In particular, agricultural landscapes—the largest land-use category in Europe (Ramankutty et al. 2008)—have been subject to significant anthropogenic stress over the past decades, characterized by habitat fragmentation, land consolidation, increased field sizes, and declines in crop diversity. This has led to a reduction in species abundance and richness (Pimm and Raven 2000; Benton et al. 2003). Animals living in these intensively-managed agricultural landscapes must not only contend with landscape degradation per se, but also with seasonally changing human-caused disturbances, e.g. the application of fertilizers and pesticides, and the sudden removal of large parts of the biomass during harvest in autumn (Ullmann et al. 2020). To adjust to the consequences of agricultural practices, animals change their behaviour and adjust their movements to survive.

Animal movement is a key life-history trait, which underlies individuals’ survival and fitness, and is influenced by landscape composition and configuration (e.g. Bennett et al. 2006; Jeltsch et al. 2013). For example, habitat fragmentation and reduced habitat diversity in agricultural landscapes lead to a reduced proportion of dispersing animals (Bonte et al. 2006), changes in habitat selection (Smith et al. 2004), and expanding home ranges and daily travel distances (Schai-Braun and Hackländer 2014; Ullmann et al. 2018). Additionally, these animal movement measures depend on seasonal changes in the available vegetation structure (Mayer et al. 2019). However, the behavioural mechanisms underlying changes in movement patterns, i.e. whether travel distances or utilization ranges become larger because the animals flee or forage more, remain largely unexplored. One promising avenue to investigate these underlying processes is through the animals’ daily behaviours. Although daily behaviours of free-ranging animals are increasingly examined (Wilson et al. 2006; Grünewälder et al. 2012; Shamoun-Baranes et al. 2012; Lush et al. 2016; Weterings et al. 2018; Tatler et al. 2021), the influence of landscape composition and configuration, as well as the corresponding seasonal dynamics on the actual animals’ behavioural modes, has rarely been studied.

The composition of a landscape (the variety of land-use types) and its configuration (the spatial arrangement of these land-use patches, as described by Fahrig et al. 2011) can be examined at different scales. For example, at the habitat scale, one can investigate an animal's utilization range and assess the habitat diversity within that range, while at the landscape scale, the overall landscape complexity (i.e., simple versus complex agricultural landscapes) in which the animal resides can be evaluated. Areas of high habitat diversity provide all the resources necessary to satisfy the animals’ needs for food and cover within a small spatial scale (Anderson et al. 2005; Saïd and Servanty 2005), thus shortening the time animals need for travelling between patches and increasing the amount of time available to them, e.g. for resting or social behaviour (Li and Rogers 2004). In contrast, areas of low habitat diversity consist mainly of agricultural crop fields and few other landscape elements, of which only certain patches may provide food, while other, distant ones provide shelter. Areas of low and high habitat diversity might be found in simple and complex landscapes simultaneously, as even simple landscapes still show small areas of high habitat diversity. In general, however, regions with a simple landscape structure consist of large agricultural crop fields that often cover vast areas, sparsely interspersed with small (semi-)natural patches. This may force animals to stay in a patch when encountering barriers, or to travel long distances between habitat patches in meeting their daily requirements (Ullmann et al. 2018). Regions with a complex landscape structure, on the other hand, generally show smaller agricultural fields and a higher crop diversity, where inhospitable areas are easier to circumvent than in simple landscapes, resulting in less travel time for animals. Hence, the decline in landscape complexity and habitat diversity seemingly leads to an increase in movement and corresponding decrease in resting, subsequently leading to higher energy expenditure (Mace and Harvey 1983) and ultimately declining physical condition and individual fitness (Daan et al. 1996).

The European brown hare (Lepus europaeus, hereafter referred to as hare) is a common open-habitat species found in agricultural landscapes throughout most of Europe and parts of Asia, as well as in introduced populations in regions such as Great Britain, South America, and Australia. The breeding season of hares lasts from January to September during which does produce on average three litters (Frylestam 1980). A reproductive pause occurs from October to December in which male hares produce less testosterone and their tests retract into their body (Simeunovič et al. 2000). Since the 1960s, hare populations have declined strongly throughout Europe, primarily due to agricultural intensification and an associated decline in landscape diversity (Smith et al. 2005). Hares in less diverse landscapes are less abundant, have lower survival rates, are lighter, and have smaller litters (Frylestam 1980; Tapper and Barnes 1986). One of the main aspects that might change with declining landscape diversity is resource availability (e.g. food and shelter), which additionally changes throughout the season and influences the hares’ spatial movement behaviour (Mayer et al. 2019). Hares forage on cultivated crops, but a high-quality diet heavily depends on non-cultivated herbs (Reichlin et al. 2006). This requirement may be greater during the breeding time of female hares, when a high energy demand to produce milk competes with the need to find safe hiding places to avoid predators (Valencak et al. 2009; Mayer et al. 2019).

Hares and specifically their daily behavioural modes have been studied with the use of acceleration sensors in the past. For example, Lush et al. (2016) found that hares spend about 25% of their time feeding when they are active. Studd et al. (2019) found that snowshoe hares spend about 50% of their time foraging, but less so during the full moon, when they move more. In a study by Majchrzak et al. (2022), accelerometers revealed that snowshoe hares reduced their foraging time in winter when they were food supplemented. Furthermore, Weterings et al. (2018) showed that forage quality and quantity have a stronger effect on foraging behaviour than predation and competition. However, none of these studies researched the link between the hares’ behavioural modes and the underlying landscape.

Here, we investigate whether changes in hares’ behavioural modes are caused by changes in habitat diversity (i.e. landscape composition) and the size of agricultural crop fields (i.e. landscape configuration). We used GPS telemetry with internal tri-axial accelerometers and subsequent classification based on random forest models (Tatler et al. 2018) to distinguish between the following behavioural modes: resting, moving, foraging, grooming and standing upright on the hind legs (i.e. vigilance behaviour). We defined foraging behaviour as the search for resources and not necessarily the process of feeding. We studied landscape complexity by choosing a structurally simple landscape with large fields in northeast Germany and a structurally complex landscape with small fields in southern Germany. We first calculated the habitat diversity within each hare's monthly utilization range in both the structurally simple and complex landscapes. We then related study area (i.e. the average field size within the monthly utilization range) and habitat diversity to the hares’ behavioural modes and investigated the relevance of seasonal changes in habitat diversity for sex-specific behavioural responses within certain life-history stages (e.g. mating during spring and summer, versus reproductive pause in late autumn and beginning winter).

Specifically, we hypothesize that:

  • Hares in diverse habitats move less and have more time to rest, as most of their resources are found within a small spatial scale. Hares in habitats of low diversity rest less and move more frequently, travelling longer distances, as they must spend more time searching for resources (especially high-quality food items).

  • Hares inhabiting areas with large fields move more than those in areas with small fields, because resources are more dispersed.

  • The behavioural modes of males and females will change throughout the season, with important life-history events like reproduction, and result in increased moving behaviour during mating and more extended resting periods in the non-reproductive period.

Methods

Study area

The study area in southern Germany, representative of a complex agricultural landscape, was characterized by intensive but small-scale agriculture with an average field size of 2.9 ± 0.04 ha (mean ± SE, Fig. 1). The area was covered to 62% by arable land, with the main production cover types: wheat, maize, and grassland (Bayerisches Landesamt für Statistik und Datenverarbeitung 2016). The study area in northeast Germany, representative of a simple agricultural landscape, was characterized by large-scale farming with an average field size of 27.5 ± 1.1 ha (mean ± SE, Fig. 1). The simple landscape consists to 73% of arable land with the following main crop types: wheat, maize, and oilseed rape. Landscape metric calculations were based on shapefile information from Landesvermessung und Geobasisinformation Brandenburg (InVeKoS 2014), Bayerische Vermessungsverwaltung (2014) and Corine Land Cover (European Environment Agency 2018a, b). More details about the two study areas (e.g. differences in temperature and precipitation) can be found in the Supplementary Material S1 and in Ullmann et al. (2018, 2020). Both study areas were classified into 11 different landscape elements: arable land, forest, grassland, loose woody vegetation, parks, quarries, streets, urban areas, water, field paths, and wetlands.

Fig. 1
figure 1

The location of Germany in Europe (upper left panel) and the study areas in northeast Germany and southern Germany (GADM http://gadm.org/). Satellite images (Google maps 2017) show representative extracts of a the simple landscape in northeast Germany and b the complex landscape in southern Germany. Both landscape representations have the same scale (1:12,000)

Full size image

Animal tracking

In spring and summer of 2014 and 2015, hares were caught by driving them into woollen nets (Rühe and Hohmann 2004). During the capture, we weighed the hares, determined their sex, and collared them. We equipped 34 adult hares (simple landscape: 12 males and 5 females, complex landscape: 9 males and 6 females) with GPS collars in both study areas simultaneously (for detailed information and deployment times see Supplementary File S2). Each collar had a weight of 69 g (Model A1, e-obs GmbH, Munich—Germany, www.e-obs.de) and included a tri-axial acceleration sensor (ACC sensor). Acceleration samples were taken every 4 min. Each accelerometer was calibrated before deployment, according to the orders proposed by the manufacturer. This ensured the comparability between the single devices. The ACC sensor was set to sample at 33 Hz, each sampling burst lasted for 3.27 s, resulting in 110 data points per burst per axis. Within 3.27 s, hares can easily conduct more than one or two different behaviours, hence we cut each burst into 1-s intervals and classified the behaviour of each of these 1-s acceleration bursts. However, as the behaviours for each 1-s interval within one burst (3.27 s) are correlated, we only used the first 1-s interval for the analyses to avoid dependency issues in the data set.

From the hourly GPS points, we calculated monthly 99% Kernel utilization distributions with the smoothing parameter href from the adehabitatHR package (Calenge 2006). We used 99%, instead of the usual 95% Kernel distribution because we also wanted to include habitat that the hare used rarely yet might be important for its behaviour and ecology. Furthermore, we used the value of each hourly step length to calculate a median step length for each month. The step length is the distance (in meters) between two consecutive GPS locations. We included step length as a confounding variable to account for (un)explained variance in the model, as the amount of resting or moving a hare conducts might depend on a temporally increased/decreased utilization range or on larger/shorter step length. We selected step length as a confounding variable instead of utilization range size because the two were highly correlated (Pearson's r = 0.72, t-value = 12.65, df = 147), and incorporating step length into our models resulted in higher R2 values compared to using utilization range size. However, we also ran the models with utilization range size to check for significant differences in comparison to step length. Both models showed very similar outcomes (see Supplementary Material S3). We received data from collared hares from April to January for both study years (2014 and 2015). One hare (ID: 3408) was removed from the data set as an outlier as the hare mainly lived in a forest. All tracking and acceleration data were stored at http://www.movebank.org/ (Wikelski and Kays 2015).

Determination of habitat diversity

To determine habitat diversity, we used the Simpson diversity index–emphasizing evenness–as the agricultural landscapes under study are frequently characterized by the dominance of extensive crop fields. To calculate the Simpson diversity index, we used each hares’ monthly utilization ranges and the underlying habitat information on crop fields and landscape element properties (Finder et al. 1999; Su et al. 2014) extracted with the R package vegan (Oksanen et al. 2013). For ‘arable land’, we added the information on the field’s current (monthly mapping) crop type or tillage state, which was included in the monthly Simpson diversity index calculation. Further, we counted the corresponding number of each behavioural mode per month and hare. We used monthly counts to account for shifts in hare behaviour and space use over the course of the year.

Behaviour classification

To classify behaviors, we directly observed four hares and correlated their exhibited behaviors with the corresponding accelerometer output. We then trained a random forest model (Fig. 2) using these observations. We observed two hares in an enclosure and two free-ranging hares, collecting 4351 acceleration bursts. All four hares exhibited all behavioural modes except for vigilance behaviour, which was not sampled for one of the free-ranging hares (Supplementary Material S4, Table S4.1). Each acceleration sample (i.e. each axis within one burst) was used to calculate the following predictors in order to be used in the random forest algorithm (R package randomforest by Liaw and Wiener 2002) for the prediction of all unobserved behaviours: standard deviation, mean, range (maximum value minus minimum value) and the mean of the burst before the current burst (more predictors were tested but the once chosen where the most parsimonious, please see Supplementary S4). We also conducted a cross validation study to test the potential of the random forest algorithm for the classification accuracy, for the results please see Supplementary Material S4.

Fig. 2
figure 2

The accelerometer output of the five different behavioural modes. The x-axis is shown in red, the y-axis in blue and the z-axis in green. Resting shows the normal position of the three axes. During a bout of moving (e.g. running) the hare executes strong movements and shows a wide variation of acceleration. When the hare forages, the head is turned to the ground, therefore the x-axis is located above the other two axes. Grooming shows a similar pattern but usually with less peaks. During vigilance behaviour the hare stands upright on its hindlegs, therefore the y-axis is located above the z-axis

Full size image

We classified the behaviour of hares into five main behavioural modes: resting, foraging, moving, grooming and vigilance behaviour (Fig. 2). We defined resting behaviour as lying down and all kinds of sitting positions that did not include any further movements of the limbs. Foraging behaviour was defined as slowly moving forward and swaying with the head from side to side or up and down. The actual feeding behaviour, however, is part of the behavioural mode ‘resting’, as feeding itself (i.e. sitting still and chewing) is a very subtle behaviour, even below the threshold of ‘fine-scale behaviours’ (Chakravarty et al. 2020) and therefore could not directly be extracted from the accelerometer attached to the animals’ neck. Hence, in our study we use the behavioural mode ‘foraging’ in the sense of searching for resources. We defined the behavioural mode ‘moving’ as all kinds of displacement, like running or hopping from one location to another. Grooming was classified when the animals were licking, scratching, shaking or stretching themselves. Vigilance behaviour in our study was defined as standing up on the hind legs (while the front paws do not touch the ground). However, we disregarded behavioural modes for the analysis when the behaviours were often misclassified. This was the case for grooming and vigilance behaviour (Supplementary Material S4). Thus, for further statistical analyses we only used resting, moving, and foraging behaviours, but did not consider grooming and vigilance.

Statistics

We used linear mixed-effects models (R package lme4; Bates et al. 2014) to test for the influence of the following fixed factors on the three main behavioural modes (resting, moving, and foraging): average field size (i.e. study area, numeric), habitat diversity (Simpson diversity index, numeric), animal sex (male or female, categorical), seasonality (month, numeric) and median monthly step length (numeric). We ran the analyses for each behavioural mode separately to assure normally distributed residuals. Animal ID was used as random intercept in each model. We included two-way interactions between average field size and habitat diversity, average field size and animal sex, habitat diversity and month, habitat diversity and animal sex, animal sex and month. The variable ‘average field size’ was log transformed due to very different field sizes in the two study regions and because the values were right skewed for the locality representing the complex landscape structure. The fixed factors were standardized using the scale function in R. Then they were tested for multi-collinearity according to Zuur et al. (2009) (GVIF was lower than three in all cases, hence we dismissed multi-collinearity) and for non-linear relationships by using generalized additive models [GAMs—package mgcv (Wood 2001)]. The variable ‘month’ was determined to be non-linear. We tested a quadratic, the third and the fourth order polynomial term to shape the curve of the variable and selected the model with the lowest AIC value. For resting and moving behaviour we used the third order polynomial for month, while the model for foraging had the lowest AIC value when month was included as a quadratic term. Scaled residuals were used [R package DHARMa (Hartig 2017)] to test for linearity and temporal autocorrelation. Each global model as well as its corresponding nested submodels were used for model selection (Dochtermann and Jenkins 2011) with an information theoretic approach built into the R package MUMIn (Barton 2013). We averaged all models with delta-AIC values below 7 (Anderson 2008; Bolker et al. 2009). Fixed effects were considered informative when the estimates’ 95% confidence interval excluded zero (Arnold 2010). We also show pseudo R2-values for the global model, represented by marginal (m) and conditional (c) R2- values using MuMIn. Throughout the text, we show estimates and their respective standard errors, while the graphs show estimates and 95% confidence intervals.

We used the monthly counts of each behavioural mode as dependent variable (Table 1) and transformed them to hours per day (within the respective month). Therefore, we multiplied the number of bursts associated with that behaviour by 24 h, and then divided the result by the total number of monthly bursts. As the acceleration data was sampled every 4 min–resulting in 360 bursts per day–over a 30-day period we collected 10,800 bursts per individual hare.

Table 1 Three averaged models analysing the number of monthly counts of resting, moving and foraging behaviour depending on average field size (FS), habitat diversity (SD for Simpson diversity), animal sex (Sex), the third polynomial of month (MM), median step length per month (Med.SL), and their respective interactions, where sex (m) stands for males
Full size table

Results

The overall model accuracy for the random forest was 80%. The precision (true positives/(true positives + false positives) for resting was 0.94, for moving 0.48 and for foraging 0.67 (for further information please see Supplementary S4). The average size of agricultural fields within the hares’ monthly utilization ranges ranged from 1.7 to 64.3 ha and was significantly smaller in the complex landscape (3.3 ± 1.1 ha) than in the simple landscape (23.8 ± 1.2 ha, F1,31 = 124, p < 0.001). Average field size was contained in all three averaged models, with an informative main term for foraging (Table 1). Females rested less and foraged more when the average field sizes were large. When the field sizes were very small (e.g. 1.7 ha) females rested 18.8 ± 0.4 h per day, but when the field sizes were very large (e.g. 64.3 ha) they rested 1.5 h less per day (17.3 ± 0.5 h, Fig. 3). On the other hand, females foraged only 2.7 ± 0.4 h when fields were small and 5.0 ± 0.5 h when the fields were large, i.e. they foraged 46% more in simple landscapes.

Fig. 3
figure 3

The number of hours per day female and male hares spent foraging and resting in relationship to the average agricultural field size in the hares’ monthly utilization range. The slope for male hares is shown in blue and for female hares in red. Please note the logarithmic scale on the x-axis

Full size image

The relationship between habitat diversity and the time hares spent for each of the three behaviours per day changed seasonally (Fig. 4, Table 1). Across both study areas, hares rested more with increasing habitat diversity during the peak breeding-season. In May, for example, resting increased by 15.5% per day from 15.3 ± 0.8 to 18.1 ± 0.4 h per day (from low to high Simpson diversity index values—please note that these values are based on the fitted model, i.e. the actual data for a certain month does not necessarily cover the entire range of the x-axis in Fig. 4). The relationship became less important over the summer and then started to decrease in winter. In December, for example, hares rested for 17.7 ± 0.2 h in areas with low habitat diversity and 16.0 ± 0.3 h in areas with high habitat diversity, which is a relative decrease of 9.6%.

Fig. 4
figure 4

The average hours hares spent resting, moving and foraging per day and season in relation to habitat diversity (measured as Simpson diversity index of landscape elements within 99% Kernel utilization ranges of each month). Both animal sexes and both study areas are pooled together in this graph. Light coloured bands display 95% confidence intervals. Please be aware of the different scales on the y axis and the data distribution for each month on the x-axis

Full size image

Non-resting implies the patterns for the behavioural modes ‘moving’ and ‘foraging’ to show opposing relationships with habitat diversity within their utilization range. Hares moved a lot during spring (e.g. in April 3.5 ± 0.6 h per day) in areas of low habitat diversity compared to areas of high habitat diversity (1.1 ± 0.3 h per day, a relative decrease of 70%). This relationship became less expressed during summer. In winter, hares moved about 40 min less (in December: 0.4 ± 0.2 h per day) in areas of low diversity than hares in areas of high diversity (1.1 ± 0.1 h per day). The relationship between the time spent foraging and habitat diversity was eminent mainly during peak breeding. Hares used up more of their time to search for resources in areas of low diversity (e.g. 5.9 ± 0.7 h per day in May) than in areas of high diversity (3.1 ± 0.4 h per day in May, a relative decrease of 47.5%).

Female and male hares showed strong seasonal differences in their resting and moving behaviours. The resting activity for female hares peaked in June (18.3 ± 0.3 h per day) and was lowest in January (16.5 ± 0.7 h per day, Fig. 5, Table 1). Male hares rested about 17% more per day in late summer (e.g. September: 18.0 ± 0.2 h per day) than in winter (e.g. January: 14.9 ± 0.3 h per day). Furthermore, males rested about an hour less than females during peak breeding in June. Both sexes moved most during spring, but males moved half an hour more than females (female: 1.2 ± 0.2 h, male: 1.7 ± 0.1 h per day). Males moved only a third of their spring levels in early autumn (0.6 ± 0.1 h per day). Daily foraging activity was highest during winter for both sexes (about 5.0 ± 0.3 h per day). In April females spent 3.1 ± 0.6 h per day foraging while males foraged for 4.7 ± 0.3 h per day, a relative difference of 34%.

Fig. 5
figure 5

The average time male (blue) and female (red) hares spent resting (upper panel), moving (middle panel) and foraging (lower panel) per day in each month from April to January. Shaded areas correspond to 95% confidence intervals. All month were used in the model, here only every second month is ticked to ease visual representation. Beware of the different scales on the y axes

Full size image

The confounding variable step length showed that hares rested more when step length decreased and moved more when step length increased. Although the confounding variable was kept in the model for ‘foraging’ the lower confidence interval for step length overlapped zero and was therefore determined as uninformative.

Discussion

In spring and early summer, we found that hares in areas of low habitat diversity were resting less, moving more, and spending more time searching for food than animals in diverse habitats. Habitat quality affects movement behaviour in most mobile animals (e.g. Turner et al. 2001; Saïd and Servanty 2005). Hares, as well as other species—such as roe deer (Capreolus capreolus), red deer (Cervus elaphus), red fox (Vulpes vulpes), wild turkey (Meleagris gallapovo silvestris), and lynx (Lynx lynx)—react to environmental changes by altering their utilization range size (Lucherini and Lovari 1996; Herfindal et al. 2005; Saïd and Servanty 2005; McLoughlin et al. 2007; Marable et al. 2012). This is in line with our results showing that, from April to June (peak breeding Click or tap here to enter text.), hares in diverse habitats rest more and spend less time moving and searching for resources. Breeding and caring for young is strenuous, and drives an increased need for nutrients (Hackländer et al. 2002; Schai-Braun et al. 2015), reflected by female hares’ body condition strongly predicting their subsequent reproductive output (Schai-Braun et al. 2021). Hence, well-fed mother hares in areas with high habitat diversity will have more and heavier offspring. We also showed that female hares are particularly sensitive to the landscape context, as they rested more and foraged less when field sizes were smaller. The female hares in those areas find their daily requirements more easily and thus can rest for up to an hour and a half more per day. Such high importance of habitat quality and overall landscape complexity during the breeding season was also found in other species, such as roe deer (C. capreolus), song sparrow (Melospiza melodia), and the little owl (Athene noctua) (McLoughlin et al. 2007; Germain et al. 2015; Grüebler et al. 2018).

In winter, however, the relationship between the behavioural modes and habitat diversity shifted. During the cold and wet season hares moved more and rested less in areas of high habitat diversity than in areas of low habitat diversity. Concurrently, during winter, areas of low habitat diversity in agricultural landscapes are characterized by vast open areas with low vegetation cover. Hares in areas with low habitat diversity might be unable to move as much as necessary to meet their daily requirements, since limiting activity is crucial for minimizing predation risk (Ellsworth et al. 2016). Additionally, the absence of tall vegetation may lead to challenges in finding thermal shelter (Laiolo 2005; Meichtry-Stier et al. 2018). Hence, areas of high vegetation are especially important as cover for hares during winter (Tapper and Barnes 1986). A study by Mayer et al. (2019) also showed that weather effects on habitat selection in hares are strongest in winter and that resting hares need high vegetation and edge structures during the cold period of the year. Furthermore, Mori et al. (2020) showed that hares are more active in winter when they have the opportunity for cover in their home ranges. Subsequently, choosing a utilization range within high quality habitat, especially in landscapes with large agricultural fields, plays an important role for individual fitness, as it subsequently results in favourable proportions of resting vs. moving and foraging behaviour, depending on the specific requirements for each season. Highly diverse habitats are also directly beneficial for the survival of the offspring (Karp and Gehr 2020; Schai-Braun et al. 2020) and adults are able to allocate more energy into reproduction as other requirements (e.g. food and shelter) are easily available (Trivers 1972; Tieleman et al. 2008). As a result, animals in areas with low habitat diversity or extensive agricultural fields must prioritize energy allocation towards self-maintenance before reproduction, leading to decreased overall abundance in such poorly structured landscapes.

Our behavioural classification model's accuracy of 80% is lower than some other ACC analyses within the literature (Nuijten et al. 2020; Yu and Klaassen 2021). After we removed grooming and vigilance behaviour from the final statistical model, the largest source of error stems from foraging, which was confused with resting or moving behaviour. We are confident that this source of error within the model is unlikely to dramatically change the results of our analysis. We would expect that the error occurred over the entire range of the independent variables (i.e. average field size, Simpson diversity index and month), which would result in a slight upwards or downwards shift (depending on the focal behaviour) of the curve, with none or minimal changes to the corresponding slope. Thus, the statistical significance of our results and the biological effect size are practically unaffected (please see also the Supplementary Material S4).

While hare populations are declining across Europe, some areas have experienced more significant decreases than others. These regions likely underwent extensive land-use changes, with Germany's land-use change over the past 75 years potentially being representative of other European areas [e.g. Poland as demonstrated in Gryz and Krauze-Gryz (2022)]. Hare populations in eastern Germany declined more sharply than in western Germany (Strauß et al. 2008; Deutscher Jagdverband 2019), which may be attributable to the large field sizes established during land consolidation in the German Democratic Republic. These larger fields led to reduced habitat diversity (Batáry et al. 2017), potentially contributing to the decline in hare numbers.

We also demonstrated that increased movement correlates with hares displaying longer step lengths and, consequently, larger utilization ranges. These changes are likely a result of alterations in the proportion of behavioral modes, potentially driven by factors such as habitat quality and landscape configuration, among others. When standing crops are high, hares in areas of small agricultural fields and high habitat quality can cross the field by tunnelling through the dense vegetation (Rühe 1999; Mayer et al. 2018; Ullmann et al. 2018). In contrast, when agricultural fields are large and the standing crops are high, they become impassable barriers to hares (Rühe 1999) and have to be circumvented to get to different foraging grounds, thus increasing the amount hares have to move per unit time.

Conclusion

Our observed changes in the behavioural modes of hares relative to landscape composition and configuration will likely apply to other mobile species. For most animals in agricultural landscapes, land-use intensification decreases habitat quality (Chamberlain et al. 2000; Burel et al. 2004). Ultimately, low habitat diversity compels animals to move more and spend more time foraging, allocating less energy to reproduction. This can lead to decreased fitness and, over time, may contribute to the local extinctions often observed in intensively managed agricultural landscapes (Benton et al. 2003).

Our study demonstrates that habitat diversity influences animal behaviour, particularly during the breeding season as well as in cold and wet seasons. To ensure the long-term survival of wildlife in agricultural areas, it is essential to increase habitat diversity and reduce field sizes. This aim may be reached by e.g. planting wild flower strips (Meichtry-Stier et al. 2014; Sliwinski et al. 2019), increasing the proportion of organic farming (Winqvist et al. 2011; Fischer et al. 2017), splitting fields to make them smaller (Tapper and Barnes 1986; Fahrig et al. 2015; Batáry et al. 2017) and implementing agri-environmental schemes (Fischer et al. 2011).