The influence of human
disturbances on the
spatio-temporal habitat
selection patterns of roe
deer near Trento (Italy)
Matthijs Hinkamp, 11007931
Research proposal Master Earth Sciences
Environmental Management track
Supervisor: Emiel van Loon
Co-assessor: Kenneth Rijsdijk
Date: 29-09-2019
1
Table of content
Introduction ... 2
Theoretical framework ... 4
Relevant characteristics of roe deer ... 4
Human disturbances ... 6
Current status of research ... 6
Behavioural sequences ... 8
Research aim and questions ... 10
Methods ... 11 Research area ... 11 Data ... 12 Tracking data ... 12 Environmental data ... 13 Temporal data ... 14 IM-SAM ... 14 Categorisation ... 15 Temporal analyses ... 17 Real trajectories... 17 Simulated trajectories ... 18 Classification ... 18 Expected results ... 20
Settlements and roads ... 20
Recreational activities ... 22 Hunting season ... 22 Tourism season ... 24 Time schedule... 25 Funding ... 26 Knowledge utilisation ... 27 Data management ... 28 References ... 29
2
Introduction
Human activities often have a great effect on the behaviour of animals. Ungulates are no exception to this rule. Urbanisation and agricultural expansion and intensification have led to diminishing and increasingly fragmented habitats (e.g. Robinson & Sutherland, 2002), which has had two main effects. On the one hand, anthropogenic landscapes, and agriculture in particular, provide high-quality food sources for many ungulates (Abbas et al., 2011), leading to an increase of both overall population density and range in Europe (Apollonio, Andersen & Putman, 2010; Morellet, Klein & Solberg, 2011). On the other hand, there have been decreases in refuge habitats and an increased exposure to human disturbances (Markovchik-Nicholls et al., 2008; Munns, 2006). These disturbances, caused by for example hunting, livestock, logging, recreation and pets, can be a chronic source of stress and can therefore be detrimental to the survival, reproduction and quality of life of wild ungulates (Knight and Gutzwiller, 1995; Tarlow and Blumstein, 2007).
Roe deer (Capreolus capreolus) are no exception to this seemingly paradoxical process of expansion under increasing habitat fragmentation (Danilkin & Hewinson, 1996). While being adapted to woodlands (Hewison et al., 1998) and having a selective feeding strategy (Tixier et al., 1997), roe deer have colonised the agricultural plains of central and western Europe (Cibien et al, 1989; Kaluzinski, 1974). Like many other ungulate species however, roe deer (Capreolus capreolus) are also negatively affected by the increased exposure to human disturbances.
There is a strong interaction between forest management and the health and size of ungulate populations (Martin et al., 2018; Partl et al., 2002). Mismanagement of forests and other habitats can lead to ungulates causing major economic losses in forestry and agriculture (Côté et al., 2004). This is because, like many other ungulates, roe deer are ecosystem engineers (Coté et al., 2004; Earl & Zollner, 2017). Gaining a greater understanding of roe deer behaviour in the modern fragmented landscapes can therefore be of value when developing management strategies (Vospernik & Reimoser, 2007).
Temporal changes in behaviour have to be taken into account when designing such strategies, as roe deer alter their feeding behaviour and general activity based on for example seasonal and daily changes (Bonnot et al., 2012). However, most studies which model animal behaviour focus mostly on the spatial aspect. Temporal differences are currently often analysed by pooling samples into classes, which means the sequential nature of behaviour cannot be taken into account (De Groeve et al., 2019). In order to fill this research gap, De Groeve et al. (2019) developed the Individual Movement Sequence Analysis Method (IM-SAM). This method is an analytical framework based on the analysis of sequences of behaviour, while taking individual variability into account (De Groeve et al., 2019).
3 In the proposed research this new and promising method will be applied to analyse the influence of human disturbances on the behaviour of roe deer near Trento in northern Italy. While it is known that in this area the roe deer populations are affected by anthropogenic pressures from hunting, recreation and expansion of built-up areas and infrastructure (Oberosler et al., 2017; Picardi et al., 2019), the precise spatial and temporal differences and implications are unknown. For the proposed research tracking data from the Fondazione Edmund Mach (FEM) will be used to determine the effect of different sources of human disturbances on the behaviour of roe deer, while taking temporal changes into account. This allows for a detailed investigation of the influence of tourism, the hunting season and the day-night cycles, providing valuable data for the local authorities and the FEM. Moreover, the proposed research will generate more widely applicable behavioural patterns and a specific framework which can easily be applied to perform similar research in other areas.
4
Theoretical framework
Relevant characteristics of roe deer
The European roe deer is one of the most widespread an well-studied ungulate species in Europe. They are ecosystem engineers through their browsing (Côté et al., 2004) and form an important prey species for many carnivores (Aanes et al., 1998). This combination of characteristics makes roe deer an essential part of many ecosystems, but their populations often have to be managed to prevent monetary losses in agriculture and forestry (Cederlund et al., 1998; Côté et al., 2004).
In general, the roe deer is a species which depends on the availability of woodlands for food and cover (Morellet et al., 2011; Pellerin et al., 2010; Ferretti et al., 2011). Their relatively small size and body shape allow them to move through densely vegetated areas (Hansson, 1994). The importance of the availability of patches of woodland seems to grow with an increasing degree of habitat fragmentation (Morellet et al., 2011). The home range of roe deer increases with habitat fragmentation and a network of woodland patches is therefore essential for providing cover in such cases (Cargnelutti et al., 2002).
Due to its small stature, roe deer feed mostly on shrubs and woody and arboreal plants up to 1.2 meters (Duncan et al., 1998). They are considered to be generalist herbivores (Hofmann, 1989), but prefer to feed on high-quality food due to their relatively small rumen (Demment & Van Soest, 1985). This causes most roe deer to feed on agricultural produce during the spring and summer months, despite the increased risk that might pose (Duncan et al., 1998; Hewison et al., 2001). Overall, seasonal changes in diet are limited (Abbas et al., 2011; Cornelis et al., 1999), but feeding behaviour and activity changes significantly throughout the year (Cederlund et al., 1981) as wel as over the diurnal cycle. Roe deer show the most activity during dawn and dusk (Pagon et al., 2013; Krop-Benesch et al., 2013). Foraging in open fields often takes place at night, most likely in order to avoid human activities (Mysterud et al., 1999). Often living in fragmented habitats, most roe deer seem to spend most of the daytime in the available patches of woodland as other habitats provide less cover and movement between patches can be risky (Fig. 1, Bonnot et al., 2012; Pellerin et al., 2010).
5 Fig. 1: Habitat use of roe deer during the day and night. Data and graph retrieved from Bonnot et al., 2012.
Individual roe deer show a high degree of environmental plasticity (Linnell, Duncan & Andersen, 1998) and they have been observed to adapt to a wide range of environments such as shrubland, moorland, coniferous, deciduous and mixed forests, arable land, pastures and even suburban gardens (Linnell, Duncan & Andersen, 1998). Their home range can vary between less than twenty and more than 200 hectares (Morellet et al., 2013). Seasonal migration can occur on an individual or population scale, depending on the circumstances (Cagnacci et al., 2011; Cagnacci et al., 2016; Peters et al., 2017).
In response to a great increase in landscape fragmentation, roe deer have colonised many European agricultural plains (Linnell, Duncan & Andersen, 1998). In fact, they seem to thrive in such fragmented landscapes (Aulak & Babinska-Werka, 1990), especially if there are many ecotonal habitats (e.g. Tufto, Andersen & Linnell, 1996). The number of mature individuals is estimated to have risen to 15 million, with populations inhabiting almost all of Europe (Lovari et al., 2017). Roe deer populations in anthropogenic areas differ significantly in social behaviour and structure from their woodland peers (Bideau et al., 1983; Hewison et al., 1998; Zejda, 1978). In fact, populations that mainly inhabit open fields are considered to be a different ecotype than the forest roe deer (San José et al., 1997). Group sizes of roe deer in open fields are significantly larger than in forests (Liberg
6 et al., 1998; Jepsen & Topping, 2004) and their movement strategies and diet compositions are altered as a response to changes in available cover and the availability of highly nutritional agricultural products (Abbas et al., 2011; Hewison et al., 1998; Tufto et al., 1996;). The two ecotypes can often coexist in the same landscape (Hewison et al., 2001). Despite their changes in behaviour, field roe deer still seem to need at least one patch of woodland in their home range (Lovari & San José, 1997).
Human disturbances
Interactions between predators and prey affect the prey populations directly, but also alter the behavioural responses of prey to perceived predation risks (Creel and Christianson, 2008; Preisser et al., 2005). Predation risk differs across space and time (Valeix et al., 2009), which creates a heterogeneous landscape of fear (Laundré et al., 2001; Searle et al., 2008). Specifically, an increase in (perceived) predation risk can alter the vigilance level (Benhaïem et al., 2008; Jayakody et al., 2008), habitat use (e.g. Heithaus & Dill, 2002), group sizes (Creel & Winnie, 2005), and foraging and movement patterns (Keuling et al., 2008; Webb et al., 2011). However, minimising predation risk often leads to trade-offs in feeding and foraging (Hernández & Laurdré, 2005; Lima & Dill, 1990). This means that the behaviour of prey species is always a balance between costs and benefits of the use of specific habitats (Kie, 1999; Verdolin, 2006). Human disturbances create new (perceived) predation risks which can greatly affect the behaviour of prey. Roe deer are no exception to this rule (Boer et al., 2004).
Current status of research
There has been a lot of research on the influence of disturbances on animals. The Population Consequences of Disturbance (PCoD) conceptual framework, established by Pirotta et al. (2018) provides an overview of the possible research focuses when investigating the impact of (human) disturbances (Fig. 2).
7 Fig. 2: The PCoD conceptual framework, summarising the impacts of disturbances on populations and individuals. Retrieved from Pirotta et al. (2018). The framework assumes ecological stressors, but it also applies to anthropogenic influences (Pirotta et al., 2018).
The proposed research will focus on one specific aspect of the factors relevant in disturbance: the behavioural changes of roe deer as a consequence of human disturbances. It has to be kept in mind that, as is indicated by the framework, linking this aspect to the other sorts of impact is essential when trying to understand the full influence of disturbances. However, before being able to research any consequence of disturbances, the nature of these disturbances (stressor in Fig. 2) must be understood (Pirotta et al., 2018).
The influence of a disturbance on a given population is determined by the aggregate reaction of the exposure of all individuals. Determining the influence on a population level therefore requires knowledge on the fraction of exposed individuals and the amount of exposure of each individual. The individual exposure is determined by the intensity of the disturbance and its duration (Pirotta et al., 2018). The overlap between the disturbance (temporal and spatial) and the population equates to the probability of exposure, which is determined by the characteristics of the disturbance and its propagation in the environment and the animals' behavioural characteristics (Costa et al., 2016; Merchant, Faulkner & Martinez, 2018). These behavioural characteristics have been discussed for roe deer in the previous section.
In general, there seem to be two ways of collecting data concerning the impact of disturbances on animal behaviour. Firstly, visual observations have often been successfully applied to determine changes in activity and occurrence (e.g. Bhattarai & Kindlmann, 2018; French et al., 2018;
8 Mammides et al., 2018). Secondly, the use of tracking data is a common way to gather data on behavioural changes (e.g. Ladle et al., 2018; De Groeve et al., 2018). In contrast to gathering visual observations, this method allows for the efficient extraction of plenty of data on the individual level. For example, tracking data can be used to infer many types of activities such as feeding, resting and fleeing (e.g. Chimienti et al., 2016; Shamoun-Baranes et al., 2012).
Behavioural sequences
While such research has greatly increased our understanding of the behaviour of many animal species, it rarely accounts for the temporal and sequential nature of behaviour. This means that without taking sequences of behaviour into account, only proportions can be established. While different individuals might have similar proportions of activity or habitat use, their actual behaviour might still differ significantly. Analysing behavioural sequences can reveal the more complex temporal aspects of animal behaviour and doing so is therefore essential (Fig. 3, De Groeve, 2018).
Fig. 3: The concept of sequential habitat use. The blue dot and red star indicate the start and end of a week-long trajectory (resolution of six hours per day). The circle indicates the home range of the individual with the trajectory and the bar indicates the habitat use sequence. Retrieved from De Groeve (2018).
In order to make use of the behavioural sequences, the differences between the sequences have to be determined. This can be done through a dissimilarity analysis. This procedure is implemented in many scientific fields and there are therefore various different dissimilarity analyses available (e.g. Magdy et al., 2015).
A method that is commonly used to analyse behavioural sequences is through fitting a Hidden Markov Model (HMM). Such a model makes use of an underlying unobservable state sequence to produce an observable series (Patterson et al., 2007). While this methodology is
9 promising and can be expanded to increase its applicability, a downside is that it needs data with a high and regular temporal resolution (Langrok et al., 2012). The tracking data that will be used for the proposed research does have regular time intervals, but the temporal resolution is probably not high enough for the application of the HMM methodology as the time interval can be six hours (De Groeve et al., 2016). A major advantage of SAM and IM-SAM, which is the method used for the proposed research, is therefore that it can be applied on data with a lower temporal resolution.
10
Research aim and questions
The overall aim of the proposed research is the analysis of the influence of human disturbances on the behaviour of roe deer near Trento (northern Italy). In order to achieve this, IM-SAM will be used, which entails the analysis of behavioural sequences of individual animals. In this case, each segment of these behavioural sequences will be the a category consisting of the distance to a source of human disturbance and the habitat. A complete sequence will therefore provide information about the avoidance of sources of human disturbance and multiple sequences can be used to establish overall behavioural patterns.
The proposed research has a spatial and temporal aspect, each with its own aim. The spatial aspect consists of determining the distance of the roe deer to sources of human disturbances and the mapping and categorisation of these disturbances (built-up areas, roads, footpaths, mountain bike trails). In addition, the availability of highly nutritional food sources such as crops and feeders has to be mapped. The temporal aspect of the proposed research consists of determining the changes in avoidance patterns over time. The aim of which is to establish the influence of each disturbance source in four temporal analyses: generally (using one entire year), during and outside of the hunting season, during and outside of the tourism season (the summer holidays) and the difference between day and night.
Finally, the proposed research also aims to establish the greater applicability of IM-SAM, which seems to be a promising new method for behavioural analysis.
There are several research questions, relating to the mentioned research aims, which will be answered through the proposed research. The main research question, addressing the overall research aim, is:
To what extent do human disturbances influence the habitat selection patterns of roe deer near Trento in northern Italy?
This main question will be answered through multiple sub-questions, relating to the spatial and temporal aspects of the proposed research:
1. To what extent do the habitat selection patterns of the roe deer change during the hunting season?
2. To what extent do the habitat selection patterns of the roe deer change during the main tourism season (summer holidays)?
3. How does the day-night cycle of the disturbance sources influence the habitat selection patterns of the roe deer?
11
Methods
Figure 4 is a schematic overview of the methodology of the entire proposed research. The core research method is IM-SAM, as developed by De Groeve et al. (2019). The adaptations that will be made for the method to fit the research are described below. Spatial data concerning human settlements, roads, trails (hikers and mountain bikers) and roe deer habitat suitability is necessary. By using tracking data and temporal information the different temporal periods, the behavioural reactions to human disturbances can be indicated through the hunting season, tourism season, general behaviour and day-night changes.
Figure 4: A schematic overview of the structure of the proposed research.
Research area
The research area consists of the of the mountainous area to the west of the city of Trento in North-Eastern Italy. More specifically, the area covers the Monte Bondone and Monte Stivo ranges, the Valle dei Laghi and Adige Valley (Fig. 5). The elevation varies between 200 and 2300 meters above sea level, resulting in various climatic conditions ranging from Alpine to semi-Mediterranean. The land cover consists mainly of various types of forest, with agricultural fields, pastures and villages in the valleys. The higher elevation range consists mostly of protected areas, which means that limited human activities take place there (De Groeve et al., 2016).
12 Fig. 5: Overview map of the general research area.
Data
The data that will be used for the proposed research consists of three components. First, there is the tracking data of the roe deer which are to be studied. This is the main dataset which will be used to study the roe deer behaviour through their positioning in the studied landscape. Second, the data which provides spatial context is environmental data concerning the habitat suitability for roe deer, roads, trails and settlements. Third, the temporal component consists of information concerning the hunting season, tourism season and day-night cycles. This temporal component will be used to extract temporally reliant patterns based on the environmental data.
Tracking data
The tracking data which is used for the proposed research will be retrieved from the EURODEER database, which is the EUropean ROe DEER Information System. It is an open database which contains tracking data of roe deer from many populations all over Europe. It is part of the larger EUROMAMMALS project, which aims at gathering tracking data for several mammalian species in Europe. Currently, the database contains tracking data from 27 research groups. The data is stored in a Spatial Relational DataBase Management System, which means that the data is stored in multiple indexed tables that are linked through common fields. The database is built on open source software, mainly PostgreSQL, PostGIS and R and can be accessed through various means such as GIS
13 applications, R, PgAdmin and an online portal, which means that it allows users to efficiently perform complex computations through convenient software (De Groeve, 2018; EURODEER, 2019).
For the proposed research data will be extracted from the EURODEER database for the roe deer population which is tracked by the Fondazione Edmund Mach near Trento (Italy). Specifically, data will be extracted for 26 roe deer the institute is tracking (De Groeve et al., 2016). The sampling period will be 2015-2018, with 4 or 6 hour time-intervals, depending on the availability. Fawns (less than one year) will be excluded from the sample.
Environmental data
The environmental data for the proposed research will consist of habitat data and the presence of the different human disturbance sources. Combining these will provide the categories which will be used to establish the different behavioural patterns. These categories are further discussed in the IM-SAM section.
Habitat
Habitat data is obviously essential for investigating the habitat selection patterns of roe deer. Research has indicated that roe deer choose their preferred habitat partly based on the presence of human disturbances (e.g. Bonnot et al., 2012). The two habitat factors that seem to be the most important are openness and food availability (Aulak & Babinska, 1990; De Groeve et al., 2019), which means that it should be possible to extract this information from the used dataset. The presence of feeding stations might also influence the habitat choice (Ossi et al., 2015; Ossi et al., 2017; Steyaert et al., 2014), so the location and management of these stations also has to be determined.
In principle, data from the CORINE Land Cover (CLC) dataset (version 20, 2018) will be used to determine the general land cover classes for the research area. This dataset has a spatial resolution of 100m and if this turns out to be insufficient, the Tree Cover Density (TCD) data can be used (from 2015, resolution of 20m). Finally, it will be possible to perform semi-automated land use classifications in ERDAS Imagine if more details are needed. Land cover classifications will be performed for the whole research area as well as the home range of each studied individual.
Data on the current presence and management of feeding stations will be retrieved from the EURODEER database (EURODEER, 2017).
Human disturbance sources
Data for the presence of all of the human disturbance sources will for the most part be retrieved from OpenStreetMap (OSM). OSM provides an open database of all roads, settlements, hiking trails and mountain bike trails. A quick visual inspection and comparison with other hiking maps shows that the level of detail is sufficient for the proposed research. There is no differentiation between
14 hiking trails and mountain bike trails, but that is not necessary. In addition, the map also provides information on land cover classes, which may be useful as complementary information to the land cover data mentioned before.
OSM provides a publicly accessible download service, which can be used to extract data for the research area from the OSM database. Data will be extracted for the whole research area and for the home ranges of the studied individual roe deer. The data can be downloaded as OSM data, which can be converted to shapefiles using the OSM Plugin for QGIS. Using GIS software, the zones mentioned in the Categorisation section will be added and the created dataset will be turned into a raster compatible with the environmental data raster dataset.
In addition to the aforementioned anthropogenic disturbances, there are winter sports areas in the research area. Information concerning the location of these areas will be retrieved from OpenSnowMap, an initiative from OpenStreetMap.
Temporal data
Time slots have to be established for hunting season and the tourism peak in order to be able to analyse the effects of these phenomena on the behaviour of the roe deer in the research area. Hunting season starts in the first week of September and ends at the end of November. However, it is important to note that hunting pressure is not steady throughout this period. In fact, the quotas are usually filled by the end of the first couple of weeks (Picardi et al., 2019).
The tourism peak in the area roughly takes places during the summer months of June, July and August (Bimonte & Farhalla, 2016).
In addition, female roe deer can give birth from mid-April through to mid-June. In the first few weeks the fawn remains hidden in one location and is fed by the mother (Van Moorter et al, 2009). This might alter the general home range and therefore affect the results. April, May and June are therefore excluded from all analyses.
IM-SAM
IM-SAM was developed by De Groeve et al. (2019) as an improvement on the Sequence Analysis Method that was developed a few years before. The methodology is quite similar, with all of the major steps in the analytical process being the same. These steps are roughly:
1. Creating trajectories from tracking data and establishing sequences (Fig. 6a).
2. Simulating probable movement patterns and sequences through hypotheses derived from the actual trajectories and sequences (Fig. 6b&c).
15 The Individual Movement part was added to SAM in the second step (Fig. 6b). Instead of simulating trajectories for the general research area, the home range of each studied individual is used. This allows for a more accurate classification of the behavioural patterns because the spatial structure and autocorrelation are directly accounted for. Spatial autocorrelation might still occur when using non-binary classes, which means that substitution weights have to be determined (De Groeve et al., 2019).
Fig. 6: Workflow for IM-SAM with examples. Retrieved from De Groeve et al. (2019). In this example, the trees represent dissimilarities between habitat use sequences with each leaf being a bi-weekly sequence.
Categorisation
In order to establish sequences, spatial categories have to be determined. These categories are a combination of the two types of spatial data: habitat and human disturbances. The research will have the following land use classes: woodland, crops, fields, anthropogenic. It probably would be better to include hedgerows and a difference between high and low crops (Bonnot et al., 2012), but that is not feasible due to the limitations of the CLC dataset. In addition, using too many categories makes the classification increasingly complex (De Groeve et al., 2018). De Groeve et al. (2019) used binary
16 habitat categories: open and closed. This is deemed to be insufficient for this research as the availability of high-nutrient crops might draw roe deer to settlements (Aulak & Babinska, 1990).
The categorisation for human disturbances will in several zones of likelihood of disturbance. The flight distance seems to vary greatly depending on habitat openness, wind direction and hunting activity (Boer et al., 2004; Picardi et al., 2019). Wind direction will not be taken into account, so the average flight distance for open and closed habitats will be used. A conservative flight distance for an area with a hunting season seems to be 50 meters for a closed habitat and 85 meters for an open habitat (Boer et al., 2004). After that first zone, the proposed research will define three more zones: flight distance to 300 meters, 300 meters to 800 meters and 800 meters and above (see Bonnot et al., 2012). Combined, this would lead to the following categories:
Table 1: The different spatial categories for the proposed research.
Flight range Close range Middle range Unaffected
Woodland WF WC WM WU
Feeding station SF SC SM SU
Crops CF CC CM CU
Anthropogenic AF AC AM AU
Fields FF FC FM FU
Note that the woodland category consists of areas with over ten percent tree canopy coverage (see USGS, 2016) The feeding station category consists of woodland areas with a feeding station. The determination of the extent of these areas will have to be performed during the actual research, but will probably be defined as being all woodland within 300 metres of a feeding station. If feed is not provided at any station during certain seasons, the area will be classified as woodland. The crops category is determined by the temporal variations in presence of crops throughout the growing, harvesting and fallow seasons. Anthropogenic areas are roads, winter sports areas and settlements. Trails are often too narrow to be classified as an anthropogenic feature in the landscape. Finally, pastures, fields, fallow agricultural fields and other areas with less than ten percent tree canopy cover as classified as fields.
However, the final number of categories will have to be determined during the actual research. There is a delicate balance between too few and too many categories as both too few and too many categories will result in less useful results. Too few categories means that the results will be an overly simplified version of reality, while too many categories can lead to overly complex
17 classifications of behavioural patterns. The final number of categories will be established during the exploratory analysis of the real trajectories (see Real Trajectories section) (Fig. 5a).
Temporal analyses
From the tracking data, several types of sequences will be established based on the mentioned temporal factors. First of all, there will be a sequence for determining the behaviour during and outside of the hunting season. As mentioned before, the hunting season lasts from the first week of September to the end of November. Most of the hunting pressure takes place during the first few weeks (Pircardi et al., 2019), so creating a sequence including September and October will show the influence of the period with the highest pressure. In order to determine the effects of the tourism peak during the summer holidays, sequences will be established for July and August. September can obviously not be included as that would mean also including a part of the hunting season. Finally, general avoidance patterns and day-night differences will be determined by establishing sequences for February and March as that excludes the summer and Christmas holidays tourism spikes and the hunting season. Note that each of the following steps is performed for all mentioned temporal factors. The specific effects of tourism and hunting will be estimated by comparing the behavioural patterns for these sequences with the ones from the general avoidance sequences.
Table 2: Temporal sequence types for the various temporal analyses of habitat selection behaviour.
Temporal sequence type Months
Tourism July and August
Hunting September and October
General avoidance February and March
Real trajectories
The first step in IM-SAM is creating exploratory trajectories and sequences based on the available tracking data (Fig. 6a). This is done by linking the categories of the categorisation section to the available sampling points in the tracking data dataset. Once each sample point is assigned such a category, behavioural sequences can be created. In order to assign categories to sample points, the land cover map for the whole research area is combined with the tracking data.
The sequences are consequently used to create exploratory data trees. This is done through creating a dissimilarity matrix which contains the dissimilarity between all pairs of sequences (as in De Groeve et al., 2019). In the proposed research, this will be done by using the Hamming distance (HD) algorithm. This algorithm calculates how many character substitutions are necessary to created completely similar sequences. Note that not all substitutions should of equal value and that
18 substitution weights will be added based on the similarity of categories, which corrects for spatial autocorrelation. The exploratory trees can be created from the generated dissimilarity matrices by a hierarchical clustering algorithm (De Groeve et al., 2016; Gabadinho et al., 2011). The exploratory trees are called as such, because they are used to formulate hypotheses concerning the behaviour of the roe deer.
Simulated trajectories
The next step is simulating individual trajectories for the studied roe deer (Fig. 6b). In order to do this, first the home range has to be calculated for each individual, which will be done through creating kernel density estimates based on the available tracking data. Overlaying these minimum convex polygons with the land cover classification map for the entire research area creates a realistic area in which each individual lives (De Groeve et al., 2019).
Consequently, a stochastic movement model will be created to simulate likely behaviour in these home ranges. In order to do this, the model created by De Groeve et al. (2019) will be adapted for this research. The model is based on patterns derived from the exploratory trees and previous research. From these patterns selection rules can be extracted, which can be implemented in the model. For example, the alternating use of closed and open habitat based on the day-night cycle (Bonnot et al., 2012; De Groeve et al., 2019). In addition, varying day lengths have to be accounted for. The starting position for each simulation has to be a random place in the home range and the simulation length and sample times are the same as for the real trajectories.
For each home range, simulations are run for all selection rules and for three selection coefficient ratios (after De Groeve et al., 2019). These ratios account for varying individual behaviour. Finally, each simulation (selection rule + selection coefficient ratio) is replicated 50 times. The total number of simulations therefore is: number of selection rules * 3 (selection coefficient ratios) * 50 (repetitions).
Classification
The final step in the process is the classification of the actual sequences based on the simulated sequences. The first part of this step is applying the HD dissimilarity algorithm to the simulated sequences (Fig. 6b&c). The same substitution weights will be applied here as with the actual sequences (De Groeve et al., 2019).
The resulting dissimilarity trees (one per studied individual roe deer) have the sequences as leaves and nodes which represent behavioural clusters. Classifying this tree into several relevant behavioural clusters therefore means defining a cut-off point. This cut-off point should provide the
19 most robust tree (Hennig et al., 2007) and robustness will be tested through bootstrapping. More specifically, the Jaccard bootstrapping index (BJ) (Hennig et al., 2007; De Groeve et al., 2016) will be determined for all clusters generated using all possible cut-off values on all simulated trees. Next, a combined bootstrapping index will calculated per cut-off point per individual. The following formula will be used (De Groeve et al., 2019; Hennig et al., 2007):
In this formula BJMD is the median BJ for all clusters per cut-off point and the BJIQ the
interquartile range for all clusters per cut-off point. Plotting all BJIQMD values for all cut-off points and
all individuals provides an indication of the best cut-off value (De Groeve et al., 2019). It has to be noted that not necessarily the cut-off point with the highest average BJIQMD is the best cut-off point.
De Groeve et al. (2016) noticed that the highest peak just corresponded to a low cut-off point which only leads to general dissimilarities. Potential inconsistencies with the best cut-off point will be checked by visually assessing all trees (De Groeve et al., 2019).
The resulting tree with an optimal cut-off point will be used to determine which relevant behavioural clusters are present. This will be done through checking characteristics of the clusters using the rules of the stochastic movement model. Once all clusters have been properly identified and tagged, the real sequences can be introduced. The HD algorithm will be used on the combined simulated and real sequences with the determined optimal cut-off point. The simulated sequences are therefore used to classify the real sequences and when leaving only the real sequences in the tree the general actual behavioural patterns can be identified (Fig. 6d).
20
Expected results
While there are still some gaps in the current research concerning the specific spatio-temporal behaviour of roe deer when faced with human disturbances, enough data is available to provide some rough estimations of probable results. In the following sections, the expected results for each proposed analysis are discussed.
Settlements and roads
In general, roe deer avoid being near settlements (Coulon et al., 2008). They also seem to alter their habitat usage when in close proximity to settlements. They tend to avoid open habitats such as meadows and croplands near settlements. If roe deer are in close proximity to a settlement, the probability of woodland habitat use increases (Fig. 4, Bonnot et al., 2012).
Fig. 4: Estimated habitat use probability of different types of habitat during the day at certain distances from settlements. Data and graph retrieved from Bonnot et al., 2012.
Roads seem to have a similar effect on the habitat selection and foraging behaviour of roe deer as the proximity of settlements. As with settlements, roe deer seem to avoid being close to and
21 crossing roads as much as possible (Coulon et al., 2008). They also seem to prefer habitats with more cover when in close proximity to roads (Fig. 5, Bonnot et al., 2012).
Fig. 5: Estimated habitat use probability of different types of habitat during the day at certain distances from roads. Data and graph retrieved from Bonnot et al., 2012.
We expect responses similar to those found in France (Bonnot et al. 2012) in the Italian study sites we included in this research: roe deer will probably prefer to stay away from settlements and roads as much as possible. If roe deer decide to get close to settlements or roads, they will likely prefer a more closed habitat such as woodland (see Bonnot et al., 2012). It is likely that the roe deer will be in the vicinity of settlements and roads more often at night than during the day (Bonnot et al., 2012) and avoidance will probably be greater during hunting (Benhaiem et al., 2008) and tourism season (e.g. Belotti et al., 2012). However, depending on the availability of food resources, it might be the case that the roe deer show a different behaviour. In the end, it seems to depend on the tradeoffs between perceived risk and food availability (Kie, 1999; Verdolin, 2006).
22
Recreational activities
There seems to be less conclusive research on the influence of recreational activities on the behaviour of roe deer. Wyttenbach et al. (2016) showed that mountain biking can scare off roe deer. Interestingly enough, off-trail simulations generated far greater flight distances, which indicates that roe deer can get more or less used to encountering mountain bikes on trails.
Scholten, Moe and Hegland (2017) showed that red deer (Cervus elaphus) attempt to avoid mountain bike trails as much as possible, which makes it reasonable to expect the same from roe deer.
In general, it can be concluded that wildlife is affected by recreational use of wildlife areas. This applies to mountain biking, but also to hiking and for example horseback riding and skiing (Naylor et al., 2009; Stankowich, 2008). Trails of any kind are avoided if possible and otherwise crossed as quickly as possible (Belotti et al., 2012; Rogala et al., 2011). Extensive activities can lead to complete avoidance (Sibbald et al., 2011) and off-trail activities seem to cause greater stress (Reimers & Coleman, 2006), while wildlife can more or less get used to predictable disturbances (Kloppers et al., 2005).
Hunting season
Hunting seems to have a great effect on the flight behaviour of roe deer. The overall vigilance of roe deer increases during hunting season and roe deer often select less risky but also less nutritious feeding sites (Fig. 6, Benhaiem et al., 2008; Bonnot et al., 2012).
23 Fig. 6: Estimated habitat us probability of different types of habitat during the day in and outside of hunting season. Data and graph retrieved from Bonnot et al., 2012.
The flight distance of roe deer has been observed to increase by two-fold in areas where there is a regular hunting regime compared to areas where there is no hunting (Boer et al., 2004). However, the magnitude of this effect depends on the vegetation density and the availability of cover (Boer et al., 2004; Jeppesen, 1987). Areas where hunting has taken place are usually avoided by roe deer for several days (Jeppesen, 1987). In addition, the overall activity and movement decreases significantly in hunted populations (Picardi et al., 2019).
Given the fact that changes in the general behaviour of roe deer have been observed during the hunting season in other areas (e.g. Bonnot et al., 2012), it is likely that the same phenomenon will be observed in the results of the proposed research. Especially avoidance of certain areas due to the presence of hunters (Jeppesen, 1987) and a switch to less-risky food sources (Benhaiem et al., 2008) will probably be observable. This would translate to a general greater avoidance of roads, settlements and trails and a greater preference for closed habitat types during the hunting season. Finally, it is reasonable to assume that hunting causes both a greater occurrence and severity of
24 disturbances, especially in areas where disturbances normally would not take place as much (as hunters probably do not stick to roads and trails). As discussed before, off-trail disturbances might have a much greater effect than regular disturbances (Wyttenbach et al., 2016).
Tourism season
There seems to be no specific data on changes in behavioural patterns during tourism spikes, but it seems reasonable to assume that there will be some changes. First, the number and severity of disturbances seems to be linked to the extent of avoidance of certain areas (Sibbald et al., 2011). For the proposed research, this would mean that avoidance behaviour in general is more common during tourism season and perhaps that more often closed habitats are preferred. The general effects as observed in the proposed research would likely be close to those observed during hunting season. However, in contrast to the hunting season, the increase of disturbances will most likely mainly take place along the current trails, roads and settlements.
25
Time schedule
The whole of the research will take from mid October 2019 to mid July 2020. From October 2019 to May 2020 the research will be performed part-time, with 15 hours per week. During the last two months, the research will be full-time. The full schedule can be seen below, split up into the part-time and full-part-time periods.
Table 3: Part one of the time schedule of the proposed research: the part-time research period. October November December January February March April May
Data collection Data pre-processing Altering scripts Running model Generating results Writing thesis Publicize data and scripts
Table 4: Part two of the time schedule of the proposed research: the full-time research period.
June July Data collection Data pre-processing Altering scripts Running model Generating results Writing thesis Publicize data and scripts
26
Funding
The proposed research will be quite affordable as there are no lab costs, no costs for equipment, excursions and licenses. The time spent by the master student is also free, so the only incurred costs would be that of the time spent on the project by the examiner and assessor. See the following table for an overview.
Table 5: Funding required for the proposed research per month in Euros.
Oct Nov Dec Jan Feb Mar Apr May June July Total
Examiner 100 100 100 100 100 100 100 100 200 200 1200 Assessor - - - 150 - - 200 350 Student - - - - Equipment - - - - - Licenses - - - - Total 100 100 100 100 100 100 250 100 200 400 1550
27
Knowledge utilisation
The proposed research is mainly aimed at expanding the applicability of IM-SAM through the investigation of the influence of human disturbances on the behaviour of roe deer near Trento and thereby providing some more insight into the behaviour of these roe deer. The generated knowledge is useful from a theoretical point of view as it further explores this new and promising method, which allows other researchers in the field of ecology to implement it themselves. The created scripts and specific methodology will be applicable to other regions and other species, which means that it will be valuable for many scientists who are researching the influence of human disturbances on the behaviour of animals. In addition, the generated information could be useful for the local authorities managing the research area as it provides insight into the specific consequences of human activities there. Adjusting the infrastructure to minimise the influence could help managing the populations and thereby creating a more healthy ecosystem.
28
Data management
The data used for the proposed research will be retrieved only from open databases, which means that the research can be easily replicated. In addition, the data and scripts will be published and publicly available. Depending on the size of the tracking data dataset, the queries that were used for the EURODEER dataset will be published instead. All data will either be published on Zenodo or will be publicly available on open databases. The types of data that will be used for and created during the proposed research can be viewed in Table 6.
Table 6: The types of data used for and created through the proposed research.
Type of data Description Format Availability
Tracking
data Tracking data for the studied roe deer, retrieved from the EURODEER database.
.csv Queries for EURODEER and/or cleaned data will be published on Zenodo
Disturbance
sources Point and line data concerning roads, settlements, trails and winter sports areas, retrieved from .
Vector files
Available through OSM
Land cover Land cover maps, probably retrieved from the CLC database. Raster files Available through CLC Maps with research categories
Map of the general research area and for each home range with the categories as described in the Methods section.
Raster files
Will be published on Zenodo.
Habitat
sequences Real and simulated behavioural sequences. .csv Will be published on Zenodo.
Results All data produced by performing the temporal analyses, such as spreadsheets, figures and graphs.
Various Will be published on Zenodo.
Scripts Scripts for the movement model and IM-SAM analyses
R scripts Will be published on Zenodo.
The data and scripts will be published according to the FAIR principles, which means that it must be Findable, Accessible, Interoperable and Re-usable. The data and scripts will be published on one of the common online databases for scientific data and a link will be included in the research. This means that the data will be both findable and accessible. In addition, all scripts will be written in R, which is an open source programme. This means that it is affordable and achievable to replicate the research for anyone who would like to do so.
29
References
Aanes, R., Linnell, J.D.C., Perzanowski, K., Karlsen, J. & Odden, J. (1998). Roe deer as prey. In: Andersen, R., Duncan, P & Linnell, J.D.C. (eds). The European roe deer: the biology of success. Oslo: Scandinavian University Press, 337-372.
Abbas, F., Morellet, N., Hewison, A.J.M., Merlet, J., Cargnelutti, B., Lourtet, B., Angibault, J-M., Daufresne, T., Aulagnier, S. & Verheyden, H. (2011). Landscape fragmentation generates spatial variation of diet composition and quality in a generalist herbivore. Oecologia 167, pp. 401–411.
Apollonio, M., Andersen, R. & Putman, R.J. (2010). Present status and future challenges for European ungulate management. In: Apollonio, M., Andersen, R. & Putman, R. (eds.) European ungulates and their management in the 21st century. Cambridge: Cambridge University Press, pp. 578-604.
Aulak, W. & Babińska-Werka, J. (1990). Use of agricultural habitats by roe deer inhabiting a small forest area. Acta Theriologica 35(1-2), pp. 121-127.
Belotti, E., Heurich, M., Kreisinger, J. ,Šustr, P. & Bufka, L. (2012). Influence oftourism and traffic on the Eurasian lynx hunting activity and dailymovements. Anim Biodivs Conserv 53, pp. 235– 246.
Benhaïem, S., Delon, M., Lourtet, B., Cargnelutti, B., Aulagnier, S., Hewison, A.J.M, Morellet, N. & Verheyden, H. (2008). Hunting increases vigilance levels in roe deer and modifies feeding site selection. Animal Behaviour 76, pp. 611-618.
Bhattarai, B.P. & Kindlmann, P. (2018). Human Disturbance is the Major Determinant of the Habitat and Prey Preference of the Bengal Tiger (Panthera tigris tigris) in the Chitwan National Park, Nepal. European Journal of Ecology 4(1), pp. 13-21.
Bideau, E., Vincent, J.P. & Maire, F. (1983). Evolution saisonnière de la taille des groupes chez le chevreuil en milieu forestier. Revue d'Ecologie 37, pp. 161–169.
Bimonte, S. & Faralla, V. (2016). Does residents'perceived life satisfaction vary with tourist season? A two-step survey in a Mediterranean destination. Tourism Management 55, pp. 199-208. Boer, H.Y. de, Breukelen, L. van, Hootsmans, M.J.M & Wieren, S.E. van. (2004). Flight distance in roe
deer Capreolus capreolus as related to hunting and other factors. Wildlife Biology 10(1), pp. 35-41.
Bonnot, N., Morellet, N., Verheyden, H., Cargnelutti, B., Lourtet, B., Klein, F. & Hewison, A.J.M. (2012). Habitat use under predation risk: hunting, roads and human dwelling influence the spatial behaviour of roe deer. European Journal of Wildlife Research 59, pp. 185-193.
30 Cagnacci, F., Focardi, S., Ghisla, A., Van Moorter, B., Merrill, E.H., Gurarie, E., Heurich, M., Mysterud,
A., Linnell, J., Panzacchi, M. & May, R. (2016). How many routes lead to migration?
Comparison of methods to assess and characterize migratory movements. Journal of Animal Ecology 85(1), pp. 54-68.
Cagnacci, F., Focardi, S., Heurich, M., Stache, A., Hewison, A.J.M., Morellet, N., Kjellander, P., Linnell, J.D., Mysterud, A., Neteler, M. & Delucchi, L. (2011). Partial migration in roe deer: migratory and resident tactics are end points of a behavioural gradient determined by ecological factors. Oikos 120(12), pp. 1790-1802.
Cargnelutti, B., Reby, D., Desneux, L., Angibault, J.M., Joachim, J. & Hewison, A.J.M. (2002). Space use by roe deer in a fragmented landscape. Some preliminary results. Revue d'Ecologie 57, pp. 29-37.
Cederlund, G. (1989). Activity patterns in moose ad roe deer in a north boreal forest. Ecography 12(1), pp. 39-45.
Cederlund, G., Bergqvist, J., Kjellander, P., Gill, R., Gaillard, J.M., Duncan, P., Ballon, P. & Boisaubert, B. (1998). Managing roe deer and their impact on the environment: maximising benefits and minimising costs. In: Andersen, R., Duncan, P & Linnell, J.D.C. (eds.) The European roe deer: the biology of success. Oslo: Scandinavian University Press, 337-372.
Chimienti, M., Cornulier, T., Owen, E., Bolton, M., Davies, I.M., Travis, J.M. & Scott, B.E. (2016). The use of an unsupervised learning approach for characterizing latent behaviors in
accelerometer data. Ecology and evolution 6(3), pp. 727-741.
Cibien, C., Bideau, E., Boisaubert, B. & Maublanc, M.L. (1989). Influence of habitat characteristics on winter social organisation in field roe deer. Acta Theriologica 34, pp. 219–226.
Cornelis, J., Casaer, J. & Hermy, M. (1999). Impact of season, habitat and research techniques on diet composition of roe deer (Capreolus capreolus): a review. Journal of Zoology 248(2), pp. 195-207.
Côté, S.D., Rooney, T.P., Tremblay, J-P., Dussault, C. & Waller, D.M. (2004). Ecological impacts of deer overabundance. Annual Review of Ecology, Evolution, and Systematics 35, pp. 113–147. Costa, D.P., Hückstadt, L.A., Schwartz, L.K., Friedlaender, A.S., Mate, B.R., Zerbini, A.N. ... Gales, N.J.
(2016). Assessing the exposure of animals to acoustic disturbance: Towards an understanding of the population consequences of disturbance. Proceedings of Meetings on Acoustics 27, 010027.
Coulon, A., Morellet, N., Goulard, M., Cargnelutti, B., Angibault, J. & Hewison, A.J.M. (2008). Inferring the effects of landscape structure on roe deer (Capreolus capreolus) movement using a step selection function. Landscape Ecology 23, pp. 603-614.
31 Creel, S. & Christianson, D, (2008). Relationships between direct predation and risk effects. Trends in
Ecology & Evolution 23, pp. 194-201.
Creel, S. & Winnie, J.A. Jr. (2005). Responses of elk herd size to fine-scale spatial and temporal variation in the risk of predation by wolves. Animal Behaviour 69, pp. 1181-1189.
Danilkin, A. & Hewinson, A.J.M. (1996). Behavioural ecology of Siberian and European roe deer. London: Chapman and Hall.
De Groeve, J. (2018). A wildlife journey in space and time: Methodological advancements in the assessment and analysis of spatio-temporal patterns of animal movement across European landscapes. Retrieved from:
https://www.researchgate.net/publication/328851740_A_wildlife_journey_in_space_and_ti
me_Methodological_advancements_in_the_assessment_and_analysis_of_spatio-temporal_patterns_of_animal_movement_across_European_landscapes.
De Groeve, J., Cagnacci, F., Ranc, N., Bonnot, N.C., Gehr, B., Heurich, M., Hewison, A.J.M., Kroeschel, M., Linnell, J.D.C., Morellet, N., Mysterud, A., Sandfort, R., Van De Weghe, N. (2019). Individual Movement - Sequence Analysis Method (IM-SAM): characterizing spatio-temporal patterns of animal habitat use across landscapes. International Journal of Geographical Information Science.
De Groeve, J., Van de Weghe, N., Ranc, N., Neutens, T., Ometto, L., Rota-Stabelli, O. & Cagnacci, F. (2016). Extracting spatio-temporal patterns in animal trajectories: an ecological application of sequence analysis methods. Methods in Ecology and Evolution 7(3), pp. 369-379.
Demment, M.W. & Van Soest, P.J. (1985). A nutritional explanation for body-size patterns of ruminant and non-ruminant herbivores. The American Naturalist 125(5), pp. 641-672.
Duncan, P., Tixier, H., Hofmann, R.R. & Lechner-Doll, M. (1998). Feeding strategies and the physiology of digestion in roe deer. In: Andersen, R., Duncan, P. & Linnell, J.D.C. (eds.) The European roe deer: the biology of success. Oslo: Scandinavian University Press, 91-116.
Earl, J.E. & Zollner, P.A. (2017). Advancing research on animaltransported subsidies by integrating animal movement and ecosystem modeling. Journal of Animal Ecology, 86, pp. 987–997. EURODEER (2019). What is EURODEER. Retrieved from: https://eurodeer.org/what-is-eurodeer/. EURODEER (2017). EURODEER Database Dictionary. Retrieved from:
https://github.com/feurbano/eurodeer_db/blob/master/documentation/eurodeer_db_dicti onary.md.
Ferretti, F., Bertoldi, G., Sforzi, A. & Fattorini, L. (2011). Roe deer and fallow deer: are they compatible neighbours? European Journal of Wildlife Research 57(4), pp. 775-783.
32 French, R.K., Muller, C.G., Chilvers, B.L. & Battley, P.F. (2018). Behavioural consequences of human
disturbance on subantarctic Yellow-eyed Penguins Megadyptes antipodes. Bird Conservation International 29(2), pp. 277-290.
Gabadinho, A., Ritschard, G., Mueller, N.S. & Studer, M. (2011). Analyzing and visualizing state sequences in R with TraMineR. Journal of Statistical Software 40(4), pp. 1-37.
Hansson, L. (1994). Vertrebate distribution relative to clear-cut edges in a boreal forest landscape. Landscape Ecology 9, pp. 105-115.
Heithaus, M.R. & Dill, L.M. (2002). Food availability and tiger shark predation risk influence bottlenose dolphin habitat use. Ecology 83, pp. 480-491.
Hernández, L. & Laundré, J.W. (2005). Foraging in the ‘landscape of fear’ and its implications for habitat use and diet quality of elk Cervus elaphus and bison Bison bison. Wildlife Biology 11, pp. 215–220.
Hewison, A.J.M., Vincent, J.P., Joachim, J., Angibault, J.M., Cargnelutti, B. & Cibien, C. (2001). The effects of woodland fragmentation and human activity on roe deer distribution in agricultural landscapes. Canadian Journal of Zoology 79(4), pp. 679-689.
Hewison, A.J.M., Vincent, J.P. & Reby, D. (1998). Social organisation of European roe deer. In: Andersen, R., Duncan, P. & Linnell, J.D.C. (eds.) The European roe deer: the biology of success. Oslo: Scandinavian University Press. pp. 189–219.
Hofmann, R.R. (1989). Evolutionary steps of ecophysiological adaptation and diversification of ruminants: a comparative view of their digestive system. Oecologia 78, pp. 443-457.
Jayakody, S., Sibbald, A.M., Gordon, I.J. & Lambin, X. (2008). Red deer Cervus elaphus vigilance behavior differs with habitat and type of human disturbance. Wildlife Biology 14, pp. 81-91. Jeppesen, J.L. (1987). The Disturbing Effects of Orienteering and Hunting on Roe Deer (Capreolus
capreolus). Danish Review of Game Biology 13(3).
Jepsen, J.U. & Topping, C.J. (2004). Modelling roe deer (Capreolus capreolus) in a gradient of forest fragmentation: behavioural plasticity and choice of cover. Canadian journal of zoology 82(9), pp. 1528-1541.
Kaluzinski, J. (1974). The occurrence and distribution of field ecotype of roe deer in Poland. Acta Theriologica 19, pp. 291–300.
Keuling, O., Stier, N. & Roth, M. (2008). How does hunting influence activity and spatial usage in wild boar Sus scrofa L.? European Journal of Wildlife Research 54, pp. 729-737.
Kie, J.G. (1999). Optimal foraging and risk of predation: effects on behavior and social structure in ungulates. J Mammal 80, pp. 1114–1129.
Kloppers, E.L., St. Clair, C.C. & Hurd, T.E. (2005). Predator-resembling aversive conditioning for managing habituated wildlife. Ecol Soc 10, pp. 31
33 Knight, R. & Gutzwiller, K. (1995). Wildlife and recreationists: coexistence through management and
research. Washington, DC: Island Press.
Krop-Benesch, A., Berger, A., Hofer, H. & Heurich, M. (2013). Long-term measurement of roe deer (Capreolus capreolus)(Mammalia: Cervidae) activity using two-axis accelerometers in GPS-collars. Italian Journal of Zoology 80(1), pp. 69-81.
Ladle, A., Avgar, T., Wheatley, M., Stenhouse, G.B., Nielsen, S.E. & Boyce, M.S. (2018). Grizzly bear response to spatio-temporal variability in human recreational activity. Journal of Applied Ecology 56, pp. 375-386.
Langrok, R., King, R., Matthiopoulos, J., Thomas, L., Fortin, D., & Morales, J.M. (2012). Flexible and practical modelling of animal telemetry data: hidden Markov models and extensions. Ecology 93(11), pp. 2336-2342.
Laundré, J.W., Hernandez, L. & Altendorf, K.B. (2001). Wolves, elk, and bison: reestablishing the ''landscape of fear'' in Yellowstone National Park, USA. Canadian Journal of Zoology 79, pp. 1401-1409.
Liberg, O., Johansson, A., Andersen, R. & Linnell, J.D.C. (1998). Mating system, mating tactics and the function of male territoriality in roe deer. In: Andersen, R., Duncan, P & Linnell, J.D.C. (eds.) The European roe deer: the biology of success. Oslo: Scandinavian University Press, 221-256. Lima, S.L. & Dill, L.M. (1990). Behavioural decisions made under the risk of predation: a review and
prospectus. Canadian Journal of Zoology 68, pp. 619–640.
Linnell, J.D.C., Duncan, P. & Andersen, R. (1998). The European roe deer: a portait of a successful species. In: Andersen, R., Duncan, P. & Linnell, J.D.C. (eds.) The European roe deer: the biology of success. Oslo: Scandinavian University Press, 11-22.
Lovari, S. & San José, C. (1997). Wood dispersion affects home range size in a monogamous large rodent. Behav. Proc. 99, pp. 42-46.
Lovari, S., Serrao, G. & Mori, E. (2017). Woodland features determine home range size of roe deer. Behavioural Processes 140, pp. 115-120.
Magdy, N., Sakr, M.A., Mostafa, T. & El-Bahnasy, K. (2015). Review on trajectory similarity measures. IEEE Seventh International Conference on Intelligent Computing and Information Systems (ICICIS), pp. 613-619.
Mammides, C., Chen, J., Goodale, U.M., Kotagama, S.W. & Goodale, E. (2018). Measurement of species associations in mixed-species bird flocks across environmental and human disturbance gradients. Ecosphere 9(7), e02324.
Merchant, N.D., Faulkner, R.C., Martinez, R. (2018). Marine noise budgets in practice. Conservation Letters 11, e12420.
34 Markovchick-Nicholls, L., Regan, H.M., Deutschman, D.H., Widyanata, A., Martin, B., Noreke, L., Hunt, T.A. (2008). Relationships between human disturbance and wildlife land use in urban habitat fragments. Conservation Biology 22, pp. 99–109.
Martin, J., Vourc'h, G., Bonnot, N., Cargnelutti, B., Chaval, Y., Lourtet, B., Goulard, M., Hoch, T., Plantard, O., Hewison, A.J.M. & Morellet, N. (2018). Temporal shifts in landscape connectivity for an ecosystem engineer, the roe deer, across a multiple-use landscape. Landscape Ecology 33(6), pp. 937-954.
Morellet, N., Bonenfant, C., Börger, L., Ossi, F., Cagnacci, F., Heurich, M., Kjellander, P., Linnell, J.D., Nicoloso, S., Sustr, P. & Urbano, F. (2013). Seasonality, weather and climate affect home range size in roe deer across a wide latitudinal gradient within Europe. Journal of Animal Ecology 82(6), pp. 1326-1339.
Morellet, N., Klein, F., Solberg, E., Andersen, R. (2011). The census and management of populations of ungulates in Europe. In: Putman R, Apollonio M, Andersen R (eds.) Ungulate management in Europe: problems and practices. Cambridge: Cambridge University Press , pp. 106–143. Munns, W.R. (2006). Assessing risks to wildlife populations from multiple stressors: overview of the
problem and research needs. Ecology and Society 11, 23.
Mysterud, A.., Larsen, P. K., Ims, R. A. & Ostbye, E. (1999). Habitat selection by roe deer and sheep: does habitat ranking reflect resource availability? Canadian Journal of Zoology 77, pp. 776-783.
Naylor, L.M., Wisdom, M.J. & Anthony, R.G. (2009). Behavioral responses ofnorth american elk to recreational activity. Journal of Wildlife Management 73(3), pp. 328–338.
Oberosler, V., Groff, C., Iemma, A., Pedrini & Rovero, F. (2017). The influence of human disturbance on occupancy and activity patterns of mammals in the Italian Alps from systematic camera trapping. (2017). Mammalian Biology 87, pp. 50-61.
Ossi, F., Gaillard, J., Hebblewhite, M. & Cagnacci, F. (2015). Snow sinking depth and forest canopy drive winter resource selection more than supplemental feeding in an alpine population of roe deer. European Journal of Wildlife Resources 61, pp. 111-124.
Ossi, F., Gaillard, J., Hebblewhite, M., Morellet, N., Ranc, N., Sandfort, R., ... Cagnacci, F. (2017). Plastic response by a small cervid to supplemental feeding in winter across a wide environmental gradient. Ecosphere 8(1), e01629.
Pagon, N., Grignolio, S., Pipia, A., Bongi, P., Bertolucci, C. & Apollonio, M., (2013). Seasonal variation of activity patterns in roe deer in a temperate forested area. Chronobiology international 30 (6), pp. 772-785.
Partl, E., Szinovatz, V., Reimoser, F. & Schweiger-Adler, J. (2002). Forest restoration and browsing impact by roe deer. Forest Ecology and Management 159(1-2), pp. 87-100.
35 Patterson, T.A., Thomas, L., Wilcox, C., Ovaskainen, O. & Matthiopoulos, J. (2007). State-space
models of individual animal movement. Trends in Ecology and Evolution 23(2), pp. 87-94. Pellerin, M., Calenge, C., Said, S., Gaillard, J., Fritz, H., Duncan, P. & Van Laere, G. (2010). Habitat use
by female western roe deer (Capreolus capreolus): Influence of resource availability on habitat selection in two contrasting years. Canadian Journal of Zoology 88(11), pp. 1052-1062.
Peters, W., Hebblewhite, M., Mysterud, A., Spitz, D., Focardi, S., Urbano, F., Morellet, N., Heurich, M., Kjellander, P., Linnell, J.D.C. & Cagnacci, F. (2017). Migration in geographic and ecological space by a large herbivore. Ecological Monographs 87(2), pp. 297-320.
Picardi, S., Basille, M., Peters, W., Pociano, J.M., Boitani, L. & Cagnacci, F. (2019). Movement Responses of Roe Deer to Hunting Risk. The Journal of Wildlife Management 83(1), pp. 43-51. Pirotta, E., Booth, C.G., Costa, D.P., Fleishman, E., Kraus, S.D., Lusseau, D., Moretti, D., New, L.F., Schick, R.S., Schwartz, L.K., Simmons, S.E., Thomas, L., Tyack, P.L., Weise, M.J., Wells, R. & Harwood, J. (2018). Understanding the population consequences of disturbance. Ecology and Evolution 8, pp. 9934-9946.
Preisser, E., Bolnik, D. & Benard, M. (2005). Scared to death? The effects of intimidation and consumption in predator-prey interactions. Ecology 86, pp. 501-509.
Reimers, E. & Colman, J.E. (2006). Reindeer and caribou (Rangifer tarandus)response towards human activities. Rangifer 26(2), pp. 55–71.
Robinson, R.A. & Sutherland, W.J. (2002). Post-war changes in arable farming and biodiversity in Great Britain. Journal of Applied Ecology 39, pp. 157–176.
Rogala, J.K., Hebblewhite, M., Whittington, J., White, C.A., Coleshill, J. & Musiani, M. (2011). Human activity differentially redistributes large mammals in the Canadian Rockies National Parks. Ecol Soc 16(3), pp. 16
San Jose, C., Lovari, S. & Ferrari, N. (1997). Grouping in roe deer: an effect of habitat openness or cover distribution? Acta Theriologica 42(2), pp. 235-239.
Scholten, J., Moe, S.R. & Hegland, S.J. (2017). Red deer (Cervus elaphus) avoid mountain biking trails. European Journal of Wildlife Research 64(1).
Searle, K.R., Stokes, C.J & Gordon, I.J. (2008). When foraging and fear meet: using foraging hierarchies to inform assessments of landscapes of fear. Behavioural Ecology 19, pp. 475-482.
Shamoun-Baranes, J., Bom, R., van Loon, E.E., Ens, B.J., Oosterbeek, K. & Bouten, W. (2012). From sensor data to animal behaviour: an oystercatcher example. PloS one 7(5), e37997.
Sibbald, A.M., Hooper, R.J., McLeod, J.E. & Gordon, I.J. (2011). Responses of reddeer (Cervus elaphus) to regular disturbance by hill walkers. Eur JWildl Res 57(4), pp. 817–825.
