**4.1 Home range**

The annual average MCP100 home range of all bears of the sample was 213,77 ± 35,8 (SE) and ranged from 58,13 – 362,12 Km2 (Table 1). The mean MCP100 for males (*n*=5) was 271.075 Km2 ± 26.12 and for females (*n*=3) 118.245 Km2 ± 48.85. The male annual home-ranges were significantly larger than female using also the other three estimating methods: 95%MCP, Fixed Kernel Method 95% and Core Areas 50% (Mann–Whitney U test: Z=−-2,236, P=0.025).

Home range sizes of males (*n*=5) differ significantly between all seasons (Friedman test, Monte-Carlo simulation for exact P<0.05). Home range sizes of females (*n*=3) did also differ significantly between all seasons only for MCP100 and FKM (Friedman test, Monte-Carlo

(descriptive variables). Here again we performed a group of diagnostic tests in order to examine the efficiency of the produced rules from the aforementioned analysis. This type of analysis is based on artificial intelligence methods (machine learning techniques principle)

Telemetry data from the twenty two (22) radio-collared bears of the sample have yielded up to **42,849** GPS radiolocations. Part of the sampled radiolocations in relation to the highway

(Vayssieres κ.α. 2000, De'ath & Fabricius 2000, Thuiller κ.α. 2003, Mazaris κ.α. 2006).

alignment and the highway buffer corridor and the study area are shown on map 3.

Map 3. Radiolocations of seven (7) different bears from the sample in the study area

Kernel Method 95% and Core Areas 50% (Mann–Whitney U test: Z=−-2,236, P=0.025).

The annual average MCP100 home range of all bears of the sample was 213,77 ± 35,8 (SE) and ranged from 58,13 – 362,12 Km2 (Table 1). The mean MCP100 for males (*n*=5) was 271.075 Km2 ± 26.12 and for females (*n*=3) 118.245 Km2 ± 48.85. The male annual home-ranges were significantly larger than female using also the other three estimating methods: 95%MCP, Fixed

Home range sizes of males (*n*=5) differ significantly between all seasons (Friedman test, Monte-Carlo simulation for exact P<0.05). Home range sizes of females (*n*=3) did also differ significantly between all seasons only for MCP100 and FKM (Friedman test, Monte-Carlo

**4. Results** 

**4.1 Home range** 

simulation for exact P<0.05). Males home range sizes were significantly larger than females with all estimate methods (Kruskal-Wallis test: MCP100: χ2=11, P=0.012, MCP95: χ2=10.76, P=0.013, FK95: χ2=9.6, P=0.022 and CA50: χ2=9.33, P=0.025).(Giannakopoulos et al.2011). (see also table 1 and 2).

In addition we found that the bears (2 males and 2 females) who kept collars more than one year seemed to maintain the same territories. Moreover the spatial patterns and distribution of home ranges between males and females were delineated in most of the cases by natural barriers and landmarks such as rivers, big streams, county roads and in some cases according with the topographic complexity (Giannakopoulos et al. subm.).

Map 4. Home ranges of 22 bears of the sample versus highways network in the study area


Table 1. Annual home range sizes of GPS collared bears (2007-2009) estimated with (MCP100, MCP95, FKM and CA50) in Northeastern Pindos mountains Greece (n=8)

Data from the above table (1) refer only to the bears of the sample (males and females) that have kept their collar for an entire year cycle. A more analytical presentation of data on seasonal home range sizes on the overall sample are presented in table (2).

Telemetry as a Tool to Study Spatial Behaviour and Patterns of

α b

c d

Brown Bears as Affected by the Newly Constructed Egnatia Highway – N. Pindos - Greece 321

Fig. 1. Distribution of habitat units clusters (from a to d) for male bears in spring. Diagrams from **a** to **d** correspond to different increment units of activity areas (surfaces)

Regarding **potential differences in movement distance and patterns:** we found no statistically significant differences in dispersal patterns of bears with respect to the time (hourdaytime/night time) of activity and/or distance from the highway. Analyses showed no statistically significant differences between the maximum and minimum distances travelled by male and female individuals during the day or night (in most of the cases up to P>0.05) in relation to the distance from the highway. Nevertheless significant differences were observed in specific cases when bears movements were studied individually and seasonally but even in these cases there was not enough evidence of a specific pattern regarding spatial behavior of the bears versus the distance of the highway. No significant differences were observed between the average and maximum distances travelled by bears in relation to their distances from the highway under construction. Similarly, we found only limited evidence to support an effect of the highway upon bears movement angles when approaching the highway corridor.

Regarding **habitat suitability** analyses in relation to bear presence and habitat use: *distance from highway* was recognized as one of the statistically significant variables affecting both

**4.3 Potential differences in movement distance and patterns** 

within the home range

**4.4 Habitat suitability** 


From map (4) we observe a high level of home range overlap among most bears. Fifty-nine of the 82 possible pairings of bears indicated overlapping areas according to the MCP95 and FK methods. For areas of high intensity of use (CA50) 40.24% pairings of bears indicated overlapping areas (Giannakopoulos et al. subm.).

Table 2. Seasonal Home ranges (km2) for Brown bears (n=20) in Northeastern Pindos mountain range in Greece, 2007-2009

#### **4.2 Potential changes in habitat use amplitude and range**

Regarding the **potential changes in habitat use amplitude and range (surface units)** in relation to the distance from the highway, the results of our analysis demonstrated that the size of the habitat units (within spring and summer male bears home range) significantly increased with the distance from the highway while their number (of used habitat units) decreased as the distance from highway increased.

The differences between the number of habitat units and their size (surface) used inside the home range in relation to distance from the highway were statistically significant in all cases of males bears in spring and summer (spring: F=5.419, P<0.01; summer: F=6.52, P<0.01) and for females in summer (F=18.735; P<0.01). An example of this differentiation is given on fig.1 in the case of the male bears of the sample.

More specifically in the case of all male individuals of the sample the number of used habitat units (perceived through clustered radiolocations) is significantly higher as their surface size decreases and subsequently their distance from the highway decreases as well (spring: x2=96.63, P<0.01; summer: x2= 20.204, P<0.01). This means that larger ranges in surface and limited distinct ranges (in clustered radiolocations) were observed as the distance from the highway increases.

Fig. 1. Distribution of habitat units clusters (from a to d) for male bears in spring. Diagrams from **a** to **d** correspond to different increment units of activity areas (surfaces) within the home range

### **4.3 Potential differences in movement distance and patterns**

Regarding **potential differences in movement distance and patterns:** we found no statistically significant differences in dispersal patterns of bears with respect to the time (hourdaytime/night time) of activity and/or distance from the highway. Analyses showed no statistically significant differences between the maximum and minimum distances travelled by male and female individuals during the day or night (in most of the cases up to P>0.05) in relation to the distance from the highway. Nevertheless significant differences were observed in specific cases when bears movements were studied individually and seasonally but even in these cases there was not enough evidence of a specific pattern regarding spatial behavior of the bears versus the distance of the highway. No significant differences were observed between the average and maximum distances travelled by bears in relation to their distances from the highway under construction. Similarly, we found only limited evidence to support an effect of the highway upon bears movement angles when approaching the highway corridor.

### **4.4 Habitat suitability**

320 Modern Telemetry

From map (4) we observe a high level of home range overlap among most bears. Fifty-nine of the 82 possible pairings of bears indicated overlapping areas according to the MCP95 and FK methods. For areas of high intensity of use (CA50) 40.24% pairings of bears indicated

> **Summer MCP100**

**Autumn MCP100**

**Winter MCP100** 

**MCP100**

MELIS MALE 159,641 251,34 178,094 - AL PATSINO MALE 34,657 - - - KOYTALAINOS MALE 153,127 54,46 - - TETRADAKTYLOS MALE 194,729 45,646 - - STRATIGOS MALE 22,054 48,908 - - KALLISTO FEMALE 13,1 - - - KATERINA FEMALE 7,09\* 28,457 54,729 7,459 TOBIAS MALE 148,126 64,061 119,532 17,46 MONAXH FEMALE - - 26,171 0 ALEKA FEMALE 1,381\* 38,18 204,066 1,744 KAPETANIOS MALE 190,297 322,4 207,807 22,789 KLEOPATRA FEMALE 75,004 50,473 30,542 2,035 ARIS MALE 59,076 39,424 - - DIAS MALE 21,662 46,594 - - SOFOKLIS MALE 196,023 337,71 14,74 - HLIAS MALE 170,466 248,97 76,969 7,861 LIGNOS MALE 195,927 140,053 53,447 28,349 PETHEROS MALE 174,917 155,129 147,834 54,901 TYXERH FEMALE - 54,764 9,197 - POLIMYLOS MALE 2.684 731,68 992,101 126,409 Table 2. Seasonal Home ranges (km2) for Brown bears (n=20) in Northeastern Pindos

overlapping areas (Giannakopoulos et al. subm.).

**Bear Sex Spring** 

mountain range in Greece, 2007-2009

**4.2 Potential changes in habitat use amplitude and range** 

decreased as the distance from highway increased.

fig.1 in the case of the male bears of the sample.

distance from the highway increases.

Regarding the **potential changes in habitat use amplitude and range (surface units)** in relation to the distance from the highway, the results of our analysis demonstrated that the size of the habitat units (within spring and summer male bears home range) significantly increased with the distance from the highway while their number (of used habitat units)

The differences between the number of habitat units and their size (surface) used inside the home range in relation to distance from the highway were statistically significant in all cases of males bears in spring and summer (spring: F=5.419, P<0.01; summer: F=6.52, P<0.01) and for females in summer (F=18.735; P<0.01). An example of this differentiation is given on

More specifically in the case of all male individuals of the sample the number of used habitat units (perceived through clustered radiolocations) is significantly higher as their surface size decreases and subsequently their distance from the highway decreases as well (spring: x2=96.63, P<0.01; summer: x2= 20.204, P<0.01). This means that larger ranges in surface and limited distinct ranges (in clustered radiolocations) were observed as the

Regarding **habitat suitability** analyses in relation to bear presence and habitat use: *distance from highway* was recognized as one of the statistically significant variables affecting both

Telemetry as a Tool to Study Spatial Behaviour and Patterns of

selection or avoidance by bears of a given pixel (habitat unit).

Altitude coefficient variation within 5

Average slope coefficient variation

Average slope coefficient variation

Number of different vegetation types

(%) of contribution of dominant

(%) of contribution of the 2nd rank

(%) of contribution of the 3rd rank

α) the type of data used in the analysis

β) a possible adaptive "shift" in bears behavior. In our case we may have two possible explanations:

Brown Bears as Affected by the Newly Constructed Egnatia Highway – N. Pindos - Greece 323

notice the importance of "altitude" and "slope" and their range of variations as prediction indicators. It comes out that the combination of landscape ruggedness with the characteristics of certain vegetation types and the distance from the highway influence

Average altitude within 5 pixels radius. -0,004 15,199 0,000

Average slope within 5 pixels radius. 0,039 58,315 0,000

**Variable Coefficient Wald Level of importance** 

pixels radius. 0,067 7,810 0,005

within 5 pixels radius. -0,003 16,071 0,000

within 15 pixels radius. -0,015 181,321 0,000 **Vegetation types variability** -0,098 0,641 0,423 Vegetation Type 23,492 0,001 Τype (1) -1,207 7,244 0,007 Τype (2) -1,112 11,511 0,001 Τype (3) -1,102 11,466 0,001 Τype (4) -1,117 8,990 0,003 Τype (5) -1,186 13,805 0,000 Τype (6) -,957 5,891 0,015 Τype (7) -,585 2,712 0,100

**Distance from highway** -0,000114 1196,691 0,000

within a 5 pixel radius -0,477 60,570 0,000

vegetation type within a 5 pixels radius -0,005 0,682 0,409

vegetation type within a 5 pixels radius 0,001 0,611 0,435

vegetation type within a 5 pixels radius 0,003 1,978 0,160

phenomenon might also be related to other parameters such as:

Table 4. Results from the LR analysis for the prediction model on bear presence/absence. The negative sign of variable "distance from highway" indicates that presence or absence of bears decreases as distance from the highway increases. In a recent study by Roever et al. (2008) it was found that grizzlies showed a relatively high frequency of occurrence in areas nearby forest roads despite the relatively high mortality probability rate in these areas (also McLellan, 1998, Benn and Herrero, 2002, Johnson κ.α., 2004 και Nielsen κ.α., 2004). But this

Aspect 65,844 0,000 Slope -0,007 10,344 0,001

analyses: the relative *abundance* (GLM) (table 3) and the *bear presence/absence* (LR) (table 4) thus influencing in both scenario cases the selection and the frequency of use of the different sites (habitat units) within the study area and in relation to the presence of the highway under construction. Bears seem to appear more often at distant sites from the highway. For the first analysis: **bear abundance and frequency of habitat pixels use,** of the set of 13 variables selected, seven (7) could be used as reliable prediction tools. Results of this analysis are presented in table (3).


Table 3. General Linear Models parameters as predictors of bear abundance in relation to the presence of the highway

For P values < 0.01, the related variables are considered to effectively contribute in the prediction model. We notice that vegetation types, altitude and aspect are recognized as important variables for the prediction of areas (habitat units) with more abundant/frequent bear presence and use. We also notice that the slope variance in neighbouring pixels also plays a role in the spatial distribution of the signs of presence. As stated above distance from the highway is the key variable with high statistical value in the model thus influencing site selection by bears. The negative value of the related coefficient indicates that the number of the most frequent bear occurrences in specific sites increases as the distance from the highway decreases.

Our analysis showed that there are no specific habitat parameters close to the highway corridor that hinder bears movements. Bears utilize the same habitat types within the overall landscape but move in a much more "conservative" pattern (in terms of duration and habitat surface used) when found in proximity of the highway corridor.

The second analysis regarding **presence/ absence** data (by means of LR & CART- *predictive accuracy of models which was high*) demonstrated a series of topographical and vegetation characteristics (habitat features) as important predictors for bear presence or absence. Here again **distance from highway** was recognized, as mentioned above, as one of the critical factors affecting the presence of an animal in a given point (pixel) of its home range. According to table (4) we may notice that a group of variables remains effective in the model for the prediction of bear presence in pixels with specific characteristics. We once again

analyses: the relative *abundance* (GLM) (table 3) and the *bear presence/absence* (LR) (table 4) thus influencing in both scenario cases the selection and the frequency of use of the different sites (habitat units) within the study area and in relation to the presence of the highway under construction. Bears seem to appear more often at distant sites from the highway. For the first analysis: **bear abundance and frequency of habitat pixels use,** of the set of 13 variables selected, seven (7) could be used as reliable prediction tools. Results of this

Diversity of vegetation types 8,961 0.003 60,570 0.000

Table 3. General Linear Models parameters as predictors of bear abundance in relation to the

For P values < 0.01, the related variables are considered to effectively contribute in the prediction model. We notice that vegetation types, altitude and aspect are recognized as important variables for the prediction of areas (habitat units) with more abundant/frequent bear presence and use. We also notice that the slope variance in neighbouring pixels also plays a role in the spatial distribution of the signs of presence. As stated above distance from the highway is the key variable with high statistical value in the model thus influencing site selection by bears. The negative value of the related coefficient indicates that the number of the most frequent bear occurrences in specific sites increases as the distance from the

Our analysis showed that there are no specific habitat parameters close to the highway corridor that hinder bears movements. Bears utilize the same habitat types within the overall landscape but move in a much more "conservative" pattern (in terms of duration

The second analysis regarding **presence/ absence** data (by means of LR & CART- *predictive accuracy of models which was high*) demonstrated a series of topographical and vegetation characteristics (habitat features) as important predictors for bear presence or absence. Here again **distance from highway** was recognized, as mentioned above, as one of the critical factors affecting the presence of an animal in a given point (pixel) of its home range. According to table (4) we may notice that a group of variables remains effective in the model for the prediction of bear presence in pixels with specific characteristics. We once again

and habitat surface used) when found in proximity of the highway corridor.

CV of mean slope within 104,065 0.000 181,321 0.000

CV of mean slope within 19,902 0.000 16,071 0.000

CV of altitude within 1.5km 23,198 0.000 7,810 0.005

Mean altitude within 1.5km 21,683 0.000 15,199 0.000

Distance from road. 288,652 0.000 1196,691 0.000 Aspect 80,444 0.000 65,844 0.000

**Variables Wald Chi-Square P-value Wald statistic P-value**

analysis are presented in table (3).

within 1.5km radius

7.5km radius

1.5km radius

radius

presence of the highway

highway decreases.

radius.

notice the importance of "altitude" and "slope" and their range of variations as prediction indicators. It comes out that the combination of landscape ruggedness with the characteristics of certain vegetation types and the distance from the highway influence selection or avoidance by bears of a given pixel (habitat unit).


Table 4. Results from the LR analysis for the prediction model on bear presence/absence.

The negative sign of variable "distance from highway" indicates that presence or absence of bears decreases as distance from the highway increases. In a recent study by Roever et al. (2008) it was found that grizzlies showed a relatively high frequency of occurrence in areas nearby forest roads despite the relatively high mortality probability rate in these areas (also McLellan, 1998, Benn and Herrero, 2002, Johnson κ.α., 2004 και Nielsen κ.α., 2004). But this phenomenon might also be related to other parameters such as:

α) the type of data used in the analysis

β) a possible adaptive "shift" in bears behavior.

In our case we may have two possible explanations:

Telemetry as a Tool to Study Spatial Behaviour and Patterns of

qualitatively affected by the existence of the highway.

habitat characteristics.

occurrences of brown bears.

frequency of occurrences of brown bears.

highway operation in the critical areas.

**6. Acknowledgements** 

**7. References** 

appropriate fencing).

Brown Bears as Affected by the Newly Constructed Egnatia Highway – N. Pindos - Greece 325

• Distance from the highway does not seem to influence independently bear habitat selection activity and abundance (presence/absence), but co-acts in synergy with other

• Findings from all three models agree on the importance of the "distance from the highway" as a critical variable for the prediction of bears spatial behavior in relation to the highway. Therefore the new highway represents a critical parameter that significantly affects distribution, habitat use, movement selection and frequency of

• The frequent presence of brown bears within the vicinity of the road network highlights the need for direct and effective protection measures in the area. (i.e adequate and

Considering previous results we suggest that animal (bear) activity is not reduced but rather

Overall we suggest that the new highway functions as a critical landscape parameter (barrier) that seems to significantly affect distribution, habitat use, movement patterns and

The results of our study will essentially contribute in further adjustment of mitigation measures along the highway as well as in close monitoring of their efficiency during

Telemetry research was possible in the framework of the two "Monitoring projects on impact evaluation of Egnatia highway construction (stretch 4.1 "Panagia-Grevena" and stretch "Panagia-Metsovo") on large mammals in the area of Grevena-Ioannina and Trikala (2006-2009). This project was co-funded by EGNATIA ODOS SA, Hellenic Ministry of Environment, Planning & Public Works and the EU (DG Regio). We thank the Forestry Services of Kastoria, Grevena & Kalambaka for forestry data provision and the NGO CALLISTO field team : Sp. Galinos, M.Petridou, H. Pilidis, Y. Tsaknakis and local assistant Y.Lazarou for their precious help. Special thanks go also to Dr. John Beecham, from Idaho

Austin, M. P. 2002. Spatial Prediction of Species Distribution: an Interface Between Ecological Theory and Statistical Modelling. Ecological Modelling 157:101-118. Benn, B., Herrero, S., 2002. Grizzly bear mortality and human access in Banff and Yoho

Bergman, C. M., J. A. Schaefer, and S. N. Luttich. 2000. Caribou Movement as a Correlated

Bontadina, F., H. Schofield, and B. Naef-Daenzer. 2002. Radio-Tracking Reveals That Lesser

Debeljak, M., S. Dzeroski, K. Jerina, A. Kobler, and M. Adamic. 2001. Habitat Suitability

Horseshoe Bats (Rhinolophus Hipposideros) Forage in Woodland. Journal of

Modelling for Red Deer (Cervus Elaphus L.) In South-Central Slovenia With

Fish & Wildlife Service, U.S and to Yorgos Iliopoulos for their help and advice.

National Parks, 1971–1989. Ursus 13, 213–221.

Classification Trees. Ecological Modelling 138:321-330.

Random Walk. Oecologia 123:364-374.

Zoology 258:281-290.


The **CRT analysis** showed also that the variable "distance from highway" was used to separate two central "branches" of the classification tree in the early analysis stages. Two differentiated branches are defined according to a limit value of 4.996 m of distance from the highway. When this distance is <4.996 m then a combination of topographic characteristics in relation to high slope values and medium altitude values are characterizing the pixels used by bears.

In the second case d > 4.996 m, vegetation types but also certain combinations of topographic characteristics define the habitat use patterns in each pixel. It also came out from this analysis that pixels at a distance > 8.434m have lower use frequencies by the sampled bears.
