Behavioral Ecology and Sociobiology

, Volume 67, Issue 11, pp 1731–1743

Simultaneous GPS tracking reveals male associations in a solitary carnivore

Authors

  • Mia Lana Lührs
    • Department of Sociobiology/Anthropology, Johann-Friedrich-Blumenbach-Institute of Zoology and AnthropologyGeorg-August-University
    • Department of Sociobiology/Anthropology, Johann-Friedrich-Blumenbach-Institute of Zoology and AnthropologyGeorg-August-University
    • Behavioral Ecology and Sociobiology Unit, German Primate Center, Leibniz Institute for Primate Research
Original Paper

DOI: 10.1007/s00265-013-1581-y

Cite this article as:
Lührs, M.L. & Kappeler, P.M. Behav Ecol Sociobiol (2013) 67: 1731. doi:10.1007/s00265-013-1581-y

Abstract

In hypercarnivorous species, females have large spatial requirements to meet their nutritional needs, and food competition among females is intense. As a result, females are typically solitary and territorial, and solitary males compete for access to dispersed females. Yet, largely anecdotal reports indicate that facultative male sociality may be more common in solitary carnivores than previously thought. We studied spatial interactions among fossas (Cryptoprocta ferox), Madagascar's largest carnivore, using simultaneous GPS tracking of 13 adult individuals to determine patterns of sex-specific spatial distribution and sociality. Male home ranges were larger than those of females, male home ranges overlapped more with those of other males than those of females with other females. Whereas some males were solitary, a subset of adult males was found to have very high home range overlap, high rates of co-location within <50 m, low minimum inter-individual distances, and significantly positive “dynamic interaction”. These associated dyads sometimes, but not always, were close relatives. The fact that solitary and associated males coexist in this population raises interesting questions concerning constraints and flexibility of social tolerance. This study yielded preliminary indications that female distribution appears to be primarily structured by resource competition, whereas male sociality seems to depend on demographic chance events, yet unknown proximate determinants of social tolerance, and it is associated with somatic and reproductive advantages. Male associations among carnivores are therefore more widespread and appear to be based on a wider range of factors than previously thought.

Keywords

Social organizationCarnivoresCryptoprocta feroxGPS telemetryMale socialityDynamic interaction test

Introduction

According to socioecological theory, a species' social organization is determined by the spatial and temporal distribution of risks and resources (Crook and Gartlan 1966; Crook 1970; Emlen and Oring 1977; Macdonald 1983; Terborgh and Janson 1986). Since female fitness is limited by access to food resources, and male fitness is mainly determined by access to receptive females (Williams 1966; Clutton-Brock and Parker 1992), female social organization is assumed to represent the primary response to ecological factors, whereas male social organization is thought to be a response to female distribution (Altmann 1990) and therefore only secondarily affected by ecological factors. Even though socioecological models were developed with a focus on group-living species (Clutton-Brock and Janson 2012), their main principles could be shown to be also applicable to solitary and pair-living species (Schülke 2003; Dammhahn and Kappeler 2009).

One explicit socioecological model is represented by the resource dispersion hypothesis (RDH; Macdonald 1983; Carr and Macdonald 1986; Johnson et al. 2002), which asserts that territory size is determined by resource dispersion, whereas group size of the territory holders is determined by resource abundance and richness. This model makes general predictions about conspecific tolerance in a territory, irrespective of other factors affecting sociality, such as costs and benefits of group-living (reviewed in Krause and Ruxton 2002). The RDH has been tested in a number of carnivores [e.g., badgers (da Silva et al. 1993; Johnson et al. 2001) and foxes (Geffen et al. 1992; Baker et al. 2000)] and also in rodents (Herrera and Macdonald 1989) and birds (Davies et al. 1995), and, despite the diversity of factors influencing sociality, enjoyed general support.

Suitable model species to illuminate determinants of social organization can be found in the mammalian order Carnivora because the number of decisive factors, such as top-down control (predation), intra-guild competition, or food resource diversity may be reduced in this clade. In top predators feeding exclusively on meat (live prey or carcasses), so-called hypercarnivorous species, the number of such confounding factors is reduced, unless intra-guild competition is intense or intra-guild predation is common, as in many continental African large carnivores (Creel and Creel 1996; Durant 1998; Caro and Stoner 2003). Since carnivore prey are a resource that is unpredictable both in space and time, spatial requirements to meet a sufficient food supply increase with the proportion of meat in the diet (McNab 1963; Gittleman and Harvey 1982; Gompper and Gittleman 1991). Females of hypercarnivores and omnivores therefore often inhabit large territories and rarely exhibit gregarious tendencies; the most prominent exception being hypercarnivorous lions and spotted hyaenas (Kruuk 1972; Schaller 1972; Bertram 1975). As a result, more than 85 % of carnivores live solitarily (Bekoff et al. 1984; Gittleman 1984).

Because male spatial distribution is not only determined by food resources but also by the distribution of females, solitary carnivore males usually inhabit large home ranges that overlap with those of several females (Sandell 1989). However, in some carnivores with solitary females, males exhibit tendencies to associate with other males. Such permanent associations appear to serve joint defense of a larger territory, i.e., an exclusively used area with a stable location, entailing access to more females (Macdonald 1983), as, for example, in cheetahs (Caro 1994) and several mongooses (Rood and Waser 1978; Cavallini and Nel 1990; Waser et al. 1994; Hays and Conant 2003; Rathbun and Cowley 2008). In other species, males may form temporary groups in response to short-term female aggregation, as e.g., in raccoons (Gehrt and Fritzell 1998), or permanent groups with little social interaction, as e.g., in striped hyenas (Wagner et al. 2008) or slender mongooses (Waser et al. 1994), to form coalitions against other males. Finally, associations may also form to enhance predation efficiency because several predators might be able to subdue larger prey (Creel and Creel 1995; Lührs and Dammhahn 2010).

The existence of such male associations in species with solitary females does not necessarily contradict socioecological theory. If associated males enjoy improved access to females, male coalitions may ultimately represent an adaptive response to female distribution. In cheetah, fitness costs of reproductive competition among allies are compensated by indirect fitness benefits because associates are often littermates (Caro and Collins 1986). Evidence for male associations in mongooses with solitary females is so far only anecdotal because solitary carnivores are difficult to study. Information about relatedness of associates as well as the stability and cohesion of male associations is therefore currently lacking.

GPS tracking methods have added a helpful tool for studying enigmatic carnivore species (e.g., Soisalo and Cavalcanti 2006; Bandeira de Melo et al. 2007; Barlow et al. 2011). In the present study, we used this technology to investigate the social organization (sensu Kappeler and van Schaik 2002) of a top predator endemic to Madagascar, the fossa (Cryptoprocta ferox), in which male associations have been observed anecdotally (Hawkins 1998; Lührs and Dammhahn 2010). Fossas are the largest members of Madagascar's mongooses (Eupleridae), weighing up to 11 kg. They are hypercarnivorous, feeding mainly on endemic primates (lemurs), other vertebrates and occasionally on invertebrates (Rasoloarison et al. 1995; Hawkins and Racey 2008; Lührs and Dammhahn 2010). As top predators in their ecosystem, fossas neither face intra-guild predation risk nor resource competition. Factors influencing the social organization of fossa are therefore largely limited to intraspecific dynamics; potential effects of recent ecological equilibria (van Schaik and Kappeler 1996) can presently not be evaluated. A previous study using radio-telemetry revealed a solitary life-style, large ranges and low population density (Hawkins 1998; Hawkins and Racey 2005). Limitations of radio-tracking posed problems on sample size, however, and likely led to underestimation of home range sizes. Furthermore, interactions, and thus the degree of sociality, could not be assessed, and the details of male associations remained obscure.

We therefore studied fossas using simultaneous GPS tracking of males and females to determine their spatial organization and to evaluate their degree of sociality. Based on the fossa's diet, we predicted females to be solitary and territorial, i.e., to rarely or never associate with conspecifics and to maintain an exclusive home range whose location is stable from month to month. With respect to males, we aimed to describe spatial association dynamics, to compare ranging behavior during and outside the mating season and to study patterns of their social interactions. If male sociality is influenced by their reproductive strategies, the ranges of associated males might be larger and should exhibit less overlap with ranges of other males than the ranges of solitary males. If male sociality serves other purposes, such as cooperative hunting, such a difference in home range overlap is not predicted.

Methods

GPS tracking

Spatial data presented here were obtained via GPS tracking of 13 adult (>5 kg; for other criteria see Hawkins et al. 2002) fossas (nine males, four females) trapped between 2008 and 2010 in Kirindy Forest / CNFEREF (44°39′ E, 20°03′ S). Kirindy is a dry deciduous forest with pronounced seasonality. The local dry season with little to no rainfall spans half a year from May to October, while the wet season from November until April is characterized by high humidity and frequent rainfall (Sorg and Rohner 1996). Fossas were trapped annually during the dry season with 10 live-traps (42 × 15 × 20-in. bobcat trap, Tomahawk, USA) along transects shifted weekly over a 9-km2 area in the forest center. Traps were set along roads, paths, or rivers, and were controlled every 2 h during the day and twice at night. Trapped animals were briefly anaesthetized using an individually adjusted dosage of ketamine, medetomidine, and a benzodiazepine. They were then measured and equipped with GPS collars (150–200 g, e-obs GmbH, Germany), making up less than 5 % of the individuals' body mass (Gannon and Sikes 2007). After recovery from anesthesia, animals were released at the site of capture. None of the individuals that were re-trapped to remove the GPS collar exhibited injuries or significant weight loss.

Since trappability was highest towards the end of the dry season, most tracking data were obtained from September onwards. GPS-tags logged positions once per hour for an average of 21 consecutive days (range, 10–31 days; Table 1), with occasional drop out values (about 40 %), presumably due to atmospheric disturbance, high forest density or weak batteries. Average horizontal accuracy of locations was ±12 m. GPS data were downloaded via remote download stations located at waterholes or via hand-held remote devices.
Table 1

Overview of spatial data obtained for females (F1–F4) and males (M1–M9)

ID

Year

Month

N days tracked

N positions

LoCoH area (ha)

MCP area (ha)

50 % kernel area (ha)

90 % kernel area (ha)

Arithmetic mean shift (m)

Weighted mean shift (m)

F1

2008

Sep

14

252

788

1,203

273

1,214

F1

2008

Oct

31

569

1,381

1,881

480

1,722

833

1,049

F1

2008

Nov

30

489

501

1,174

337

1,216

903

1,039

F1

2008

Dec

28

500

282

1,127

507

1,486

501

1,119

F1

2009

June

21

298

1,250

1,972

173

1,469

F1

2009

July

31

473

1,136

2,024

591

2,015

378

1,901

F1

2009

Aug

31

431

1,188

2,158

571

2,007

362

282

F1

2009

Sep

30

282

1,269

2,286

398

1,981

292

940

F1

2009

Oct

31

294

1,834

2,210

544

2,177

257

652

F1

2009

Nov

22

74

794

1,853

401

1,376

412

695

F2

2009

Sep

19

294

875

1,390

459

1,435

F2

2009

Oct

21

280

997

1,343

590

1,687

117

288

F3

2009

Oct

28

355

879

1,383

424

1,317

F3

2010

Aug

19

317

969

1,733

529

1,643

F3

2010

Sep

16

250

1,500

2,526

488

1,961

433

617

F4

2010

Nov

30

431

1,092

2,130

740

2,308

M1

2008

Sep

10

166

3,152

3,477

267

2,163

M1

2008

Oct

31

488

4,947

7,739

2,239

6,909

1,651

2,664

M1

2008

Nov

30

513

3,808

6,690

1,463

5,672

1,494

3,090

M2

2009

Sep

15

180

3,143

3,676

139

1,409

1,682

666

M2

2009

Oct

31

297

5,403

7,619

858

6,602

1,326

830

M3

2009

May

11

152

1,540

4,607

1,920

6,489

M3

2009

June

19

251

3,554

8,457

2,886

8,627

1,347

1,885

M3

2009

Sept

11

145

1,145

4,405

1,715

5,921

2,013

761

M3

2009

Oct

24

321

3,019

5,488

2,539

7,394

134

1,787

M4

2009

May

11

165

1,795

4,579

1,765

5,896

M4

2009

June

17

197

3,001

8,065

2,794

8,295

1,690

1,468

M4

2009

Sep

12

150

2,067

4,842

1,951

6,877

2,223

1,635

M4

2009

Oct

25

317

2,982

5,568

2,772

7,597

333

2,267

M4

2010

Sep

29

503

2,710

5,278

1,844

5,783

M4

2010

Oct

14

176

3,179

6,552

3,861

10,156

1,010

1,541

M5

2009

Oct

31

425

3,449

4,245

718

3,140

M6

2010

Oct

17

239

2,449

4,234

1,025

4,032

M6

2010

Nov

30

367

4,259

8,961

2,182

8,882

3,649

3,974

M6

2010

Dec

31

353

4,133

8,386

3,040

9,253

2,935

194

M6

2011

Jan

19

143

2,253

6,736

2,909

9,029

2,007

1,099

M7

2010

Oct

17

248

4,481

6,774

2,143

7,771

M7

2010

Nov

20

163

3,007

6,081

2,938

9,158

2,277

2,761

M8

2010

Sep

30

486

1,936

3,589

997

3,350

M8

2010

Oct

31

485

1,934

3,301

754

2,604

279

811

M8

2010

Nov

30

462

2,346

4,211

1,284

3,970

1,206

188

M8

2010

Dec

31

424

958

2,089

629

2,007

699

238

M8

2011

Jan

24

266

924

1,876

678

2,084

830

353

M9

2010

Oct

23

355

1,392

2,183

851

2,334

M9

2010

Nov

30

420

2,503

4,213

975

3,301

1,155

816

Static spatial data analyses

GPS locations were used to calculate three types of home range size estimates (minimum convex polygons (MCPs; Mohr 1947), kernel home ranges (Worton 1989), and local nearest-neighbor convex-hull home ranges (Getz and Wilmers 2004)). Furthermore, we determined spatial stability of home range areas (stability of the area's centroid, referred to as the “arithmetic mean”) and home range use (stability of the mean of all tracking locations, referred to as the “weighted mean”). The maximum tracking frequency was one location per hour and fossas were potentially able to cross their entire range within this period (based on observed hourly travel distances) so that we consider data points to be statistically independent (Swihart and Slade 1985; De Solla et al. 1999).

Exploratory data analyses revealed that the removal of outliers led to loss of biological meaning and only little reduction of MCP home range size. Similar problems were noticed for the use of local convex-hull home range estimates. We therefore chose 100 % MCPs as appropriate to quantify maximum space cover and range overlap among individuals. In order to avoid overestimates of home range size, use, and overlap associated with MCPs, we additionally determined kernel estimates between 50 and 90 % probability density (Börger et al. 2006). Since 90 % kernel home range areas exceeded areas of 100 % MCPs, thereby creating even more false overlap than MCPs, we provide 90 % kernel areas only as a reference and based our analyses exclusively on MCPs and 50 % kernels. The combination of MCPs, 50 % kernels and shifts in home range means were used to estimate static interaction between individuals as a proxy for the degree of territoriality and sociality. For site fidelity and true territoriality, we assumed stable arithmetic means of a range as an indication for the individual being “trapped” within a given space, and weighted means, i.e., preferably used regions within this stable range, to represent variation in home range use. Home range areas were calculated in R (R Development Core Team 2011) using the package “adehabitatHR” (Calenge 2006). Weighted home range means were determined in ArcView 3.3 (ESRI, California, USA) using the “weighted means” extension (Jenness 2004). All spatial analyses were performed month-wise to create comparable, temporally overlapping units among different individuals.

Dynamic spatial data analyses

Because static interaction between individuals is insufficient to characterize the degree of sociality, we further used dynamic spatial data analysis to examine direct interaction between simultaneously tracked individuals. Simultaneous positions were obtained for a total of 36 dyads (three female–female (F–F), 17 male–female (M–F), and 16 male–male (M–M) dyads) (see Table 2 for details). Only those positions were considered for interaction tests where both individuals could be located simultaneously. Dyadic analyses were performed over the entire period of simultaneous tracking (4–6 weeks), except for two dyads that could be tracked long enough to create separate data sets for two months.
Table 2

Overview of spatial data obtained from dyadic comparison. Statistically supported deviation from Doncaster's (1990) dynamic interaction test for a critical distance of ≤ 50 m are marked in bold

Year

ID1

ID2

Dyad

N positions

N days tracked

Month

MCP area ID1 (ha)

MCP area ID2 (ha)

MCP overlap ID1 (%)

MCP overlap ID2 (%)

Overlap MCP Area (ha)

50 % kernel area ID1 (ha)

50 % kernel area ID2 (ha)

Kernel overlap ID1 (%)

Kernel overlap ID2 (%)

50 % kernel overlap (%)

Tracked <50 m (%)

DIA test: χ2

P

Average inter- individual distance (m)

Average distance expected (m)

Distance observed/expected

Minimum distance observed (m)

Minimum distance observed/average

Relatedness coefficient (TrioML)a

2009

F1

F2

F–F

335

40

Oct

2,209.9

1,342.4

1.5

2.4

32.8

544.4

589.8

0.0

0.0

0.0

0

3,744.0

3,770.0

1.0

727

0.2

0.00

2009

F1

F3

F–F

235

37

Oct

2,209.9

1,391.6

58.5

92.8

1,291.8

544.4

486.6

7.1

8.3

7.7

0

0.09

0.760

2,642.8

2,498.0

1.1

494

0.2

0.69

2009

F2

F3

F–F

161

18

Oct

1,342.4

1,391.6

8.7

8.4

116.7

589.8

486.6

0.0

0.0

0.0

0

4,013.1

4,031.8

1.0

996

0.2

0.05

2008

F1

M1

M–F

1,092

30

Oct

1,997.3

7,739.4

78.4

20.2

1,566.6

544.4

2,239.0

64.3

15.5

39.9

0

0.77

0.381

4,043.9

3,939.1

1.0

260

0.1

0.00

2008

F1

M1

M–F

944

28

Nov

1,242.2

7,292.4

84.1

14.3

1,044.7

400.8

1,561.8

27.3

7.1

17.2

0

0.03

0.861

4,603.7

4,522.6

1.0

548

0.1

2009

F1

M1

M–F

320

32

June

1,971.8

1,082.9

26.2

47.6

515.9

173.1

189.2

10.0

10.0

10.0

0

10.89

<0.001

2,670.6

2,575.3

1.0

157

0.1

2009

F1

M2

M–F

406

43

July

2,024.0

1,652.0

44.4

54.3

897.8

590.9

103.2

0.0

0.0

0.0

0

3.15

0.076

2,614.8

2,545.0

1.0

111

0.0

0.00

2009

F1

M2

M–F

275

46

Oct

2,209.9

7,618.7

62.3

18.1

1,376.8

544.4

858.5

14.3

9.1

11.7

0

0.02

0.904

3,517.2

3,456.1

1.0

546

0.2

2009

F1

M3

M–F

274

35

Oct

2,209.9

5,488.4

29.8

12.0

659.2

544.4

2,538.8

14.3

3.0

8.7

0

0.05

0.828

5,052.4

5,019.2

1.0

345

0.1

0.17

2009

F1

M4

M–F

284

37

Oct

2,209.9

5,567.5

30.2

12.0

668.0

544.4

2,771.7

71.4

13.9

42.7

0

0.05

0.824

4,921.4

4,914.7

1.0

463

0.1

0.06

2009

F1

M5

M–F

287

36

Oct

2,209.9

4,244.7

86.8

45.2

1,919.2

544.4

718.4

71.4

55.6

63.5

0

0.20

0.653

2,545.5

2,479.1

1.0

101

0.0

0.00

2009

F2

M2

M–F

305

36

Oct

1,342.4

7,618.7

100.0

17.6

1,342.4

589.8

858.5

26.7

18.2

22.4

0

0.29

0.593

2,574.8

2,579.1

1.0

133

0.1

0.05

2009

F2

M3

M–F

329

31

Oct

1,342.4

5,488.4

35.1

8.6

470.9

589.8

2,538.8

0.0

0.0

0.0

0

0.02

0.893

5,518.7

5,469.0

1.0

427

0.1

0.06

2009

F2

M4

M–F

312

32

Oct

1,342.4

5,567.5

35.1

8.5

471.2

589.8

2,771.7

0.0

0.0

0.0

0

0.05

0.815

5,111.5

5,145.4

1.0

420

0.1

0.05

2009

F3

M2

M–F

195

30

Oct

1,391.6

7,618.7

75.3

13.8

1,047.6

486.6

858.5

25.0

13.6

19.3

0

0.03

0.873

3,866.6

3,869.3

1.0

409

0.1

0.00

2009

F3

M3

M–F

194

21

Oct

1,391.6

5,488.4

19.2

4.9

267.7

486.6

2,538.8

0.0

0.0

0.0

0

7,286.1

7,301.7

1.0

1,109

0.2

0.20

2009

F3

M4

M–F

204

22

Oct

1,391.6

5,567.5

19.6

4.9

272.3

486.6

2,771.7

8.3

1.4

4.9

0

0.02

0.889

6,977.0

6,974.2

1.0

1,134

0.2

0.00

2010

F3

M4

M–F

187

15

Sep

2,526.4

5,278.4

25.9

12.4

653.3

488.4

1,844.4

83.3

21.7

52.5

0

0.06

0.800

3,812.7

3,878.1

1.0

202

0.1

2009

F3

M5

M–F

281

32

Oct

1,391.6

4,244.7

76.3

25.0

1,062.3

486.6

718.4

0.0

0.0

0.0

0

0.04

0.836

3,495.7

3,486.2

1.0

309

0.1

0.00

2010

F3

M8

M–F

209

18

Sep

2,526.4

3,589.3

52.4

36.9

1,323.7

488.4

996.6

29.2

14.0

21.6

0

0.01

0.922

3,314.4

3,266.9

1.0

281

0.1

0.01

2010

F4

M6

M–F

298

36

Nov

2,129.6

8,961.4

100.0

23.8

2,129.6

740.4

2,182.0

80.0

26.7

53.3

3.4

58.11

<0.001

4,319.7

4,128.4

1.0

2

0.0

0.00

2010

F4

M7

M–F

134

25

Nov

2,129.6

6,080.6

88.3

30.9

1,880.3

740.4

2,937.9

85.0

21.3

53.1

7.5

71.65

<0.001

4,328.2

4,066.0

1.1

5

0.0

0.19

2010

F4

M8

M–F

358

35

Nov

2,129.6

4,211.4

81.5

41.2

1,735.1

740.4

1,283.5

35.0

20.0

27.5

0.3

2.98

0.084

3,115.6

2,990.1

1.0

4

0.0

0.00

2010

F4

M9

M–F

346

35

Nov

2,129.6

4,212.5

12.9

6.5

273.9

740.4

975.4

0.0

0.0

0.0

0

0.11

0.744

4,522.7

4,650.7

1.0

113

0.0

0.00

2009

M1

M2

M–Mb

271

24

June

1,082.9

711.3

65.6

99.9

710.9

189.2

86.9

50.0

100.0

75.0

43.2

416.33

<0.001

420.3

1,488.8

0.3

0

0.0

0.72

2009

M1

M3

M–M

291

30

June

1,082.9

8,456.6

67.6

8.7

731.8

189.2

2,886.4

20.0

1.2

10.6

0

5,200.3

5,203.2

1.0

374

0.1

0.01

2009

M1

M4

M–M

262

27

June

1,082.9

8,065.4

60.0

8.1

649.5

189.2

2,793.7

20.0

1.3

10.6

0

5,059.2

5,067.7

1.0

291

0.1

0.00

2009

M2

M3

M–M

273

34

Oct

7,618.7

5,488.4

31.6

43.9

2,411.0

858.5

2,538.8

0.0

0.0

0.0

0.4

0.75

0.388

6,359.9

6,247.5

1.0

44

0.0

0.05

2009

M2

M4

M–M

281

37

Oct

7,618.7

5,567.5

31.9

43.7

2,433.3

858.5

2,771.7

9.1

2.8

5.9

0.4

0.16

0.689

5,870.7

5,776.9

1.0

24

0.0

0.00

2009

M2

M5

M–M

284

42

Oct

7,618.7

4,244.7

15.4

27.7

1,175.5

858.5

718.4

0.0

0.0

0.0

0

0.01

0.918

4,766.4

4,809.3

1.0

815

0.2

0.00

2009

M3

M4

M–Mb

241

27

June

8,456.6

8,065.4

93.3

97.8

7,890.4

2,886.4

2,793.7

83.6

86.3

84.9

92.9

18,360.91

<0.001

20.6

4,242.1

0.0

1

0.0

0.00

2009

M3

M4

M–Mb

347

35

Oct

5,488.4

5,567.5

99.1

97.6

5,436.5

2,538.8

2,771.7

89.4

81.9

85.7

85.3

43,281.21

<0.001

206.7

4,438.4

0.0

0

0.0

0.00

2009

M3

M5

M–M

319

33

Oct

5,488.4

4,244.7

15.3

19.8

840.9

2,538.8

718.4

3.0

11.1

7.1

0

0.07

0.798

5,258.8

5,201.6

1.0

369

0.1

0.02

2009

M4

M5

M–M

310

34

Oct

5,567.5

4,244.7

15.3

20.0

851.0

2,771.7

718.4

11.1

44.4

27.8

0

0.30

0.586

5,194.8

5,141.3

1.0

221

0.0

0.11

2010

M4

M8

M–M

515

43

Sep

5,278.4

3,589.3

8.3

12.2

439.6

1,844.4

996.6

16.3

30.0

23.2

0

0.01

0.914

5,435.1

5,322.4

1.0

471

0.1

0.00

2010

M6

M7

M–Mb

267

37

Nov

8,961.4

6,080.6

65.3

96.2

5,851.8

2,182.0

2,937.9

93.3

70.0

81.7

20.6

2,212.21

<0.001

2,302.9

4,920.5

0.5

2

0.0

0.52

2010

M6

M8

M–M

393

41

Nov

8,961.4

4,211.4

41.5

88.4

3,722.0

2,182.0

1,283.5

55.0

94.3

74.6

0

0.66

0.417

3,857.8

3,783.3

1.0

53

0.0

0.13

2010

M6

M8

M–M

348

55

Dec

6,909.4

2,089.4

29.6

97.8

2,044.3

3,040.3

628.9

13.0

64.3

38.7

0

0.06

0.815

4,336.1

4,315.4

1.0

490

0.1

2010

M6

M9

M–M

488

51

Nov

8,961.4

4,212.5

17.7

37.6

1,585.7

2,182.0

975.4

23.3

53.8

38.6

3.5

68.57

<0.001

4,791.1

4,849.2

1.0

3

0.0

0.00

2010

M7

M8

M–M

279

37

Nov

6,080.6

4,211.4

57.8

83.4

3,512.2

2,937.9

1,283.5

40.0

91.4

65.7

0.4

2.65

0.103

4,203.1

4,269.0

1.0

3

0.0

0.01

2010

M7

M9

M–M

287

37

Nov

6,080.6

4,212.5

17.0

24.5

1,033.8

2,937.9

975.4

17.5

53.8

35.7

0.7

8.80

0.003

5,644.5

5,546.7

1.0

34

0.0

0.00

2010

M8

M9

M–M

302

27

Oct

3,301.1

2,183.3

21.4

32.3

705.2

754.4

851.0

10.5

9.1

9.8

0

0.06

0.802

3,627.8

3,592.9

1.0

126

0.0

0.20

2010

M8

M9

M–M

321

32

Nov

4,211.4

4,212.5

35.3

35.3

1,488.6

1,283.5

975.4

31.4

42.3

36.9

0.3

0.01

0.933

2,958.4

3,689.9

0.8

46

0.0

aTriadic likelihood estimator TrioML based on 16 microsatellite markers applied to a population of 33 individuals (see Lührs et al. 2013 for details). Close relatedness (r > 0.5) is marked in bold

bMale associations

In order to determine whether two individuals associate, avoid each other, or simply move randomly in relation to each other, we applied Doncaster's method of dynamic interaction (Doncaster 1990). This test compares n observed inter-individual distances with expected ones calculated from all n2 distances possible within the given set of spatial points (i.e., within an n × n matrix). A critical distance assuming awareness of each other's presence can then be taken to compare observed cumulative probability of occurrence of two individuals within the critical distance with expected probabilities using a chi-squared test. Based on behavioral observations of encounters, we chose 50 m as the critical distance, which is a rather short distance for a carnivore of this size. Any interaction effects found within 50 m can therefore be assumed to be conservative. We compared observed probabilities with expected values with a 4 × 4 contingency table containing counts below and above 50 m, respectively, and evaluated whether two individuals could be located within 50 m more often or less often than expected by chance based on the distribution of given locations.

In order to further describe inter-individual dynamics, we evaluated minimum distances as well as average distances between individuals of a dyad. We divided minimum distances by dyadic average distance and divided observed distances by expected distances calculated for Doncaster's model to account for effects of differences in home range overlap.

All calculations associated with spatial data were performed in R. Graphical presentation of spatial data was produced in ArcView 3.3 (“Animal Movement” extension; Hooge and Eichenlaub 2000). Statistical graphs were produced in Statistica 10.

Since relatedness among individuals contributes crucial information to the study of social organization, we additionally provide information on dyadic relatedness from an earlier study based on 16 microsatellite markers applied to a total of 33 fossas (see Lührs et al. 2013 for details). Relatedness coefficients for dyads represented here are provided in Table 2.

Statistical analyses

Since home range areas were gamma distributed, we used a generalized linear mixed model (GLMM) based on a gamma-distribution to determine whether the variance in monthly home range sizes was best explained by sex, social organization (solitary vs. associated), ecological season (dry vs. wet), or reproductive season (mating vs. non-mating season). Equivalent analyses were performed for monthly arithmetic and weighted mean shifts using linear mixed-effects models (LMMs). Dry season was assigned to months May to October; wet season to November until January within the tracking period. October and November represented the mating season because mating activity peaked in November during all years, but males started long-distance excursions in search for mating trees in October.

Due to inflated type I error rates of stepwise model selection procedures (Mundry and Nunn 2009), we retained all those predictor variables in our mixed models, which were of potential biological significance for the response variable. Models differed only in response variables and the fixed factors sex and social organization, which were alternate depending on the question addressed. Seasons were included in every model and an animal's identity was included as a random factor throughout. Variables and residuals were tested for normal distribution using a Shapiro–Wilk test (Shapiro and Wilk 1965). Arithmetic mean shifts were log-transformed to fit a normal distribution. Since shifts in weighted means were correlated with home range size, only fitted residuals were taken for analyses of weighted mean shifts.

P values for LMMs were calculated from 1,000 Monte Carlo simulations and significance level was accepted at P ≤ 0.05. All statistical analyses were performed in R.

Results

Home range size, overlap, and shifts

Static home range analyses revealed that males consistently used larger MCP ranges than females (GLMM: t11 = −9.43, P < 0.001; ESM Table 1, Fig. 1, ESM Fig. 1). Whereas female range size appeared to remain stable over the course of the study period (Table 1, Fig. 1, ESM Fig. 1), male range sizes tended to vary with reproductive season (LMM: t = 1.76, P = 0.091, n = 9; ESM Table 2, Fig. 1). Male ranges increased by a factor of 1.4 ± 0.6 SD during the mating season and decreased to their previous size in December (Table 1, Fig. 1), but this variation was not statistically significant. Neither male nor female range sizes were affected by ecological season (GLMM: t30 = −0.72, P = 0.475; ESM Table 1).
https://static-content.springer.com/image/art%3A10.1007%2Fs00265-013-1581-y/MediaObjects/265_2013_1581_Fig1_HTML.gif
Fig. 1

MCP home range areas of males (squares) and females (circles) per reproductive season. Plotted are means and 95 % confidence intervals (whiskers). *P < 0.05, statistical differences from a GLMM (see ESM, Table 1)

Home range overlap differed as a function of sex and social organization (ESM Fig. 1). Whereas female ranges appeared not to overlap with those of unrelated individuals of the same sex (see Table 2 for information on relatedness), male monthly MCP ranges overlapped extensively both with those of females (23 ± 12 % SD) and with those of other males (48 ± 15 % SD). More so, a subset of males (M1–M2, M3–M4, and M6–M7) shared ranges by 81 to 98 % (Table 2). In females, high MCP overlap (58 and 93 %, respectively) could only be found for a mother–daughter dyad (F1–F3; Table 2). Yet, overlap in 50 % kernels was generally low for female–female (<9 %) and male–female dyads (19 ± 14 % SD). Among male–male dyads, there was high variability, ranging from 0 to 100 % overlap (40 ± 11 % SD), with the same subset of six males (M1–M2, M3–M4, and M6–M7) showing highest overlap (50–100 %; Table 2).

Arithmetic and weighted means of male ranges shifted from month to month by on average (mean ± SD) 1,522 ± 700 and 1,429 ± 799 m, respectively. In contrast, female ranges shifted only in their weighted means by 838 ± 476 m, whereas average shifts in arithmetic means were negligible (499 ± 404 m; Table 1). As a result, males and females differed in arithmetic mean shifts (LMM: t = −3.02, P = 0.006, n = 11; ESM Table 3), but not in weighted shifts (LMM: t = 0.23, P = 0.823, n = 11; ESM Table 4, Fig. 2), when range size was controlled for. Weighted mean shifts were neither affected by ecological season (LMM: t = −0.30, P = 0.765, n = 11), nor by reproductive season (t = 1.30, P = 0.205, n = 11; ESM Table 4). Arithmetic mean shifts remained unaffected by reproductive season (LMM: t = −1.48, P = 0.152, n = 11), but tended to increase towards the wet season (t =2.05, P = 0.051, n = 11; ESM Table 3), possibly as a result of enduring shifts by males beyond the mating season.
https://static-content.springer.com/image/art%3A10.1007%2Fs00265-013-1581-y/MediaObjects/265_2013_1581_Fig2_HTML.gif
Fig. 2

Monthly MCP home range shifts of arithmetic means (left) and weighted means (right) for males (squares) and females (circles). Plotted are means and 95 % confidence intervals (whiskers). Weighted means were controlled for difference in absolute home range size and only residuals were plotted. *P < 0.05 statistical differences from LMMs (see ESM, Table 3 and Table 4)

Dyadic interaction

Doncaster's dynamic interaction test revealed that the only non-exclusive female–female dyad interacted randomly (χ21 = 0.09, P = 0.760) despite 76 % of MCP overlap (Table 2). Females generally avoided conspecifics at a higher contact radius than the critical distance of 50 m (Table 2). A comparison of observed inter-individual distances with those predicted by the model revealed that female–female dyads maintained larger distances than expected, and they differed in this respect from all other dyads (Mann–Whitney U test: Z = 2.81, P = 0.005, n = 36; ESM Fig. 2). Moreover, females maintained higher minimum distances to conspecifics of either sex than males (427 ± 286 vs. 234 ± 257 m; Z = 2.06, P = 0.038, n = 53), with highest separations to other females (Table 2).

Male–male dyads showed variability in accordance with variability in extent of kernel overlap. Whereas 13 male–male dyads exhibited random interaction (χ21 ≤ 2.65, P ≥ 0.103; Table 2, ESM Fig. 3), the three male dyads with exceptionally high kernel overlap showed high attraction (χ21 ≥ 416.33, P < 0.001) and were found within 50 m at 21, 43, and 88 %, respectively, of the positions tracked (Table 2). Figure 3 shows the respective deviation of observed inter-individual distances from expected values for these three dyads. They showed lower average inter-individual distances than expected and differed in this respect from all other dyads (Mann–Whitney U test: Z = 2.81, P = 0.005, n = 36; Fig. 3). Because these males had stable social partners, we refer to them as “associated males” as opposed to “solitary males”. The association M6–M7 was found in close proximity to M9 more often than expected (χ21 ≥ 8.80, P < 0.003; Table 2). This may, however, represent an artifact of temporary close proximity at a mating site where males aggregated.
https://static-content.springer.com/image/art%3A10.1007%2Fs00265-013-1581-y/MediaObjects/265_2013_1581_Fig3_HTML.gif
Fig. 3

Cumulative probabilities of observed (red circles) and expected (black line) inter-individual distances for three simultaneously tracked male–male dyads classified as “associated”. Expected values were calculated according to Doncaster (1990) from all possible n2 distances between n positions of both individuals. As opposed to solitary males (cf. ESM, Fig. 4), associated males can be recognized by large deviation of observed proximity from random expectation

Male–female dyads interacted randomly (χ21 ≤ 3.42, P ≥ 0.076; Table 2), except for one male association (M6–M7) being found close to a female (F4) more often than expected (χ21 ≥ 58.11, P < 0.001; Table 2, ESM Fig. 4). This attraction was mainly driven by close proximity during the mating season. Only one male–female dyad (M1–F1) actively avoided each other at the critical distance (χ21 = 10.89, P < 0.001). This dyad, however, could not be shown to avoid each other in comparisons of observed and expected average distances (Table 2). MCP overlap between these two individuals was relatively low (36.9 %) and interaction assessment may therefore be less reliable.

Differences between solitary and associated males

In comparison to solitary males, associated males inhabited larger MCP ranges (LMM: t = 5.35, P < 0.001, n = 9; ESM Table 2). Differences in range overlap could not be globally evaluated because two dyads of associates ranged far outside the study area, where information about the number of resident males and females was unavailable. The only associated dyad within the main study area did not show signs of territorial defense, however. Individual-based comparison revealed that associated males overlapped with ranges of non-associated males by 30.2 ± 17.4 %, which was comparable to solitary males' overlap with other males (31.5 ± 12.7 %; Table 2). Absolute 50 % kernel areas overlapping with females did not differ between associated and solitary males (Mann–Whitney U test: Z = −0.75, P = 0.456, n = 9). A difference emerged, however, in kernel overlap of male associations among each other when compared to other male–male dyads, with different associations avoiding each other's kernel areas (Mann–Whitney U test: Z = −2.03, P = 0.042, n = 20; Table 2). This pattern could not be confirmed for inter-individual distances, where factors of observed distances between associated males were comparable to those of solitary male–male dyads (Z = 0.42, P = 0.671, n = 12).

Permanence and stability of male associations

While our GPS-based tracking data documented male associations only for up to a month, stable compositions of male associations were observed over several years. Association 1 (M1–M2) had first been observed in 2007 and was since observed every year until the death of M2 in 2011. Association 2 (M3–M4) had first been observed in 2009 and was regularly re-sighted until 2011. Association 3 (M6–M7) could not be trapped before 2010, but an early photo showed both males roaming together in 2007. Because there has neither been any documentation of a change in association partners nor of secondary separation (apart from death), we assume male associations in fossas to be of a long-term or even permanent nature.

Discussion

This preliminary study of a small number of females suggested no indication for female sociality, but indications for female territoriality. Furthermore, males exhibited a tendency to expanded their ranges with the onset of the mating season and shifted their ranges in space, whereas those of females remained stable in space and time. Male associations were characterized by extensive dyadic home range overlap and frequent close proximity. In contrast to solitary males, associated males inhabited larger ranges and therefore potentially overlapped with more females, even though we were unable to quantify the latter. There were no indications for male territoriality, suggesting that male association serves purposes other than territorial defense. Thus, female fossas are solitary, whereas males can either live solitarily or permanently associated with other males. Such intraspecific social variability is a rare phenomenon among mammals that we discuss in more detail below.

Female social organization

The observed pattern of female social organization corresponds to general predictions for a hypercarnivorous species, where females do not compete with other females for unpredictably distributed resources. Fossa females appear to be territorial (see also Hawkins and Racey 2005). Although the present sample size only permits preliminary conclusions, and future studies with larger sample size may document more variation, we found an unrelated female dyad to have exclusive ranges, and both range locations and sizes of four females were stable in space and time. The only exception was a mother–daughter dyad (F1–F3), which shared a common territory but still avoided social contact, as indicated by their exclusive 50 % kernels, random pattern of dynamic interaction and an average inter-individual distance of 2.6 km. We also found no indication for an effect of a second female's presence on territory size. This is generally predicted by the resource dispersion hypothesis if resource richness is high enough to allow tolerance of conspecifics at low cost. Relatedness may reduce the costs of tolerance in this case (Lindström 1986), especially when both females preferentially use different areas within their territory. The stability of female territory size across seasons and years suggests that territory size is adapted to resource dispersion and to the minimum territory size needed to sustain its inhabitant (and potential dependent offspring) during lean periods.

What remains to be clarified, however, is why females seem to be socially intolerant, whereas some males tolerate associations even with unrelated males. Recent research on wolverines (Dalerum 2005; Dalerum et al. 2006) suggested that solitary carnivores may be much more socially tolerant than previously assumed (cf. Kleiman and Eisenberg 1973). In fact, carnivores show high social flexibility in grouping tendencies when resource availability increases (e.g., red foxes: reviewed in Cavallini 1996; badgers: reviewed in da Silva et al. 1993; Woodroffe and Macdonald 1993). In accordance with the RDH, carnivore group size is therefore likely limited by resource productivity alone, suggesting that a change in resource patch richness would promote sociality even in otherwise strictly solitary species (such as the wolverine). If the biggest prey type (a 3 kg lemur of the genus Propithecus) or the highest density prey (Tenrecidae) in Kirindy Forest do not permit association of more than two individuals, fossa females with dependent offspring cannot afford tolerance of other individuals. Food competition and potential reproductive failure as a result of social stress, which has been documented in other solitary carnivores (da Silva et al. 1994; Dalerum et al. 2006), may therefore explain why fossa females are socially intolerant.

Male social organization

Ranging patterns of male fossas were only partly consistent with general patterns found in other polygynous carnivores. Males had larger ranges than females, presumably because they compete not only for food but also for females. Because the fossa's mating system is characterized by seasonal aggregation at specific mating trees (Hawkins and Racey 2009), it may be more important for males to keep track of the locations of mating trees than to defend female territories year-round. Males are therefore more likely to benefit from seasonal expansion of their range to visit several mating trees than from maximizing exclusive overlap with females in this seasonal breeder. In this respect, fossa differ from other solitary carnivores, where female territoriality promotes year-round male defense of females (Sandell 1989).

The most striking result of this study concerns facultative male sociality. As revealed by Doncaster's dynamic interaction test, some males associated with stable partners, even though the degree of cohesion varied extensively, whereas other males were solitary. Associated males' ranges overlapped much more than those of any other dyad, and these males clearly deviated in their interaction pattern from random expectation. The coexistence of two male tactics of sociality in fossa is unlikely to reflect an artifact of habitat loss or a consequence of local heterogeneity in habitat quality because both patterns occurred inside continuous and homogeneous parts of the forest. Instead, it may be explained by one of the following hypotheses that are not mutually exclusive and consider both ultimate and proximate factors: kin associations, bachelor groups, reproductive coalitions, or constraints on group formation suffered by solitary males.

First, male associations in fossa may represent a convergent behavioral adaptation to associations of littermates in cheetah. Similarities between these two species exist in (1) association size, which in both cases corresponds to the maximum litter size of three offspring, (2) the proportion of associated males in a population, as would be expected if litter sex ratio was a crucial determinant (about 60 %; Caro and Collins 1986), and the apparent (3) temporal stability of association composition. If males were socially flexible and coalition formation was temporary, some change in the composition of associations would have been likely during the course of our study. Because two males that lost their partners due to fatalities since the completion of this study have not paired up with another male, associations may only form once, but long-term observations of a larger number of dyads are required to verify this notion. We found that two out of five male associations consisted of non-relatives (Lührs et al. 2013, see Table 2 for relatedness of associates referred to in this study). Kinship alone is therefore insufficient to explain male associations in this species.

Second, groups of bachelor males have been reported from a number of group-living ungulates (e.g., Jarman 1974; Feist and McCullough 1975; Klingel 1975; Clutton-Brock et al. 1982), primates (e.g., Pusey and Packer 1987) and other taxa. These groups typically consist of young males lacking physical strength to compete with older males for access to females, providing them with benefits of group-living without the costs of reproductive competition. Associated fossa males are unlikely to represent an example of bachelor groups, however. First, male associations seem to be stable year-round and not limited to a particular reproductive season; in fact, associated males were actually observed to mate more often than solitary males during the mating season (unpublished data). Second, solitary males do not monopolize access to females. Third, there is no indication for an age difference between solitary and associated males (Lührs et al. 2013).

Third, associated males enjoy greater mating success than solitary males during female peak fertility (unpublished data). This mating advantage is correlated with superior body mass, which enhances competitive ability. Coalitionary support among associated males during or before mating is much less important, even though challenging rivals prematurely terminate more than two thirds of all copulations. Thus, associated males only benefit indirectly in reproductive competition because the year-round advantages of cooperative hunting provide them with a competitive physical edge during the short mating season (Lührs et al. 2013).

Finally, the persistence of solitary males, in combination with the observed stability in composition of associations, indicates limited male social tolerance that might permit more social flexibility. Whereas male littermates may easily associate due to familiarity, unrelated associates may depend on a sensitive developmental phase of social tolerance to ally. Associates do not differ in age from solitary males and members of a given association appear to be of similar age (Lührs et al. 2013). Furthermore, the unrelated members of one association were still subadult (Lührs et al. 2013). It may thus be this phase of early independence where males may either form associations or stay alone for the rest of their life. Again, additional observations are required more a rigorous test of this hypothesis.

Acknowledgments

This study was funded by the Deutsche Forschungsgemeinschaft (DFG KA 1082/17-1), the Fossa Fund of Zoo Duisburg AG and the German Primate Center GmbH (DPZ). We thank Rémy Ampataka, Tianasoa Andrianjanahary, Nielsen Rabarijaona, and Jean-Pierre Tolojanahary for field assistance; Léonard Razafimanantsoa, Rodin Rasoloarison, and Heike Klensang for administrative support; Elise Huchard and Sandra Langer for veterinary assistance; Franz Kümmeth from e-obs GmbH for technical support; Melanie Dammhahn, Cornelia Kraus, Peter Waser, and three anonymous referees for very helpful comments. We thank the Département de Biologie Animale de l'Université d'Antananarivo, the Commission Tripartite CAFF, and CNFEREF Morondava for their authorization and support of this study.

Ethical standards

This study is in compliance with animal care regulations and applicable national laws of Germany and Madagascar. All research protocols were approved by the responsible authorities in Germany (Bundesministerium für Naturschutz, BfN) and Madagascar (Ministère de l'Environnement et des Eaux et Forêts, MINEEF).

Supplementary material

265_2013_1581_Fig4_ESM.jpg (75 kb)
Fig. 1

Selected plots of MCP home ranges per sex and month. Home ranges (colored lines) are presented within the Kirindy Forest (green contours) and marked individually (females F1–F4 on the left, males M1–M9 on the right). Ranges of the month stated are in thicker lines than those from other years (males) or months (females), which were included as additional information. For female ranges, resource information such as the course of a river (blue) and a central unforested area (green contour) was additionally provided. Female F4 (lower left) was tracked singly and therefore plotted with potential ranges of other females and its 50–90 % kernel ranges. For male ranges, additional information is provided for known and assumed locations of mating trees (red hash) in the area. Members of a male association are denoted with a “+” between their IDs. (JPEG 74 kb)

265_2013_1581_MOESM1_ESM.tif (732 kb)
High-resolution image (TIFF 732 kb)
265_2013_1581_Fig5_ESM.jpg (25 kb)
Fig. 1

Selected plots of MCP home ranges per sex and month. Home ranges (colored lines) are presented within the Kirindy Forest (green contours) and marked individually (females F1–F4 on the left, males M1–M9 on the right). Ranges of the month stated are in thicker lines than those from other years (males) or months (females), which were included as additional information. For female ranges, resource information such as the course of a river (blue) and a central unforested area (green contour) was additionally provided. Female F4 (lower left) was tracked singly and therefore plotted with potential ranges of other females and its 50–90 % kernel ranges. For male ranges, additional information is provided for known and assumed locations of mating trees (red hash) in the area. Members of a male association are denoted with a “+” between their IDs. (JPEG 74 kb)

265_2013_1581_MOESM2_ESM.tif (206 kb)
High-resolution image (TIFF 206 kb)
265_2013_1581_Fig6_ESM.jpg (115 kb)
Fig. 1

Selected plots of MCP home ranges per sex and month. Home ranges (colored lines) are presented within the Kirindy Forest (green contours) and marked individually (females F1–F4 on the left, males M1–M9 on the right). Ranges of the month stated are in thicker lines than those from other years (males) or months (females), which were included as additional information. For female ranges, resource information such as the course of a river (blue) and a central unforested area (green contour) was additionally provided. Female F4 (lower left) was tracked singly and therefore plotted with potential ranges of other females and its 50–90 % kernel ranges. For male ranges, additional information is provided for known and assumed locations of mating trees (red hash) in the area. Members of a male association are denoted with a “+” between their IDs. (JPEG 74 kb)

265_2013_1581_MOESM3_ESM.tif (829 kb)
High-resolution image (TIFF 829 kb)
265_2013_1581_Fig7_ESM.jpg (120 kb)
Fig. 1

Selected plots of MCP home ranges per sex and month. Home ranges (colored lines) are presented within the Kirindy Forest (green contours) and marked individually (females F1–F4 on the left, males M1–M9 on the right). Ranges of the month stated are in thicker lines than those from other years (males) or months (females), which were included as additional information. For female ranges, resource information such as the course of a river (blue) and a central unforested area (green contour) was additionally provided. Female F4 (lower left) was tracked singly and therefore plotted with potential ranges of other females and its 50–90 % kernel ranges. For male ranges, additional information is provided for known and assumed locations of mating trees (red hash) in the area. Members of a male association are denoted with a “+” between their IDs. (JPEG 74 kb)

265_2013_1581_MOESM4_ESM.tif (1 mb)
High-resolution image (TIFF 1,074 kb)
265_2013_1581_MOESM5_ESM.pdf (49 kb)
ESM Tables(PDF 49 kb)

Copyright information

© Springer-Verlag Berlin Heidelberg 2013