Population Study on Astrochelys Radiata

Home , Astrochelys, Geochelone, Kinixys, Radiated tortoise

Population study on Astrochelys radiata

SALAMANDRA 45 4 219-232 Rheinbach, 20 November 2009 ISSN 0036-3375

Population study on Astrochelys radiata (Shaw, 1802) in the Tsimanampetsotsa National Park, southwest Madagascar

Jutta M. Hammer & Olga Ramilijaona

Abstract. During three months of studying a population of Astrochelys radiata in the Tsimanampet- sotsa National Park, data for 32 tortoises were recorded. On two parallel transects, carapace length and weight of each animal were carried out and the exact location of each animal discovered during surveys recorded. Population densities of 32 to 47 animals per km² were estimated. A four-day survey to the eastern boundary of the park was carried out to gain data from a less human inﬂuenced region. Even though the study took place during the dry season the tortoises were found in abundance. During this study a great number of large and therefore old tortoises were found on the transects. In contrast, the comparison study near the eastern edge of the Park exclusively yielded large tortoises. This fact could not be suit- ably analysed within this study due to time constraints but should be taken into consideration in future studies. Transect T2 showed greater population densities of Astrochelys radiata compared to transect T. Even though low numbers of tortoises were recorded during the transect study a viable population of A. radiata at Tsimanampetsotsa can be inferred. Keywords. Testudinidae, Astrochelys radiata, Madagascar, Tsimanampetsotsa, morphology, ecology, population structure.

Introduction diata have been found in the region of Mo- rondava, about 300 km north of its current The island of Madagascar is well known for range area (Pedrono & Smith 2003). Today its species richness and high endemism, espe- the distribution of the radiated tortoise is re- cially among its herpetofauna. Species diver- stricted to a narrow band of about 00 km sity reaches an endemism rate of 99% in rep- along the south-western coast of the island of tiles and 93% among amphibians (Nussbaum Madagascar (O’Brien 2002, Fig. ). North of & Raxworthy 2000, Glaw & Vences 2007). Toliara and to the east of Ambondro the spe- The family Testudinidae is represented by ﬁve cies is almost extinct nowadays (Juvik 975). tortoise species in Madagascar, four of which Lewis (995) reported a decline of the tor- are endemic to the Indian Ocean island (Leu- toises’ range and population densities with- teritz 2005, Glaw & Vences 2007). Where- in the last ten years. The International Union as tortoise species in the north (Kinixys belli- for the Conservation of Nature and Natural ana), in the northwest (Astrochelys yniphora) Resources (IUCN) classiﬁed the radiated tor- and in the western part of Madagascar (Pyxis toise as a vulnerable species in 982. Only planicauda) each show a restricted distribu- after a recent tortoise specialist workshop tion area, both southern tortoise species (As- hosted by the IUCN, Wildlife Conservation trochelys radiata, Pyxis arachnoides) distribu- Society and Conservation International, the tion ranges stretch along the south-western status of A. radiata was upgraded to critically and southern coast of Madagascar. The lat- endangered (IUCN 2008). ter species has been divided into three sub- Typical habitat where A. radiata can be species inhabiting adjacent regions (Glaw found comprises the dry deciduous forest in & Vences 2007). Subfossil remains of A. ra- the Mahafaly Plateau south of Toliara (Nuss-

Fig. 1. Schematic map of southern Madagascar showing estimated ranges of Astrochelys radiata in the years 1865, 1975 and 2000 (data from O’Brien 2002). baum & Raxworthy 2000, Seddon et al. recommend putting additional areas of the 2000). Adult specimens may reach up to 38 Mahafaly Plateau under protection. The ra- cm carapace length and weigh up to 3 kilo- diated tortoise is amongst the most common grams. Sexual dimorphism in adult tortoises reptiles within the national park even though is indicated by a concave plastron and a long- its population density was calculated to be er tail in male specimens (Pedrono & Smith only 27 animals per km² (Leuteritz 2002), 2003). Carapace colour varies from bright whereas Lewis (995) found densities of 262 yellow to nearly black animals displaying a animals per km² in Tsimanampetsotsa. Tsi- pattern of stripes radiating from the centre manampetsotsa first received protected sta- of each carapace scute. In old specimens the tus in 927 when a nature reserve was in- colouration may change to a more uniform stalled. In 966 the protection status was con- colour. verted to that of a national park embracing Lying within the core distribution area of 43,200 ha of forest and wetland (Goodman A. radiata, the Tsimanampetsotsa Nation- et al. 2002). The residents of surrounding vil- al Park draws scientific attention due to its lages consider the park, and especially Lake unique deciduous dry forest. As this forest Tsimanampetsotsa, sacred. Within its con- type belongs among the most endangered fines they both collect medicinal plants and habitats, Du Puy & Moat (998) strongly bury their dead below the Mahafaly Plateau.

220 Population study on Astrochelys radiata

The flora of the national park is dominated geological and floral structure three differ- by xerophytic plants, of the Didieraceae and ent types of habitat are discernible within Euphorbiaceae families, and contains a high the study area. On the northern banks of percentage of endemic plant species (AN- the Manampetse salt lake, a forest forma- GAP et al. 999). The management of the Na- tion consists of a thick thornbush scrub and tional Park is being carried out by Madagas- trees of up to seven meters in height. Main- car National Parks (formerly called Associa- ly on sandy ground, the most common spe- tion Nationale pour la Gestion des Aires Pro- cies within this forest formation are Salva- tégées or ANGAP), which is represented by dora angustifolia, Euphorbia laro, Gyrocarpus an office in the nearby village of Efoetse. americanus and Grewia grevei. Aligning with The removal of animals seems to remain the eastern shore of the salt lake a limestone the greatest threat to wild populations (Dur- cliff stretches in a northwestern to southeast- rell et al. 989, Behler 2002, Lewis 995, ern direction. On this limestone cliff and the O’Brien et al. 2003, Glaw & Vences 2007). adjacent plateau Alluaudia comosa, Diospyros ANGAP (200) reports harvesting of A. ra- manampetse and Commiphora mahafalen- diata in the western part of Tsimanampet- sis are the most common plants. The thick sotsa, both outside and inside the protected thornbush reaches only four meters high and area. Within reach of the seashore tortoises forms a dense, nearly impenetrable wall of can be easily shipped to Toliara. Collections vegetation. East of the calcite plateau a sink seem to supply both national food markets formation containing red sands predomi- and the international pet trade (O’Brien nates. Large trees are rarely found in this type et al. 2003). Therefore, a decline of tortoise of habitat which is dominated by Didiera populations in the national park is at severe madagascariensis. Other common plants are risk, and the protection of A. radiata is a spe- Delonix andansonioïdes, Euphorbia plagian- cific declared aim of the programme of the tha and Gyrocarpus americanus. Tsimanampetsotsa National Park (ANGAP Two former zebu-cart tracks were used 200). Still, a national park can only help pro- as transects in the western study area and tect an existing population by preserving its henceforth are referred to as transect  (T) natural habitat. Therefore this study aims to and transect 2 (T2). Running approximately determine more information about the sta- parallel, both transects cross all three habitat tus of A. radiata within the National Park of types. They are of 2.9 (T) and 2.4 (T2) kilom- Tsimanampetsotsa. Does a viable population eters in length, separated by an average dis- of radiated tortoises exist in Tsimanampet- tance of .5 kilometres (Fig. 2), which allows sotsa? How can this population be character- us to assume independent samples in tortoise ised? Is this national park of importance for data acquisition. Both transects were marked tortoise protection, and does it contain a vi- using flagging tape, and waypoints along the able reserve (or assurance) population of A. route were recorded using GPS-technology radiata? Which human activities within the (geko 30, Garmin). national park influence tortoise habitat or be- The survey was carried out during the dry haviour? What preservation measures should season between 23 May and 9 August 2006. be enacted to ensure the survival of A. radia- Transects were walked twice daily from 8:00 ta in Tsimanampetsotsa? to 2:00 h and from 4:00 to 8:00 h in four- day periods starting on 29 May 2006, 5 June 2006, 24 June 2006, 0 July 2006, 24 July Material and methods 2006 and 09 August 2006. Both transects were connected and could be walked in a cir- The main study area (which will be referred cle. During the survey all four possibilities of to as the western study area) lies in the north- starting positions on the transects were var- western part of Tsimanampetsotsa. Due to ied to allow equal daytime research during

221 Jutta M. Hammer et al.

Fig. 2. Location of the permanent transects and survey squares within Tsimanampetsotsa National Park compiled from GPS data mapped in the study area. the survey. With one exception these four- same number was notched into the marginal day-periods of transect walking were evenly scutes using a saw blade (Cagle 939, Leute- distributed during the study period. Find- ritz 2002). Juveniles were marked only with ing a tortoise the perpendicular distance of the marker pen. To estimate the visual field the animal to the transect was measured us- on the transects the perpendicular distance ing a tape measure (30 m). The exact loca- to a tortoise dummy was measured using a tion of each tortoise was also recorded with a tape measure (30 m) placing the dummy at GPS. Measurements of carapace and plastron greatest visibility. This measurement was car- length, body height and width of each animal ried out every 20 m along both transects. The were carried out using a measuring tape (.5 m) and a folding rule (2.0 m). Body mass was measured on a digital scale (My Weigh Ul- tra Ship 55). Where possible the animal’s sex was determined based on plastron characteristics and tail length as explained. All animals whose gender could not be ascertained were recorded as juveniles. Age estimation of each tortoise was carried out. To quanti- fy sexual dimorphism in A. radiata the anal fork width (in the following ASG) of each animal was measured as well as the posterior shell opening between carapace and plastron (LAP) (Fig. 3) using calipers (INOX Wisent, 5 cm). These measurements were carried Fig. 3. Ventral view of plastron of a male radiated tortoise. The concave shape is clearly to be seen, out even if there was no unequivocal sexual as well as the thickened anal plastron scutes. The classification. In order of their discovery all picture shows the parts of anal scutes measured in animals were marked with a number on the the study (LAP: posterior shell opening between fifth central scute using a black felt pen. The plastron and carapace; ASG: anal fork width).

222 Population study on Astrochelys radiata visual field calculated from the visibility and are fixed at the initial sighting position; they transect length served to estimate population don’t move before being detected and they densities of A. radiata. To record mean daily are not counted twice; (3) distances and an- temperature in the survey area minimum and gles from the transect line to the animal are maximum temperature were measured twice measured exactly; (4) sightings of individual daily at 7:00 and 9:00 h local time. Precipita- animals are independent events. tion was recorded reading a pluviometer dai- Calculation of population densities were ly at 7:00 h local time. carried out using the King formula (Brower Besides the transects, three squares of 00 et al. 989): × 00 m were marked, choosing one exami- 104 n2 nation square in each habitat type (see Fig. D = 2). These squares were chosen at random 2L ∑ di to compare results to the transect data and searched four times during the study period D: the populations’ density (number of in- between 8:00 to :00 h when tortoise activ- dividuals per hectare); n: number of animals ity was considered to be still quite low to en- sighted; L: transect length (in metres); di: dis- sure the tortoises would not move out of the tance of animals sighted to the transect (in study area before being detected. All data of metres); 04: factor for converting m² to ha. tortoises found on both transects, the con- This method assumes that there is a ran- nections in between and on the three marked dom distribution of animals over the area squares were used in the population study, sampled and all animals are equally likely to be including those animals that were found out- counted, the sighting of one animal does not side the six standardised transect surveys. influence sightings of other animals and no During a four day trip to the eastern animal is counted more than once (Brower boundary of the national park, short transects et al. 989). Additionally, and for comparison were installed along the route. Every two kil- purposes, population densities were calculat- ometres a transect of 200 m was carried out ed using the computer program DISTANCE. right-angled to the route, alternately to the A hazard-rate key function was chosen, com- right and the left side. Data on all tortoises plemented by a cosine series expansion to found in these short transects was record- ensure the best reproduction of tortoise dis- ed. Additionally a transect of 2.5 km length persal data. In the programme, the integrat- was walked at the eastern part of the park to ed AIC term (Akaike Information Criterion) estimate tortoise density in this inland, less has been used to select the key function for human influenced area. All data on tortois- best data representation. Additionally a right es found on the walk back was also record- truncation of the data at 5% was carried out ed. All results from this trip are referred to (Buckland et al. 200). To perform statisti- as data from eastern study area. However cal analysis of the data the following tests and tortoise data found on the four- day trip to calculations were used: To test the temporal the eastern boundary of the park is excluded and local dispersion of recaptures for sexual in the population study and serves only for differences a Kruskal Wallis analysis was cal- comparison purposes. culated. The Mann Whitney U-Test was per- Tortoise population density estimations formed to interpret different questions con- refer to the animals found in the standard- cerning tortoise data: Sexual differences in ised transect surveys. To estimate population weight, age, carapace length, anal fork width density from transect surveys the follow- (ASG) and the posterior shell opening (LAP) ing assumptions have to be taken into con- were analysed using the Mann-Whitney-U- sideration (Krebs 998): () animals on the Test. Mean carapace length and weight on the transect line will never be missed; (2) animals two transects were compared as well as the

223 Jutta M. Hammer et al. mean age dispersion of the tortoises in the vey days, whereas another 22 tortoises were western study area and that of animals found detected outside the standardised catching in the east of the National Park. A Spearman periods. In the survey squares a total of 9 tor- correlation was performed to implicate the toises A. radiata were found and another 20 carapace length of the tortoises from their first sightings occurred on connecting paths weight, their age and the anal fork width of between transect T and T2. the animals (ASG). This test was also used to compare the number of tortoises found during the four-day-periods of transect walking. Behaviour The age dispersion of male and female- tor toises in the western study area as well as in About 56% of the tortoises were active upon the eastern and western part of the park were detection, which means they were walking or compared using a Chi²-test. Furthermore eating. The other 44% of animals were resting this test was used to examine the activity and in the shade, sunbathing or sleeping. There recapture data of the tortoises towards sex- was no gender based difference in activity at ual influenced differences in tortoise disper- the time of detection (p = 0.973, Chi² test val- sal. All statistical tests were performed using ue = 0,055). Activity of tortoises is assumed the computer program SPSS 3.0. For further to be restricted to daytime. During surveys descriptions of all tests see Bühl (2006). All before 9:00 a.m. and after 5:00 p.m. little tor- GPS data mapped in the study area were ed- toise movement was detected. Animal sight- ited using the computer program ArcMap. ings occurred during all daytime hours.

Results Population structure

During the three-month survey in Tsiman- Out of 03 radiated tortoises found in the ampetsotsa National Park we made a total western study area, we confirmed the sex of 69 sightings of A. radiata, two of Pyx- of 3 males and 27 females giving a sex ra- is arachnoides as well as finding two empty tio (male: female) of .5. The sex ratio in the shells of A. radiata in Tsimanampetsotsa Na- eastern study area was calculated to 0.63 (0 tional Park (Tab. ). The A. radiata observa- males, 6 females). tions in the western study area were represented by 03 initial sightings, 37 recaptures and a single empty shell. In the areas outside Morphology of the main study site, we observed three A. radiata near the village of Efoetse, 26 speci- The mean carapace length and weight of mens in the eastern study area and another male tortoises were significantly higher than empty shell. Within the 03 first sightings of females’ (weight: p = 0.022, carapace length: A. radiata in the western study area a total of p = 0.007; Mann-Whitney-U-Test; see Tab. 52 animals were found during transect sur- 2). Furthermore the age difference between

Tab. 1. Tortoise sightings divided by study area, sort of sighting and tortoise species.

western study area eastern study area village of Efoetse ﬁrst sightings A. radiata 03 26 3 recaptures A. radiata 37 – – empty shells A. radiata   – ﬁrst sightings P. arachnoides 2 – –

224 Population study on Astrochelys radiata

Tab. 2. Mean value and standard deviation of diﬀerent carapace measurements, as well as estimated age and weights of A. radiata, sorted by sex. Minimum and maximum values are given in parenthesis. Animals that could not be classiﬁed by their sex where recorded as juveniles. ASG: anal fork width, LAP: posterior shell opening between carapace and plastron.

weight [g] carapace length [cm] ASG [cm] LAP [cm] age [years] all tortoises 3424 ± 2229 23. ± 7.0 5.6 ± 2.0 2.7 ± .0 2.3 ± .6 (24 - 7680) (4.5 - 34.5) ( - 0.5) (0.5 - 4.9) (2 - 50) male 5520 ± 28 29.6 ± 2.7 7.8 ± .3 3.3 ± 0.7 3.8 ± 0.6 (320 - 7680) (24.5 - 34.5) (5.9 - 0.5) (2. - 4.7) (3 - 50) female 4578 ± 482 26.9 ± 3.7 5.8 ± 0.8 3.6 ± 0.8 26.0 ± 7.5 (2080 - 780) (20.5 - 32.5) (4.2 - 7.4) (2.2 - 4.9) (5 - 40) juvenile 264 ± 708 6.5 ± 4. 3.9 ± .0 .8 ± 0. .3 ± 3.3 (24 - 2680) (4.5 - 22.5) ( - 5.9) (0.5 - 2.9) (2 - 6) male and female tortoises was significant (p bers of individuals are between 22.6 cm and = 0.034) with a greater number of older male 25 cm of carapace length. A linear correlation tortoises found, while the anal fork width dif- between the anal fork width and the cara- fers highly significantly in male and female pace length could be observed (r = 0.882; p < animals (p < 0.00). No differences could be 0.00; Spearman-Correlation). observed between the LAP-data of male and Compared to transect T2 tortoise data female radiated tortoises (p = 0.40). Both of males and females collected on transect groups differ clearly from the data of juvenile T shows greater mean carapace length and tortoises. Body mass of the tortoises within weight measures (Tab. 3). In contrast the ju- the western study area was bimodal, show- venile tortoise data shows greater means on ing peaks between -2 kg and 5-6.5 kg (Fig. the second transect. A significant difference 4). The weight and carapace length of the tor- between weight and carapace length of the toises are significantly correlated; with juve- adult animals could only be detected in the nile and adult animals treated separately in data of male carapace lengths (Mann-Whit- data examination (juvenile: r = 0.937, adult: ney-U-Test: p = 0.03). The data of the females r = 0.85 with p < 0.00 in both cases). The (weight: p = 0.058; carapace length: p = 0.) dispersion of carapace length also shows a and the weights of the males (p = 0.238) did bimodal distribution. Observed lower num- not show significant differences between the

14 12 10 8 6 4 2 number of tortoises 0 500 - 1500 2000 2500 3000 3500 4000 4500 5000 5500 6000 6500 7000 7500 8000 1000 ------0 - 501 1001 1501 2001 2501 3001 3501 4001 4501 5001 5501 6001 6501 7001 7501 weight [g] Fig. 4. Distribution of weights of Astrochelys radiata from Tsimanampetsotsa National Park, subdivided into classes of 500 g.

225 Jutta M. Hammer et al.

Tab. 3. Mean values of weights and carapace length of Astrochelys radiata including their standard deviation, divided into the two transects and the animals’ sex.

T T2 weight [g] carapace length number weigth [g] carapace length number (std.-dev.) [cm] (std.-dev.) (std.-dev.) [cm] (std.-dev.) male 587 ± 05 30.8 ± 2. 3 535 ± 367 28.5 ± 2.9 3 female 5044 ± 982 28. ± 2.9 9 3673 ± 686 24.8 ± 4.5 9 juvenile 857 ± 544 5.0 ± 3.2 4 737 ± 629 8.6 ± 2.3 6 total 36 38 two transects. In contrast the data of juvenile ed to be between ten and twenty years old, tortoises shows a significant difference be- about 65% of the tortoises in the east were es- tween the two transects (weight: p = 0.002; timated to be over 40 years old. In compar- carapace length: p = 0.006). ing these data it must be emphasized that the Looking at the tortoises with an estimat- trip to the east of the Park lasted only four ed age of over 30 years a greater number of days. The total of 26 animals found on this males were detected during the study period. excursion provide a database too small for a A total of 6 males were found with an esti- detailed comparison. However, the findings mated age of 30 years or older whereas only 7 suggest regional differences of tortoise popu- females of this age class were detected in the lations in the national park and highlight fur- study area. This discrepancy emerges even ther questions for future studies. more clearly by looking at the tortoises with Considering the mean estimated age of an estimated age of over 40 years. Four males the tortoises in both study areas (p < 0.00, were thus classified, but no female older than Mann-Whitney-U-Test) as well as for the 40 years was found. In the statistical compar- age distribution (p < 0.00, Chi²-homoge- ison of the age distribution of males and fe- neity-test) there are significant differences males in the western study area no significant between the two regions. The correlation of difference could be detected using the Chi²- age and carapace length can be described as a test (p = 0.8). Figure 5 shows the age distri- curve with a steep gradient that flattens with bution of A. radiata in both western and east- the age of the animals (Fig. 6). In this analy- ern study areas. While over 50% of tortoises sis the curve has been cut into two sections found in the western study area were estimat- with an age limit of 23 years as pivot point.

5 number of tortoises 0 1 5 9 13 17 21 25 29 33 37 >40 age [years] main study area east of the Park Fig. 5. Age distribution of radiated tortoises found in the main study area and during the trip to the east of the Park.

226 Population study on Astrochelys radiata

40 35 30 y = 0,1501x + 24,505 25 20 15 10 y = 1,0666x + 4,8194

carapace length [cm] 5 0 0 10 20 30 40 50 60 age [years] until 23 years over 23 years Fig. 6. Carapace lengths (in cm) and age distribution of Astrochelys radiata. The data is divided in animals up to the age of 23 and those older than 23 years. The formulas for the lines of best ﬁt are given in the diagram.

The group of animals up to 23 years shows test: p = 0.358). About 90% of all recaptures a strong linear correlation between the two were within 200 m of the original capture variables (r = 0.826, p < 0.00), whereas the spot. group of animals older than 23 years shows During the four-day periods of transect only a weak correlation (r = 0.494, p = 0.00; survey each transect was walked twice daily. Spearman-Correlation). The carapace length With a total length of 2853 m on transect T measurement of animals over 23 years ranges and 2349 m on transect T2; a total distance of from 24 to 35 cm. 25.9 km has been walked during the study Out of those 03 tortoises found in the period. With a total of 52 first captures of ra- western study area, a total of 25 animals were diated tortoises a theoretical distribution of recaptured. As nine animals were recaptured one tortoise every 4843.4 m or 0.2 tortois- more than once the total number of recap- es per walked km can be calculated. Figure tures adds up to 37. A well-balanced ratio of 7 clearly shows a constant increase of cumu- males, females and juveniles was detected in lative first catches of A. radiata during the the recaptures, and no sex-based component study period. This curve reaches no plateau. could be examined within recapture data A coefficient of correlation of r =  has been (Kruskal-Wallis-ANOVA: p = 0.53; Chi²- calculated for the interrelation of the number

30 25 20 15 10 5 number of tortoises 0 1 2 3 4 5 6 transect survey period Transect 1 Transect 2 Fig. 7. Cumulative tortoise catches during transect surveys. The diagram displays the number of ﬁrst sightings.

227 Jutta M. Hammer et al.

Tab. 4. Population densities of Astrochelys radiata in animals per km², calculated using the King formula, with the computer program DISTANCE and estimated via the visual ﬁeld of each transect, calculated from the sighting distance and transect length. The program DISTANCE gives a limit of conﬁdence with it’s calculations. In the category ‘sightings up to 12 m’ those tortoises with a transect distance of over 12 m were not included in the calculation. This corresponds to the truncation of 5% of all tortoise counts in the computer program DISTANCE.

King DISTANCE visual field all tortoises 39.0 37.9 (22.2 - 64.8) 34.3 sightings up to 2 m 47.0 4.9 (22.8 - 76.9) 32. T 34.0 23. (6.2 - 32.8) 23.2 T2 50.0 64.7 (23.0 - 82.2) 50.2 of capture periods and the number of tortois- served. In contrast Leuteritz et al. (2005) es found on transect T (p = 0.0; Spearman- and O’Brien (2002) both state to have ob- Correlation). The coefficient of correlation served increased tortoise activity during between the capture periods and the number morning and afternoon hours. These two sur- of tortoises on transect T2 reaches r = 0.986 veys were carried out during the rainy season (p = 0.0). The tortoise data from the stand- which leads to the assumption of a season- ardised transect surveys was used to calculate ally changing activity scheme of A. radiata. tortoise population densities (Tab 4). The re- Nussbaum & Raxworthy (2000) observed sults of the three methods of calculation show A. radiata throughout the year without per- little differences in result values. In compar- ceiving a distinct hibernation period of the ing the two transects all calculation methods tortoises. A seasonal comparison of tortoise display a greater tortoise population density activity is still lacking. Several authors men- on transect T2. tion having observed greater tortoise activity after rainfall (Nussbaum & Raxworthy 2000, Pedrono & Smith 2003), which this Discussion study also indicates. When comparing the two transects, only During a three-month survey in the Tsiman- the carapace length of male radiated tor- ampetsotsa National Park a total of 03 radi- toises shows significant differences between ated tortoises were detected in the main study animals detected on T and T2. The female area. As the diagram of cumulative tortoise tortoise data and the weights of the male tor- captures shows, only a part of the tortoise toises do not differ. Among adult tortoises no population has been recorded (Fig. 7). The significant population differences attributa- great number of tortoise finds during the first ble to the transects could be observed. There- transect period could have been a function of fore all animals detected during the transect it being the end of rainy season. The increase survey apparently belong to the same tortoise of tortoise captures on transect T2 during fifth population. and sixth transect period suggests increasing During a survey in seven different study tortoise activity at the end of dry season. Pos- areas Leuteritz (2002) provides mean cara- sibly the animals had been unable to store up pace lengths for A. radiata between 8.5 and enough fat during rainy season to outlast a 30.9 cm. Compared to Leuteritz’ (2002) data long winter break, and thus had to leave their the mean carapace length detected in this wintering grounds. Astrochelys radiata could study is quite low (23. ± 7.0 cm). This could be seen on transects during all daytime hours be interpreted as a function of greater juve- and no specific daytime activity could be ob- nile tortoise activity during the dry season, or

228 Population study on Astrochelys radiata could indicate the paucity of large tortoises their sex could also be related to the fact of in the study area. Maximum carapace lengths a male domination in old animals and there- of A. radiata are stated to lie between 38 cm fore be misinterpreted in this study as a sign (Pedrono & Smith 2003) and 40 cm (Lewis of sexual dimorphism in radiated tortoises. 995, Nussbaum & Raxworthy 2000), The characteristic shape of the anal scutes whereas Lewis (995) derives this figure from and plastron best exemplify sexual dimor- only one specific study area and gives a maxi- phism in A. radiata. A significant difference mum carapace length of 38 cm for other re- between male and female tortoises concern- gions, too. This study’s maximum carapace ing the distance of the LAP was not found. length of 34.5 cm is clearly lower than those However the data show significant differences observed in other studies. between anal fork width (ASG) in male and Correlation of carapace length and weight female tortoises. The narow but tapering anal permit an estimation of the age at which the fork of the females probably offers advantag- tortoises reach maturity. Beyond an estimat- es in egg deposition. A tendency towards a ed age of 23 years, the carapace length of radi- narrow versus wide anal fork is observable in ated tortoises increases very little. Adults ex- subadult animals and therefore the animals’ amined in this study had carapace lengths be- sex may possibly be detected thus before oth- tween 24 and 35 cm. Therefore, the carapace er sexual characteristics in plastron shape length of mature tortoises varies by up to 0 and tail length become distinct (see Fig. 3). cm; given the maximum carapace length of In the course of the study period, an even 40 cm the variation in carapace length in ful- ratio of male and female tortoises could be ly grown tortoises may be 6 cm. Lower num- ascertained. However, a remarkable propor- bers of individuals detected in the weight tion of male tortoises having an estimated classes of 250 to 4000 g and 450 to 5000 g age in excess of 30 years was detected in this could be related to lower activity of these ani- study. The age distribution of tortoises in the mals during the dry season, but could also re- east of the National Park seems unbalanced, flect the loss of tortoises during a long, severe though a relatively brief period (only four dry period a few years ago whose impact can days) was devoted to focused study there. In now be observed in lower counts of specific the less anthropogenically influenced inland weight and therefore age classes. area, exclusively animals with an estimated In this study, male tortoises appear to have age of over 30 years and carapace lengths a greater mean carapace length and weight over 30 cm were found. Both areas sur- than females in the study area. This conclu- veyed by Goodman et al. (2002) are close to sion mirrors that of O’Brien (2002) who the survey area of this study. Four years ago tabulates mean carapace lengths observed in these areas showed an uneven distribution in eight different study areas in southwestern sex ratio of A. radiata which differs from the Madagascar. However, O’Brien does not clar- one detected in this study. Goodman et al. ify whether a significant difference exists in (2002) stated a dominance of male tortoises carapace length between males and females, in the west of the Park (Mitoho) and an even as this data had originated from other anal- sex ratio in the east (Bemananteza). Among yses and statistical testing has not been car- 3 adult tortoises found in Mitoho there was ried out. Within this actual study a significant only one female A. radiata. Furthermore no difference in the age distribution of male and tortoise with an estimated age of over 40 female tortoises was observed with more old years was found (Goodman et al. 2002). The males than females, probably caused by selec- lack of old tortoises in the west of the Nation- tive collection through poachers. As older tor- al Park was confirmed in this study. toises usually attain a greater carapace length Comparing the actual study to that of the differences in tortoise size considering Goodman et al. (2002), the varying domina-

229 Jutta M. Hammer et al. tion of male and female tortoises in the study the actual study took place during the dry pe- areas with uneven sex-ratio is striking. The riod and should not be hastily compared to area of Bemananteza in the east of the Park other studies. Low population densities could shows an evenly distributed sex ratio with an be attributed to lower tortoise activity during index of .7 (Goodman et al. 2002) whereas the dry season. Considering that both Lewis in the actual study a dominance of female an- (995) and Leuteritz (2002) detected their imals was detected calculating a sex-index of lowest tortoise densities in Tsimanampetsot- 0.63 in the east of the National Park. Good- sa, a low abundance of radiated tortoises in man et al. (2002) attributed the observed sex the National Park may be inferred. Given the ratio in the stronger human-influenced area balanced proportion of male and female tor- of Mitoho to the removal of A. radiata. Fe- toises and the still impressive number of tor- male tortoises are more often collected as toises found in this dry season study, Tsiman- large females sometimes contain eggs (Leu- ampetsotsa still seems to have a viable popu- teritz et al. 2005). The continuous collec- lation of A. radiata. Goodman et al. (2002) tion of large tortoises could select for small also conclude that a viable population exists animals, resulting in the reduced carapace in Tsimanampetsotsa and emphasise that the lengths detected in the actual study. Likewise National Park provides an important con- the female-biased sex ratio found in the east servation opportunity for A. radiata. On the of the park suggests lower collection pres- other hand, ANGAP (200) warns of a highly sures beyond the Mahafaly Plateau. fragmented tortoise population due to col- Because the survey of Goodman et al. lection of animals from shoreline areas out- (2002) took place during rainy season, a com- side the national park. parison to the actual study must be framed So far, a detailed comparison of tortoise against the background of an annual cycle activity correlated to seasonal differences to of varying tortoise activity (Leuteritz et al. explain apparent variations in population 2005, O’Brien 2002). Leuteritz (2002) pro- structure of Astrochelys radiata is needed. vides population densities of 28 to 5744 ani- A detailed study might also shed light on mals per km² from seven survey areas in the whether the eastern tortoise population con- southwestern Madagascar. This was also the tains only old specimens which no longer re- biggest range in radiated tortoise densities produce. This would explain the absence of observed in last year’s studies. Lewis (995) juvenile tortoises there in the course of our calculated densities between 262 and 077 study. However, the lack of juveniles in the tortoises per km² in five survey areas. Both eastern survey area could also be explained studies stated the lowest tortoise densities by a better food supply in this region allowing in Tsimanampetsotsa National Park which juveniles to build up enough reserves for sus- leads tortoise densities ranging from 28 to tained hibernation throughout the dry sea- 262 animals per km² in this area. Surveying son. Moreover, a comparative survey during three study areas O’Brien (2002) calculated the rainy season could show clearly whether tortoise densities between 34 and 535 ani- there are differences in male and female tor- mals per km². In comparison to these studies toise activity which could thereby explain an population densities determined in the actu- annual variation in the sex ratio. al survey are considerably lower, consisting Until now, A. radiata has been protected of 32 to 47 radiated tortoises per km². Only by a ‘fady,’ a taboo which forbids local peo- the population density of 27 tortoises per ple to eat or even to touch the tortoises. This km² stated by Leuteritz (2002) in Tsiman- ‘fady’ is only respected by some local tribes ampetsotsa National Park of is lower still, in southwestern Madagascar while others whereas Leuteritz found only one radiated seem not to be constrained by this taboo. As tortoise during a three-day-trip to the Na- people from other parts of the island migrate tional Park. Again it must be emphasized that to the south the ‘fady’ weakens and tortoise

230 Population study on Astrochelys radiata populations become increasingly exploited Acknowledgements by poachers. Killing tortoises for crop protection seems to occur, however tribes who Research permission was kindly provided by respect the local ‘fady’ do not seem to coop- Madagascar National Parks (formerly ANGAP or Association Nationale pour la Gestion des Aires erate with poachers. Protégées) and the University of Antananarivo. Tortoises are still considered sacred and For the motivated and dedicated supervision of believed to contain the ancestors’ spirit when my diploma thesis, many thanks to T. Schmitt found close to a grave (Nussbaum & Rax- (University of Trier). I owe the financial support worthy 2000, Juvik 975). Several graves of this study to the DGHT (Deutsche Gesellschaft can be found along the transect T2. During für Herpetologie und Terrarienkunde), SIGS (Tor- the study period a new grave was construct- toise-Interest-Association of Switzerland), DFG ed below the limestone cliff of the Mahafa- (Ga 342/5), and WWF Germany. Many thanks ly Plateau. Therefore local people still bury to J. Jeglinski for good collaboration and com- panionship in everyday camp life. For his support their dead within the National Park, which and helpfulness I wish to thank M. Barson D. D. may possibly explain why there is a greater Rakotomanga, the director of the Tsimanampet- number of tortoises on the second transect. sotsa National Park. Many thanks to Monja, park A reduced collection of tortoises due to re- guide in Tsimanampetsotsa, who accompanied us luctance to disturb ancestors could be a fac- on the trip to the east of the Park. The project was tor here. During the construction of a new supported by ANGAP Antananarivo and Toliara, grave a considerable amount of natural veg- in particular D. Rakotomalala greatly support- etation was cut down which probably has ed our work. Many thanks to WWF Madagascar negative impacts on tortoise populations. A and especially to J. Ganzhorn and S. Goodman for their help. For their hospitality and further in- broad and sensible conservation of A. radia- formation on A. radiata, I wish to thank H. Ran- ta therefore must take into consideration the driamahazo from WCS (Wildlife Conserva- impact of human land use. tion Society) and R. E. Lewis from JWPT (Jersey Whereas there are current plans to extend Wildlife Preservation Trust) in Antananarivo. the National Park, adequate accessibility by park rangers must be given priority in order to exercise control so as to prevent poaching References of the tortoises, for despite their protection by national and international law (UNEP- ANGAP, DFS & EEDR (999): Etude pour WCMC 2009, Lewis 995), they are still col- l’élaboration d’un plan d’aménagement et de lected and sold for food and as pets. ANGAP gestion au niveau de la Réserve Naturelle In- (200) reports poachers operating even with- tégrale de Tsimanampetsotsa: Diagnostic phys- io-bio-écologique. – ANGAP, DFS, EEDR. An- in the National Park. Moreover they suspect tananarivo. some local farmers of collaborating with the poachers as they consider the tortoises pests ANGAP (200): Plan de gestion du réseau national des Aires protégées de Madagascar, Antan- that destroy their crops (Lewis 995). Left out anarivo. – unpublished report. of the ANGAP scenario is the fact that tor- Behler, J. L. (2002): Madagascar tortoise crisis: toises are wild animals that can choose their letter to CITES animal committee and con- whereabouts. By concentrating protection cerned parties, 9 January 2002. – Turtle and activities within the National Park it might Tortoise Newsletter, 5: 8-9. be forgotten that the animals may leave the Brower, E. J., J. H. Zar & C. N. von Ende (989): sanctuary and then be removed by a farm- Field and laboratory methods for general ecol- er or a poacher. Therefore, the difficult chal- ogy: third edition. – Wm. C. Brown Publish- lenge remains to raise awareness of the local ers. people and to educate them in favour of na- Buckland, S. T., D. R. Anderson, K. P. Burn- ture conservation, essential to assure the pro- ham, J. L. Laake, D. L. Borchers & L. Tho- tection of A. radiata in its natural habitat. mas (200): Introduction to distance sampling:

231 Jutta M. Hammer et al.

estimating abundance of biological popula- Geochelone radiata (Shaw, 802), in southwest tions. – Oxford University Press, New York. Madagascar. – PhD thesis, George Mason Uni- Bühl, A. (2006): SPSS 4: Einführung in die mo- versity, Fairfax, Virginia, USA. derne Datenanalyse. – Pearson Studium, Mu- Leuteritz, T. E. J., T. Lamb & J. C. Limberaza nich. (2005): Distribution, status and conservation Cagle, F. R. (939): A system of marking turtles of radiated tortoises (Geochelone radiata) in for future identification. – Copeia, 939: 70- Madagascar. – Biological Conservation, 24: 73. 45-46. Du Puy, D. J. & J. F. Moat (998): Vegetation Lewis, R. E. (995): Status of the radiated tortoise mapping and classification in Madagascar (us- (Geochelone radiata). – unpublished report, ing GIS): implications and recommendations Antananarivo, WWF Madagaskar. for the conservation of biodiversity. – pp. 97- Nussbaum, R. A. & C. J. Raxworthy (2000): 7 in Huxley, C. R., J. M. Lock & D. F. Cut- Commentary on conservation of „Sokatra“, ler (eds.): Chorology, taxonomy and ecology the radiated tortoise (Geochelone radiata) of of the African and Madagascan floras. – Royal Madagascar. – Amphibian and Reptile Conser- Botanical Gardens, Kew. vation, 2: 6-4. Durrell, L., B. Goombridge, S. Tonge & Q. O’Brien, S. H. (2002): Population dynamics and Bloxam (989): Geochelone radiata. – The con- exploitation of the radiated tortoise Geochelone servation biology of tortoises. – Occasional Pa- radiata in Madagascar. – PhD thesis, Darwin pers of the IUCN Species Survival Commision, College, University of Cambridge and Institute 5: 96-98. of Zoology, Zoological Society of London. Glaw, F. & M. Vences (2007): A field guide to the O’Brien, S. H., E. R. Emahalala, V. Beard, R. amphibians and reptiles of Madagascar. Third M. Rakotondrainy, A. Reid, V. Raharisoa edition. – Vences & Glaw Verlags GbR, Co- & T. Coulson (2003): Decline of the Madagas- logne. car radiated tortoise Geochelone radiata due to Goodman, S. M., M. J. Raherilalao, D. Rako- overexploitation. – Oryx, 37: 338-343. tomalala, D. Rakotondravony, A. P. Rase- Pedrono, M. & L. L. Smith (2003): Testudinae, limanana, H. V. Razakarivony & V. Soar- land tortoises. – pp. 95-956 in Goodman, S. imalala (2002): Inventaire des vertébrés du M. & J. P. Benstead (eds.): The Natural His- Parc National de Tsimanampetsotsa (Toliara). tory of Madagascar. – The University of Chi- – Akon’ny Ala, 28: -36. cago Press. IUCN (2008): 2008 IUCN Red List of Threatened Seddon, N., J. Tobias, J. W. Yount, J. R. Ram- Species. – www.redlist.org, 5 February 2009. anampamonjy, S. Butchart & H. Randri- Juvik, J. O. (975): The radiated tortoise of Mada- anizahana (2000): Conservation issues and gascar. – Oryx, 3: 45-48. priorities in the Mikea Forest of south-west Krebs, C. J. (998): Ecological methodology: sec- Madagascar. – Oryx, 34: 287-304. ond edition. – Benjamin, Cummings. UNEP-WCMC (2009): UNEP-WCMC Species Leuteritz, T. E. J. (2002): Distribution, status and Database: CITES-listed species – http://www. reproductive biology of the radiated tortoise, unep-wcmc.org, 5.2.2009.

Manuscript received: 8 May 2007 Authors’ addresses: Jutta Hammer, Department of Animal Ecology and Conservation, Biocenter Grindel, University of Hamburg, Martin-Luther-King-Platz 3, 20146 Hamburg, Germany, E-Mail: jutta. [email protected]; Olga Ramilijaona, Department of Animal Biology, University of Antananarivo, BP 906, Antananarivo 101, Madagascar, E-Mail: [email protected].

232