Tortoises in arid, unpredictable regions may use a bet-hedging strategy of regularly laying small clutches regardless of actual environmental conditions so that some hatchlings will emerge when conditions are favourable. Karoo dwarf tortoises (Chersobius boulengeri) are endangered in their arid, unpredictable range in South Africa, yet studies of reproduction are lacking. I used radiography and captive breeding to study reproduction in wild (n = 22, 2018–2022) and captive (n = 2, 2020–2023) individuals, expecting to find bet-hedging characteristics. My expectation was that females would regularly lay eggs in summer when most rain falls. In addition, I expected that females would produce consistently large eggs and hatchlings capable of surviving their harsh environment. Indeed, females produced multiple clutches of large single eggs per year, some even during drought. However, the egg-laying season had a unique bimodal pattern in autumn and spring, possibly dispersing hatching events throughout the unpredictable rainfall season. Fecundity was low, approximately 5 eggs per year when resources were unlimited. The large eggs were usually wider than pelvic width, but egg size was nevertheless constrained by female size. Hatchling size correlated positively with egg size, confirming a potential survival benefit. Surprisingly, egg size decreased in successive clutches in a season, apparently to increase offspring frequency while resources were limited. Reproduction of Karoo dwarf tortoises is unlikely to offset current predation levels that contribute to their endangered status, thus predation reduction should be a conservation priority. Projected regional aridification due to climate change may pose a previously underestimated threat because females lay fewer eggs during drought, females that lay eggs during drought may suffer increased mortality when body reserves cannot be restored, and hatchlings (particularly from late clutches) may have insufficient reserves for survival under aridified conditions. Conservation plans might focus on sites least affected by aridification.

Study site and recordings in the wild

The study site consisted of a rocky mountain slope (altitude ~1,400 m) with adjacent plateaus, exposed dolerite sills, and dry river beds in the Northern Cape Province of South Africa (Loehr, 2023b; due to the sensitivity of this population the coordinates are withheld here, but are recorded on the biodiversity database of CapeNature, Western Cape Province, South Africa). Long-term modelled (30 years of hourly weather simulations) climatic conditions show that the site is arid (annual rainfall 162 mm, largely falling in summer), with mean minimum and maximum monthly temperatures ranging between 1–14 and 14–30 °C, respectively (www.meteoblue.com). I placed data loggers (Hobo S-THB-M00x, RS3-B, and S-RGF-M002 connected to H21-USB; Onset Computer Corporation, Bourne, MA, USA) in the centre of the study site to record temperature and rainfall during this study.

Between 20 February and 8 March 2018, I captured 16 female tortoises through inspection of holes under and among rocks. Females were equipped with radio transmitters (type RI-2B with internal antenna; Holohil Systems, Carp, ON, Canada), which were glued onto the posterior costal scutes (equipment mass ≤9.7% of body mass) to not obstruct mating. In summer (20 February and 8 March 2018, and 12 March 2019) and in spring (2 October, 19 October and 7 November 2018, and 9 October 2019), females were tracked, transported by hiking and driving to the nearest radiography facility (~185 km), radiographed dorso-ventrally at 50 kV for 0.25 s at 50 mA with 6–10 tortoises placed on each radiograph, and released at the location of capture within 24 h after tracking. Prior to release, females were offered drinking water to replenish voided urine. On 8 March 2018, two opportunistically captured females were also radiographed, whereas two females with transmitters were found dead on 9 October 2019, and the signal of one female was lost on 7 November 2018. In autumn (13 April 2022), after all transmitters had been removed in March 2020, I captured and radiographed four females, one of which had been radiographed previously. All radiographed females appeared mature, based on the lack of male characteristics (e.g., large tail, concave plastron) and straight carapace lengths that exceeded the minimum male straight carapace length.

Each year in February, and for each female not captured by radiotelemetry, I used digital sliding calipers to record the nearest 0.01 mm straight carapace length (SCL), maximum shell width (SW), maximum shell height (SH), and plastron length (PL). Shell volume (SV, cm³) was calculated as π × SCL × SW × SH / 6,000. Because all females appeared mature, I assumed that growth would be negligible within years. Body mass (BM) was recorded on each radiography date and used to calculate body condition (BC) as BM / SV. Radiographs were inspected for females gravid with shelled eggs, and successive clutches were distinguished based on the level of egg calcification (e.g., successive radiographs of a female showing a heavily and lightly calcified egg indicated different clutches) and egg dimensions. For each female on the radiographs, I measured PL and pelvic width (PW, mm, at the largest gap between the ilia), and for each egg, I measured egg length (EL, mm) and egg width (EW, mm), to the nearest 0.01 mm. Egg volume (EV, cm³) was calculated as π × EL × EW² / 6,000. I estimated egg mass (EM) based on the egg density of captive eggs in this study and subtracted EM from female BM to calculate BC of gravid females excluding egg mass. Measurements of radiographs were performed using either DICOM files (MiViewer 10.25.0, MillenSys, Cairo, Egypt) or bitmaps with scale bar (ImageJ v1.50i, LOCI, University of Wisconsin, Madison, WI, USA). I verified recordings by comparing PL on radiographs with actual PL. One radiograph (13 April 2022) was not available as DICOM or bitmap with a scale bar, so I used the ratio between PL on the radiograph and actual PL to determine the scale.

Husbandry and recordings in captivity

In March 2019, I captured two females and two males outside the core of the study site and transported them to indoor facilities in the northern hemisphere (Netherlands), for study until 2023. Both females had very low body masses and appeared emaciated by drought. Consequently, I provided a northern hemisphere seasonal cycle (i.e., March representing spring), delaying unfavourable winter conditions (Loehr, 2023a). Females were individually housed in well-structured enclosures, fed ad libitum on a fibre-rich diet low in sugars and proteins, and seasonal cycles of temperatures and photoperiod mimicked those in the wild. Enclosure temperatures were monitored with Pt100 sensors (Yageo Nexensos, New Taipei City, Taiwan) connected to a controller (Siemens LOGO!, Munich, Germany). Males were occasionally introduced in female enclosures to mate.

When an egg had been produced, it was excavated immediately. I measured EL and EW to the nearest 0.01 mm with digital sliding calipers, and recorded egg mass (EM) to the nearest 0.1 g with a digital balance. Egg dimensions were used to calculate EV. Eggs were transferred to an incubator (IPP30plus, Memmert, Schwabach, Germany), where they incubated at a diurnal temperature cycle, interrupted by a 20-day period in a constant-temperature incubator (INCU-Line IL 10, VWR International, Amsterdam, Netherlands, retrofitted with a CAREL ir33 temperature controller, CAREL Industries, Padova, Italy) to control offspring sex assuming temperature-dependent sex determination. Initially, the diurnal temperature cycle ranged from 28.0–33.0 °C, but this was changed to 26.0–31.0 °C from 10 September 2021 onwards to reduce the development of supernumeral scutes. During the constant-temperature phase (incubation days 30–50 for the first two eggs, 23–43 for all other eggs), eggs were incubated at 30 °C (to produce males) or 33.0 °C (to produce females), but the latter temperature was reduced to 32.5 °C from 4 April 2023 onwards. I ensured that incubation treatments were evenly distributed among eggs from the two females. After hatching, hatchlings were placed on wet tissue for 12–24 h, after which I measured SCL, SW, SH, and BM, enabling calculation of SV using the equation that had been used for adults. At the end of this study, husbandry and breeding were continued in a studbook under the auspices of Dwarf Tortoise Conservation (IJsselstein, Netherlands) and the European Studbook Foundation (Sneek, Netherlands), with the aim to develop an ex-situ assurance population.