The investigation and sampling of the baobabs was approved and authorised by the Forestry Direction of the Ministry of Environment, Ecology and Forestry of Madagascar and by the Madagascar National Parks. The Madagascar National Parks, Tsimanampetsotse National Park and PARRUR project (ECOBAO) provided scientific assistance and support in the investigation. The baobabs were not endangered by the sampling. After each coring, the increment borer was disinfected with methyl alcohol. The small coring holes were sealed with Steriseal (Efekto), a special polymer sealing product, for preventing any possible infection of the trees.

The two fony baobabs are located in the Tsimanampetsotse National Park, at a distance of only 1.5 km apart from each other. The mean annual rainfall in this very arid area is around 350 mm.

The”Grandmother” (La grand-mère) is generally considered the oldest A. rubrostipa. Its GPS coordinates are 24°02.707' S, 043°45.266' E and the altitude is 36 m. It has a height of 7.47 m and a circumference at breast height (cbh) of 9.67 m. The overall wood volume is 25 m³. The tree, which has a pot shape, consists of 3 perfectly fused stems (Figs. 2–4). However, some fusion lines of stems can be identified at the upper part of the trunk.

The ”polygamous baobab” (Le baobab polygame), the largest known A. rubrostipa, is located at 24°03.060' S, 043°45.418' E and the altitude is 27 m. It has a maximum height of 14.2 m and a restored circumference (cbh) of 13.50 m. The overall wood volume is 60 m³. It exhibits a cluster structure with 6 partially fused stems. The stem from the northern extremity collapsed some time ago, but it is still alive (Figs. 5 and 6).

Several wood samples were collected from the outer part of the 3 stems of Grandmother, at convenient heights between 1.27–1.55 m. The longest samples from each stem (labelled GM-1, GM-2, GM-3) were investigated. The stem numbering (I, II, III) and the sampling positions of the longest samples are shown in Fig. 4.

Additional 5 wood samples (labelled PB-1 to PB-5) were collected from 5 stems of the polygamous baobab, at heights between 1.30–1.53 m. The stem numbering (from I to VI) is displayed in Figs. 5 and 6.

The samples were collected by using Haglöf increment borers (0.80–1.00 m long, 0.010–0.012 m inner diametre). Several small pieces/segments of the length of 0.001 m were extracted from determined positions of the 10 samples and investigated by AMS radiocarbon dating.

The standard acid-base-acid pretreatment method [13] was used for removing soluble and mobile organic components. The resulting samples were combusted to CO₂ via the closed tube combustion method [14]. Next, CO₂ was reduced to graphite on iron catalyst, under hydrogen atmosphere [15]. The graphite samples were analysed by AMS.

AMS radiocarbon measurements were performed at the NOSAMS Facility of the Woods Hole Oceanographic Institution by using the Pelletron Tandem 500 kV AMS system [16,17]. The obtained fraction modern values were ultimately converted to a radiocarbon date.

Fraction modern values were calibrated and converted into calendar ages with the OxCal v4.2 for Windows [18], by using the SHCal13 atmospheric data set [19].

Fraction modern values and radiocarbon dates bp (before present, referring to ad 1950) of 10 wood samples/segments which originate from the Grandmother (GM) and the polygamous baobab (PB) are displayed in Tables 1 and 2. Radiocarbon dates and errors were rounded to the nearest year. The accession numbers in the NOSAMS database of the 10 radiocarbon analyses are also listed in Tables 1 and 2.

The 1σ probability distribution (68.2%) was typically selected to derive calibrated age ranges. For 4 sample segments, the 1σ distribution is consistent with one range of calendar years, while for other 4 segments, the 1σ distribution corresponds to two or three ranges of calendar years. For these segments, the confidence interval of one range is considerably greater than that of the others; therefore, it was selected as the cal ad range of the sample for the purpose of this discussion. For 2 sample segments, there are two 1σ ranges with close confidence intervals. In these special cases i.e., GM-1(a) and GM-1(c), we used the higher 2σ probability distribution (95.4%) for calibration.

For obtaining single age values, we derived a mean calibrated (cal) age of each segment from the selected range (marked in bold). Calibrated ages of samples/segments represent the difference between ad 2014 and the mean ad value of the selected range, with the corresponding error. All calibrated sample/segment ages, as well as stem ages and tree ages are expressed in (calendar) yr in ad 2014. Calibrated ages and errors were rounded to the nearest 5 yr.

Even if the borer penetration in the soft wood was quasi-complete in all cases, the collected samples were much shorter. This shows that more than half of the tree is hollow, due to decay. We extracted and dated 3 segments (labelled a, b, c) from sample GM-1, which originate from the largest stem I. The innermost and oldest segment GM-1(c) corresponds to the sample end located at 0.46 m from the sampling point. Its radiocarbon date was found to be 1136 ± 16 bp, which corresponds to a calibrated age of 1070 ± 50 yr. The second oldest dated segment GM-2(b), which corresponds to a depth of 0.20 m in the wood, has an age of 650 yr (neglecting the error). The ages and distances of segments GM-1(b) and GM-1(c) indicate that stem I grew 0.26 m in 420 yr in the sampling direction. This value indicates a mean radial increase of 0.62 x 10^-3 m yr^-1 over the time frame ad 945–1365.

The outermost segment GM-1(a), which was adjacent to the bark, is not as young as it would be expected. Its calibrated age of 250 yr shows that the large stem I stopped growing around ad 1765. One should, however, mention that the calibration of such low radiocarbon dates is very difficult and also somewhat uncertain, particularly with the southern data sets. According to ages and distances of segments GM-1(a) and GM-1(b), stem I grew 0.20 m in 400 yr in the sampling direction. The value indicates a slightly lower mean radial increase, of 0.50 x 10^-3 m yr^-1 over the time frame ad 1365–1765.

We dated the deepest segments from samples GM-2 and GM-3, which were collected from the second stem II and from the smallest stem III. These segments correpond to a depth in the wood of 0.52 and 0.55 m and have ages of 735 ± 10 and 535 ± 20 yr.

The 5 samples PB-1, PB-2, PB-3, PB-4 and PB-5 originate from the five standing stems, i.e., I, II, III, IV and V. The 5 samples were also shorter than the penetration distance of the increment borer, showing the presence of significant hollow parts in each stem. In all cases, we extracted and dated the innermost/ deepest segment of each sample.

The oldest dated segment PB-1 originates from stem I, at a distance of 0.52 m from the sampling point. Its radiocarbon date of 852 ± 28 bp corresponds to a calibrated age of 785 ± 25 yr. The other 4 sample segments exhibit a continuous age decrease from PB-2 to PB-5 having ages between 565 and 375 yr.

For detemining the true age of Grandmother, which is identical to the age of stem I, we used the age of the oldest segment GM-1(a), the estimated distance to the calculated pith of stem I from the sampling point (1.15 m) and from the oldest segment (0.69 m) and also the mean growth rates for 650 and 420 yr (see Table 1 and Fig. 4). In a conservative estimate, the age of stem I and, consequently, of Grandmother is 1,500–1,700 yr, i.e., 1,600 ± 100 yr. According to this value, the Grandmother started growing around ad 400.

The ages of stems II and III were determined from the ages of samples GM-2 and GM-3 and the distances from these sample ends to the presumptive piths of stems II and III. According to these values, the estimated ages of stems II and III are of ca. 1,100 and 750 yr.

The age of the polygamous baobab, which is identical to the age of its oldest stem I, was determined by extrapolating the calibrated age of sample PB-1 to the corresponding pith (see Table 2). Therefore, we consider that the polygamous baobab and its stem I are 1,000 ± 100 yr old. The ages of the other 4 sampled stems were estimated in a similar manner. Thus, stem II and stem III, which are fused completely up to a height of 2.60 m, are both around 750 yr old. Stems IV and V have estimated ages of 600 and 500 yr. The fallen stem VI, which broke partially in the collapse, was not sampled. According to its size, we estimate that it is around 300 yr old.

The two large and old A. rubrostipa specimens started growing as single-stemmed and became over time multi-stemmed, thus developing different shapes. This evolution is due to the general ability of baobabs, which we identified for the first time for A. digitata [5,9,10], to produce stems during their life cycle, such as other trees produce branches.

The Grandmother started growing its main stem I around ad 400. The tree produced new stems in ad 900 and 1250, developing over time its well-known pot shape with 3 perfectly fused stems. The main stem I stopped growing in ad 1750, because of old age.

The main stem I of the polygamous baobab started growing around ad 1000. Stems II and III emerged probably together in ad 1250 and the fony baobab gained a triangular structure. Next, 3 new stems emerged succesively toward north in ad 1400, 1500 and 1700. Thus, the polygamous baobab developed its cluster structure, with 6 very distinct bottle-shaped stems, which are fused only in the proximity of their base and have separate canopies.