Genetic data suggest a natural prehuman origin of open habitats in northern Madagascar and question the deforestation narrative in this region

Abstract

The impact of climate change and anthropogenic deforestation on biodiversity is of growing concern worldwide. Disentangling how past anthropogenic and natural factors contributed to current biome distribution is thus a crucial issue to understand their complex interactions on wider time scales and to improve predictions and conservation strategies. This is particularly important in biodiversity hotspots, such as Madagascar, dominated by large open habitats whose origins are increasingly debated. Although a dominant narrative argues that Madagascar was originally entirely covered by woodlands, which were destroyed by humans, a number of recent studies have suggested that past climatic fluctuations played a major role in shaping current biome distributions well before humans arrived. Here, we address the question of the origin of open habitats in the Daraina region in northern Madagascar, using a multiproxy approach combining population genetics modeling and remote-sensing analyses. We show that (i) contrary to most regions of Madagascar, the forest cover in Daraina remained remarkably stable over the past 60 y, and (ii) the golden-crowned sifaka (Propithecus tattersalli), a forest-dwelling lemur, underwent a strong population contraction before the arrival of the first humans, hence excluding an anthropogenic cause. Prehuman Holocene droughts may have led to a significant increase of grasslands and a reduction in the species' habitat. This contradicts the prevailing narrative that land cover changes are necessarily anthropogenic in Madagascar but does not preclude the later role played by humans in other regions in which recent lemur bottlenecks have been observed.

Fig. 1.

Trends in forest cover change in the Daraina region between 1949 and 2002. (A) Circled letters on the satellite imagery of the north of Madagascar indicate the locations of the four places named in the main text and in Fig. 4: Ankarana massif (A), Irodo (B), Vohémar (C), and Andavakoaera (D). (B) Black dots on the map of the Daraina region show the sampling locations of the 105 P. tattersalli social groups genotyped. This figure shows that trends in forest cover change radically in the northwestern (deforestation) and southeastern (forest regeneration) parts of the region as exemplified by the two focus zones [detail (a) and detail (b)] for the periods 1949–1972, 1972–1994, and 1994–2002. The two focus zones are detailed in Fig. S1.

Fig. 2.

BF values for the four tested scenarios under the different sampling schemes. H1 corresponds to a contraction attributable to anthropogenic effects following the arrival of Europeans (0–500 Cal. YBP); H2 (500–1,000 Cal. YBP) corresponds to a contraction attributable to anthropogenic effects during the period for which first cities appeared in far northern Madagascar; H3 (1,000–2,000 Cal. YBP) corresponds to the first millennium of human presence, during which human densities were probably very low; and H4 (2,000–10,000 Cal. YBP) corresponds to a contraction caused by environmental factors before the arrival of humans in Madagascar. In the “global sample” analyses (green), samples were taken across the whole sampled area, ignoring any substructure identified in a previous study. In the K_A (red) and K_B (blue) analyses, the samples were taken only considering the K_A and K_B genetic clusters identified in a previous study (details are provided in Materials and Methods).

Fig. 3.

Most likely period for the start of the population decrease. This figure shows the value of the BF for fixed-length intervals of 500 y between 0 and 50,000 y ago and between 0 and 10,000 y ago (i.e., the Holocene period). The BF measures the weight of evidence for the hypothesis that the population started to decrease in a given interval (H_a) vs. any other time period (H_b). The different solid lines correspond to different sampling strategies described in Materials and Methods: red (genetic unit K_A), blue (genetic unit K_B), and green (global sample). BF = 1 indicates that H_a and H_b are equally likely. BF values greater (or smaller) than 1 support H_a (or H_b). We considered BF values >7 as significant evidence for H_a. We indicated, along the time frame, the intervals corresponding to the four tested scenarios (H1–H4) and the historical, palynologicial, and paleontological evidence discussed in the main text.