Lemurs Dependent On Bamboo Risk Climate Caused Extinction

A characteristic of bamboo-feeding mammals is that most, if not all, of them are considered threatened by extinction [8]. In Asia, both giant (Ailuropoda melanoleuca) and red (Ailurus fulgens) pandas have much diminished geographical ranges when compared with their historical and paleontological records [9, 10]. Similarly, in Madagascar, the two larger bamboo lemurs, the greater bamboo lemur (Prolemur simus) and the golden bamboo lemur (Hapalemur aureus), have highly restricted distributions within the island and are listed as critically endangered [5, 8]. Both the endangerment and the dietary specializations of bamboo-feeding taxa suggest that they may be especially sensitive to changes in climate.

Here we focus on P. simus (Figure 1A) because of abundant paleontological and neontological evidence. First, we evaluated whether feeding on woody bamboos, which are grasses belonging to the subfamily Bambusoideae, can be diagnosed from the P. simus dentition. Mammalian dentitions are especially informative about the diet of extinct and taxonomically distant taxa. Relevant to a bamboo diet, feeding on fibrous vegetation in seasonally variable environments is linked in recent and past mammal communities to high-crowned molars [11, 12], and grass specialists also tend to have complex molar surfaces [13]. For example, horses, which are specialized grazers, have high-crowned molars with high fine-scale dental complexity resulting from complex folding of enamel bands [14]. Although bamboo specialists lack high-crowned molars, they appear to have high dental complexity [13] (Figure 1B). To measure fine-scale complexity using orientation patch count rotated (OPCR; see STAR Methods and Figure 1B; Figures S1A and S1B), here we contrasted the complexity of the bamboo specialists P. simus, A. melanoleuca, and A. fulgens with the dental complexity of the brown bear (Ursus arctos), the raccoon (Procyon lotor), and the black lemur (Eulemur macacao). These latter taxa, which represent the phylogenetic contrasts to the bamboo specialists, have generalized diets (STAR Methods and Figure 1B).

With increasing resolution, the OPCR values of bamboo specialists show progressively higher values compared to the non-bamboo feeders (Figure S1C and Table S1). Both upper and lower dentitions of P. simus and A. melanoleuca show high fine-grained complexity, whereas the relatively low OPCR values of A. fulgenslower dentition are likely to reflect the highly selective foraging of different bamboo parts by red pandas (Figure S1C) [15]. The average cheek tooth complexities of A. melanoleuca (108.8) and P. simus (84.0) are comparable to the values reported previously for the horse (109.5) and the early equid Eohippus (88.6), respectively (calculated for upper teeth at 50-row resolution) [14]. Moreover, similar to equids, but exceptional for primates, P. simus has a high degree of molarization of the premolars [16]. All of these dental features indicate that P. simus has a dentition that is at least as diagnostic of bamboo feeding as that of A. melanoleuca. Because advancing wear removes surface features from low-crowned teeth, bamboo-feeding taxa should be especially sensitive to, and informative about, changes in precipitation patterns.

Next, we used data on living P. simus feeding behavior to link the details of bamboo feeding to precipitation. The durations of rainy and dry seasons have been proposed to have especially strong effects on P. simus communities [5, 17], and although they feed on most parts of bamboo, the preferred parts are new ground shoots (Figure 1A). The availability of shoots, however, is seasonally restricted to the rainy season [3, 18], and feeding on mechanically demanding [19] and nutritionally poor culm can be considered a regular fallback food for P. simus. This preference for shoots, which are the nutritionally high-quality parts of bamboo, appears to be shared by all bamboo specialists [15, 18]. Furthermore, the birth and weaning seasons of P. simus are during peak bamboo shoot availability [3, 5].

An 18-month focal-animal follow of a P. simus group was carried out during 2006–2008 in Ranomafana National Park (RNP), Madagascar, a location in which P. simuswas first found in 1987 and the first site at which P. simus was habituated [3, 17]. From 2,387 feeding observations, we tabulated the proportion of feeding time spent on eating bamboo culms, leaves, and shoots (see STAR Methods). In RNP, 95% of feeding time was spent on a single species of woody bamboo (Cathariostachys madagascariensis), and individual members of P. simus groups typically had synchronous activities. One exception is that juvenile P. simus cannot break the bamboo culm and rely on adults in obtaining the softer inner parts of the culm [3].

The results show that culm feeding is restricted to the months of August to November (Figure 2A and Table S2). Shoot feeding begins abruptly with the full onset of the rainy season, and shoots are the primary diet when rainfall exceeds 250 mm/month (Figure 2B). The distinct pattern of feeding either on shoots or culm (Figure 2) was reported for the same locality 10 years previously [3], suggesting robust feeding patterns. Feeding on shoots continues in the austral winter when an increasing amount of time is spent feeding on leaves (Figures 2A and 2B).

Because the dental morphology (Figure 1) of P. simus appears ecologically highly diagnostic and because RNP has distinct seasonal and precipitation patterns of bamboo feeding (Figure 2), we next estimated the durations of wet-season and dry-season feeding for the whole of Madagascar. This allows us to integrate data on precipitation in present-day localities with the paleontological localities of P. simus, which are scattered throughout Madagascar (except for the southwest) and which are relatively rich in dental material (Table S3).

Using climate data from Worldclim.org [20] in 10-min degree grid cell resolution (approximately 18 × 18 km at the equator, mean precipitation from 1950 to 2000), we first determined both the wet-season and dry-season precipitation thresholds for the grid cell containing RNP. The wet-season feeding threshold was set at >250 mm monthly precipitation, encompassing the months of December to March. We consider this a reasonable threshold because it covers the four peak months of feeding on shoots in our feeding data at RNP (Figures 2A and 3A ). For dry-season feeding, we used two grid cell thresholds. To account for the strict culm-feeding season, we used a <50 mm monthly precipitation threshold that encompasses the months August to October, only excluding November when the transition from culm to shoot feeding happens as the rainy season begins (Figures 2A and 3A). To cover the whole dry-season feeding period, which includes bamboo culm and a mixture of leaves and shoots (Figure 2A), we used a <80 mm monthly precipitation threshold, encompassing the months of May to October (Figure 3A). Similarly, the precipitation reanalysis data (ERA-Interim) from the RNP region spanning 2005 to 2015 show that the months December to March and May to October are the periods of the highest and the lowest precipitation, respectively (see STAR Methods and Figure S2).

The island-wide maps show the projected wet-season and dry-season feeding durations in months (Figure 3B), revealing that grid cells containing only paleontological P. simus finds are consistently in regions where dry-season feeding would be required for at least five consecutive months in the present-day climate (Figures 3B and 3C; randomization tests between fossil and recent means are p = 0.0025 for <50 mm threshold and p = 0.0174 for <80 mm threshold). In contrast, there is no marked difference in the number of potential wet-season months between the sites (Figures 3B and 3C; randomization test between fossil and recent means is p = 0.2394). These patterns indicate that if P. simus were still living in the areas of P. simus paleontological localities, they would have to persist either on bamboo leaves and culm for the extended dry season or on alternative diets. Woody bamboos have been reported to have present-day distribution in the regions that contain at least some of the P. simus paleontological localities [21]. Furthermore, relatively high present-day palm species richness in northeastern Madagascar has been reconstructed to be a relict pattern influenced by high precipitation in that region during the Last Glacial Maximum (LGM) [22]. Our combined analyses (Figures 2, 3, and S3) support wetter conditions for the northern half of Madagascar in the past, and some P. simus populations may have persisted even beyond the LGM in the north [4].

In China, compounded with human caused deforestation, changing climate has been suggested to affect bamboo distribution in the 21^st century, thereby causing food shortage for the giant panda [23]. Our data suggest that rapidly changing climate may also endanger bamboo feeders in a subtle way by affecting seasonal availability of preferred bamboo parts, and the giant panda may be similarly vulnerable [24]. Southwestern Madagascar is devoid of P. simus fossil localities, and it also has the longest and the shortest projected durations of dry-season and wet-season feeding, respectively (Figure 3B). We note that the present P. simus localities with exceptionally long projected dry-season feeding even at the threshold <50 mm are the westernmost sites at the edge of the high plateau (for example, Andringitra; Figure S4A). These exceptionally long dry-season feeding estimates are likely to be artifacts of grid cells spanning beyond the relatively small patches of forest, but they may also signify the precarious nature of these sites (a close up of the localities is shown in Figure S4A).

To examine the potential effect of climate change on P. simus, we used representative concentration pathway (RCP; see STAR Methods and [25]) scenarios for the future. The climate model data were downscaled to the same resolution as the present-day data and were bias-corrected using the present-day as a baseline (see http://worldclim.org/downscaling for details). We calculated a model ensemble mean value, using all available climate model data for each RCP to obtain a more robust estimate of the future conditions than using individual climate model runs.

For the grid cells in which P. simus is currently present, our future projections predict a relatively small drop in the duration of wet-season feeding by 2070, with the median duration shortening from 4 to 3 months (Figures 4B and 4C ; randomization test between recent and 2070 sites is p = 0.1942). However, the <50 mm threshold dry seasons in particular are predicted to be prolonged for many of the present-day P. simus sites (Figure 4A, 4B, and S4; randomization tests between recent and 2070 means are p = 0.0261 for <50 mm threshold and p = 0.1438 for <80 mm threshold). This projection is not restricted to the high-greenhouse-gas-concentration scenario (RCP 8.5), as all of the RCPs show that the dry-season durations become longer (Table S4 and Figure S3). These analyses indicate that regions in which P. simus is currently found are projected to experience precipitation patterns more comparable to those in the regions in which P. simus has gone extinct.

Overall, the projected lengthening of the dry-season feeding period would cause prolonged usage of bamboo culm and other mechanically demanding and drier plants, which in addition to nutritional challenges can be predicted to increase the rate of tooth wear, leading to loss in dental function and premature dental senescence [26]. In wild giant pandas, which are obligate bamboo specialists, tooth wear is prevalent, and poor dental condition has been found in dead animals [27]. In comparison to human-induced habitat loss directly driving populations to extinction, climate change drives prolongation of the dry season. Hence, P. simus populations in both protected and unprotected areas would be affected, suggesting that protecting the existing habitats might not be enough to guarantee the continued existence at the face of ongoing climate change (see also [7]). An additional factor in Madagascar is the frequent occurrence of tropical cyclones that can defoliate large swathes of forest. Whereas cyclone damage may benefit fast-growing bamboo, and thus open new areas for P. simus to persist, the toughness of plants during the dry season would still remain a challenge. Increasing dryness of bamboo would most likely threaten all bamboo specialists, despite their physiological adaptations to have low daily energy expenditure and to cope with chemical defenses of bamboo [15, 18, 28, 29, 30, 31].

In general, linking morphology, resource use, and precipitation data with distribution data allows new ways to make specific predictions about the response of biodiversity to future climate change and also to interpret paleontological data [6, 7]. Specifically, fossil assemblages consisting of species with low-crowned teeth have been used to reconstruct generally mesic habitats [6, 11, 12]. Our analyses suggest that the combination of low-crowned teeth with high dental complexity indicates mechanically demanding diets but with limited duration and severity of dry seasons (Figure 3). Taxa with complex but low-crowned teeth are known from the fossil record [32], as is also incipient evolution of taller tooth crowns in lineages that eventually go extinct as the environment dries [33].

Finally, although lemurs are considered to be one of the most endangered mammals on earth, due to habitat destruction [8], this study indicates that climate change has become an additional threat to the lemurs and has most likely contributed to the critical endangerment of P. simus. Considering bamboo lemurs and other bamboo specialists, the price of their specialization to an abundant but demanding food source appears to be their undoing in a changing climate.