Resolving the biome debate: ecological and palaeo‐data concur that the extensive tree‐limited grassy biomes of central and north‐western Madagascar are largely anthropogenic

Abstract

ABSTRACT For millennia, human activities have transformed ecosystems globally. Determining which landscapes are predominantly anthropogenic is crucial to conservation and habitat restoration. Madagascar, a critically‐threatened biodiversity hotspot, is a prominent example. Currently 80% of non‐cropland habitat is tree‐limited savanna falling under the tropical grassy biome (savanna/TGB). Savanna/TGB is particularly prevalent in the Dry Deciduous Forest (DDF) in the north/northwest and below the treeline in the Central Highlands (CH). Relative savanna/TGB cover when agropastoralism arrived <1300 years ago (ya) has been intensely debated. Two competing hypotheses have emerged. The Mixed Mosaic Perspective argues current savanna/TGB is mostly anthropogenic, and differs from endemic, tree‐rich savannas, because it has been produced by frequent agropastoralist fires, grazing by 7–10 million zebu cattle, and habitat clearance. These novel disturbances replaced a mosaic of savanna/TGB, comprised of C 3 and C 4 graminoids (both grasses and sedges), forest, heathland and wetland, with non‐endemic fire‐ and grazing‐adapted pantropical C 4 grasses. This perspective is founded on multiple lines of evidence: >90% of endemic vertebrates are forest‐adapted, Malagasy people named many locations currently in savanna/TGB after forest, and only 1000 ya, abundant tree‐adapted fauna with predominantly C 3 diets inhabited the CH and DDF. The Vast Grassland Perspective (VGP) contends that late Holocene aridity favoured vast savanna/TGB, and that cattle and human‐lit fires maintain savanna/TGB previously grazed by now‐extinct megafauna. This view contends that ecological models predict ‘treeless grassland’ across the CH and DDF, with savanna/TGB woodland naturally limited to <5%, and it questions palaeoecological evidence as spatially‐limited. Clarifying which viewpoint is correct is fundamental to conservation; if human‐lit fires and cattle grazing are not surrogates for past ecological processes, they may be detrimental to ecosystem function. We address this debate by applying ecological models that predict grassy habitat likelihood based on precipitation, and, for the first time, map these likelihoods across the entire CH and DDF. We review all available literature and apply novel interpolations of palaeo‐data to highlight existing spatial coverage and broad spatiotemporal trends. We then assess if predicted habitats from our ecological models concur with reconstructed habitat at the onset of pastoralism. Ecological models align with palaeo‐data at >90% of sites but diverge markedly from current conditions. Today, savanna/TGB is three to seven times more extensive than predicted, with >183,000 km 2 less forest than models suggest should exist, given current precipitation. Multiple lines of evidence indicate widespread anthropogenic transformation started with the arrival of agropastoralism ~1300 ya, and began shifting biomes to tree‐limited savanna by ~1100–1000 ya. Charcoal records show increased fire activity preceded decline of forest, tree‐rich savanna woodlands, heathland, and populations of large, now‐extinct, endemic vertebrates. Paradoxically, this happened when DDF climate was relatively wet, which should have promoted, not reduced, forest, and limited fire. Expansion of savanna/TGB dominated by C 4 grasses followed the increase in fire coinciding with the onset of pastoralism. In the context of a biome transition that is still unfolding, we present a critical pre‐agropastoralist baseline that can guide ecological restoration efforts to reinstate lost ecological functions and preserve Madagascar's remaining extraordinary endemic biota.