Response of Amazonian forests to mid‐Holocene drought: A model‐data comparison

• Published in: Global change biology 2022-01, Vol.28 (1), p.201-226
• Main Authors: Smith, Richard J., Singarayer, Joy S., Mayle, Francis E.
• Format: Article
• Language: English
• Subjects: Amazonia; Anthropogenic factors; biome resilience; Climate Change; Climate Models; Deforestation; Dieback; Drought; Droughts; dynamic global vegetation models; Forests; Gases; Greenhouse effect; Greenhouse gases; Holocene; Humans; Intercomparison; middle Holocene; Palaeoclimate; palaeoecology; Paleoclimate; Rain; Rainfall; Rainforests; Sensitivity; Simulated rainfall; Simulation; Trees; Tropical climate; Tropical forests; Vegetation
• ISSN: 1354-1013; 1365-2486
• DOI: 10.1111/gcb.15929

There is a major concern for the fate of Amazonia over the coming century in the face of anthropogenic climate change. A key area of uncertainty is the scale of rainforest dieback to be expected under a future, drier climate. In this study, we use the middle Holocene (ca. 6000 years before present) as an approximate analogue for a drier future, given that palaeoclimate data show much of Amazonia was significantly drier than present at this time. Here, we use an ensemble of climate and vegetation models to explore the sensitivity of Amazonian biomes to mid‐Holocene climate change. For this, we employ three dynamic vegetation models (JULES, IBIS, and SDGVM) forced by the bias‐corrected mid‐Holocene climate simulations from seven models that participated in the Palaeoclimate Modelling Intercomparison Project 3 (PMIP3). These model outputs are compared with a multi‐proxy palaeoecological dataset to gain a better understanding of where in Amazonia we have most confidence in the mid‐Holocene vegetation simulations. A robust feature of all simulations and palaeodata is that the central Amazonian rainforest biome is unaffected by mid‐Holocene drought. Greater divergence in mid‐Holocene simulations exists in ecotonal eastern and southern Amazonia. Vegetation models driven with climate models that simulate a drier mid‐Holocene (100–150 mm per year decrease) better capture the observed (palaeodata) tropical forest dieback in these areas. Based on the relationship between simulated rainfall decrease and vegetation change, we find indications that in southern Amazonia the rate of tropical forest dieback was ~125,000 km2 per 100 mm rainfall decrease in the mid‐Holocene. This provides a baseline sensitivity of tropical forests to drought for this region (without human‐driven changes to greenhouse gases, fire, and deforestation). We highlight the need for more palaeoecological and palaeoclimate data across lowland Amazonia to constrain model responses. The mid‐Holocene (around 6000 years ago) is an approximate analogue for future rainfall projections over Amazonia. Using multiple computer models of climate and vegetation, we simulate the response of the Amazon rainforest to the drier conditions of the mid‐Holocene to understand its sensitivity to reduced rainfall. We find that, while the central Amazonian rainforest was unaffected by drying, southern and eastern Amazonia likely experienced significant dieback. Moreover, climate and vegetation models that performed better