The resilience of Amazon tree cover to past and present drying
T Kukla, A Ahlström, SY Maezumi, M Chevalier, Z Lu, MJ Winnick, CP Chamberlain (2021). Global and Planetary Change.
The Problem
The fundamental resilience of Amazon tree cover to drying is difficult to constrain because human-driven deforestation obscures natural climate-vegetation relationships. However, the Amazon rainforest is likely tens of millions of years old—could its longevity inform its resilience? A key question in the rainforest’s survival is: has rainfall ever dipped below ~2,000 mm/yr and if so, by how much? 2,000 mm/yr is the threshold where a dry-season water deficit can amplify feedbacks that threaten tree cover (e.g. Guan et al., 2015), and it should apply as long as tropical temperatures are within 5-10°C of modern. However, paleo-precipitation reconstructions are mostly qualitative. We know Amazonia has experienced wetter and drier intervals in the past, but we don’t know by how much.
Our hypothesis
Using the mid-Holocene as a case study, we hypothesize that, in the absence of intensive land use, tree cover is resilient to drying below the 2,000 mm/yr threshold because feedbacks that threaten tree cover remain weak. An emerging consensus among models supports this hypothesis, but it has not been tested with paleoclimate data.
What did we find?
(1) Amazon rainfall dipped well-below the 2,000 mm/yr threshold in the mid-Holocene, but proxy data show no evidence of strong moisture recycling or fire feedbacks, and tree cover remained high. (2) These proxy results are corroborated by Dynamic Global Vegetation Model (DGVM) experiments which show high tree cover with low, mid-Holocene rainfall. (3) Modern tree cover appears as resilient to drying as the mid-Holocene when human land use is ignored.
-
![]()
Tree cover response to drying depends on the balance of forcings and feedbacks
Forcings and feedbacks can have different effects on how tree cover responds to drying. This means we can infer their relative contributions by constraining how tree cover changed with drying in the past. Given a large amount of drying, a small decrease in tree cover indicates weak feedbacks and a large decrease points to strong feedbacks.
-
![]()
Mid-Holocene was drier, but no clear evidence for a stronger fire feedback
We use a vapor transport model of oxygen isotopes (Kukla et al., 2019) to calculate rainfall from the change in δ18O from east to west, and find that mid-Holocene rainfall was likely much lower than 2,000 mm/yr. However, the charcoal index (a metric for fire) does not significantly correlate with the oxygen isotope gradient. This indicates that fire is not very sensitive to drying in this range, or there are competing effects where anthropogenic fire decreases as natural fires increase (e.g. Riris & Arroyo-Kalin, 2019; Montoya et al., 2020). In either case, natural fires did not increase enough to cause widespread tree cover loss.
-
![]()
Dynamic Global Vegetation Modeling supports high tree cover at mid-Holocene rainfall
We force a DGVM (LPJ-GUESS v4.0) with our mid-Holocene rainfall reconstructions within the rectangle domain. Tree cover remains high even as rainfall decreases below 1,300 mm/yr. Comparing tree cover with and without the fire module in the model shows that the sensitivity of tree cover to fire is low in this rainfall range.
-
![]()
Is modern tree cover as resilient to drying as the mid-Holocene?
The answer is complicated. Our analysis (building on Ahlström et al., 2017) suggests that, in the absence of human deforestation, modern tree cover resilience is similar to the mid-Holocene. The state of tree cover (the silver ball) remains forested at mid-Holocene rainfall only when high land use regions are filtered out. While we cannot constrain the resilience with modern deforestation in this study, our findings demonstrate that human land use is a more urgent threat to Amazon tree cover than drying alone.
