Forest and savanna ecosystems naturally exist as alternative stable states. The maximum capacity of these ecosystems to absorb perturbations without transitioning to the other alternative stable state is referred to as ‘resilience’. Previous studies have determined the resilience of terrestrial ecosystems to hydroclimatic changes predominantly based on space‐for‐time substitution. This substitution assumes that the contemporary spatial frequency distribution of ecosystems’ tree cover structure holds across time. However, this assumption is problematic since ecosystem adaptation over time is ignored. Here we empirically study tropical forests’ stability and hydroclimatic adaptation dynamics by examining remotely sensed tree cover change (ΔTC; aboveground ecosystem structural change) and root zone storage capacity (S_r; buffer capacity towards water‐stress) over the last two decades. We find that ecosystems at high (>75%) and low (<10%) tree cover adapt by instigating considerable subsoil investment, and therefore experience limited ΔTC—signifying stability. In contrast, unstable ecosystems at intermediate (30%–60%) tree cover are unable to exploit the same level of adaptation as stable ecosystems, thus showing considerable ΔTC. Ignoring this adaptive mechanism can underestimate the resilience of the forest ecosystems, which we find is largely underestimated in the case of the Congo rainforests. The results from this study emphasise the importance of the ecosystem's temporal dynamics and adaptation in inferring and assessing the risk of forest‐savannah transitions under rapid hydroclimatic change.

Global forest loss impacts climate, biodiversity and sustainable development goals. Deforestation footprinting attributes forest loss to commodity production and consumption, identifying global trends, drivers and hot spots to inform zero-deforestation policies. In this Review, we provide an overview of global deforestation footprinting approaches and their trends. Major economies, including Brazil, Indonesia, China, the United States and Europe, are responsible for most commodity-linked deforestation, with agriculture-linked deforestation in Brazil alone reaching over 12.8 million hectares between 2005 and 2015. Agriculture is a dominant driver of deforestation. For example, 86% of global deforestation occurring between 2001 and 2022 can be attributed to crop and cattle production. Footprinting of commodity-linked deforestation has contributed to the scope and implementation of supply chain regulation to mitigate forest loss. For example, footprint estimates have been used in risk assessments for EU and UK due diligence regulations. Although forest loss to agriculture is relatively well documented, a lack of data on non-agricultural drivers — such as mining and mangrove clearance for aquaculture — limits the scope of footprints in fully attributing total global forest loss to human activities. Future research should focus on methodological and data harmonization, transparency and sharing to enable footprinting approaches to cover a wider range of deforestation drivers.