Tropical forests play a critical role in the hydrological cycle and can influence local and regional precipitation¹. Previous work has assessed the impacts of tropical deforestation on precipitation, but these efforts have been largely limited to case studies². A wider analysis of interactions between deforestation and precipitation—and especially how any such interactions might vary across spatial scales—is lacking. Here we show reduced precipitation over deforested regions across the tropics. Our results arise from a pan-tropical assessment of the impacts of 2003–2017 forest loss on precipitation using satellite, station-based and reanalysis datasets. The effect of deforestation on precipitation increased at larger scales, with satellite datasets showing that forest loss caused robust reductions in precipitation at scales greater than 50 km. The greatest declines in precipitation occurred at 200 km, the largest scale we explored, for which 1 percentage point of forest loss reduced precipitation by 0.25 ± 0.1 mm per month. Reanalysis and station-based products disagree on the direction of precipitation responses to forest loss, which we attribute to sparse in situ tropical measurements. We estimate that future deforestation in the Congo will reduce local precipitation by 8–10% in 2100. Our findings provide a compelling argument for tropical forest conservation to support regional climate resilience.

Tropical forests have an important regulating influence on local and regional climate, through modulating the exchange of moisture and energy between the land and the atmosphere. Deforestation disrupts this exchange, though the climatic consequences of progressive, patch-scale deforestation of formerly intact forested landscapes have not previously been assessed. Remote sensing datasets of land surface and atmospheric variables were used to compare the climate responses of Amazon forests that lost their intact status between 2000 and 2013. Clear gradients in environmental change with increasing disturbance were observed. Leaf area index (LAI) showed progressively stronger reductions as forest loss increased, with evapotranspiration (ET) showing a comparative decline. These changes in LAI and ET were related to changes in temperature (T), with increased warming as deforestation increased. Severe deforestation of intact Amazon forest, defined as areas where canopy cover was reduced below 70%, was shown to have increased daytime land surface T by 0.44°C over the study period. Differences between intact and disturbed forests were most pronounced during the dry season, with severely deforested areas warming as much as 1.5°C. Maintenance of canopy cover was identified as an important factor in minimizing the impacts of disturbance. Overall, the results highlight the climate benefits provided by intact tropical forests, providing further evidence that protecting intact forests is of utmost importance.

Water recycled through transpiring forests influences the spatial distribution of precipitation in the Amazon and has been shown to play a role in the initiation of the wet season. However, due to the challenges and costs associated with measuring evapotranspiration (ET) directly and high uncertainty in remote-sensing ET retrievals, the spatial and temporal patterns in Amazon ET remain poorly understood. In this study, we estimated ET over the Amazon and 10 sub-basins using a catchment-balance approach, whereby ET is calculated directly as the balance between precipitation, runoff, and change in groundwater storage. We compared our results with ET from remote-sensing datasets, reanalysis, models from Phase 5 and Phase 6 of the Coupled Model Intercomparison Projects (CMIP5 and CMIP6 respectively), and in situ flux tower measurements to provide a comprehensive overview of current understanding. Catchment-balance analysis revealed a gradient in ET from east to west/southwest across the Amazon Basin, a strong seasonal cycle in basin-mean ET primarily controlled by net incoming radiation, and no trend in ET over the past 2 decades. This approach has a degree of uncertainty, due to errors in each of the terms of the water budget; therefore, we conducted an error analysis to identify the range of likely values. Satellite datasets, reanalysis, and climate models all tended to overestimate the magnitude of ET relative to catchment-balance estimates, underestimate seasonal and interannual variability, and show conflicting positive and negative trends. Only two out of six satellite and model datasets analysed reproduced spatial and seasonal variation in Amazon ET, and captured the same controls on ET as indicated by catchment-balance analysis. CMIP5 and CMIP6 ET was inconsistent with catchment-balance estimates over all scales analysed. Overall, the discrepancies between data products and models revealed by our analysis demonstrate a need for more ground-based ET measurements in the Amazon as well as a need to substantially improve model representation of this fundamental component of the Amazon hydrological cycle.

A tree ring oxygen isotope (δ18OTR) chronology developed from one species (Cedrela odorata) growing in a single site has been shown to be a sensitive proxy for rainfall over the Amazon Basin, thus allowing reconstructions of precipitation in a region where meteorological records are short and scarce. Although these results suggest there should be large-scale (> 100 km) spatial coherence of δ18OTR records in the Amazon, this has not been tested. Furthermore, it is of interest to investigate whether other, possibly longer -lived, species similarly record interannual variation of Amazon precipitation, and can be used to develop climate sensitive isotope chronologies. In this study, we measured δ18O in tree rings from seven lowland and one highland tree species from Bolivia. We found that cross-dating with δ18OTR gave more accurate tree ring dates than using ring width. Our “isotope cross-dating approach” is confirmed with radiocarbon “bomb-peak” dates, and has the potential to greatly facilitate development of δ18OTR records in the tropics, identify dating errors, and check annual ring formation in tropical trees. Six of the seven lowland species correlated significantly with C. odorata, showing that variation in δ18OTR has a coherent imprint across very different species, most likely arising from a dominant influence of source water δ18O on δ18OTR. In addition we show that δ18OTR series cohere over large distances, within and between species. Comparison of two C. odorata δ18OTR chronologies from sites several hundreds of kilometres apart showed a very strong correlation (r = 0.80, p < 0.001, 1901–2001), and a significant (but weaker) relationship was found between lowland C. odorata trees and a Polylepis tarapacana tree growing in the distant Altiplano (r = 0.39, p < 0.01, 1931–2001). This large-scale coherence of δ18OTR records is probably triggered by a strong spatial coherence in precipitation δ18O due to large-scale controls. These results highlight the strength of δ18OTR as a precipitation proxy, and open the way for temporal and spatial expansion of precipitation reconstructions in South America

Land–atmosphere interactions have an important influence on Amazon precipitation (P), but evaluation of these processes in climate models has so far been limited. We analysed relationships between Amazon P and evapotranspiration (ET) in the 5th Coupled Model Intercomparison Project models to evaluate controls on surface moisture fluxes and assess the credibility of regional P projections. We found that only 13 out of 38 models captured an energy limitation on Amazon ET, in agreement with observations, while 20 models instead showed Amazon ET is limited by water availability. Models that misrepresented controls on ET over the historical period projected both large increases and decreases in Amazon P by 2100, likely amplified by unrealistic land–atmosphere interactions. In contrast, large future changes in annual and seasonal-scale Amazon P were suppressed in models that simulated realistic controls on ET, due to modulating land–atmosphere interactions. By discounting projections from models that simulated unrealistic ET controls, our analysis halved uncertainty in basin-wide future P change. The ensemble mean of plausible models showed a robust drying signal over the eastern Amazon and in the dry season, and P increases in the west. Finally, we showed that factors controlling Amazon ET evolve over time in realistic models, reducing climate stability and leaving the region vulnerable to further change.

Key Message Radiocarbon dating shows that Cedrela trees from Bolivia, Ecuador and Venezuela form one ring per year but Cedrela trees from Suriname form two rings per year. Abstract Tropical tree rings have the potential to yield valuable ecological and climate information, on the condition that rings are annual and accurately dated. It is important to understand the factors controlling ring formation, since regional variation in these factors could cause trees in different regions to form tree rings at different times. Here, we use ‘bomb-peak’ radiocarbon (¹⁴C) dating to test the periodicity of ring formation in Cedrela trees from four sites across tropical South America. We show that trees from Bolivia, Ecuador and Venezuela have reliably annual tree rings, while trees from Suriname regularly form two rings per year. This proves that while tree rings of a particular species may be demonstrably annual at one site, this does not imply that rings are formed annually in other locations. We explore possible drivers of variation in ring periodicity and find that Cedrela growth rhythms are most likely caused by precipitation seasonality, with a possible degree of genetic control. Therefore, tree-ring studies undertaken at new locations in the tropics require independent validation of the annual nature of tree rings, irrespective of how the studied species behaves in other locations.