Tropical deforestation causes large reductions in observed precipitation

The study investigates how tropical deforestation affects precipitation across local to regional spatial scales, addressing a key gap beyond previous case studies. Tropical forests strongly influence energy, water and carbon cycles and control regional rainfall patterns through evapotranspiration and moisture recycling. Rapid forest loss across the tropics raises concerns about feedbacks that could destabilize evergreen tropical forests, particularly under increasing drought stress. Although small-scale deforestation can sometimes increase precipitation frequency via thermally induced circulations, larger-scale forest loss is expected to reduce precipitation recycling and rainfall. A pan-tropical, observational assessment across spatial scales has been lacking; this work aims to quantify precipitation responses to forest loss across the tropics, determine scale dependence, reconcile differences among data products, and project impacts of future deforestation on rainfall.

Prior studies include regional and case-specific analyses showing mixed responses: small-scale deforestation in the southern Amazon linked to increases in precipitation frequency, while larger-scale deforestation reduces precipitation via diminished moisture recycling. Over Indonesia/SEA, deforestation has been associated with declining precipitation and amplified El Niño impacts. Climate models generally predict substantial precipitation declines under extensive Amazonian deforestation (for example, ~8% reduction by mid-century), but model estimates vary widely, especially for the Congo. Previous estimates for complete deforestation suggest 16–70% precipitation reductions in the Amazon and 18–50% in the Congo, highlighting uncertainty. Sparse in situ observations in tropical forests challenge station-based datasets and reanalyses, which can mask land-cover-driven precipitation changes due to interpolation and limited constraints, motivating the use of satellite-based precipitation products for robust assessment.

The authors conducted a pan-tropical, observation-based analysis of how forest cover loss (2003–2017) affected precipitation, focusing on evergreen broadleaf forests in the Amazon, Congo, and Southeast Asia (SEA). Forest cover change was derived from a satellite dataset (Hansen et al.) to identify pixels experiencing loss. They analyzed 18 precipitation datasets: 10 satellite-based, 4 station-based, and 4 reanalysis products. For each spatial resolution (0.05°, 0.1°, 0.25°, 0.5°, 1.0°, 2.0°; roughly 5–200 km), they compared precipitation changes over pixels with forest loss to neighboring control pixels with less loss to isolate the impact of deforestation from broader climate variability. Mean precipitation was compared between multi-year periods (e.g., 2003–2007 vs 2013–2017), and statistical significance assessed (P < 0.05). Seasonal analyses partitioned months into wettest three, driest three, and transition months. Robustness tests varied analysis period, choice of start/end windows, and the spatial extent of control regions, and evaluated responses during El Niño vs non–El Niño years. The study emphasized satellite products where in situ data are sparse and assessed inter-dataset consistency. For future impacts, observed sensitivities of precipitation to forest loss were combined with projected land-cover change under a high-deforestation scenario (SSP3–RCP4.5) to estimate 2015–2100 rainfall changes at 2.0° resolution, including sensitivity analyses that (a) cap responses to the well-sampled 0–30% forest loss range and (b) assume nonlinear response functions, recognizing potential thresholds and feedbacks.

• Satellite datasets show statistically significant declines in annual mean precipitation over deforested regions at all analyzed scales, with effects strengthening at larger scales (>0.5°). At 2.0° (~220 km), each percentage point of forest loss reduces precipitation by 0.25 ± 0.1 mm per month across the tropics.
• Regional 2.0° satellite-based sensitivities: SEA ~0.48 ± 0.36 mm month−1 per percentage point forest loss; Amazon ~0.23 ± 0.12; Congo ~0.21 ± 0.19. At least 8/10 satellite datasets agree on the sign of the response within each region.
• Station-based and reanalysis products generally do not show significant changes and sometimes indicate small increases, diverging from satellite results; this is attributed to sparse in situ observations and limitations in interpolation and data assimilation over tropical forests.
• Seasonal responses are nearly uniformly negative. Absolute reductions are greatest in the wet season for the tropics; in the Amazon, the largest declines occur in transition months. Some nonsignificant dry-season increases are observed in the Amazon (2°) and Congo (1–2°), whereas SEA shows dry-season reductions at all scales.
• Responses are robust across methodological choices and remain negative during both El Niño and non–El Niño years; reductions are stronger during El Niño in the Amazon and SEA, consistent with higher transpiration and greater sensitivity of rainfall to moisture recycling during drought.
• Comparison with models shows consistency in percentage declines per percentage point forest loss (e.g., Amazon ~0.25% per percentage point observed vs model meta-analysis ~0.16 ± 0.13%).
• Future projections (SSP3–RCP4.5): by 2100, Congo could experience up to 16.5 ± 6.2 mm month−1 reduction (≈8–10% decline) in annual precipitation due to projected forest loss, with strongest impacts in western and southern Congo. Capping analysis to 0–30% forest loss yields smaller projected reductions (Congo 6.5 ± 2.6; SEA 6.2 ± 2.5 mm month−1), likely underestimates in heavily deforested areas.
• The analysis up to 200 km likely underestimates broader downwind impacts given moisture recycling length scales of 500–2,000 km.

Findings demonstrate that tropical deforestation robustly reduces precipitation, with impacts intensifying at larger spatial scales consistent with reduced moisture recycling becoming the dominant mechanism. Satellite products provide a more reliable constraint over tropical forests than station-based or reanalysis datasets due to sparse in situ measurements. The observed sensitivities align with, and support, climate model projections that predict rainfall declines with regional-scale deforestation, lending confidence to models at these scales. Seasonal and ENSO-dependent patterns suggest interactions between forest transpiration, atmospheric moisture availability, and drought conditions modulate the magnitude of rainfall responses. Because the analysis is limited to scales up to ~200 km and a 15-year loss period, it likely underestimates total impacts, especially downwind over hundreds to thousands of kilometers where moisture recycling is important, and does not capture potential nonlinear thresholds or tipping points. The projected precipitation reductions from continued deforestation are of a magnitude comparable to or exceeding those anticipated from climate change alone over similar periods, underscoring the critical role of forest conservation in sustaining regional hydroclimate resilience.

This pan-tropical, observation-based study shows that deforestation significantly reduces precipitation across the tropics, with stronger effects at larger spatial scales; at ~200 km scales, each percentage point of forest loss reduces rainfall by about 0.25 mm per month. Seasonal analyses confirm predominantly negative impacts, and ENSO conditions can amplify reductions. Projections indicate substantial future declines in rainfall, particularly in the Congo, under high-deforestation scenarios. The results provide empirical support for climate models predicting rainfall declines with deforestation and highlight that current estimates may be conservative due to analysis scale and record length. The study underscores the urgency of conserving and restoring tropical forests to maintain regional rainfall, support agriculture and hydropower, and enhance climate resilience. Future research should extend observational records, improve in situ coverage, resolve larger-scale downwind impacts and moisture recycling pathways, and investigate nonlinearities and tipping points in forest–rainfall feedbacks.

• Sparse in situ precipitation measurements across tropical forests undermine the reliability of station-based datasets and reanalyses, leading to disagreement with satellite-based results.
• The observational record (2003–2017) is relatively short, and large temporal and spatial variability in precipitation introduces uncertainty in inferred sensitivities.
• Analysis is limited to spatial scales up to ~200 km; impacts at larger downwind scales tied to moisture recycling (500–2,000 km) are not captured, likely underestimating total effects.
• Sensitivities for high levels of forest loss (>30%) are poorly constrained by observations; capping responses to 0–30% likely underestimates impacts where deforestation exceeds this range.
• Potential nonlinear responses and climate system tipping points are not fully resolved by the linear or simplified nonlinear scaling approaches used.
• Projections depend on the chosen land-cover scenario (SSP3–RCP4.5) and assumptions about linearity/nonlinearity of responses.