The scientific team characterized the event using several complementary indicators. Drought conditions were quantified with the three-month Standardized Precipitation-Evapotranspiration Index, or SPEI-3, which captures anomalies in the surface water balance by comparing precipitation against potential evapotranspiration estimated with the Penman-Monteith formulation. Heat stress was measured with the Wet Bulb Globe Temperature, or WBGT, a metric that integrates temperature, humidity, and evaporative constraints and is widely regarded as one of the most relevant indicators of human thermal strain. During September to November 2020, WBGT values in the region persistently exceeded the 95th percentile of the 1979-2022 climatology, while SPEI values revealed drought conditions stretching back through 2019 and into 2020.

What made the 2020 event particularly unusual was its timing relative to the seasonal cycle. The extreme period was not confined to the climatological dry season but emerged during a transition between anomalously dry rainy seasons, indicating a disruption of the typical pattern in which rainfall recharges the landscape from October through April. Observations from Brazil’s national meteorological service, analyzed with the CTX90pct heatwave index, showed that heatwave frequency peaked in 2020, the hottest year on record for the region. The prolonged dry conditions during 2019 likely enhanced atmospheric evaporative demand, and the delayed onset of the following rainy season further amplified heat accumulation and moisture deficits by prolonging land-atmosphere feedbacks associated with reduced surface moisture.

To quantify the human fingerprint, the researchers turned to two independent modeling frameworks. The first was HadGEM3-A, the Met Office’s high-resolution atmosphere-only attribution model, which runs at roughly 60-kilometer resolution over Brazil and provides large ensembles of up to 525 members. The model was run in two configurations: a realistic world including both anthropogenic and natural forcings, and a counterfactual world driven only by natural factors such as solar variability and volcanic eruptions. When the observed 2020 WBGT anomaly was placed against the probability distributions from these two worlds, the result was stark. The observed event sat in the upper tail of the anthropogenically forced simulations, while the probability density under natural-only forcing was close to zero at that threshold, yielding a Fraction of Attributable Risk of 0.99 with a 95 percent confidence interval of 0.97 to 1.00.

The second framework drew on simulations from the Coupled Model Intercomparison Project Phase 6, using three structurally different models: MIROC6, CanESM5, and ACCESS-ESM1-5. All three showed the same directional pattern, with simulations including anthropogenic forcing systematically shifted toward higher WBGT anomalies relative to natural-only runs. CanESM5 gave event probabilities under natural conditions that were effectively zero, producing a probability ratio greater than 1000. ACCESS-ESM1-5 showed a median probability ratio of 860, and MIROC6 produced a lower but still substantial value of 129. Because the two frameworks capture different sources of uncertainty, with HadGEM3-A constraining internal variability through large ensembles and CMIP6 quantifying structural uncertainty across model physics, the consistency of their results strengthens the conclusion that human-caused warming dramatically reshaped the odds of the event.

The climatic anomalies did not remain an abstract statistical finding; they translated into fire on the ground. Burned area estimates derived from the MODIS MCD64A1 satellite product, intersected with the MapBiomas land cover dataset, revealed a pronounced asymmetry in where the fires burned. Approximately 80 percent of the total burned area, about 1920 square kilometers, occurred in non-forest land cover classes dominated by pasture, while forest formations accounted for roughly 20 percent, about 468 square kilometers. Pasture fires showed a clear seasonal signal, increasing sharply from August to October before declining in November, reflecting the rhythms of fire use in land management. Forest fires were more spatially heterogeneous and less seasonal, with hotspots in municipalities such as Feijó, Tarauacá, and Porto Walter concentrated in September and October.

The smoke from these fires, combined with the extreme heat, left a measurable mark on public health. Between September and November 2020, 39,453 respiratory disease notifications were recorded across Acre, corresponding to 5.4 percent of the state’s population. The capital, Rio Branco, concentrated the largest absolute burden with more than 15,000 cases and 159 reported deaths, but the highest proportions of affected residents appeared in smaller, more vulnerable municipalities such as Assis Brasil, Santa Rosa do Purus, Acrelândia, and Porto Walter, where lower population density, geographic isolation, and limited healthcare infrastructure amplified the relative impact.

To understand the environmental drivers behind these health outcomes, the team fitted generalized linear models with a Poisson error distribution for each month of the peak fire season, using predictors including fine particulate matter concentrations, burned area, precipitation, air temperature, wind speed, wind direction, and aerosol optical depth. The models performed well overall, with September showing the strongest fit and most predictors statistically significant. A geographically weighted regression analysis then revealed that the influence of environmental variables was anything but uniform. Particulate matter exhibited by far the strongest positive average association with respiratory notifications, though with substantial local variability, while wind speed also showed a positive average effect. Temperature, precipitation, aerosol optical depth, and burned area showed weaker or mixed average relationships, and their coefficients varied considerably from one municipality to another.

This spatial heterogeneity carries an important lesson: municipalities exposed to similar environmental conditions did not necessarily experience similar health outcomes. Relative risk values from the regression framework ranged from approximately 0.18 to 1.57, indicating localized areas with substantially fewer or more notifications than expected. The authors interpret these patterns as evidence that respiratory outcomes were shaped not only by smoke exposure but also by social and territorial factors, including socioeconomic conditions, healthcare accessibility, demographic composition, and housing quality. The study period also coincided with the COVID-19 pandemic, which may have influenced healthcare-seeking behavior, reporting practices, and clinical overlap of respiratory symptoms, adding an acknowledged layer of uncertainty to the interpretation of the notification counts.

The researchers also examined the policy landscape and found striking gaps. No heatwave-specific records appeared in Brazil’s official disaster registry for the study period, suggesting that heatwaves are not yet consistently classified as a formal disaster category in the country, where drought remains the primary designation for heat-related impacts. In contrast, all municipalities in Acre reported wildfire-related disaster events in 2020, classified as fires affecting air quality. The authors argue that reducing future risks will require integrated adaptation strategies that combine climate-risk management, wildfire prevention, air-quality monitoring, and public health preparedness, including early warning systems, fire-free land management techniques, water resource conservation, community resilience building, and the intersectoral coordination now mandated by recent Brazilian climate adaptation legislation.

Taken together, the findings from Acre illustrate how a single climatic anomaly can propagate through environmental and social systems to produce uneven and cascading impacts. Anthropogenic climate change did not merely raise temperatures; it contributed to the drought conditions, vegetation drying, and fire-weather environment that allowed burning to expand and smoke to accumulate over populated areas. The 2020 event, the researchers conclude, should be understood not as an isolated anomaly but as part of an ongoing shift in the probability distribution of climate extremes driven by human forcing, a shift that is already reshaping the risks faced by vulnerable communities across the western Amazon.

Subject of Research: Attribution of the 2020 compound heatwave, drought, wildfire, and respiratory health event in the western Brazilian Amazon to anthropogenic climate change

Article Title: Western amazon 2020 extreme heatwave

Article References: Dutra, D. J., Veiga, R. Q., Abreu, R. C., Huang, W. T. K., dos Santos, J. C., Cunha-Zeri, G., de Luna Francisco, Q.-H. R., Silva de Lima, Y. M., Leyton, A. O., da Silva, L. A., Quevedo, R. P., Goulart Fiscina, L. F., Sousa, M., Calado, B. N., Cunningham, C., Carneiro da Silva, I. S., Noronha, P., de Meneses, E., Leão, H., … Anderson, L. O. (2026). Western amazon 2020 extreme heatwave. Regional Environmental Change, 26(4), Article 192. https://doi.org/10.1007/s10113-026-02681-0

Image Credits: AI Generated

DOI: 10.1007/s10113-026-02681-0

Keywords: climate change, heatwave, Amazon, drought, wildfires, event attribution, heat stress, WBGT, public health, respiratory disease, Acre, CMIP6