Loss has fallen, but risk persists.
The newly reported rate is a record low for this monitoring series. Degradation, fire, warming and earlier cumulative clearing still influence how forests function.
ENVIRONMENT / FIELD NOTE 001 / UPDATED 16 OCT 2026
One rainforest.
A planetary
water system.
The Amazon isn't just a collection of trees. It helps move moisture across a continent, stores carbon and sustains millions of lives. Explore what satellite measurements show—and what scientists' models warn could happen next.
Evidence and model projections are deliberately separated. No single loss percentage confirms that the Amazon has crossed a biome-wide tipping point.
2026 / PRODES PRELIMINARY
3,335 km²
Forest clearance in the Brazilian Legal Amazon, August 2025–July 2026.
01 / ACTUAL OBSERVATIONS
What the satellites have seen.
Annual deforestation in the Brazilian Legal Amazon—not the entire Amazon rainforest and not the percentage of original forest remaining.
SELECTED REPORTING YEAR
3,335 km²
2026 · Preliminary satellite estimate
Use arrows to change the year; the chart updates in response.
Compared with 2025: −2,396 km² (−41.8%).
19881995 peak: 29,059 km²2026 preliminary
Annual area cleared, km²
Read this correctly: 3,335 km² is land cleared in the 2026 reporting year. It is not “3,335 km² of forest left” and does not translate directly into a biome-wide loss percentage.
Different boundaries, different answers.
PRODES reports the Brazilian Legal Amazon; scientific tipping studies often model the larger Amazon biome. They should not be treated as identical datasets.
A threshold is not a stopwatch.
The warning is about interacting pressures and modelled risk. It does not mean that exactly 26% of the forest is already gone or that collapse automatically occurs at 26%.
02 / GEOGRAPHY AND MECHANISMS
Explore the living network.
Select a physical process, then choose a state to explore the latest annual satellite figures.
Amazon focus areaApproximate schematic outline
REGIONAL RECORDS AVAILABLE
Showing 2026 preliminary state figures
77°W / 4°NSOUTH AMERICA
Choose a state
Borders use simplified Natural Earth geometry. The shaded Amazon footprint and process arrows are illustrative, not satellite measurements. State markers denote approximate locations, not mapped state boundaries.
VERIFY THE PIXELS
See the measured landscape, not just our diagram.
This illustration is a teaching aid. Open original satellite-derived maps to compare actual land-cover classifications and mapped clearings. MapBiomas and PRODES measure different quantities; do not equate their percentages.
03 / FUTURE SCENARIOS · NOT OBSERVATIONS
What happens when the forest can no longer maintain itself?
Change two hypothetical drivers. Compare your selections against the conditions reported in a Nature 2026 dynamical-systems study—not a live forecast of the Amazon.
YOUR SCENARIO
Move the conditions.
0%22–28% research band40%
Assumed proportion of original forest removed for this exercise, not a measured current rate. Different datasets define original forest and study extent differently.
+1.0°C+1.5 to 1.9°C model range+4.0°C
This is an assumed warming level in the model, not a measurement of the temperature at a particular rainforest location.
TRY A SCENARIO
INTERPRETATIONRESEARCH WARNING BAND
Conditions within a published model's warning range.
Under the study's scenarios, a combination of 22–28% deforestation and 1.5–1.9°C global warming was associated with a possible near system-wide ecological transition. This is not a measured tipping point.
ASSUMED FOREST REMAINING74%Arithmetic complement of the scenario loss input, not observed cover
AMAZON AREA AT RISK IN STUDY62–77%Modelled vulnerable area under specified combined pressures
Scientific boundary: sliders explore a published risk range; they cannot calculate rainfall in millimetres, future CO₂ emissions, or when/where forest might irreversibly change.
INTERACTIVE MODEL REFERENCE / NOT A FORECAST
Where does your scenario sit?
The outlined area marks one published combination of forest loss and warming. No colours here predict the probability of collapse outside that combination.
Published joint range Your assumption
Swipe the graph horizontally to inspect the full axes. The sliders above remain available for precise control.
Your assumption: 26% forest loss · +1.7°C
Inside the study’s highlighted joint-parameter window; not a quantitative risk forecast.
Select a point by tapping the grid, moving the sliders, or using arrow keys when the plot is focused (← / → forest loss, ↑ / ↓ warming). A separate warming-only threshold of approximately 3.7–4.0°C was reported for a different model condition; it is not drawn as a universal hazard region. Read the study ↗
04 / THE CHAIN OF CONSEQUENCES
From leaves to rivers to the atmosphere.
These causal pathways are grounded in published research. Their exact magnitude depends on location, season, warming, degradation and ecosystem responses.
ILLUSTRATED FEEDBACK LOOP / CONCEPTUAL
Follow one molecule of water.
- 01Soil & rootsAbsorb available water
- 02TranspirationRelease water vapour
- 03AtmosphereMove and recycle moisture
- 04PrecipitationReturn some water as rain
- 05Forest responseSupport regrowth or face stress
Conceptual pathway only: rain does not necessarily fall where evaporation occurs. Flow strength and direction vary in space and time; the animation is not a transport measurement.
Less forest can mean less dry-season rain.
HISTORICAL / SOUTHERN BASIN 1980–20198–11% less annual rainfallEstimated observed trend; study attributes roughly 52–72% to deforestation in southern/upwind regions.
TREESEVAPOTRANSPIRATIONMOISTURE / RAINFALL
Forest loss reduces local evapotranspiration and can weaken moisture recycling. A 2025 Nature study found important seasonal differences: rain may increase locally in cleared grids during wet months, even while dry-season rainfall falls.
Read the rainfall study ↗Forest damage can shift the carbon balance.
ESTIMATED ABOVEGROUND CARBON / 2010–2020−0.37 ± 0.17 PgCNet stock change in one research reconstruction—not future CO₂ emissions or a current annual rate.
BIOMASSFIRE + DECAYATMOSPHERIC CO₂
Living forest stores carbon. Clearing, fire and forest degradation can release carbon, while regrowing or intact stands may absorb it. The net carbon balance depends on multiple competing processes—not just percent cleared.
Read the carbon study ↗Stress can reinforce more stress.
2050 / 2024 SCIENTIFIC PROJECTION10–47% exposedShare of Amazon forests exposed to compounding disturbances under scenarios; different outcome from the 2026 transition model.
DROUGHTFIRE + TREE LOSSMORE HEAT / DRYING
Warming, forest loss and fire can form reinforcing feedbacks. Modelling published in 2026 indicates that deforestation-driven drying can lower the warming conditions under which large-scale transitions are possible.
Read the threshold model ↗05 / WHAT CAN CHANGE
A forest still worth protecting.
Actions should be assessed against measurable outcomes, not slogans. Progress in annual clear-cutting does not erase prior loss, warming, fire, degradation or the need to support affected communities.
Detect clearing and act on evidence.
Use satellite alerts, land records and enforceable safeguards to stop illegal clearing before it becomes permanent.
What would show progress?
Verified reductions in annual PRODES rates, alongside credible inspection, enforcement and leakage tracking. DETER alerts and PRODES annual estimates are different instruments.
Support stewardship and secure rights.
Protected areas, Indigenous territories and local expertise require durable rights, funding and governance rather than symbolic designations.
What would show progress?
Tenure security, local participation, forest integrity and independent measures of social and ecological outcomes over time.
Restore connections, not just tree counts.
Prioritize degraded landscapes, native biodiversity, watershed continuity and long-term survival instead of one-day planting totals.
What would show progress?
Native forest recovery, persistent canopy and carbon gains, hydrological indicators and field-verified biodiversity—not only hectares pledged.
Lower the warming pressure.
Reducing global greenhouse-gas emissions complements forest protection because warming and forest loss interact in published model experiments.
What would show progress?
Verified emissions reductions and climate trajectories, considered alongside local forest degradation and drought indicators.
06 / REPRODUCIBLE RESEARCH
Evidence behind every claim.
Primary monitoring data, peer-reviewed studies and the original map geometry are linked so visitors can check the definitions themselves.
01 / DATA · 16 OCT 2026MMA / INPE · PRODES 2026 preliminary rate and state tableAnnual cleared area / Brazilian Legal Amazon02 / DATA SERIESINPE TerraBrasilis · official annual rates1988–2026 annual time-series monitoring03 / MODEL · MAY 2026Wunderling et al. · Deforestation-induced drying lowers Amazon climate thresholdModelled combined deforestation and warming thresholds04 / OBSERVATIONS · 2025Qin et al. · Wet/dry-season precipitation responsesScale-dependent regional rainfall effects05 / OBSERVATIONS · 2026Historical deforestation and southern-basin rainfall1980–2019 trends and moisture transport06 / OBSERVATIONS · 2025Aboveground carbon dynamics and disturbanceCarbon losses and gains across forest types
Methods and limits
The bar chart uses annual PRODES area-cleared estimates for Brazil's Legal Amazon. The 2026 number is preliminary; the 2025 number is consolidated. Changes in revised historical numbers are documented in the source manifest. The mapped state figures for 2025 are consolidated and for 2026 preliminary; only these two regional years are available in this release. Selecting earlier national years does not create unavailable state observations. The map uses simplified Natural Earth geometry taken from the existing Learning Atlas. The translucent Amazon outline is schematic, not an exact biome boundary or loss raster. Scenario sliders are qualitative literature interpretation only, not an operational climate model. The published model parameter rectangle is a literature reference, not a calibrated risk surface. No real-time feeds or site visitor information are collected. Download the 39-year CSV ↧ · Download state data ↧