Fire: Fires

MAAP #250: Amazon Tipping Point – Where Are We? (2026 Update)

Posted on by Matt Finer
Figure 1. Amazon primary forest loss. Data: Amazon Conservation/MAAP, UMD.

It is often reported that the largest rainforest in the world, the Amazon, is rapidly approaching a tipping point.

Generally defined, this tipping point is where parts of the rainforest will convert into drier ecosystems due to disrupted precipitation patterns caused by increasing deforestation.

The Amazon uniquely generates much of its own rainfall by recycling water from its source in the east (Atlantic Ocean) all the way west to the Andes Mountains and beyond, in a phenomenon known as flying rivers (see MAAP #232 and the follow-up white paper). Thus, high deforestation in the eastern Amazon may lead to downwind impacts (reduced precipitation) in the western Amazon. Importantly, these flying rivers follow different pathways during the wet season (northern route) vs. the dry season (southern route), implying that the tipping point is not a uniform event across the Amazon biome, as some areas are more vulnerable to atmospheric moisture (flying rivers) disruptions than others.

The scientific literature indicates this tipping point could be triggered at around 25% Amazon forest loss, in conjunction with climate change impacts (see References).

In a previous report published in 2022 (MAAP #164), we reconstructed the likely extent of the original Amazon biome forest prior to European colonization, and then estimated accumulated total Amazon forest loss (through 2021). Based on this analysis, we estimated that 13.2% of the original Amazon biome primary forests were lost to deforestation and fires.

Here, for our 250th report, we update this estimate by incorporating forest loss data through 2025. From 2022 to 2025 we found an additional forest loss of 13 million hectares, including the record-breaking year of 2024 for fires. Our updated estimate is that 15% of the original Amazon biome primary forest has been lost, mostly due to deforestation and fires (Figure 1).

In addition, breaking the results by seasonal moisture flow routes (see Figure 3 below), we find that moisture flows during the dry and transition seasons are most affected and disrupted (9.1% and 8.6% forest loss, respectively). This finding is critical because during the critical dry and transition seasons, the role of forest-driven moisture recycling is especially important to maintain rainforests in the southwest Amazon. In contrast, we find that the wet season moisture flow is more intact, with just 3.2% total forest loss.

Original Amazon Forest

As first presented in MAAP #164, we merged several datasets to estimate the extent of the original Amazon forest (Figure 2). We define this original forest as the extent of the Amazon forest before European colonization (and thus may include changes from pre-European indigenous populations). Note that we use a broader biogeographical definition of the Amazon that covers nine countries (Amazon biome) rather than the strict Amazon watershed (see Methodology).

The map has just three classes: Original Amazon forest, Original non-forest (such as natural savannah), and Water. 

We found that the original Amazon primary forest covered 647.6 million hectares (equivalent to 1.6 billion acres).

Of this total, 61.4% occurred within the current border of Brazil, followed by Peru (12%), Colombia (7%), Venezuela (6%), and Bolivia (5%). The remaining four countries (Ecuador, Guyana, Suriname, and French Guiana) make up the final 8%.

Figure 2. Original Amazon biome forest. Data: Amazon Conservation/MAAP. Source MAAP#164.

Amazon Forest Loss

In MAAP #164, we reported the historical accumulated Amazon primary forest loss, as of the end of 2021. We documented the total loss of 85.5 million hectares, or 13.2% of the original Amazon.

For this report, we updated the data as of the end of 2025. This data is broken down into three time periods:

By 2000, we estimate that the original Amazon forest had lost 36 million hectares (MAAP #164).

The 2001-2021 estimate of primary forest cover loss was reduced from 49.5 million to 48 million hectares due to adjustments in the forest loss methodology used by the University of Maryland (UMD). It is important to highlight that this finding includes forest loss from both fire (7.3M ha) and non-fire (40.7M ha) causes, and does not include forest gain/regeneration.

A novel 2022-2025 estimate was generated based on the most recent UMD data, adding 13.04 million hectares, including the record-breaking year of 2024 that was mainly driven by forest fires following the strong El Niño event of 2023 (see MAAP #229 and #248). Similarly, this includes forest loss from both fire (5.43M ha) and non-fire (7.61M ha) causes. This total loss over just four years is the equivalent size of the US state of Louisiana.

Combining these three periods gives an updated total of 97.06 million hectares lost (equivalent to 240 million acres). This includes forest loss from both fire (12.78 M ha) and deforestation (84.28 M ha) causes.

The largest loss occurred in Brazil (79.4%), followed by Bolivia (6.2%),  Peru (5.8%), Colombia (4.5%), Venezuela (1.7%), Guyana (0.8%), Ecuador (0.7%), Suriname (0.6%), and French Guiana (0.2%).

Taking this updated forest loss total of 97.06 million hectares in relation to the original 647.6 million hectares gives our key finding of 15% of the original Amazon forest lost, mostly due to deforestation and fire (Figure 1). This estimate also includes natural causes such as landslides and windstorms.  

Figure 3 further illustrates the results broken down by seasonal moisture flow routes (as presented in MAAP #232). We found the transition season moisture flow is most threatened and impacted (9.1% total forest loss), followed by the dry season moisture flow (8.6%).

This finding is important because during the critical dry and transition seasons, the role of flying rivers to carry moisture to water-stressed forest in the southwest Amazon is especially important. In other words, extensive deforestation in the eastern and central parts of these pathways could negatively impact the downwind western part, which means that the tipping point event is not a single and homogenous one across the Amazon biome and there will be geographical areas that will be impacted sooner and harder.

In contrast, we find that the wet season moisture flow is more intact, with just 3.2% total forest loss.

Figure 3. Amazon forest loss in flying rivers pathways. Data: Amazon Conservation/MAAP

Methodology

At the core of this work, we generated two major estimates: original Amazon forest and total historical Amazon forest loss.

For both of these estimates, we used the biogeographical boundary of the Amazon (as determined by RAISG 2020), which encompasses nine countries. Thus, we used a broader definition of the Amazon (Amazon biome) rather than the strict Amazon watershed, which omits part of the northeastern Amazon biome.

As presented in MAAP #164, for original Amazon forest, we defined three major classes: Forest, Non-Forest, and Water. This analysis was based on data from MapBiomas Brazil (collection 2 from 1990) with some additional modifications. As noted above, we define this original forest as the extent of the Amazon forest before European colonization (and thus may include changes from pre-European indigenous populations).

Original Forest was made up of these MapBiomas categories: Forest Formation, Mangrove, Flooded Forest, Mosaic of Agriculture and Pasture. Non-Forest was made up of these MapBiomas categories: Savanna Formation, Natural Non-Forest Flood Formation, Grassland, and Other non-Forest Formations. Water was made up of these MapBiomas categories: River, Lake, Ocean and Glacier. 

We then made a number of modifications with manual edits based on data from the University of Maryland, INPE (Terrabrasilis), ArcGis satellite images, Planet mosaics, Google Earth Engine Landsat images from 1984-1990, and official government data for several countries (Ministry of the Environment of Ecuador (MAE) and Peru (GeoBosques/MINAM), Forest and Carbon Monitoring System/IDEAM of Colombia, National Institute for Space Research of Brazil (INPE/Terrabrasilis), General Directorate of Forest Management and Development of Bolivia (DGGDF), and the National Service of Protected Areas of Bolivia (SERNAP). As an example of a major modification, deforested areas and historic dam reservoirs were changed to Original Forest based on an analysis of the oldest available satellite image for the area (1984-1990). We also corrected some misclassifications, such as forest patches in clearly non-forest areas were changed to Non-Forest (and vice versa) and mountain forest areas found as water were changed to Forest. Also, agriculture and urban areas in likely savannah areas were changed to Non-Forest. Additional Water data from MapBiomas based on 1985 was incorporated. Overall, our focus was defining Original Forest as best as possible; data confusions between Non-Forest and Water categories were not worked on as thoroughly. 

For original Amazon forest,For total historical Amazon forest loss, we used data from the University of Maryland. Specifically, we first used their data layer ‘Tree Cover 2000″ (>30% canopy density) to estimate historical (pre-2000) forest loss.

We then used the Global Forest Change (GFC) and Global forest loss due to fire datasets developed by the University of Maryland (UMD) for the period 2001–2025. We then derived annual forest loss from the loss year layer of GFC. Subsequently, we separated fire-related from non-fire-related forest loss using the annual Global forest loss due to fire dataset. Areas classified with medium to high certainty of fire attribution (codes 3 and 4) were considered fire-related forest loss, whereas codes 1 and 2 were used to identify non-fire-related forest loss, representing areas of forest loss attributed to other drivers. This classification was applied to minimize potential overestimation of fire-related forest loss. All data processing and analyses were conducted using Google Colaboratory.

It is important to note that for the more modern post-2000 era (2001-25), data is available for forest loss due to both fire and non-fire causes. This data represents our best estimate or proxy for primary forest loss.  Non-fire loss can be considered more permanent, such as deforestation, while fire loss may be considered more as degradation. The results above combine both types of data to be consistent with the original results presented in MAAP #164.

It is also important to note that our estimates do not include forest gain/regeneration.

Finally, we estimate the amount of primary forest loss in the major seasonal moisture flow pathways (wet, dry, and transition seasons), as presented in  MAAP #232.

References

Sampaio, G., Nobre, C., Costa, M. H., Satyamurty, P., Soares‐Filho, B. S., & Cardoso, M. (2007). Regional climate change over eastern Amazonia caused by pasture and soybean cropland expansion. Geophysical Research Letters, 34(17).

Hansen, M. C. et. al. (2013) High-Resolution Global Maps of 21st-Century Forest Cover Change. Science 342.

Nobre et al. (2016) Land-use and climate change risks in the Amazon and the need of a novel sustainable development paradigm. PNAS, 113 (39).

Turubanova S., Potapov P., Tyukavina, A., and Hansen M. (2018) Ongoing primary forest loss in Brazil, Democratic Republic of the Congo, and Indonesia. Environmental Research Letters.

Lovejoy, T. E., & Nobre, C. (2018). Amazon Tipping Point. Science Advances, 4(2).

Lovejoy, T. E., & Nobre, C. (2019). Amazon tipping point: Last chance for action. Science Advances, 5 (12).

Bullock et. al. (2019) Satellite-based estimates reveal widespread forest degradation in the Amazon. Glob Change Biol., 26.

Amigo, I. (2020) The Amazon’s fragile future. Nature, 578.

Killeen (2021) A Perfect Storm in the Amazon Wilderness

Berenguer E. et. al. (2021) Ch 19. Drivers and ecological impacts of deforestation and forest degradation. In: Nobre C, Encalada et al. (Eds). Amazon Assessment Report 2021. United Nations Sustainable Development Solutions Network, New York, USA. Available from https://www.theamazonwewant.org/spa-reports

Hirota M et. al (2021) Science Panel for the Amazon, Ch 24. Resilience of the Amazon Forest to Global Changes: Assessing the Risk of Tipping Points. In: Nobre C, Encalada et al. (Eds). Amazon Assessment Report 2021. United Nations Sustainable Development Solutions Network, New York, USA. Available from https://www.theamazonwewant.org/spa-reports/

Wunderling et al (2022) Recurrent droughts increase risk of cascading tipping events by outpacing adaptive capacities in the Amazon rainforest. PNAS 119 (32) e2120777119.

Acknowlegments

This work was supported by NORAD (Norwegian Agency for Development Cooperation).

Citation

Finer M, Mamani N, Ariñez A, Castillo H (2026) Amazon Tipping Point – Where Are We? (2026 Update).

MAAP #248: Implications of upcoming major El Niño event on Amazon fires

Posted on by Matt Finer
Figure 1. El Niño 2023. Data: Google Gemini

In our annual series on forest loss trends in the Amazon (MAAP #244 & MAAP #229), we observed that the two most recent intense fire seasons, 2016 and 2024, immediately followed major El Niño events. 

For example, extreme drought conditions created by a strong El Niño in late 2023 created extreme drought conditions the following year, enabling widespread fires across the Amazon in 2024. In contrast, the reduced fires of 2025 were associated with the moister conditions of La Niña.

For background, the definition of El Niño is based on warm sea surface temperature in a specific zone of the Pacific Ocean several hundred kilometers off the coast of South America (Figure 1).

Specifically, a Sea Surface Temperature Anomaly (SSTA) above 0.5 degrees Celsius in the El Niño 3.4 Zone is classified as an El Niño event (See Methods section for details). Anomalies above 1 degree C are a moderate event, and above 2 degrees C, a very strong event.

We took this SSTA data annually from 2002 to 2025 for the 3-month period of October-November-December, which typically correlates with peak temperature anomalies.

We then analyzed this data in relation to the following year’s fire-caused primary forest loss (data from the University of Maryland). We focused the analysis on years with at least a weak El Niño (SSTA > 0.5).

Results

Figure 2. El Niño and Amazon Fires. Data: UMD, NOAA

Filtering the dataset for El Niño years (SSTA > 0.5 degrees C) isolates 8 specific historical events (2002, 2004, 2006, 2009, 2014, 2015, 2018, and 2023). 

Of these events:

Four were Weak El Niño (SSTA 0.5- 0.9): 2004, 2006, 2014, and 2018;

Two were Moderate (SSTA 1.0 – 1.4): 2002, 2009;

One was Strong (SSTA 1.5 to 1.9): 2023;

And one was Very Strong (SSTA ≥ 2.0): 2015.

To analyze the delayed (lagged) impact, the SSTA from each year (Year t) is paired directly with the Fire Primary Forest Loss Area from the following year (Year t+1). 

See Figure 2 for the visual correlation between El Niño events (blue) and primary forest loss from fires (red).

Note the initial observation that prompted this study: the major peak in fires in 2016 following the Very Strong El Niño of 2015, followed by the record-breaking fire season in 2024 following the Strong El Niño of 2023.

Despite a limited sample size (only 8 El Niño events since 2000) and limited statistical power, the analysis reveals a clear, strong positive relationship between the strength of El Niño events (that is, magnitude of the sea surface temperature anomaly) and the severity of the following fire season. 

The strength of this relationship approaches the traditional margin of statistical significance (probability value under  0.05), with a probability value of 0.08 (see the Annex for details).

Thus, the data do indicate a potential escalating relationship between El Niño and Amazon fires.

Most notably, the two highest sea surface temperature anomalies —2015 (+2.34) and 2023 (+1.49)—directly preceded the two largest fire seasons, 2016 (1.76 million hectares) and 2024 (2.79 million hectares), respectively.

Low-tier warm anomalies (like +0.52 or +0.73) correspond to lagged fire sizes under 160,000 hectares.

Discussion

As detailed above, the available data indicates a strong positive correlation between the severity of El Niño events and the intensity of the Amazon fire season the following year.

This correlation is anchored by two key events that prompted this analysis in the first place: the peak fire season in 2016 following the Very Strong El Niño of 2015; and the record-breaking fire season in 2024 (see MAAP #229) following the Strong El Niño of 2023.

There is currently an urgency regarding these findings due to the predicted Very Strong El Niño event (that is, SSTA > 2 degrees C) in late 2026 and implications for the 2027 fire season. Some predictions have the SSTA as the highest on record, even approaching 3 degrees C by November to December 2026. 

The current SSTA (for May-June-July 2026) is +0.98 degrees C (Weak to Moderate El Niño), rapidly coming out of last year’s La Niña phase, and warming fast.

Thus, the best recent analogue is the Very Strong El Niño of 2015, and the subsequent burning of nearly 1.8 million hectares of primary Amazon forest the following year in 2016. The most recent analogue is 2023 (see Figure 1), and the subsequent record-breaking fire season of 2024.

We also raise the possibility that the impact on fires carries over to a second year following a major El Niño – this could account for the severe fire seasons of 2017 (1 million hectares) and 2025 (1.5 million hectares).

The weaker El Niño years show a consistent directional pattern but without the same magnitude response.

We also note a major structural break in the data: in general, fires are much greater starting in 2016 (average fire area pre-2016: ~150k ha; Post-2016: ~1.0M ha). This time frame slightly precedes, overlaps, and then continues after the President Bolsonaro administration (2019-2023), which weakened illegal deforestation enforcement policies. Thus, there may be a new modern elevated baseline of fire activity in the Brazilian Amazon.

Since El Niño typically peaks in November to December, which corresponds to the onset of the wet (monsoon) season, the mechanism for this correlation may be shutting down the convection engine driving the rainy season. The disruption of the wet season moisture flow could extend the dry and transition seasons, reducing precipitation, creating drier conditions, and preventing the soil from recovering its moisture, setting up the Amazon for an intense fire season.

It is worth noting that another recent study found that the years with the largest burned areas in the Peruvian Amazon between 2013 and 2024 coincide with the occurrence of El Niño events and drought conditions (Reference 10).

How ready are Amazonian countries for the upcoming 2027 El Niño fire season?

Experience of 2024

The 2023 El Niño and subsequent 2024 fire season exposed the depth of the preparedness gap across the Amazon basin. The responses that emerged were largely reactive, deployed well into a crisis that monitoring systems had anticipated months earlier. Brazil, Peru, and especially Bolivia were most affected among the Amazonian countries, and this review of institutional preparedness focuses on these three countries. Drier conditions created by El Niño, however, created an increased fire risk across the entire biome, affecting all countries.

Brazil mounted the most substantial national response, though it arrived belatedly. By August 2024 — almost three months since the wildfires started — a state of emergency was decreed in 45 municipalities and 48 Brazilian cities were put on high alert (Reference 1). In the end, fires directly impacted 1.9 million hectares, the highest on record (MAAP #229).

Peru also declared a national emergency during the 2024 crisis, unlocking emergency coordination powers and international assistance. In the end, however, fires impacted 47,574 hectares, more than double the previous high (MAAP #229).

Bolivia also had its worst fire season on record in 2024, as fires directly impacted  779,960 hectares, far surpassing the previous record (MAAP #229). The main cause of deforestation and subsequent fires was likely agrarian speculation. Two supreme decrees in 2024 granted zero tariffs and tax incentives to agribusiness, which triggered a surge in demand for land for monocultures of soy, sugarcane, and seed oils. This expansion was supported by existing regulations that expedite the clearing of up to 20 hectares.  The government declared a national emergency and later a national disaster, but the response was hampered by weak enforcement: Bolivia’s fines for illegal forest burning (under $20/hectare) are roughly 2% of Brazil’s, and tensions between national and local government further slowed emergency coordination. To contain the disaster, the government declared an Ecological Environmental Pause, which suspended burning permits and placed the burned public lands under quarantine for five years. The Inter-American Commission on Human Rights’ special rapporteur documented how tensions between national and local governments hindered emergency responses, while policies aimed at aggressively expanding industrial agriculture had dried out large areas of the country, leaving ecosystems vulnerable to fire.

Changes implemented in 2025

Fire impact was much lower in 2025 relative to 2024 (1.5 million vs 2.8 million hectares of primary forest burned, respectively) (MAAP #244). Several Amazonian countries implemented stronger fire management policies in response to the 2024 fire season. The reduced burnt area, however, cannot be solely attributed to better institutional preparedness, as climate conditions were also more favorable, with La Niña bringing relief through higher humidity levels across the biome. 

In February 2025, Brazil’s Environment Minister Marina Silva decreed a pre-emptive environmental emergency ahead of the fire season, giving authorities extra powers and resources to contain wildfires before they spread. The federal government committed to hiring an additional 250 federal firefighters and allocated 45 million reais to bolster state-level fire brigades across six Amazonian states (Reference 3). Other structural measures followed: Brazil’s Supreme Court ordered the federal government and all Amazon and Pantanal states to draw up emergency fire management plans, while the Ministry of Environment announced “governance offices” in 70 Amazon municipalities — supported by FUNBIO and UNDP — to provide vehicles, boats, drones, and training for fire prevention (Reference 4). The Prevfogo program, managed by IBAMA, was expanded as a platform integrating satellite-based detection with ground response. These combined efforts contributed to a measurable decline in fire activity in 2025 compared with the previous year (MAAP #244).

In Peru, in response to the 2024 events, the Ministry of Environment proposed a Forest Fire Prevention and Control Law to establish regional firefighting brigades and prohibit land-use change in areas affected by wildfires, though recent amendments to the Forest Law (Law No. 31973) may undermine this approach (Reference 5). Seven of Peru’s 24 regions have published forest fire prevention and risk reduction plans, and a Multisectoral Plan for Wildfire Response 2025–2027 is currently being developed, drawing on risk assessments that classify nearly 20% of national territory as high or very high fire risk.

In Bolivia, the legislative trajectory since 2024 has moved in the opposite direction. Agribusiness lobbies have pressured to roll back burning fines, and a land reform that would have accelerated deforestation in Santa Cruz and Beni, the departments that burned most severely in 2024, was enacted in the first half of 2026, and then repealed in the face of social protests (Reference 6). The country has been navigating its worst economic and political crisis in decades, and institutional preparedness for the 2027 fire season could be severely challenged under these conditions. 

In terms of regional cooperation, the most significant recent development is the Operational Understanding for Preparedness and Response, approved by ACTO member countries in early 2026. This non-binding regional cooperation framework establishes mechanisms for coordinating mutual assistance among Amazonian countries when the magnitude of forest fires requires a joint response, facilitating the articulation of national capacities, exchange of operational information, and technical support during fire-related emergencies. Concurrent with its adoption, member countries agreed to establish the Forest Fire Response Committee (CRIF), a regional body tasked with developing technical instruments, operational procedures, and coordination mechanisms to improve preparedness and joint response (Reference 7).

In conclusion, the reduction in burnt area observed across the three countries in 2025 occurred under favorable climate conditions, which, as our analysis shows, are unlikely to be seen in 2027. The true test of institutional preparedness will come under El Niño-driven stress. The regional architecture, meanwhile, can only perform as well as its weakest national link. Under a degraded political landscape and without policy improvements since 2024, Bolivia could become the epicenter of the 2027 fire season.

Outlook for 2026 – 2027

From the FAO’s El Niño Fire Readiness checklist (Reference 8), there are five major components of readiness for next year’s 2027 fire season.

First is “Fire Awareness”, which includes understanding historical trends that distinguish an average season from an anomaly season, and what drives major fires (such as fuel load from recently deforested areas) that could be incorporated into Risk Reduction strategies.

As noted above, annual MAAP reports (MAAP #229, MAAP #244) provide historical trends for the Amazon regarding fire impact to primary forests. These reports allow for the identification of intense, or anomalous, fire seasons (such as 2024).

In terms of what drives major fires, previous MAAP reports have revealed the tight link between deforestation and fires in the Amazon (MAAP #189). That is, most major fires are burning recently deforested areas, and then may escape into surrounding forests, especially with extended dry conditions such as 2016 and 2024. For example, over 70% of major fires in the Brazilian Amazon burn recently deforested areas  (MAAP #189). There is also a strong link between deforestation and fire in the Bolivian Amazon, where deforestation also often precedes fires, which may then escape into surrounding forests or savannahs.

Thus, one of the key strategies to minimize major fires in 2027 is to minimize new deforestation in 2026.

Second is the establishment of “Fire Danger Rating and Early Warning” systems and the provision of this information to stakeholders.

As the El Niño intensifies throughout 2026, this could include the monitoring of moisture or drought conditions and issuing respective warnings.

Third is “Fire Preparedness,” and includes guidance for prevention, detection, and suppression of fires based on the above rating and warning system.

In terms of prevention, fire-free agricultural practices could be promoted in 2026 and more strictly enforced in 2027, coordinating closely with local rural unions and cooperatives (Reference 9). Local community initiatives represent a critical component of this preparedness, serving as tangible examples of prevention, detection, and suppression mechanisms.

For instance, the Territorial Strengthening Actions implemented across the Amazon basin demonstrate this integrated approach: communities in Colombia, Peru, Bolivia and Ecuador are actively developing local capacities for fire detection, sustainable management (safe fire use) and early response, while promoting alternative practices to slash-and-burn agriculture. 

This community-based approach is further strengthened by broader regional efforts, such as the CoRAmazonia program. This initiative promotes an Integrated Fire Management strategy that integrates technical, scientific, and traditional indigenous knowledge. Through platforms like ExpoMIF and the Amazon Network for Integrated Fire Management (RAMIF), the program highlights the crucial role of Indigenous peoples and traditional communities in fire governance, emphasizing the importance of community brigades, territorial monitoring, and the rescue of affected wildlife.

Fourth is “Pre-fire Season Activities,” which highlights whether agencies are adequately engaging stakeholders and setting up national agreements.

Fifth is “Fire Detection, Communication, and Dispatching,” which forms the real-time fire Response procedures and mechanisms.

In terms of detection, real-time fire monitoring should be integrated into national response protocols and field-level coordination.

For communication and dispatching, there could be the establishment of contingency plans to extinguish forest fires rapidly before they spread, focusing heavily on fires starting in recently deforested areas and expanding towards protected areas or Indigenous territories.

In Peru specifically (Reference 10), the most critical districts are concentrated in the Ucayali, Madre de Dios, and Huánuco regions. Regarding public budget allocation, there are gaps in the prevention of and response to forest fires, resulting in limited operational capabilities and significant disparities in the allocation and execution of public resources for disaster risk management. Overall, there is the need to strengthen systems for monitoring, early warning, planning, and institutional response regarding forest fires, prioritizing areas with the greatest historical vulnerability.

Methods

The analysis combined two primary data sources: 1) Relative Oceanic Niño Index (RONI) based on sea surface temperature anomaly (SSTA) and 2) annual primary forest loss from fires.

RONI, the standard used by NOAA for classifying El Niño (warm) and La Niña (cool) events in the eastern tropical Pacific, is the running 3-month average sea surface temperature (SST) for the Niño 3.4 region, minus the overall average tropical 3-month sea surface temperature anomaly (SSTA). For this study, we used the 3-month period of October-December for each year. 

Events are defined as 5 consecutive overlapping 3-month periods at or above the +0.5 anomaly for warm (El Niño) events and at or below the -0.5 anomaly for cool (La Niña) events.  The threshold is further broken down into Weak (with a 0.5 to 0.9 SST anomaly), Moderate (1.0 to 1.4), Strong (1.5 to 1.9), and Very Strong (≥ 2.0) events. 

In addition to warm SSTA in the Pacific Ocean, Amazon drought conditions may also be influenced by warm SSTA in the North Tropical Atlantic Ocean (Reference 10), but our study did not account for this.

The fire analysis was based on 30-meter resolution annual forest loss data produced by the University of Maryland and also presented by Global Forest Watch. This Global Forest Loss due to fire dataset is unique in terms of being consistent across the Amazon (in contrast to country-specific estimates) and distinguishes forest loss caused directly by fire (note that virtually all Amazon fires are human-caused). The values included were ‘medium’ and ‘high’ confidence levels (code 3-4).

For the baseline, it was defined to establish areas with >30% tree canopy density in 2000. Importantly, we applied a filter to calculate only primary forest loss by intersecting the forest cover loss data with the additional dataset “primary humid tropical forests” as of 2001 (Turubanova et al 2018). For more details on this part of the methodology, see the Technical Blog from Global Forest Watch (Goldman and Weisse 2019).

The analysis was supported and reviewed by AI: Gemini and Claude

Annex

Because we have a constrained subset (N = 8), the statistical power is limited, but a clear, strong positive pattern is evident across several major metrics.

The Pearson Linear Correlation indicates a strong positive linear relationship between the magnitude of the warm anomaly and the next year’s fire area (r = +0.6146), approaching the traditional margin of statistical significance (p-value = 0.1049). This correlation is even stronger when we take the natural log of the fire area to account for natural fire activity  (p-value = 0.0827).

Spearman Rank Correlation reaffirms that higher SSTA values rank alongside worse fire outcomes in the succeeding year ( +0.5476; p-value= 0.1600).

  • Log-Transformed Pearson Correlation (r log): +0.6474
  • Wildfire spread typically behaves exponentially rather than linearly. When we take the natural log of the fire area (ln{Fire}_t+1) to correct for this, the correlation strengthens to nearly +0.65, approaching the traditional margin of statistical significance (p < 0.10).

References

1. Lancet Regional Health – Americas / PMC. (2025). The 2024 South America ablaze: Health impacts and policy imperatives for protecting population health in an era of wildfires

2. IACHR (2025) REDESCA publishes report on wildfires in Bolivia and calls for urgent action to address impacts on human rights and ecosystems

3. Mongabay (2025). Brazil declares environmental emergency ahead of 2025 fire season.

4. Muggah, R., & Szabo, I. / Mongabay. (2025). Brazil is speeding-up forest fire prevention to avoid dangerous tipping points in the Amazon. 

5. Lancet Regional Health – Americas / PMC. (2025). The 2024 South America ablaze. (See footnote 3.) Peru-specific governance section

6. Copa Pabón, M. V. (2026, May 15). Indigenous protest forces repeal of land privatization law in Bolivia.

7. ACTO. (2026, February–March). ACTO member countries approve Operational Understanding for Preparedness and Response to Forest Fires in the Amazon Region.

8. FAO El Niño Fire Readiness checklist (2024) As presented in the Global Fire Management Hub and informed by its Integrated Fire Management Voluntary Guidelines

9. Brown F, et al (2026). Alert related to the coming severe dry period, heat waves, fires, smoke and climate in the Madre de Dios-Peru, Acre-Brazil, Pando-Bolivia (MAP) Region – 30 April 2026. Foster Brown

10. ACCA (2026) El Niño y los incendios forestales en la Amazonía peruana, ¿qué podemos esperar para la temporada 2026-2027?

Citation

Finer M, Bodin B (2026) Implications of upcoming major El Niño event on Amazon fires. MAAP: 248.

Acknowledgements

We thank colleagues from the following organizations for helpful comments on the report: Conservación Amazónica – ACEAA in Bolivia and Conservación Amazónica – ACCA in Peru.

This work was supported by Norad (Norwegian Agency for Development Cooperation).

MAAP #244: Amazon Deforestation & Fire Hotspots 2025

Posted on by Matt Finer
Base Map. Deforestation and fire hotspots across the Amazon in 2025. Data: UMD/GLAD, Amazon Conservation/MAAP

Continuing our annual series, we present a detailed look at the major 2025 Amazon forest loss hotspots and trends, based on the annual data recently released by the University of Maryland and featured on Global Forest Watch. As in other reports of the series, we take this global dataset and analyze it specifically for the Amazon.

This dataset, which serves as a consistent source across all nine countries of the Amazon, distinguishes forest loss from fire and non-fire causes. We use the non-fire forest loss as a proxy for human-caused deforestation, although it also includes some natural loss. In addition, we apply a filter to focus just on primary forest loss.

With this context, we are able to identify the primary forest loss hotspots from fire and non-fire (deforestation proxy) causes across the Amazon in 2025 (see Base Map). 

The non-fire (proxy for deforestation) hotspots were largely due to agriculture and gold mining across the Amazon. These hotspots were concentrated in the:

  • Soy frontiers of southeast Brazil (Area A; see MAAP #161) and southern Bolivia (Area B; MAAP #179),
    .
  • Along major roads in Brazil, such as the Trans-Amazonian Highway (Area C) and BR-364 (Area D). There is also agricultural expansion along an expansive road network in northern Brazil (Area K).
    .
  • Agricultural areas in central Peru (Area E), including lands occupied by Mennonite colonies (MAAP #222),
    .
  • Arc of deforestation in northwest Colombia (Area F) associated with roads, land grabbing (and associated cattle pastures), and coca cultivation (MAAP #224),
    .
  • Gold mining areas in southern and central Peru (Area G; MAAP #233, MAAP #241), northern Ecuador (Area H)  (MAAP #230, MAAP #227, MAAP #219), northeast Amazon (Venezuela, Guyana, Suriname – for example, Area I), and Indigenous territories in Brazil (for example, Area J; MAAP #239).

 

The fire hotspots were concentrated in the soy and cattle frontiers of the southeast Brazilian Amazon and southeast Bolivian Amazon (including the important ecosystem of the Chiquitano dry forests), and also northeast Bolivia. This fire data may be interpreted as forest degradation, in contrast to the more permanent impacts of deforestation.

Amazon Primary Forest Loss, 2002-2025

In 2025, the major story was that fires were down from the record-breaking year of 2024 (Graph 1). Fires were still historically high, however (1.5 million hectares), marking the 3rd-highest since 2002 (behind only the peak fire seasons of 2016 and 2024).

Non-fire forest loss was also down from 2024 (Graph 1). While still just above 1 million hectares, it was the lowest total over the past 10 years and the 5th-lowest since 2002.

Cumulatively, we estimate the non-fire forest loss of 34.8 million hectares of primary forest since 2002, about the size of Germany or the U.S. state of Montana. An additional 12.2 million hectares have been impacted by fires.

Note that Graph 1 is interactive: The reader can click items in the legend (Non-fire and Fire-caused forest loss), and on the circles for each year to visualize the data point.

Amazon Primary Forest Loss, 2025

In 2025, the majority of non-fire primary forest loss occurred in Brazil (55%), followed by Bolivia (20%), Peru (14%), and Colombia (6%) as the clear top four (Graph 2a; Annex 1).

Notably, Brazil had the lowest annual loss on record since 2002, at around 560,475 hectares. 

Bolivia‘s non-fire primary forest loss (200,000 ha) was still historically high (4th highest on record), but lower than the previous peak three years of 2022-24.

Peru’s non-fire primary forest loss was the 5th highest on record (147,480 ha), and the highest over the past 5 years.

Colombia’s non-fire primary forest loss (66,310 ha) was the second lowest since the FARC peace agreement in 2016.

The vast majority (97%) of fire-caused primary forest loss occurred in just two countries: Brazil and Bolivia. Peru added 2% (26,580 ha). All three countries’ fire impact was much lower than last year’s record-breaking fire season.

Note that Graph 2a is interactive: The reader can click the bars for each country for non-fire forest loss (purple bars) and fire forest loss (orange bars). To see the data for the countries with less forest loss, click on the “Log” option in the upper right (or see Annex 1, further below).

Amazon Primary Forest Loss Rate, 2025

Standardizing for area, we show that Bolivia has the highest non-fire primary forest loss rate, followed by Peru, Colombia, then Brazil (Graph 2b).

Bolivia also has, by far, the highest fire-caused primary forest loss rate, followed by Brazil, and more distantly, Peru.

Note that Graph 2b is interactive: The reader can click the bars for each country for non-fire forest loss (purple bars) and fire forest loss (orange bars).

Amazon Deforestation 2025

In a novel analysis, we directly estimate Amazonian deforestation for the first time. As noted above, the primary forest loss data described above is a good proxy for deforestation, but also includes loss associated with natural events, such as landslides, windstorms, and meandering rivers.

Using the “WRI Google Drivers of Tree Cover Loss” dataset, we estimate the primary forest loss directly caused by agriculture, mining, and infrastructure. That is, directly estimate human-caused deforestation.

In 2025, we estimate the deforestation of 736,484 hectares across the Amazon (Graph 3). The vast majority (94.6%) of this deforestation came from agriculture (both permanent and shifting). An additional 5.3% came from hard commodities, mostly gold mining. The remaining 0.1% was caused by roads and infrastructure.

Over half (55.2%) of this deforestation occurred in Brazil, followed by Peru (16.8%), Bolivia (13.8%), and Colombia (8.5%).

Peru had the most mining deforestation, followed by Brazil, Guyana, Suriname, and Venezuela. However, we note that Amazon Mining Watch indicates that Brazil had higher mining deforestation than Peru in 2025.

Note that Graph 3 is interactive: The reader can click on the bars for each country. To see the data for the countries with less forest loss, click on the “Log” option in the upper right

Amazon Deforestation 2025 in Protected Areas & Indigenous Territories

Of the 2025 Amazon deforestation noted above, nearly 132,000 hectares (18%) occurred in protected areas and Indigenous territories (Graph 4). This may be considered a general estimate for illegal deforestation.

Agriculture accounted for 89% of this deforestation, and mining for the remaining 11%.

Brazil had the most deforestation in protected areas and Indigenous territories (33%), followed by Bolivia (25%), Peru (20%), Colombia (10%), Venezuela (6%), and Ecuador (4%).

Specifically for gold mining, Brazil had the most deforestation in protected areas and Indigenous territories, followed by Peru and Venezuela.

Note that Graph 4 is interactive: The reader can click items in the legend (Agriculture and Mining, by designation), and on the bars to visualize the data for each country.

Annex 1

Note that Annex 1 is interactive: The reader can click on the countries in the legend, and on the circles for each year to visualize the data point. To see the data for the countries with less forest loss, click on the “Log” option in the upper right.

Policy Implications

Following the record-breaking fire season of 2024, fire impact in 2025 was still historically high (3rd highest on record) but much reduced from the previous year’s peak. As detailed in MAAP #229, the 2024 fire season was associated with a strong El Niño event, creating extremely dry conditions across the Amazon. In contrast, 2025 was associated with the moister conditions of La Niña. This correlation has major implications for the predicted upcoming super El Niño season and will be the subject of an upcoming report.

Instead of fires, the major story in 2025 was relatively positive: the lowest non-fire primary forest loss over the past 10 years, and the 5th lowest on record.

However, in 2025 an additional 1 million hectares of primary forest was lost, bringing the cumulative total lost to 34.8 million hectares since 2002, the size of Germany or Montana.

As in previous years, the countries with the highest primary forest loss were Brazil, Bolivia, Peru, and Colombia, respectively. 

Notably, Brazil had the lowest annual loss on record since 2002, and Colombia was the second lowest since the FARC peace agreement in 2016. In contrast, Bolivia and Peru were both relatively high, but with different trends: Bolivia was lower than the previous peak years, while Peru was the highest over the past 5 years.

Standardizing for area, Bolivia had the highest primary forest loss rate, followed by Peru, Colombia, and then Brazil.

In terms of spatial patterns, non-fire primary forest loss hotspots were detected in all countries. Major agricultural deforestation areas occurred in southeast Brazil, southern Bolivia, central Peru, and northwest Colombia. Major mining areas were detected in southern and central Peru, northern Ecuador, the northeast Amazon (Venezuela, Guyana, Suriname), and Indigenous territories in Brazil.

Finally, in a novel analysis, we directly estimate Amazonian deforestation for the first time using a new dataset from WRI and Google. In 2025, we estimate the deforestation of 736,484 hectares across the Amazon. The vast majority (94.6%) of this deforestation came from agriculture (both permanent and shifting). An additional 5.3% came from hard commodities, mostly gold mining. The remaining 0.1% was caused by roads and infrastructure.

Over half (55.2%) of this deforestation occurred in Brazil, followed by Peru, Bolivia, and Colombia. Peru had the most mining deforestation, followed by Brazil, Guyana, Suriname, and Venezuela.

While agriculture accounts for the greatest impact in terms of total number of hectares deforested, much of this impact occurs in expanding deforestation zones and along roads. Key examples include expanding deforestation along the major roads of the eastern and southern Brazilian Amazon, expanding soy deforestation in the southern Bolivian Amazon, expanding deforestation by Mennonite colonies in the central Peruvian Amazon, and the arc of deforestation in the northwest Colombian Amazon. 

Gold mining, on the other hand, has the greatest impact in terms of targeting sensitive areas. In contrast to agricultural deforestation following roads, gold mining, particularly illegal gold mining, often targets the most remote and intact areas, such as protected areas and Indigenous territories. Key examples include the southern Peruvian Amazon, northern Ecuadorian Amazon, border between the Colombian and Brazilian Amazon, Indigenous territories of the Brazilian Amazon, and the northeast Amazon (Venezuela, Suriname, and Guyana).

It is important to note that the data presented here may differ from national data presented by governments. This difference may be due to methodology (we focus on impact on primary forests), spatial resolution (30 meters in our case), and Amazon boundaries (we employ a hybrid boundary designed for maximum inclusion of both watershed and biogeography). Due to these potential differences among sources, it is best to focus on the convergence of overall trends and patterns, and not overly focus on the absolute numerical difference.

Methodology

The analysis was based on 30-meter resolution annual forest loss data produced by the University of Maryland and also presented by Global Forest Watch.

This data was complemented with the Global Forest Loss due to fire dataset that is unique in terms of being consistent across the Amazon (in contrast to country specific estimates) and distinguishes forest loss caused directly by fire (note that virtually all Amazon fires are human-caused). The values included were ‘medium’ and ‘high’ confidence levels (code 3-4). This fire data may be interpreted as forest degradation, in contrast to the more permanent impacts of deforestation.

The remaining forest loss serves as a likely close proxy for deforestation, with the only remaining exception being natural events such as landslides, wind storms, and meandering rivers. The values used to estimate this category were ‘low’ certainty of forest loss due to fire (code 2), and forest loss due to other ‘non-fire’ drivers (code 1).

For the baseline, it was defined to establish areas with >30% tree canopy density in 2000. Importantly, we applied a filter to calculate only primary forest loss by intersecting the forest cover loss data with the additional dataset “primary humid tropical forests” as of 2001 (Turubanova et al 2018). For more details on this part of the methodology, see the Technical Blog from Global Forest Watch (Goldman and Weisse 2019).

Our geographic range for the Amazon is a hybrid designed for maximum inclusion: biogeographic boundary (as defined by RAISG) for all countries, except for Bolivia and Peru, where we use the watershed boundary, and Brazil, where we use the Legal Amazon boundary.

Protected areas and Indigenous territory data from RAISG and official sources. In case of an overlap, data was included in the protected areas category. Note that Suriname does not have titled Indigenous territories.

To identify the deforestation hotspots, we conducted a kernel density estimate. This type of analysis calculates the magnitude per unit area of a particular phenomenon, in this case, forest cover loss. We conducted this analysis using the Kernel Density tool from the Spatial Analyst Tool Box of ArcGIS. We used the following parameters:

Search Radius: 15000 layer units (meters)
Kernel Density Function: Quartic kernel function
Cell Size in the map: 50 x 50 meters (0.25 hectares)
Everything else was left to the default setting.

For the Base Map, we used the following concentration percentages: High: 3-14%; Very High: >14%. These percentages correspond to the concentration of forest loss pixels, with a pixel size of 50 x 50 meters (0.25 hectares).

Using the “WRI Google Drivers of Tree Cover Loss” dataset, we then estimated human-caused deforestation. The main challenge was analyzing this 1 km resolution dataset in relation to the 30 m resolution annual forest loss dataset described above.

Building upon the annual forest loss by confidence level )—from which fire-related loss was excluded based on confidence levels 3 and 4—a second layer, designated “Forest Loss Non-Fire” (confidence levels 1 and 2), was generated; onto this layer, the cumulative “Drivers” layer (2001–2025) was overlaid to analyze which underlying causes were associated with the recorded loss.

The result was an artificial scale 30 m resolution; it should be noted here that the spatial correlation is not exact. Since Driver’s original data has a resolution of 1 km—which encompasses multiple 30-meter pixels—this value has been replicated (downscaled). Through layer merging, a layer was obtained containing forest loss pixel values ​​accompanied by a confidence level and an assigned driver—meaning the probable primary cause of the loss has been identified.

To estimate human-caused deforestation, we focused on just four of the drivers: agriculture (both permanent and shifting), hard commodities, and roads & infrastructure. In other words, we did not include Natural forest loss, Wildfires, or Logging.

Acknowledgements

We thank colleagues from the following organizations for helpful comments on the report: Conservación Amazónica – ACEAA in Bolivia, Conservación Amazónica – ACCA in Peru, and  Fundación EcoCiencia in Ecuador.

This work was supported by Norad (Norwegian Agency for Development Cooperation).

 

Citation

Finer M, Ariñez A, Bodin B, Santana A (2026) Amazon Deforestation & Fire Hotspots 2025. MAAP: 244.

MAAP #229: Amazon Deforestation & Fire Hotspots 2024

Posted on by Matt Finer
Base Map. Deforestation and fire hotspots across the Amazon in 2024. Data: UMD/GLAD, Amazon Conservation/MAAP.

Continuing our annual series, we present a detailed look at the major 2024 Amazon forest loss hotspots, based on the final annual data recently released by the University of Maryland and featured on Global Forest Watch. As in other reports of the series, we take this global dataset and analyze it for the Amazon specifically.

This forest loss dataset serves as a consistent source across all nine countries of the Amazon, distinguishing forest loss from fire and non-fire causes. We use the non-fire forest loss as a proxy for human-caused deforestation, although it also includes some natural loss (such as landslides and windstorms). Previous research has confirmed that nearly all Amazon fires are human-caused (MAAP #189).

With this context, we are able to present both estimated deforestation and fire hotspots across the Amazon (see Base Map and Graph 1). 

In 2024, the deforestation was the fifth highest on record (since 2002), at over 1.7 million hectares (4.3 million acres) across the Amazon. This value represents a major increase (34%) from 2023, but a decrease (12%) from the recent peak in 2022 (1.98 million hectares). The majority of the deforestation occurred in Brazil (54.7%), followed by Bolivia (27.3%), Peru (8.1%), and Colombia (4.7%) as the clear top four in 2024.

The big story in 2024, however, was the record-breaking impact of fires on primary forests, with a total of 2.8 million hectares (6.9 million acres). This total shattered the previous record of 1.7 million hectares in 2016. The vast majority (95%) of this fire impact occurred in just two countries: Brazil and Bolivia, which both set annual fire records of their own (along with Peru, Guyana, Suriname, and French Guiana). Overall, this fire data may be interpreted as forest degradation, in contrast to the more permanent impacts of deforestation.

In terms of spatial patterns, the Base Map indicates that most of the intense forest loss hotspots were due to fire. These fire hotspots were especially concentrated in the soy and cattle frontiers of the southeast Brazilian Amazon and southeast Bolivian Amazon. The deforestation hotspots (without major associated fires) were largely due to agriculture and gold mining across the Amazon, notably in Bolivia, Brazil, Colombia, Ecuador, and Peru. See Annex 1 for the overall forest loss hotspots (without the specific fire loss data).

Previous research has revealed the tight link between deforestation and fires in the Amazon (MAAP #189). That is, many major fires are burning recently deforested areas, and sometimes escape into surrounding forests, especially with extended dry conditions.

In total, 4.5 million hectares (11.2 million acres) of primary forest were impacted by deforestation and fire combined. This total is the highest on record by far, surpassing 2016 (3.4 million hectares) by over a million hectares.

Since 2002, we estimate the deforestation of 33.7 million hectares (83.4 million acres) of primary forest, greater than the size of the state of New Mexico. An additional 10.6 million hectares (26.2 million acres) have been impacted by fires.

Below, we zoom in on the four countries with the highest deforestation (Brazil, Bolivia, Peru, and Colombia), plus additional highlights from around the Amazon (Guyana, Venezuela, and Ecuador).

Amazon Primary Forest Loss, 2002-2024

Graph 1 shows the historical trend of Amazonian primary forest loss, from 2022 to present.

Graph 1. Amazon Forest Loss, 2002-24. Data: UMD/GLAD, Amazon Conservation/MAAP.

Amazon Primary Forest Loss (By Country), 2002-2024

Graph 2a shows 2024 Amazon primary forest loss for all nine countries. In Annex 2, Graph 2b removes Brazil and Bolivia to see the other countries in greater detail.

Graph 2a. Amazon primary forest loss for all nine countries. Data: UMD

Brazilian Amazon

Figure 2. Deforestation and fire hotspots in the Brazilian Amazon. Data: UMD.

In 2024, the Brazilian Amazon lost 954,126 hectares (2.4 million acres) of primary forest to deforestation. Although this total marked a 13.6% increase from 2023, it was historically relatively low (16th highest overall since 2002).

The bigger story is that fires directly impacted an additional 1.9 million hectares (4.6 million acres). This fire impact was the highest on record, surpassing the previous high of 2016 (1.6 million hectares).

All of the most intense forest loss hotspots were characterized by intense fires. Many of these hotspots were concentrated in the southeast Brazilian Amazon (Figure 2). These areas include along the major north-south road in the state of Pará (BR-163), and further to the east of this road. The hotpots also expanded south into the soy frontier of Mato Grosso state.

Fire hotspots were also located in the northern state Roraima, and along the other major road networks, especially road BR-230 (Trans-Amazonian Highway) in the states of Pará and Amazonas, and road BR-364 in the state of Acre.

Previous research has revealed that over 70% of major fires in the Brazilian Amazon are burning recently deforested areas (MAAP #189). In extended dry conditions, such as 2016 and 2024, these major fires escape into surrounding forests.

 

 

Graph 3. Deforestation and fire trends in the Brazilian Amazon. Data: UMD.

Bolivian Amazon

Figure 3. Deforestation and fire hotspots in the Bolivian Amazon. Data: UMD.

In 2024, the Bolivian Amazon lost 476,030 hectares (1.2 million acres) of primary forest to deforestation. This total was the highest on record, surpassing the previous high of 2022 (245,177 hectares).

In an even bigger shattering record, fires directly impacted an additional 779,960 hectares (1.9 million acres). This total crushed the previous record of 2023 (250,843 hectares).

Like Brazil, the most intense forest loss hotspots were characterized by intense fires.

These fires were concentrated in the soy frontier located in the southeastern department of Santa Cruz (Figure 3). We emphasize that this particular hotspot is further north than in previous years, indicating a northern expansion of soy plantations.

There was also a concentration of fire hotspots along the Beni and Pando departments’ border, and closer to the Andes Mountains in the departments of La Paz and Beni.

Deforestation hotspots (without fires) were concentrated in the soy frontier in the southeast.

 

 

 

 

Graph 4. Deforestation and fire trends in the Bolivian Amazon. Data: UMD.

Peruvian Amazon

Figure 4. Deforestation and fire hotspots in the Peruvian Amazon. Data: UMD.

In 2024, the Peruvian Amazon lost 141,781 hectares (350,341 acres) of primary forest to deforestation. This total marks the 6th highest on record since 2002.

Breaking a record, fires impacted an additional 47,574 hectares (117,554 acres). This total more than doubled the previous high of 2023 (20,042 hectares). As noted above, this fire data may be interpreted as forest degradation, in contrast to the more permanent impacts of deforestation.

Like both Brazil and Bolivia, all of the most intense forest loss hotspots were characterized by intense fires. These fires were concentrated in the central and southeast Amazon (Ucayali and Madre de Dios regions, respectively) (Figure 4).

Deforestation hotspots were concentrated in the gold mining frontier in the southern Amazon and throughout the central Amazon. The very high hotspot in central Peru corresponds to the latest deforestation by Mennonite colonies (see MAAP #222 for context).

 

 

 

 

 

 

 

Graph 5. Deforestation and fire trends in the Peruvian Amazon. Data: UMD

Colombian Amazon

Figure 5. Deforestation and fire hotspots in the Colombian Amazon. Data: UMD.

In 2024, the Colombian Amazon lost 81,396 hectares (201,129 acres) of primary forest to deforestation. This total marked a striking 82.5% increase from the recent low recorded in 2023. It was the 7th highest on record, continuing the trend of elevated forest loss since the FARC peace agreement in 2016 (all top seven deforestation annual totals have occurred since 2016).

Major fires were less of an issue in the Colombian Amazon, but did directly impact an additional 5,184 hectares (8th highest on record).

As described in previous reports (see MAAP #120), Figure 5 shows that there continues to be an “arc of deforestation” in the northwest Colombian Amazon (Caqueta, Meta, Putumayo and Guaviare departments). Most notably, there is a pair of very high deforestation areas surrounding Chiribiquete National Park, and high deforestation areas within Tinigua and Macarena National Parks.

This arc impacts numerous Protected Areas (particularly Tinigua and Chiribiquete National Parks) and Indigenous Reserves (particularly Yari-Yaguara II and Nukak Maku).

 

 

 

 

Graph 6. Deforestation and fire trends in the Colombian Amazon. Data: UMD.

Rest of the Amazon

Other important highlights from around the Amazon include:

Deforestation and fire hotspots in northeast Guyana. In total, Guyana lost 25,858 hectares of primary forest to deforestation, and an additional 38,314 hectares to fires, both of which shattered previous records.

Deforestation in the Venezuelan Amazon was the highest on record (32,240 hectares), and additional fire impacts were the second highest (36,471 hectares).

Deforestation in the Ecuadorian Amazon was the second highest on record (18,615 hectares), just behind 2022 (18,902 hectares). Deforestation hotspots were concentrated in the northern Amazon, areas characterized by high gold mining activity (MAAP #227, MAAP #221, MAAP #219). As in Colombia, major fires were less of an issue in the northwest Amazon overall, but did directly impact an additional 1,540 hectares (4th highest on record).

Fires were the highest on record in Suriname (7,926 ha) and French Guiana (635 ha).

Policy Implications

The dominant policy development of 2024 was the record-breaking fire season across the Amazon. These fire records were not only region-wide but also country-specific, occurring in Brazil, Bolivia, Peru, Guyana, Suriname, and French Guiana.

The 2024 records are particularly significant given that the Amazon has experienced several intense fire years over the past two decades. The most notable, and previous record-breaker, occurred in 2016, following the “Godzilla” El Niño event of 2015-16. However, the extreme drought conditions of 2023 and 2024, also associated with El Niño,  exceeded those past benchmarks, creating extreme conditions for widespread fires across the Amazon.

As a result, fire policy is now emerging as a central pillar of Amazon conservation, alongside longstanding efforts to curb deforestation. This growing prominence is directly related to climate change, both in terms of intensifying dry seasons and the projected increase in the frequency, length, and severity of El Niño events.

These policies must be centered on how to avoid fires in the first place, and then how to effectively respond to major fires once they appear.

Previous research has revealed the tight link between deforestation and fires in the Brazilian Amazon (MAAP #189). That is, many major fires are burning recently deforested areas, and sometimes escape into surrounding forests, especially with extended dry conditions. Therefore, strengthening deforestation control remains one of the most effective fire mitigation strategies in Brazil and other Amazonian countries.

There is also a strong link between deforestation and fire in the Bolivian Amazon. While deforestation often precedes fires, as in Brazil, there is also a second round of deforestation following the fires

In both contexts, real-time fire monitoring, such as the MAAP Fire Tracker,  should be integrated into national response protocols and field-level coordination. 

Beyond fire, high rates of forest loss continue to be driven by deforestation across other parts of the Amazon. In Colombia’s arc of deforestation, we detected very high deforestation around Chiribiquete National Park, as well as high deforestation within Tinigua and Macarena National Parks. Although the national government is engaged on the issue of deforestation, these losses are closely tied to the presence and influence of armed groups in the country, which exert substantial control over land-use and deforestation dynamics (explaining shifts such as the dip in 2023 and the rise again in 2024). The major deforestation drivers in Colombia are roads, land grabbing (and associated cattle pastures), and coca cultivation.

Other high forest loss areas include gold mining fronts in northern Ecuador and southern Peru, Mennonite colonies in central Peru, the soy frontier in southeast Bolivia, and along major existing roads in Brazil. Although there is gold mining in both Venezuela and Guyana, the most intense forest loss hotspots were associated with fires surrounding agricultural areas.

It is important to note that the data presented here may differ from national data presented by governments. This difference may be due to methodology (we focus on impact on primary forests), spatial resolution (30 meters in our case), and Amazon boundaries (we employ a hybrid boundary designed for maximum inclusion of both watershed and biogeography). Due to these potential differences among sources, it is best to focus on the convergence of overall trends and patterns, and not overly focus on absolute numerical differences. 

Annex 1

Annex 1. Forest loss hotspots in relation to fire specific hotspots across the Amazon in 2024. Data: UMD/GLAD, ACA/MAAP.

 

Annex 2

Graph 2b. Amazon primary forest loss for seven countries (except Brazil and Bolivia). Data: UMD

Methodology

The analysis was based on 30-meter resolution annual forest loss data produced by the University of Maryland and also presented by Global Forest Watch.

This data was complemented with the Global Forest Loss due to fire dataset that is unique in terms of being consistent across the Amazon (in contrast to country specific estimates) and distinguishes forest loss caused directly by fire (note that virtually all Amazon fires are human-caused). The values included were ‘medium’ and ‘high’ confidence levels (code 3-4). This fire data may be interpreted as forest degradation, in contrast to the more permanent impacts of deforestation.

The remaining forest loss serves as a likely close proxy for deforestation, with the only remaining exception being natural events such as landslides, wind storms, and meandering rivers. The values used to estimate this category was ‘low’ certainty of forest loss due to fire (code 2), and forest loss due to other ‘non-fire’ drivers (code 1).

For the baseline, it was defined to establish areas with >30% tree canopy density in 2000. Importantly, we applied a filter to calculate only primary forest loss by intersecting the forest cover loss data with the additional dataset “primary humid tropical forests” as of 2001 (Turubanova et al 2018). For more details on this part of the methodology, see the Technical Blog from Global Forest Watch (Goldman and Weisse 2019).

Our geographic range for the Amazon is a hybrid designed for maximum inclusion: biogeographic boundary (as defined by RAISG) for all countries, except for Bolivia and Peru, where we use the watershed boundary, and Brazil, where we use the Legal Amazon boundary.

To identify the deforestation hotspots, we conducted a kernel density estimate. This type of analysis calculates the magnitude per unit area of a particular phenomenon, in this case, forest cover loss. We conducted this analysis using the Kernel Density tool from the Spatial Analyst Tool Box of ArcGIS. We used the following parameters:

Search Radius: 15000 layer units (meters)
Kernel Density Function: Quartic kernel function
Cell Size in the map: 50 x 50 meters (0.25 hectares)
Everything else was left to the default setting.

For the Base Map, we used the following concentration percentages: High: 3-14%; Very High: >14%. These percentages correspond to the concentration of forest loss pixels, with a pixel size of 50 x 50 meters (0.25 hectares).

Acknowledgements

We thank colleagues at Global Forest Watch (GFW), an initiative of the World Resources Institute (WRI) for early access to data.

We also thank colleagues from the following organizations for helpful comments on the report: Conservación Amazónica – ACEAA in Bolivia, Conservación Amazónica – ACCA in Peru, Fundación EcoCiencia in Ecuador, and Instituto Centro de Vida (ICV) in Brazil.

This work was supported by Norad (Norwegian Agency for Development Cooperation).

Citation

Finer M, Ariñez A, Mamani N, Cohen M, Santana A (2025) Amazon Deforestation & Fire Hotspots 2024. MAAP: 229.

MAAP #189: Amazon Fire Season Heats Up

Posted on by Ana Folhadella
Image 1. Example of 2023 (June 29) major fire in Brazilian Amazon.

The Amazon fire season is well under way: to date, we have detected over 260 major fires thus far in 2023 (see Base Map below).

This year is of special concern because scientists indicate we have entered a new El Niño episode. The most intense Amazon fire seasons on record, 2016 and 2017, immediately followed the last major El Niño event.

Most of the fires (54%) this year have occurred in the Brazilian Amazon.

Of these, the vast majority (73%) have burned r­­­­ecently deforested areas. This high number is consistent with previous years (see MAAP #168) and once again highlights the critical link between deforestation and fires in the Brazilian Amazon. That is, most major fires are burning the remnants of a recent deforestation event.

It is also worth noting that many of the fires in the Brazilian Amazon (42%) were burning areas recently deforested specifically for new soy plantations.

We have thus far detected 40 major fires in the Bolivian Amazon. The vast majority (88%) have been burning areas recently deforested specifically for new soy plantations.

We have detected an additional 30 major fires in the Peruvian Amazon, mostly burning high elevation grasslands.

Earlier in the year, between January and March, we detected 50 major fires in the Colombian Amazon. Notably, 100% of them were in burning recently deforested areas.

These findings are based on the unique data from the real-time Amazon Fires Monitoring app developed by our partner organization in Peru, Conservación Amazónica ACCA. In a novel approach, the app combines data from the atmosphere (aerosol emissions in smoke) and the ground (heat anomaly alerts) to quickly and precisely detect major fires, defined as fires burning abundant biomass. In short, the app filters out smaller fires (such as routine burning an old field) and highlights major fires (such as burning recently deforested areas, standing forest, or natural grasslands).

2023 Major Amazon Fires Base Map

Base Map. 2023 major Amazon fires (through July 2023). Data: ACCA, ACA/MAAP.

Amazon Fires Dashboard

We also present our new Amazon fires dashboard, which currently shows results for the 2022 fire season. The dashboard highlights a number of the key findings from last year:

  • We detected 983 major fires.
  • The vast majority (72%) were in Brazil, followed by Bolivia, Peru, and Colombia.
  • Importantly, 73% of the major fires burned recently deforested areas, followed by grasslands, forest fires, and pasture.

The dashboard was developed by the SAS Institute’s Data for Good Program.

Methodology

The reported results are based on an analysis of data generated by a unique real-time Amazon Fires Monitoring app during the year 2023, through July 13.

The app, hosted by Google Earth Engine, was developed and updated daily by the Peru-based organization Conservación Amazónica (ACCA). The resulting data was analyzed and recorded daily by the US-based organization Amazon Conservation. The app was created in 2019 and upgraded in 2020, with the current version launching in May 2021.

When fires burn, they emit gases and aerosols (aerosol definition: Suspension of fine solid particles or liquid droplets in air or another gas) as part of the outgoing smoke. A relatively new satellite (Sentinel-5P from the European Space Agency) detects these aerosol emissions.

The aerosol data, which has a spatial resolution of 7.5 sq km, is not impacted by cloud cover, thus enabling near real-time monitoring during all weather conditions. The app is typically updated each day in the late afternoon/early evening with data for that same day. Thus, there is a high potential for authorities and civil society to also use this app to respond to major fires in the field.

Importantly, the app distinguishes small fires (such as from clearing old fields and thus burning little biomass) from larger fires (such as burning recently deforested areas or standing forests and thus burning high amounts of biomass).

We define a “major fire” as one showing elevated aerosol emission levels on the app, thus indicating the burning of elevated levels of biomass. This typically translates to an aerosol index (AI) of >1 (or cyan-green to red on the app).

In a novel approach, the app combines this aerosol data from the atmosphere with heat anomaly data from the ground.

For all detected major fires, we cross-referenced the aerosol emissions pattern with the ground heat-based data to pinpoint the exact location of the fire source. Typically for major fires, there is a large cluster of heat-anomaly alerts aiding the process.

In a final step, the detected major fires are then analyzed with high-resolution optical satellite imagery from Planet Explorer. With this imagery, we can confirm the major fire (by observing smoke on the day of the fire or a burned area scar in the days following the fire) and estimate its size.

Moreover, with Planet’s extensive satellite imagery archive, we can determine the fire type. That is, by comparing imagery from the fire date to previous dates, we can determine whether the fire was burning a) a recently deforested area (defined as fires in areas recently deforested during the past three years), b) an older deforested area (typically long-standing pasture areas), c) standing forest (that is, a forest fire), or natural savannah.

In the app, we can also cross-reference if a major fire has occurred within a protected area or titled indigenous territory.

Note that the high values in the aerosol indices may also be due to other reasons such as emissions of volcanic ash or desert dust so it is important to cross-reference elevated emissions with heat data and optical imagery.

Acknowledgements

This work was supported by Norad (Norwegian Agency for Development Cooperation) and ICFC (International Conservation Fund of Canada).

Citation

Finer M, Costa H, Villa L (2023) Amazon Fire Season Heats Up. MAAP: 189.

MAAP #187: Amazon Deforestation & Fire Hotspots 2022

Posted on by Ana Folhadella
2022 Amazon Forest Loss Base Map. Deforestation and fire hotspots across the full Amazon biome. Data: UMD/GLAD, ACA/MAAP.

We present a detailed look at the major 2022 Amazon forest loss hotspots, based on the final annual data recently released by the University of Maryland (and featured on Global Forest Watch).

This dataset is unique in that it is consistent across all nine countries of the Amazon, and distinguishes forest loss from fire, leaving the rest as a proxy for deforestation (but also includes natural loss).

Thus, we are able to present both deforestation and fire hotspots across the Amazon.

The Base Map (see right) and Results Graph (see below) reveal several key findings:

  • In 2022, we estimate the deforestation of 1.98 million hectares (4.89 million acres). This represents a major 21% increase from 2021, and is the second highest on record, behind only the peak in 2004.
    k
  • Deforestation hotspots were especially concentrated along roads in the Brazilian Amazon, the soy frontier in the southeast Bolivian Amazon, and near protected areas in northwest Colombian Amazon.
    k
  • The vast majority of the deforestation occurred in Brazil (72.8%), followed by Bolivia (12.4%), Peru (7.3%), and Colombia (4.9%). Note that deforestation in Bolivia was the highest on record, and in Brazil the highest since the early 2000s.
    k
  • Fires impacted an additional 491,223 hectares (1.2 million acres) of primary forest. This total represents a 1.6% increase from 2021, and the 4th highest on record (behind only intense fire seasons of 2016, 2017, and 2020). Moreover, each of the seven most intense fire seasons has occurred in the past seven years. Nearly 93% of the fire impact occurred in just two countries: Brazil and Bolivia.
    k
  • In total, 2.47 million hectares (6.1 million acres) of primary forest were impacted by deforestation and fire. This total represents the third highest on record, only behind the post-El Niño years of 2016 and 2017.
    k
  • Since 2002, we estimate the deforestation of 30.7 million hectares (75.9 million acres) of primary forest, greater than the size of Italy or the U.S. state of Arizona.

Below, we zoom in on the six countries with the highest deforestation (Brazil, Bolivia, Peru, Colombia, Ecuador, and Venezuela) with additional maps and analysis.

Amazon Primary Forest Loss (Combined), 2002-2022

Amazon Forest Loss Results Graph, 2002-22. Data: UMD/GLAD, ACA/MAAP.

Amazon Primary Forest Loss (By Country), 2002-2022

Brazilian Amazon

Brazil Base Map, 2022. Deforestation and fire hotspots in the Brazilian Amazon in relation to major roads. Data: UMD/GLAD, ACA/MAAP.

In 2022, the Brazilian Amazon lost 1.4 million hectares (3.56 million acres) of primary forest to deforestation. Fires directly impacted an additional 348,824 hectares.

The deforestation rose 20.5% from 2021, and was the highest on record since the peak years of 2002 – 2005.

The fire impact was the 4th highest on record, only behind the intense fire years of 2016, 2017, and 2020.

The deforestation was concentrated along the major road networks, especially roads 230 (Trans-Amazonian Highway), 364, 319, and 163 in the states of Amazonas, Pará, Rondônia, and Acre (see Brazil Base Map).

The direct fire impacts were concentrated in the soy frontier, located in southeastern state of Mato Grosso

 

 

 

 

 

 

Bolivian Amazon

Bolivia Base Map, 2022. Deforestation and fire hotspots in Bolivian Amazon. Data: UMD/GLAD, ACA/MAAP.

In 2022, the Bolivian Amazon lost 245,177 hectares of primary forest to deforestation. Fires directly impacted an additional 106,922 hectares.

We highlight that this deforestation was 47% higher than 2021, and the highest on record (by far).

The fire impact was also up from last year, and the second-highest on record behind just the intense year of 2020.

Both the deforestation and fires were concentrated in the soy frontier located in southeastern department of Santa Cruz (see Bolivia Base Map).

 

 

 

 

 

 

 

 

 

 

Peruvian Amazon

Peru Base Map, 2022. Deforestation and fire hotspots in the Peruvian Amazon. Data: UMD/GLAD, ACA/MAAP.

In 2022, the Peruvian Amazon lost 144,682 hectares of primary forest to deforestation. Fires directly impacted an additional 16,408 hectares.

Deforestation increased 6.7% from 2021, and was the 5th highest on record. Fire impact decreased from last year, but was still relatively high.

The deforestation was concentrated in the central and southern Amazon (Ucayali and Madre de Dios regions, respectively) (see Peru Base Map).

In the central Amazon, we highlight the rapid deforestation for a new Mennonite colony (see MAAP #166).

In the southern Amazon, gold mining deforestation continues to be an issue in indigenous communities and within the official Mining Corridor.

 

 

 

 

 

 

 

Colombian Amazon

Colombia Base Map, 2022. Deforestation and fire hotspots in northwest Colombian Amazon. Data: UMD/GLAD, ACA/MAAP, FCDS.

In 2022, the Colombian Amazon lost 97,417 hectares of primary forest to deforestation. Fires directly impacted an additional 12,880 hectares.

Deforestation decreased 2% from 2021, but it was still relatively high (5th highest on record), continuing the trend of elevated forest loss since the FARC peace agreement in 2016.

Fire impact increased from last year and was actually the highest on record, edging out 2018 and 2019.

As described in previous reports (see MAAP #120), the Colombia Base Map shows there continues to be an “arc of deforestation” in the northwest Colombian Amazon (Caqueta, Meta, and Guaviare departments).

This arc impacts numerous Protected Areas (particularly Tinigua and Chiribiquete National Parks) and Indigenous Reserves (particularly Yari-Yaguara II and Nukak Maku).

 

 

 

 

Ecuadorian Amazon

Ecuador Base Map, 2022. Deforestation and fire hotspots in the Ecuadorian Amazon. Data: UMD/GLAD, ACA/MAAP.

Although accounting for just 1% of total loss across the Amazon, deforestation in the Ecuadorian Amazon was the highest on record in 2022 (18,902 hectares), up a striking 80% since 2021.

There are several deforestation hotspots caused by gold mining (see MAAP #182), oil palm plantation expansion, and small-scale agriculture.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Venezuelan Amazon

In the Venezuelan Amazon, deforestation was on par with last year (12,584 hectares).

There is a deforestation hotspot caused by gold mining in Yapacana National Park (see MAAP #173, MAAP #156, MAAP #169).

There are also hotspots in the Orinoco Mining Arc caused by mining and agriculture.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Methodology

The analysis was based on 30-meter resolution annual forest loss data produced by the University of Maryland and also presented by Global Forest Watch.

This data was complemented with the Global Forest Loss due to fire dataset that is unique in terms of being consistent across the Amazon (in contrast to country specific estimates) and distinguishes forest loss caused directly by fire (note that virtually all Amazon fires are human-caused). The values included were ‘medium’ and ‘high’ confidence levels (code 3-4).

The remaining forest loss serves as a likely close proxy for deforestation, with the only remaining exception being natural events such as landslides, wind storms, and meandering rivers. The values used to estimate this category was ‘low’ certainty of forest loss due to fire (code 2), and forest loss due to other ‘non-fire’ drivers (code 1).

For the baseline, it was defined to establish areas with >30% tree canopy density in 2000. Importantly, we applied a filter to calculate only primary forest loss by intersecting the forest cover loss data with the additional dataset “primary humid tropical forests” as of 2001 (Turubanova et al 2018). For more details on this part of the methodology, see the Technical Blog from Global Forest Watch (Goldman and Weisse 2019).

Our geographic range for the Amazon is a hybrid designed for maximum inclusion: biogeographic boundary (as defined by RAISG) for all countries, except for Bolivia and Peru, where we use the watershed boundary, and Brazil, where we use the Legal Amazon boundary.

To identify the deforestation hotspots, we conducted a kernel density estimate. This type of analysis calculates the magnitude per unit area of a particular phenomenon, in this case, forest cover loss. We conducted this analysis using the Kernel Density tool from the Spatial Analyst Tool Box of ArcGIS. We used the following parameters:

Search Radius: 15000 layer units (meters)
Kernel Density Function: Quartic kernel function
Cell Size in the map: 200 x 200 meters (4 hectares)
Everything else was left to the default setting.

For the Base Map, we used the following concentration percentages: High: 3-14%; Very High: >14%.

Acknowledgements

We thank colleagues at Global Forest Watch (GFW), an initiative of the World Resources Institute (WRI) for comments and access to data.

This work was supported by Norad (Norwegian Agency for Development Cooperation) and ICFC (International Conservation Fund of Canada).

Citation

Finer M, Mamani N (2023) Amazon Deforestation & Fire Hotspots 2022. MAAP: 187

MAAP #183: Protected Areas & Indigenous Territories Effective Against Deforestation Across Amazon

Posted on by Ana Folhadella
Base Map. Primary forest loss (2017-21) across the Amazon, in relation to protected areas and indigenous territories.

As deforestation continues to threaten primary forest across the Amazon, key land use designations are one of the best hopes for the long-term conservation of critical remaining intact forests.

Here, we evaluate the impact of two of the most important: protected areas & indigenous territories.

Our study looked across all nine countries of the Amazon biome, a vast area of 883.7 million hectares (see Base Map).

We calculated primary forest loss over the past 5 years (2017-2021).

For the first time, we were able to distinguish fire vs non-fire forest loss. For non-fire, while this does include natural events (such as landslides and wind storms), we consider this our best proxy for human-caused deforestation.

We analyzed the results across three major land use categories:

1) Protected Areas (national and state/department levels), which cover 197 million hectares (23.6% of Amazon).

2) Indigenous Territories (official), which cover 163.8 million hectares (19.6% of Amazon).

3) Other (all remaining areas outside protected areas and indigenous territories), which cover 473 million hectares (56.7% of Amazon).

In summary, we found that deforestation was the primary driver of forest loss, with fire always being a smaller subset. Averaged across all 5 years, protected areas and indigenous territories had similar levels of effectiveness, reducing primary forest loss rate by 3x compared to areas outside of these designations.

Below, we show the key results across the Amazon in greater detail, including a breakdown for the western Amazon (Bolivia, Colombia, Ecuador, and Peru) and the Brazilian Amazon.

Key Findings

Amazon Biome

We documented the loss of 11 million hectares of primary forests across all nine countries of the Amazon biome between 2017 and 2021. Of this total, 71% was non-fire (deforestation and natural) and 29% was fire.

For the major land use categories, 11% of the forest loss occurred in both protected areas and indigenous territories, respectively, while the remaining 78% occurred outside these designations.

To standardize these results for the varying area coverages, we calculated annual primary forest loss rates (loss/total area of each category). Figure 1 displays the results for these rates across all nine countries of the Amazon biome.

Figure 1. Primary forest loss rates across the Amazon, 2017-21.

Broken down by year, 2017 had the highest forest loss rates, with both a severe deforestation and fire season. In addition, 2021 had the second highest deforestation rate, while 2020 had the second highest fire loss rate.

Averaged across all five years, protected areas (green) had the lowest overall primary forest loss rate (0.12%), closely followed by indigenous territories (0.14%).

Interestingly, indigenous territories (orange) actually had a slightly lower deforestation rate compared to protected areas (0.7 vs 0.8%), but higher fire loss rate (o.7 vs .04%), resulting in the overall higher forest loss rate noted above.

Outside of these designations (red), the primary forest loss rate was triple (.36%), especially due to much higher deforestation.

Western Amazon

Breaking the results down specifically for the western Amazon (Bolivia, Colombia, Ecuador, and Peru), we documented the loss of 2.6 million hectares of primary forests between 2017 and 2021. Of this total, 80% was non-fire (deforestation and natural) and 20% was fire.

For the major land use categories, 9.6% occurred in protected areas, 15.6% in indigenous territories, and the remaining 74.8% occurred outside these designations.

Figure 2 displays the standardized primary forest loss rates across the western Amazon.

Figure 2. Primary forest loss rates across the Western Amazon, 2017-21.

Broken down by year, 2017 had the highest deforestation rate and overall forest loss rates. But 2020 had the highest fire loss rate, mainly due to extensive fires in Bolivia. 2021 also had a relatively high deforestation rate. Also, note the high level of fires in protected areas in 2020 and 2021, and indigenous territories in 2019.

Averaged across all five years, protected areas had the lowest overall primary forest loss rate (0.11%), followed by indigenous territories (0.16%).

Outside of these designations, the primary forest loss rate was .30%. That is, triple the protected areas rate and double the indigenous territories rate.

Brazilian Amazon

Breaking the results down specifically for the Brazilian Amazon, we documented the loss of 8.1 million hectares of primary forests between 2017 and 2021. Of this total, 68% was non-fire (deforestation and natural) and 32% was fire.

For the major land use categories, 9.4% occurred in indigenous territories, 11.2% occurred in protected areas, and the remaining 79.4% occurred outside these designations.

Figure 3 displays the standardized primary forest loss rates across the Brazilian Amazon.

Figure 3. Primary forest loss rates in the Brazilian Amazon, 2017-21.

Broken down by year, 2017 had the highest forest loss rate recorded in the entire study (.58%), due to both elevated deforestation and fire. Note that indigenous territories were particularly impacted by fire in 2017.

2020 had the next highest forest loss rate, also driven by an intense fire season. Fires were not as severe the following year in 2021, but deforestation increased.

Averaged across all five years, indigenous territories had the lowest overall primary forest loss rate (0.14%), closely followed by protected areas (0.15%).

Interestingly, indigenous territories had a lower deforestation rate compared to protected areas (0.5 vs 0.11%), but higher fire impact (0.09 vs 0.04%).

Outside of these designations (red), the primary forest loss rate was triple (.45%).

Methodology

To estimate deforestation across all three categories (protected areas, indigenous territories, and other), we used annual forest loss data (2017-21) from the University of Maryland (Global Land Analysis and Discovery GLAD laboratory) to have a consistent source across all countries (Hansen et al 2013).

We obtained this data, which has a 30-meter spatial resolution, from the “Global Forest Loss due to Fires 2000–2021” data download page. It is also possible to visualize and interact with the data on the main Global Forest Change portal.

The annual data is disaggregated into forest loss due to fire vs. non-fire (other disturbance drivers). It is important to note that the non-fire drivers include both human-caused deforestation and forest loss caused by natural forces (landslides, wind storms, etc.).

We also filtered this data for only primary forest loss, following the established methodology of Global Forest Watch. Primary forest is generally defined as intact forest that has not been previously cleared (as opposed to previously cleared secondary forest, for example). We applied this filter by intersecting the forest cover loss data with the additional dataset “primary humid tropical forests” as of 2001 (Turubanova et al 2018). Thus, we often use the term “primary forest loss” to describe this filtered data.

Data presented as primary forest loss rate is standardized per the total area covered of each respective category per year (annual). For example, to properly compare raw forest loss data in areas that are 100 hectares vs 1,000 hectares total size respectively, we divide by the area to standardize the result.

Our geographic range extends from the Andes to the Amazon plain and reaching the transitions with the Cerrado and the Pantanal. This range includes nine countries of the Amazon (or Pan-Amazon region as defined by RAISG) and consists of a combination of the Amazon watershed limit, the Amazon biogeographic limit and the Legal Amazon limit in Brazil. See Base Map above for delineation of this hybrid Amazon limit, designed for maximum inclusion.

Additional data sources include:

  • National and state/department level protected areas: RUNAP 2020 (Colombia), SNAP 2022 (Ecuador), SERNAP & ACEAA 2020 (Bolivia), SERNANP 2022 (Peru), INPE/Terrabrasilis 2022 (Brazil), SOS Orinoco 2021 (Venezuela), and RAISG 2020 (Guyana, Suriname, and French Guiana.)
  • Indigenous Territories: RAISG & Ecociencia 2022 (Ecuador), INPE/Terrabrasilis 2022 (Brazil), RAISG 2020 (Colombia, Bolivia, Venezuela, Guyana, Suriname, and French Guiana), and MINCU & ACCA 2021 (Peru). For Peru, this includes titled native communities and Indigenous/Territorial Reserves for indigenous groups in voluntary isolation.

For analysis, we categorized Protected Areas first, then Indigenous Territories to avoid overlapping areas. Each category was disaggregated by year created/recognized to match the annual report of forest loss, for example. If a Protected area was created in December 2018, it would be considered within the analysis for the year 2019.

Acknowledgements

This work was supported by the Andes Amazon Fund (AAF), Norwegian Agency for Development Cooperation (NORAD), and International Conservation Fund of Canada (ICFC).

We thank M. MacDowell and M. Cohen for helpful comments on this report.

Citation

Finer M, Mamani N (2023) Protected Areas & Indigenous Territories Effective Against Deforestation Across Amazon. MAAP: 176.

MAAP #161: Soy Deforestation in the Brazilian Amazon

Posted on by Ana Folhadella
Example of fires burning an area recently deforested for a new soy plantation. Data: Planet.

The Amazon Soy Moratorium has often been credited with significantly reducing soy-related deforestation in the Amazon over the past 15 years.

The Moratorium is a voluntary zero-deforestation agreement in which traders agree not to purchase soy grown on land cleared after 2008.

However, increasing soybean prices may be driving a resurgence of the problem of direct soy deforestation. That is, direct conversion of primary deforestation to soy plantation without passing an initial period as cattle pasture.

A recent report by Global Forest Watch estimated the direct soy deforestation of 29,000 hectares in the Brazilian Amazon in 2019 (Schneider et al 2021).

Here, we report the additional direct soy deforestation of at least 42,000 hectares in the Brazilian Amazon since 2020. All of these areas occurred in the state of Mato Grosso, located on the southeast edge of the Amazon.

We detected all of these soy plantations based on recent major fire activity (84 major fires), in which the recently deforested area was burned in preparation for the upcoming planting season (see Methodology below for more details).

Below, we show a base map of these recently deforested and then burned areas in the Mato Grosso state of the Brazilian Amazon followed by a series of examples from the satellite imagery.

Base Map – Recent Soy Deforestation in Brazilian Amazon

The Base Map below shows the areas, indicated by red dots, of recent direct deforestation for new soy plantations that we detected by monitoring major fire activity in 2022.

Between May 2021 and June 2022, we detected 84 major fires that corresponded to burning areas recently deforested for new soy plantations. These 84 areas, all of which occurred in the state of Mato Grosso, cover an area of 42,000 hectares.

Our geographic focus was the Brazilian Amazon biome in the state of Mato Grosso, as covered by the Amazon Soy Moratorium. For example, we also documented extensive direct soy deforestation and fire in the Bolivian Amazon (Santa Cruz department), but we did not include that information here.

Base Map. Recent Soy Deforestation in Brazilian Amazon. Data: ACA/MAAP, NICFI.

Examples of Deforestation & Fire for New Soy Plantations

As noted above, we detected the direct deforestation for new soy plantations by monitoring major fire activity in 2022. It is assumed that fires are preparing the recently deforested area for upcoming soy planting.

Methodology

We first tracked major fires in 2021 and 2022 using our novel real-time fire monitoring app. See MAAP #118 for more background information about the app and general methodology for detecting major fires based on aerosol emissions. The first major fires were detected in May of each year (2021 and 2022) and we continued collecting data on a daily basis through early July of each year. We monitored fires across the entire Amazon, but this report focuses on Brazil.

For all major fires detected with the app, we confirmed them with high-resolution satellite imagery from Planet. This confirmation was accomplished by visualizing either smoke plumes the day of the fire or burned areas in subsequent days after the fire.

All confirmed fires were assigned a category based on likely direct fire type or driver. These categories include 1) burning area recently deforested for new soy plantation, burning area recently deforested for new cattle pasture, and burning grasslands embedded in the larger rainforest matrix. On rarer occasions, one of these fire types may escape into the surrounding forest, making it an actual forest fire.

Specifically, the soy-related fires were defined as those burning recently deforested areas (that is, areas cleared since 2020) that had a distinctive linear pattern seemingly designed for organized crop agriculture. Most of the newly identified soy areas were also adjacent to existing soy plantations. In other words, the soy deforestation and fire pattern were visually quite distinct from cattle-related and grassland fires. Local experts have informed us that the fires are likely prepping the recently deforested area for the upcoming soy planting season. For all determined direct soy-related fires, we estimated the burned area using the spatial measurement tools in Planet Explorer and entered it into a database. We noted that in July of both years, the fires shifted away from soy and more towards cattle areas.

References

Martina Schneider, Liz Goldman, Mikaela Weisse, Luiz Amaral and Luiz Calado (2021) The Commodity Report: Soy Production’s Impact on Forests in South America. Link: https://www.globalforestwatch.org/blog/commodities/soy-production-forests-south-america/

X.-P. Song, M.C. Hansen, P. Potapov, et al (2021). Massive soybean expansion in South America since 2000 and implications for conservation. Nature Sustainability. Link: https://www.nature.com/articles/s41893-021-00729-z

Acknowledgements

We thank V. Silgueiro and R. Carvalho from the organization Instituto Centro de Vida (ICV) for helpful information and comments related to this report.

Citation

Finer M, Ariñez A (2022) Soy Deforestation in the Brazilian Amazon. MAAP: #161.

MAAP #158: Amazon Deforestation & Fire Hotspots 2021

Posted on by Ana Folhadella
2021 Amazon Forest Loss Base Map. Deforestation and fire hotspots across the full Amazon biome. Data: UMD/GLAD, ACA/MAAP.

We present a detailed look at the major 2021 Amazon forest loss hotspots, based on the final annual data produced by the University of Maryland.

This dataset is unique in that distinguishes forest loss from fire, leaving the rest as a close proxy for deforestation.

Thus, for the first time, the results include both deforestation and fire hotspots across the Amazon.

The Base Map (see right) and Results Graph (see below) reveal several key findings:p

  • In 2021, we estimate the loss of 2 million hectares (4.9 million acres) of primary forest loss across the nine countries of the Amazon biome. This total represents a slight decrease from 2020, but the 6th highest on record.
    l
  • The vast majority of this loss was deforestation (78%), accounting for 1.57 million hectares. This total represents a slight increase from 2020, and the 5th highest on record. This deforestation impacted the entire stretch of the southern Amazon (southern Brazil, Bolivia, and Peru) plus further north in Colombia.
    l
  • This deforestation was concentrated in Brazil (73%), Bolivia (10%), Peru (8%), and Colombia (6%). In Brazil and Bolivia, deforestation was the highest since 2017. In Peru and Colombia, deforestation dropped from 2020 but was still historically high. See below for maps and graphs for each country. See Annex for 2020-21 details.
    k
  • Fires directly caused the remaining primary forest loss (22%), accounting for 436,000 hectares. This total represents a decrease from the severe fire season of 2020, but was the 4th highest on record. Moreover, each of the six most intense fire seasons has occurred in the past six years. Over 90% of the fire impact occurred in just two countries: Brazil and Bolivia. Note that fire impacts were concentrated in the southeast of each country (Mato Grosso and Santa Cruz states, respectively).
    k
  • Since 2002, we estimate the deforestation of over 27 million hectares (67 million acres) of primary forest, greater than the size of the United Kingdom or the U.S. state of Colorado. On top of this, we estimate an additional impact of 6.7 million hectares due to fires.

Below, we zoom in on the four countries with the highest deforestation (Brazil, Bolivia, Peru, and Colombia), with additional maps and analysis.

Amazon Forest Loss Results Graph, 2002-21. Data: UMD/GLAD, ACA/MAAP.

For deforestation, note that in 2021 there was a slight increase across the Amazon, continuing a gradual four-year trend. 2021 had the 5th highest deforestation on record (behind just 2002, 2004, 2005, and 2017).

For fire, in 2021 there was a decrease from the severe fire season of 2020, but was the 4th highest on record (behind just 2016, 2017, and 2020). Moreover, each of the last six years is in the top six worst fire seasons across the Amazon.

For total forest loss (deforestation and fire combined), in 2021 there was slight decrease from 2020, but the 6th highest on record.

Brazil Base Map, 2021. Deforestation and fire hotspots in the Brazilian Amazon. Data: UMD/GLAD, ACA/MAAP.

Brazilian Amazon

In 2021, the Brazilian Amazon lost 1.1 million hectares of primary forest to deforestation. Fires directly impacted an additional 293,000 hectares.

The deforestation was the highest since 2017 and also the peak of the early 2000s (6th highest on record). The fire impact was relatively high (5th highest on record), but less than the peak years of 2016, 2017, and 2020.

The deforestation was concentrated along the major road networks, especially roads 163, 230, 319, and 364 in the states of Acre, Amazonas, Pará, and Rondônia (see Brazil Base Map).

The direct fire impacts were concentrated in the southeastern state of Mato Grosso.

It is also important to note that many areas experienced the one-two combination of initial deforestation followed by fire to prepare the area for agriculture or cattle.

 

 

 

Bolivia Base Map. Deforestation hotspots in Bolivian Amazon. Data: UMD/GLAD, ACA/MAAP.

Bolivian Amazon

In 2021, the Bolivian Amazon lost 161,000 hectares of primary forest to deforestation. Fires directly impacted an additional 106,000 hectares.

Deforestation was the third-highest on record, just behind the peak in 2016 and 2017. The fire impact was the second-highest on record, behind just the intense year of 2020 (thus, the last two years are the two highest on record).

Both the deforestation and fires were concentrated in the southeastern department of Santa Cruz (see Bolivia Base Map).

Much of the deforestation was associated with large-scale agriculture, while the fires once again impacted important natural ecosystems, most notably the Chiquitano dry forests.

 

 

 

 

 

 

 

Peru Base Map. Deforestation hotspots in the Peruvian Amazon. Data: UMD/GLAD, ACA/MAAP.

Peruvian Amazon

In 2021, the Peruvian Amazon lost 132,400 hectares of primary forest to deforestation. Fires directly impacted an additional 21,800 hectares.

Deforestation dropped from a record high in 2020, but was 6th highest on record. Thre fire impact was the second-highest on record (behind just 2017).

The deforestation was concentrated in the central and southern Amazon (Ucayali and Madre de Dios regions, respectively) (see Peru Base Map).

We highlight the rapid deforestation (365 hectares) for a new Mennonite colony in 2021, near the town of Padre Marquez (see MAAP #149).

Also, note some additional hotspots in the south (Madre de Dios region), but these are largely from expanding agriculture instead of the historical driver of gold mining.

Indeed, gold mining deforestation has been greatly reduced due to government actions, but this illegal activity still threatens several key areas and indigenous territories (MAAP #154).

 

 

 

 

Rapid deforestation (365 hectares) for a new Mennonite colony in 2021, near the town of Padre Marquez. Data: Planet.

Colombia Base Map. Deforestation hotspots in northwest Colombian Amazon. Data: UMD/GLAD, ACA/MAAP, FCDS.

Colombian Amazon

In 2021, the Colombian Amazon lost 98,000 hectares of primary forest to deforestation. Fires directly impacted an additional 9,000 hectares.

Deforestation and fire dropped from last year, but both were the fourth highest on record, following the trend of elevated forest loss and associated fires since the peace agreement in 2016.

As described in previous reports (see MAAP #120), the Colombia Base Map shows there continues to be an “arc of deforestation” in the northwest Colombian Amazon (Caqueta, Meta, and Guaviare departments).

This arc impacts numerous Protected Areas (particularly Tinigua and Chiribiquete National Parks) and Indigenous Reserves (particularly Yari-Yaguara II and Nukak Maku).

The main drivers of deforestation in the Colombian Amazon are land grabbing, expansion of road networks, and cattle ranching.

 

 

 

Annex

Notes and Methodology

The analysis was based on 30-meter resolution annual forest loss data produced by the University of Maryland and also presented by Global Forest Watch. For the first time, this data set distinguished forest loss caused directly by fire (note that virtually all Amazon fires are human-caused). The remaining forest loss serves as a likely close proxy for deforestation, with the only remaining exception being natural events such as landslides, wind storms, and meandering rivers.

Importantly, we applied a filter to calculate only primary forest loss by intersecting the forest cover loss data with the additional dataset “primary humid tropical forests” as of 2001 (Turubanova et al 2018). For more details on this part of the methodology, see the Technical Blog from Global Forest Watch (Goldman and Weisse 2019).

Our geographic range for the Amazon is a hybrid designed for maximum inclusion: biogeographic boundary (as defined by RAISG) for all countries, except for Bolivia where we use the watershed boundary.

To identify the deforestation hotspots, we conducted a kernel density estimate. This type of analysis calculates the magnitude per unit area of a particular phenomenon, in this case, forest cover loss. We conducted this analysis using the Kernel Density tool from the Spatial Analyst Tool Box of ArcGIS. We used the following parameters:

Search Radius: 15000 layer units (meters)
Kernel Density Function: Quartic kernel function
Cell Size in the map: 200 x 200 meters (4 hectares)
Everything else was left to the default setting.

For the Base Map, we used the following concentration percentages: Medium: >5%; High: >7%; Very High: >14%.

Acknowledgements

We thank A. Gómez (FCDS), R. Botero (FCDS)… for helpful comments on earlier drafts of the text and images.

This work was supported by NORAD (Norwegian Agency for Development Cooperation) and ICFC (International Conservation Fund of Canada).

Citation

Finer M, Mamani N (2022) Amazon Deforestation Hotspots 2021. MAAP: 153.

MAAP #153: Amazon Deforestation Hotspots 2021

Posted on by Ana Folhadella
Amazon Base Map. Deforestation hotspots across the Amazon in 2021 (as of September 18). Data: UMD/GLAD, ACA/MAAP.

We present a first look at the major 2021 Amazon deforestation hotspots.*

The Amazon Base Map illustrates several key findings:p

  • We estimate the loss of over 1.9 million hectares (4.8 million acres) of primary forest loss across the nine countries of the Amazon biome in 2021.
    k
  • This matches the previous two years, bringing the total deforestation to 6 million hectares (15 million acres) since 2019, roughly the size of the state of West Virginia.
    p
  • In 2021, most of the deforestation occurred in Brazil (70%), followed by Bolivia (14%), Peru (7%), and Colombia (6%).
    p
  • In Brazil, hotspots are concentrated along the major road networks. Many of these areas were also burned following the deforestation.
    j
  • In Bolivia, fires once again impacted several important ecosystems, including the Chiquitano dry forests.
    p
  • In Peru, deforestation continues to impact the central region, most notably from large-scale clearing for a new Mennonite colony.
    p
  • In Colombia, there continues to be an arc of deforestation impacting numerous protected areas and indigenous territories.

Below, we zoom in on the four countries with the highest deforestation (Brazil, Bolivia, Peru, and Colombia), with additional maps and analysis.

Brazil Base Map. Deforestation hotspots in Brazilian Amazon. Data: UMD/GLAD, ACA/MAAP.

Brazilian Amazon

The Brazil Base Map shows the notable concentration of deforestation hotspots along the major roads (especially roads 163, 230, 319, and 364) in the states of Acre, Amazonas, Pará, and Rondônia.

 

 

 

 

 

 

 

 

 

 

 

Bolivia Base Map. Deforestation hotspots in Bolivian Amazon. Data: UMD/GLAD, ACA/MAAP.

Bolivian Amazon

The Bolivia Base Map shows the concentration of hotspots due to major fires in the Chiquitano dry forest biome, largely located in the department of Santa Cruz in the southeast section of the Amazon.

 

 

 

 

 

 

 

 

 

 

 

Peru Base Map. Deforestation hotspots in the Peruvian Amazon. Data: UMD/GLAD, ACA/MAAP.

Peruvian Amazon

The Peru Base Map shows the concentration of deforestation in the central Amazon (Ucayali region).

We highlight the rapid deforestation (365 hectares) for a new Mennonite colony in 2021, near the town of Padre Marquez (see MAAP #149).

Also, note some additional hotspots in the south (Madre de Dios region), but these are largely from expanding agriculture instead of the historical driver of gold mining.

Indeed, gold mining deforestation has been greatly reduced due to government actions, but this illegal activity still threatens several key areas and indigenous territories (MAAP #130).

 

 

 

 

 

 

 

Colombia Base Map. Deforestation hotspots in northwest Colombian Amazon. Data: UMD/GLAD, ACA/MAAP.

Colombian Amazon

As described in previous reports (see MAAP #120), the Colombia Base Map shows there continues to be an “arc of deforestation” in the northwest Colombian Amazon (Caqueta, Meta, and Guaviare departments).

This arc impacts numerous Protected Areas (particularly Tinigua and Chiribiquete National Parks) and Indigenous Reserves (particularly Yari-Yaguara II and Nukak Maku).

 

 

 

 

 

 

 

 

 

*Notes and Methodology

The analysis was based on 10-meter resolution primary forest loss alerts (GLAD+) produced by the University of Maryland and also presented by Global Forest Watch. These alerts are derived from the Sentinel-2 satellite operated by the European Space Agency.

We emphasize that this data represents a preliminary estimate and more definitive annual data will come later in the year.

We also note that this data does include forest loss caused by natural forces and burned areas.

Our geographic range for the Amazon is a hybrid between both the biogeographic boundary (as defined by RAISG) and watershed boundary, designed for maximum inclusion.

To identify the deforestation hotspots, we conducted a kernel density estimate. This type of analysis calculates the magnitude per unit area of a particular phenomenon, in this case, forest cover loss. We conducted this analysis using the Kernel Density tool from the Spatial Analyst Tool Box of ArcGIS. We used the following parameters:

Search Radius: 15000 layer units (meters)
Kernel Density Function: Quartic kernel function
Cell Size in the map: 200 x 200 meters (4 hectares)
Everything else was left to the default setting.

For the Base Map, we used the following concentration percentages: Medium: 5-7%; High: 7-14%; Very High: >14%.

Acknowledgements

This work was supported by NORAD (Norwegian Agency for Development Cooperation) and ICFC (International Conservation Fund of Canada).

Citation

Finer M, Mamani N, Spore J (2022) Amazon Deforestation Hotspots 2021. MAAP: 153.