For a pulp producer that depends on millions of hectares of forest, a wildfire is not only a conservation problem.

It can become a fibre-supply problem.

Suzano is expanding a forest-fire monitoring system that combines satellite imagery, artificial intelligence, tower-mounted cameras and drones across its Brazilian forest operations.

The individual technologies are already familiar.

What is more important is how Suzano is connecting them.

Satellites provide broad-area surveillance.

Fixed cameras watch selected landscapes continuously.

Artificial intelligence analyzes those images for smoke.

Drones can then be sent to investigate suspected incidents more closely.

The objective is to shorten the distance between:

seeing something unusual

and

putting the right firefighting resources in the right place.

Suzano says its current average time between detecting an occurrence and mobilizing response teams is approximately 35 minutes.

The company also reports that over the past three years, 85% of recorded fires were controlled before they developed into higher-impact incidents.

Those figures are company-reported rather than independently audited wildfire-performance benchmarks.

But they point toward the more important industrial question:

How much forest loss can be avoided by detecting a fire while it is still small?

For a Pulp Producer, the Forest Is Production Infrastructure

Pulp-industry discussions usually focus on the mill.

Digesters.

Recovery boilers.

Drying machines.

Evaporation plants.

Power generation.

Those assets are visible and capital-intensive.

But for an integrated eucalyptus pulp producer, the production system begins years before wood reaches the mill gate.

Suzano says its total forest base covers approximately 2.8 million hectares, including planted forests and conservation areas.

Its sustainability data describe approximately 1.47 million hectares of eucalyptus areas, alongside roughly one million hectares of native forest in older reporting boundaries, while its current climate materials refer to around 1.1 million hectares of protected native forest within the broader estate.

That scale changes what fire protection means.

A mill can repair a damaged pump.

A burned plantation compartment cannot simply be restarted the following morning.

Depending on tree age, severity and location, fire can disrupt:

harvest plans;

wood volumes;

road access;

replanting schedules;

forest productivity;

and future fibre flows.

That makes forest protection part of asset management.

The forest is effectively a distributed biological production system feeding the pulp mill several years in the future.

Detection Speed Matters Because Fire Growth Is Nonlinear

Wildfire response has one unforgiving characteristic.

A small fire and a large fire are not simply the same incident at different sizes.

As fire spreads, the response problem changes.

More area is involved.

More fuel is burning.

More crews may be needed.

Wind can create new fronts.

Access becomes harder.

Nearby plantations, native vegetation, infrastructure and communities may become exposed.

That makes the first stages of an incident particularly important.

Suzano’s current system is therefore designed around shortening several separate time intervals:

ignition → detection

detection → confirmation

confirmation → mobilization

mobilization → suppression

The company’s published 35-minute figure covers only the period between detection and team mobilization.

It should not be interpreted as 35 minutes from ignition to extinguishment.

That distinction matters.

A fire can exist for some period before any monitoring system detects it.

Response crews also still need time to reach the site after mobilization.

So one of the next useful metrics would be the complete interval from first detectable fire signal to suppression activity on the ground.

Satellites Solve the Coverage Problem

Suzano’s forest holdings are geographically dispersed.

No network of ground cameras can observe every hectare equally well.

Satellite monitoring helps solve the scale problem.

Earth-observation systems can scan very large areas and identify thermal anomalies or other signals associated with fire.

But satellite monitoring has limitations.

Resolution matters.

Revisit frequency matters.

Cloud cover matters.

The size and intensity of a fire matter.

A satellite therefore does not eliminate the need for other detection methods.

Suzano’s model uses satellite information as one layer rather than the only layer.

The company describes the logic clearly:

satellite systems provide wide-area coverage while ground-based cameras provide continuous local monitoring.

That layered approach is more important than any individual sensor.

Towers Solve the Continuity Problem

Fixed cameras offer something satellites cannot always provide:

continuous observation of a defined landscape.

Suzano has been building this infrastructure for several years.

Earlier company disclosures described large networks of observation towers equipped with high-resolution cameras and automated smoke detection.

For example, Suzano reported in 2025 that its forest operations were using more than 100 monitoring towers across Brazil, with some systems providing 360-degree surveillance over areas extending up to 15 kilometres from a tower.

In Mato Grosso do Sul, the company said in June 2026 that an integrated system with 57 video-monitoring towers, AI and a thermal drone operated 24 hours a day and covered more than one million hectares during the dry season.

That network was described as directly protecting more than 300,000 hectares of native vegetation and conservation areas in the Cerrado.

Fixed cameras therefore provide the persistent observation layer.

The next problem is having somebody—or something—watch all of those images.

AI Solves the Attention Problem

A camera network can generate an enormous volume of visual information.

Human operators cannot continuously scrutinize every pixel from dozens or hundreds of cameras with equal attention.

That is where artificial intelligence becomes useful.

Suzano says its algorithms continuously analyze camera images for signs associated with smoke and automatically generate alerts for responsible teams.

This is a more practical industrial use of AI than many corporate AI announcements.

The system is not being asked to make a broad strategic decision.

It has a narrow task:

look at many images repeatedly and flag something that may require human attention.

That can reduce monitoring workload and potentially shorten detection time.

But AI also introduces another operational question:

How many alerts are real fires?

False positives matter.

Dust.

Fog.

Clouds.

Industrial emissions.

Controlled burns.

Other visual anomalies may resemble smoke.

Too many false alarms waste drone flights and firefighting resources.

Suzano specifically says that combining the technologies helps reduce false alarms.

That is why confirmation becomes the next step.

Drones Solve the Verification Problem

A camera several kilometres away may identify smoke.

It may not reveal precisely:

where the fire started;

how large it is;

which direction it is moving;

or what terrain surrounds it.

A drone can close that information gap.

Suzano uses drones to investigate incidents and improve location accuracy before or during the deployment of firefighting resources.

Thermal imaging makes this particularly useful.

Visible smoke can obscure a fire.

A thermal sensor can help identify heat signatures, hotspots and active edges that are harder to interpret through an ordinary camera alone.

In Mato Grosso do Sul, Suzano has already added a high-resolution thermal-camera drone to its dry-season monitoring system.

Again, the value is not the drone itself.

It is the role the drone plays inside the decision chain.

Satellite: something may be happening.

Tower camera + AI: smoke may be present.

Drone: this is the location and condition.

Brigade: send resources here.

That is an operational system.

Integration Is More Important Than Buying More Sensors

This may be the strongest lesson for other forest companies.

It is relatively easy to purchase technology.

A satellite-data subscription can be purchased.

AI software can be licensed.

A drone can be bought.

A camera can be placed on a tower.

The harder question is whether the information flows into one operating process.

Who receives the alert?

How quickly?

Can the operator see the satellite and tower data together?

Who decides whether to launch the drone?

Which brigade receives the incident?

Does the response team receive coordinates?

Can nearby crews see the same information?

How are false alarms recorded?

How is the response outcome fed back into the detection system?

Those workflow questions determine whether technology improves firefighting or simply generates more screens in a control room.

Suzano itself identifies integration as one of the main advantages of the programme.

That is more significant than saying the company “uses AI.”

The 85% Figure Needs Careful Interpretation

Suzano says 85% of fires during the past three years were controlled before becoming incidents of greater impact.

That sounds impressive.

But it should not be rewritten as:

“Suzano’s AI prevents 85% of wildfires.”

The company does not make that claim.

Several things contribute to the outcome:

early detection;

trained brigades;

access roads;

firefighting equipment;

weather;

fuel conditions;

community reporting;

and incident location.

The technology is one layer in that system.

The company also has not publicly provided, alongside the 85% figure:

a definition of “greater impact”;

the total number of incidents;

average burned area;

a control-group comparison;

or a breakdown of outcomes before and after the current technology was introduced.

So the safest conclusion is narrower:

Suzano reports that its overall prevention and response structure has contained 85% of recorded fires before they escalated into higher-impact events.

That is useful operating information.

It is not enough to isolate the causal effect of AI.

The 35-Minute Figure Also Has a Boundary

The same caution applies to the response-time metric.

Suzano says its current average is 35 minutes between detection and mobilization of teams.

Mobilization is not arrival.

And arrival is not extinguishment.

An incident two kilometres from a brigade base has a different access problem from one deep inside a remote forest block.

Future reporting would become much more useful if Suzano disclosed several measures separately:

detection-to-validation time;

validation-to-dispatch time;

dispatch-to-arrival time;

arrival-to-control time;

and final burned area.

That would allow forest managers to identify where technology is producing the greatest improvement.

Fire Risk Is Also a Logistics Problem

Even when fire does not destroy an entire plantation, it can disrupt wood logistics.

Forest roads can be closed.

Harvest crews can be moved to emergency response.

Machinery may be relocated.

Transport routes may need to change.

Smoke can affect visibility and worker safety.

Firebreaks may have to be expanded.

Harvest sequencing may need revision.

For a large pulp mill consuming wood continuously, those disruptions matter.

Pulp production prefers consistency.

The mill cannot easily absorb large fluctuations in fibre delivery simply because a forest district has a fire emergency.

This gives the monitoring programme another economic function:

reduce the probability that a local ignition becomes a large operational disruption.

That is difficult to express as a simple return on investment.

The value is partly avoided loss.

Wildfire Risk Is Becoming More Relevant in 2026

Suzano’s August announcement also comes during a period of elevated fire concern in Brazil.

Federal climate authorities warned in July that an intensifying El Niño could increase wildfire risk between August and October, particularly across the Amazon, Cerrado and parts of Brazil’s Centre-West.

Cemaden’s August monitoring later reported continuing drought conditions across several parts of Brazil and warned that higher temperatures combined with reduced rainfall could favour fire propagation, particularly in the Centre-West and southern Amazon.

Brazil’s federal government also activated its 2026 national multi-agency wildfire coordination structure at the end of July for the critical fire season.

This does not mean every Suzano plantation faces the same fire risk.

The company’s forest base extends across several states and biomes.

Local rainfall, vegetation, topography and human activity differ considerably.

But the broader operating context supports continued investment in detection and response.

Mato Grosso do Sul Shows Why the System Matters to Pulp Supply

Mato Grosso do Sul is an especially useful example.

Suzano operates major pulp assets in the state, including its Três Lagoas complex and the newer Ribas do Rio Pardo operation.

Its Três Lagoas site alone has stated production capacity of approximately 3.25 million tonnes per year across two production lines.

The same state contains hundreds of thousands of hectares of Suzano eucalyptus plantations and conservation areas.

That connects fire protection directly with industrial scale.

A modern pulp mill is built around an expected long-term wood basket.

Forest productivity, harvest radius and transportation cost feed into the mill’s economics.

If major wildfire losses change the age structure or usable volume within that basket, the effect can persist beyond one fire season.

Suzano has not disclosed a quantified relationship between its current wildfire system and avoided pulp production losses.

So that should not be invented.

But the strategic link is clear:

protecting standing eucalyptus protects part of the future raw-material base supporting the mill.

Plantation Age Makes Fire Loss More Complicated Than Hectares Burned

A hectare figure alone is also incomplete.

The economic impact of fire depends on what was standing on that hectare.

A recently planted area represents sunk establishment costs and several years of future growth.

A mature stand close to harvest represents a different inventory value.

Fire intensity matters too.

Some timber may remain recoverable after a lower-severity event.

Other damage may compromise wood quality or make salvage uneconomic.

Road and harvest conditions matter.

So a forest company’s most useful fire KPI is not necessarily:

hectares affected.

It may also need to measure:

standing volume affected;

age-class distribution;

recoverable wood;

replanting requirement;

logistics disruption;

and future harvest-plan deviation.

This is where combining wildfire monitoring with forest-inventory data could become particularly powerful.

Suzano has not publicly stated that the current AI system performs that complete optimization.

But the company’s large digital forest database creates an obvious operational context in which precise fire location becomes valuable.

Conservation Areas Add Another Dimension

Suzano is not monitoring only commercial eucalyptus.

The company says the technology is intended to protect both planted forests and permanent preservation areas.

That distinction matters.

A plantation loss can often be replanted.

Fire within native vegetation can have different ecological consequences involving:

habitat;

soil;

biodiversity;

restoration;

and landscape connectivity.

Suzano’s broader forest base contains extensive conservation areas, and the company has committed to connecting native forest fragments as part of its biodiversity strategy.

Wildfire protection therefore serves two overlapping objectives:

protect commercial fibre assets

and

protect ecological assets.

Those outcomes should not be reduced to one financial metric.

Communities Are Part of the Detection Network Too

Technology also cannot cover every ignition.

Many forest fires start near roads, farms, communities or human activity.

People therefore remain part of the prevention system.

Suzano says its programme includes awareness work with neighbouring communities and trained brigades, as well as cooperation with other organizations.

That is important because wildfire prevention begins before smoke is visible.

Education.

Controlled access.

Safe land management.

Fuel management.

Reporting.

Coordination with neighbours.

These can prevent or contain incidents before sophisticated detection systems become necessary.

The technology strengthens the network.

It does not replace it.

Mutual Assistance Expands Available Firefighting Capacity

Suzano also participates in Mutual Assistance Plans, or PAM arrangements, involving public authorities, neighbouring companies and regional institutions.

The objective is to coordinate emergency response and share resources when an incident exceeds what one organisation can handle alone.

That matters because large wildfires are inherently regional.

Fire does not respect property boundaries.

A blaze starting outside Suzano land can enter company plantations.

A fire originating within a plantation can threaten neighbours.

Equipment may be closer on another organisation’s property.

Water resources may be distributed across several sites.

Mutual assistance therefore turns the response system from a company network into a territorial network.

The value becomes most visible during the incident that is too large for the normal brigade.

Suzano Is Now Looking Lower in Orbit

The next stage of the technology strategy includes low-Earth-orbit satellite systems.

Suzano says it is assessing additional satellite-detection capabilities, including lower-orbit platforms.

The attraction is understandable.

For wildfire monitoring, useful satellite performance depends on combinations of:

revisit time;

spatial resolution;

sensor type;

latency;

and cloud conditions.

More frequent observations can potentially narrow the time between ignition and detection.

But “low Earth orbit” by itself does not guarantee earlier detection.

The actual benefit depends on the constellation and sensing capability being used.

Suzano has not publicly identified:

a satellite provider;

resolution;

detection threshold;

coverage frequency;

or deployment timetable.

So this remains a technology under evaluation, not a proven upgrade to the 35-minute response metric.

BVLOS Drones Could Change Coverage More Significantly

Suzano is also evaluating long-range drones capable of BVLOS — beyond visual line of sight — operation, equipped with thermal cameras.

That could be more consequential than simply adding additional conventional drones.

Normal visual-line-of-sight operations constrain how far a pilot can send an aircraft.

Forestry operates over huge distances.

BVLOS capability could potentially allow drones to investigate suspected incidents across a much larger radius from a control point.

That can be valuable where:

road access is slow;

terrain is difficult;

nighttime conditions reduce visibility;

or the fire location is uncertain.

But BVLOS operation introduces another layer of complexity.

Airspace regulation.

Communications reliability.

Navigation.

Command links.

Operational procedures.

Interactions with firefighting aircraft.

Suzano says it is evaluating the technology.

It should therefore not yet be described as a fully deployed autonomous long-range drone network.

Thermal Cameras Can Extend Observation Beyond Smoke

Thermal sensors are particularly relevant to fire monitoring because they detect heat rather than depending entirely on visible smoke.

They can help teams identify:

active flame fronts;

hotspots;

residual heat;

and potentially reignition points.

Suzano’s Mato Grosso do Sul system already includes thermal drone capability, while the proposed BVLOS systems would extend that concept to longer-range operations.

For firefighting, this can support both the beginning and end of an incident.

At the beginning:

Where exactly is the fire?

After suppression:

Is anything still hot enough to restart?

The latter point is easy to overlook.

A fire that appears extinguished visually may still contain hotspots capable of reignition.

Remote Firebreak Equipment Moves Technology From Detection to Intervention

The most interesting experimental technology may not involve AI at all.

Suzano says it is testing remotely operated equipment capable of creating firebreaks without requiring operators to remain directly in hazardous areas.

This represents a different stage of wildfire technology.

Satellites, cameras and drones primarily improve information.

Remote firebreak equipment changes physical intervention.

A firebreak removes or disrupts combustible material along a planned line, helping limit fire spread.

Traditionally, creating or reinforcing one can expose operators and machinery to:

smoke;

heat;

rapidly changing wind;

falling trees;

and approaching flame fronts.

Remote operation could reduce some direct human exposure.

It could also allow preventive work in areas where risk would otherwise delay deployment.

But Suzano has not provided detailed technical specifications or field-performance data for the equipment.

It remains under testing.

That is the appropriate wording.

The Technology Stack Should Ultimately Be Judged by Burned Area

For a forest manager, the most meaningful outcome is not the number of AI alerts.

It is whether forest loss declines.

Useful long-term metrics would therefore include:

annual hectares burned;

average burned area per incident;

median detection time;

response time;

percentage controlled during initial attack;

false-alert rate;

firefighting cost per incident;

commercial wood volume affected;

and restoration area required.

Suzano’s 85% early-control result begins to address one of these.

The company has also previously reported reductions in fire incidents within some operating regions following investments in monitoring and prevention.

But a complete nationwide before-and-after dataset tied specifically to the newest integrated technology has not been published.

That would be the next useful evidence.

AI Should Not Become the Headline KPI

This is particularly important because AI can distract from the operating outcome.

A forest company can add an AI model and still have poor fire performance.

The model matters only if it improves:

detection speed;

confirmation accuracy;

dispatch decisions;

or resource allocation.

The best question is not:

How much AI is Suzano using?

It is:

How many minutes did the system remove from the fire-response chain?

And after that:

How many hectares and tonnes of standing wood did those minutes protect?

Those are the numbers that connect digital technology with pulp economics.

The Fibre-Security Calculation Extends for Years

Wildfire risk also needs a longer time horizon than an annual sustainability report.

Suzano says eucalyptus maintenance typically continues until harvest at around five to seven years, depending on the production cycle and operating conditions.

That means a severe fire can disturb a forest-production plan developed years in advance.

The lost volume may not affect the mill tomorrow.

It may influence the wood basket several harvest seasons later.

Companies can respond through:

salvage;

alternative stands;

third-party wood;

changed harvest sequencing;

or greater transport distances.

Each option has different economics.

This is why wildfire prevention belongs beside:

genetics;

silviculture;

harvesting;

road design;

and wood logistics

inside the fibre-supply strategy.

What Should Other Pulp Producers Measure?

Forest-based pulp companies evaluating similar systems should begin with the response chain rather than the technology catalogue.

Measure the current baseline.

How long between ignition and detection?

How long between detection and confirmation?

How long before a brigade is dispatched?

How long until it arrives?

What percentage of alerts are false?

What is the average fire size when crews arrive?

Then determine which technology attacks which delay.

If detection is slow because the estate is too large:

satellite coverage may help.

If towers already detect smoke but operators miss it:

AI image analysis may help.

If incidents are difficult to locate:

thermal drones may help.

If crews know where the fire is but cannot reach it quickly:

another camera does not solve the problem.

The bottleneck has moved elsewhere.

That is why system integration matters.

Before Buying Another Drone, Follow One Fire From Smoke to Suppression

For pulp producers, forest managers and wood-procurement teams, Suzano’s programme offers a practical audit.

Take the last ten real fire incidents.

Build a timeline for each one.

When did ignition probably occur?

When was the first signal detected?

Which system detected it?

When was the alert verified?

Was a drone needed?

When was the brigade mobilized?

How far did it travel?

When was the fire contained?

How much plantation or conservation area was affected?

Then identify where the largest delays occurred.

That is where the next investment belongs.

Not necessarily in another satellite.

Not necessarily in another AI model.

Not necessarily in another drone.

It might be:

a new tower;

a closer brigade base;

a better road;

a water point;

additional training;

a firebreak;

or more reliable radio communications.

Suzano’s strongest idea is therefore not any individual piece of technology.

It is treating detection, verification and response as one system.

That becomes particularly important in 2026, when Brazilian authorities continue to warn of significant seasonal fire risk across several regions where major plantation-forest operations are located.

The company says its current network can mobilize teams about 35 minutes after detection and that 85% of fires over the past three years were contained before escalating.

Those are useful operational indicators.

The next level of transparency would be showing what they mean in terms of:

hectares not burned,

standing wood protected,

and ultimately

fibre supply kept on schedule.

For an integrated pulp producer, that is where wildfire technology moves from environmental protection into manufacturing strategy.

The real value of detecting smoke earlier is not the alert itself. It is preserving the trees that the mill expects to consume years from now.

Last Updated on 08/09/2026 by Jota Machinery