Landslides: How to Predict Operational Shutdowns and Interdiction by Asset

Three highways leading to São Paulo's north coast were shut down the same Sunday by fallen slope debris. The Mogi-Bertioga highway closed at kilometers 77 and 98, the Oswaldo Cruz highway was blocked at kilometer 81 in Ubatuba, and the Tamoios highway had a partial closure on the Serra Antiga stretch.

The trigger was the rain of February 22, 2026, with more than 100 mm accumulated in the region. A man of about 50 went missing after a landslide hit his property at kilometer 64 of the Oswaldo Cruz highway, between São Luiz do Paraitinga and Natividade da Serra. The same storm flooded about 400 homes in Ubatuba and displaced 15 families, by flooding, not by mass movement.

Landslides make the news as sudden events, but the physical mechanism actually has two speeds. In the slow form, what pushes a slope to failure is days of accumulated rain: infiltrated water reduces the soil's resistance to the point where the next rainfall, even a moderate one, is enough to make the slope give way. In the fast form, typical of shallow slopes and steep terrain, the rainfall of the last few hours alone is already intense enough to trigger the movement, without needing days of prior saturation.

Without enough warning that a specific point is about to reach its limit, a landslide charges on three fronts at once:

Life: whoever is in the path of the descending mass. The crew at the excavation face, the driver on the mountain stretch, the resident on the slope below the cut.

Operations: whatever depends on that point stops working. Mining, traffic, the transmission line, and the maintenance access road all halt together, because the same collapsed slope brings down the structure and blocks the road that would get the crew there.

Budget: every day of that shutdown adds to the month's operating cost, on top of slope recovery and the contractual penalty that comes with it.

Natural disasters cost Latin America US$ 2.67 billion in the first quarter of 2026, and only 20% of that value was insured. In Mexico alone, catastrophe losses topped US$ 1 billion in 2025. In Europe, weather- and climate-related extremes caused EUR 40.4 billion in economic losses in 2024 alone, with less than 20% of total losses insured across the region.

Landslides cut across several sectors that operate on terrain, from mining to highway concessions, from railways to power transmission, and they tend to hide inside every single company's risk spreadsheet.

They show up as delayed earthworks, as an unscheduled shutdown, as a blocked stretch, as a contract time extension, as a property claim. The real cost stays spread across several cost lines, so it rarely shows up consolidated in a single report labeled "landslide."

In this article

How landslides form in Mexico, Europe, Brazil, and Chile.
Turning rain on a slope into risk calculated by asset.
What the same landslide hits in every sector, starting with the table that sums it all up.
- Mining: the saturation threshold missing from the shutdown protocol.
- Heavy construction: entering the rainy season with the schedule protected.
- Highways: the landslide point has a kilometer and a probable window.
- Railways: the segment about to shut down shows up before the ground moves.
- Power transmission and distribution: the tower on the slope, and the access road that disappears with it.
- Insurance: exposure by address improves landslide pricing.
The return on investment from anticipating the hazard.
The method: three steps to apply starting tomorrow.
Where this becomes a daily decision.

How landslides form in Mexico, Europe, Brazil, and Chile

Landslide is the common term for mass movement: soil, rock, or fill material that moves down a slope when the force pulling the material downward exceeds the resistance holding it in place.

Water is the main driver of that tipping point, and it acts in two ways at once. By infiltrating, it adds weight to the mass. By filling the soil's voids, it raises pore pressure, reduces the effective stress between particles, and with it, the shear resistance that was holding the slope together.

The indicator that matters changes with the depth of the failure. In a deep-seated landslide, what matters is rainfall accumulated over days: a slope that is already saturated fails at a volume that, in dry soil, would have no effect at all. In a shallow landslide, common on steep slopes with thin soil, the intensity of rainfall over the last few hours alone is enough to trigger the movement, with little or no prior saturation. Brazil's Cemaden uses exactly those two windows, from 1 to 72 hours of antecedent rainfall, to calculate the critical moment for landslides on the São Paulo coast. i4sea follows the same principle, but without a single nationwide threshold: the methodology varies by region, calibrated by soil, slope, occurrence history, vegetation cover, and existing structures at that point, among other factors.

Mexico: two different systems, one inland and one on the coast, drive the country's landslide risk. In June 2026, the state of Hidalgo recorded 28 landslides, 33 affected roads, and seven damaged or closed bridges across 14 municipalities, with rural communities isolated. In the same period, the México-Toluca federal highway was damaged near Contadero.

Europe: the worst damage shows up where persistent rain meets an already unstable slope. In Niscemi, Sicily, the storm that hit southern Italy starting January 18, 2026, linked to Cyclone Harry, triggered a mass movement with a failure front of at least 4 km and roughly 6 meters of subsidence on the SP10 provincial road. Two provincial roads closed, the connection to the Gela-Catania state highway was blocked, and about a thousand residents were evacuated.

Brazil: mountainous terrain concentrates the risk into narrow strips of highway, railway, pipeline, and transmission line, where a single unstable cut takes an entire region's access out of operation. That was the pattern behind the February 22, 2026 event on São Paulo's north coast that opens this article.

Chile: frontal systems in the south of the country arrive linked to atmospheric rivers, which dump a large volume of rain in a few hours onto already-wet slopes, a picture of the fast form described above. On July 7, 2026, a frontal system with a category 4 atmospheric river hit Valdivia, in the Los Ríos region, with a forecast above 100 mm. Landslides occurred in coastal areas and Route T-450, connecting Valdivia to Corral, was cut off.

Four regions, four geologies, the same physical mechanism: accumulated water that reduces a slope's resistance to the point of failure. That is why anticipating accumulated rain on a slope applies to any operation built on terrain, in any of these markets.

Turning rain on a slope into risk calculated by asset

Two operations receive the same three-day rainfall. The one that knows which of its slopes will reach the limit, and in what window, pulls back equipment, reschedules the work front, and keeps the rest of the network running. The other pays for improvisation. In practice, the difference shows up as equipment pulled from the pit ahead of the event, a clearing crew positioned at the right kilometer the day before, and an earthworks front rescheduled with no overtime.

Rain accumulated on a slope is the hazard. Interdiction, shutdown, and accidents are the risk. Between the two sit factors that multiply or reduce the impact, and they decide whether the same saturation brings down a slope in an area with nobody nearby, or brings down the cut that supports the only access to a substation. i4sea follows three layers to make that conversion.

Layer 1, the data. The foundation is 10 years of historical data from a proprietary model with 1 to 3 km resolution across all of Latin America, calibrated and optimized for local climate. Where public forecasts treat an entire mountain range as a single point of roughly 25 km, i4sea's reading is more than 5 times finer, and the threshold for each registered slope is calibrated with that specific point's history.

Layer 2, the interaction with the territory and the business. The same volume of rain does not produce the same outcome at two different points. The reading crosses the hazard with:

Slope angle and geometry: the angle, height, and shape of the cut or fill at that point.
Soil type and land use: clay-rich material, poorly compacted fill, and sealed surfaces respond differently to the same accumulated rainfall.
Vegetation cover: live roots hold the topsoil together; a deforested or burned slope loses that hold and channels more water into the mass.
Prior saturation: in a deep-seated failure, the accumulated rain of prior days defines how much new rain that slope can still take. On a shallow slope, the intensity of the last few hours matters more than the accumulation, so the same slope carries two different thresholds: one by duration, one by intensity.
What sits in the path of the descending mass: a slope that fails in an area with no human or economic activity creates no business risk. The same slope failing onto the road, the operating track, the structure, or a yard full of people does.
What depends on the point that gets hit: the more people, assets, and revenue that pass through that specific stretch, the greater the risk, even with an identical climate hazard.

Geotechnical instrumentation acts as a complementary layer, and each one answers a different question:

What instrumentation answers: structural condition and movement already underway. Borehole surveys, stability reports, inclinometers, piezometers, and ground radar support the engineering decision.
What the climate reading answers: the rainfall forecast for that point, the crossing with the threshold that has already pushed that slope to its limit, and how many hours the crew has to act.

Together, the two turn a structural snapshot into a decision window.

Layer 3, the risk matrix. i4sea connects that threat's index, intensity, and hazard window at the exact point, with the impact that produces in that specific business. That connection is what turns "heavy rain is coming to the region" into a risk reading calculated by asset, ready for a decision.

These three layers are what deliver the four gains the operation feels: lower cost, less exposure of lives, crews allocated where the risk actually is, and planning locked in before the event.

What the same landslide hits in every sector

The physics is the same. The asset that fails, the indicator that suffers, and who signs off on it change completely.

Table 1 · What landslides charge, by sector

Sector	What the slope compromises	Where the cost shows up
Mining	Pit slope, waste rock pile, access road, and internal conveyor	Unplanned downtime, lost throughput
Heavy construction	Cut, fill, earthworks front, retaining structure	Day of delay, contract time extension, life-safety risk
Highways	Debris over the road, marginal slope, access bridge	Closure, concession SLA, accident
Railways	Cut slope, track bed, remote stretch	Track downtime, cascading network effect
Power T&D	Tower on a slope, right-of-way, access road	Outage, continuity, emergency crew
Insurance	Concentrated property claims by address	Pricing, provisioning, subrogation

Mining: the saturation threshold missing from the shutdown protocol

Start with the outcome. A mine that knows on Monday that the pit's north sector will reach its saturation limit in the early hours of Thursday pulls back equipment, reschedules maintenance, and alerts the internal railway days ahead. A mine that finds out once the slope is already moving pays for evacuation, recovery, and lost shipment all at once.

An unanticipated shutdown in mining costs between US$ 150,000 and US$ 500,000 per event, with an expected ROI of 25 to 50 times for those who anticipate it, in i4sea's estimate by sector. The cost is high because the chain is long: it hits the mining face, the conveyor, the internal railway, delays loading at the terminal, and reschedules a vessel already en route.

Landslides have three distinct addresses in a mine, and three different financial consequences: the pit slope, the waste rock pile, and the area around a tailings dam. All three react to the same physical trigger, rising pore pressure from accumulated water, and all three tend to be logged under separate line items, which dilutes the real size of the problem.

Calling the event force majeure has a side effect rarely discussed in committee: the investigation ends there too. When the event is filed as unpredictable, nobody asks what the previous seven days of accumulated rain looked like, nobody checks whether that part of the pit already had a saturation history for that time of year, and no learning happens. The bill reopens next rainy season, on the same slope.

The opposite path starts by registering every critical slope as a monitored asset, with coordinates, geometry, and the accumulated rainfall that has already pushed it to its limit. From there, the alert stops being meteorological and becomes operational: high saturation risk in sector X, probable window, pull back equipment, and trigger preventive maintenance days ahead.

Heavy construction: entering the rainy season with the schedule protected

At the mine, the compromised asset is the slope itself. On a construction site, what rain attacks is the schedule, and the schedule is what holds the contract together.

A day of delay on a major project costs between US$ 50,000 and US$ 150,000, with an expected ROI of 10 to 15 times for those who anticipate it. Multiplied across a rainy season on a linear project over rugged terrain, the number moves from an operating line item to a structural one.

On large projects, landslides rarely go by that name. They show up as delayed earthworks, retaining-wall rework, a time-extension claim, an accident involving a crew in a cut or fill. The climate cause dissolves into the line items before anyone adds it up.

Whoever wants to support a time-extension claim for a climate event needs to prove two things at once: that the event happened, and that the company acted with diligence given what was foreseeable. Without a record of the alert, the decision, and the action, the claim turns into a word-against-word dispute with the client and the insurer.

What changes the contractor's position is the record. Every alert issued, every decision to suspend the excavation front, every preventive mobilization gets logged with time, data source, and the person responsible. The same document protects the crew, negotiates the extension, and supports the claim.

Highways: the landslide point has a kilometer and a probable window

A highway concessionaire rarely needs convincing that landslides are manageable. The protocol exists, the operations control center exists, the field crew exists. What changes the cost is how far ahead the protocol fires, because the trigger the operation receives is regional and the risk is local.

On mountainous terrain, the scale is already measured. According to a survey by Brazil's DNIT on the May 2024 event in Rio Grande do Sul, a 24 km segment of highway BR-470 recorded more than a hundred points of debris fall, with 8 points of complete road destruction, and highway BR-116 reached 26 total or partial closures in the same event.

The effect does not stay on the network. The CNT Climate Resilience survey found that 70.6% of Brazilian transport companies suffered financial losses from climate events over the last five years, and nearly one in four reported losses above R$ 1 million.

For operators that already have a protocol, the differentiation comes down to three concrete points:

Trigger by kilometer, not by region. Knowing the risk sits at kilometer 223, not in the neighboring town, is the difference between mobilizing one crew and mobilizing ten.
Accumulation when the failure is deep, intensity when it’s shallow. An alert that only looks at prior days’ accumulation misses the shallow slope that fails from the last few hours of rain, and an alert that only looks at rain right now misses the slope that was already saturated before it started raining today.
A record for the granting authority. The decision trail supports the conversation about availability and rebalancing far better than a screenshot from a weather app.

Railways: the segment about to shut down shows up before the ground moves

On a highway, vehicles can detour. On a railway there is no detour, and a blocked track halts the entire operation behind it.

Railways are the most mature sector on this list. Using a rainfall threshold to restrict traffic is already standard practice, and AI-based instrumentation already serves major Brazilian rail operators. The rail client's question stopped being whether anticipation works and became where the climate layer adds to what they already have.

The answer starts with the cost of getting it wrong. A rail interdiction costs between US$ 50,000 and US$ 200,000 per day of blocked track, with an expected ROI of 10 to 20 times for those who anticipate it.

Sensors, radar, and InSAR measure movement already underway, with high precision, right when the decision window has already shrunk. Hyperlocal forecasting anticipates the rain that will trigger that movement, and covers remote stretches where there is no hardware installed at every kilometer. The two layers together cover the event's entire timeline.

The other gain is in the output format. The control center operates on the probability of interdiction by stretch, with a window and a recommended action, at the level of the registered slope.

Power transmission and distribution: the tower on the slope, and the access road that disappears with it

On a railway, a landslide halts traffic. On the power grid, it brings down the asset and, in the same movement, blocks the road that would get the crew there.

Nearly half of transmission-line outages in Brazil originate from climate events, according to a sector survey. The cost of an unanticipated event in energy is estimated at US$ 40,000 per day, with an expected ROI of 6 to 12 times, and that figure still does not capture the impact on regulatory continuity indicators, which is where the outcome really hurts.

In transmission and distribution, a landslide gets recorded under another name: an outage. The tower sits on a slope, the right-of-way cuts through unstable terrain, the maintenance crew's access depends on a dirt road at high elevation, and the same accumulated rain threatens all three at once.

For utilities that already monitor climate, the gain lies in linking three decisions that today are made separately:

Asset exposure: which tower sits on a slope segment with a history of instability, registered by coordinate rather than by municipality name.
Crew access: the same accumulated rain that threatens the structure also blocks the access road, and that changes where the crew should be positioned before the event.
Preventive maintenance window: intervening on the right-of-way costs far less when scheduled than when it’s an emergency.

The result is preventive crew allocation and protected availability, with the alert reaching the channel the operations center already uses, from WhatsApp and email to Teams.

Insurance: exposure by address improves landslide pricing

The construction site closes the cycle of who suffers from landslides. The insurance sector looks at the same event from the other side of the table, and reaches the same conclusion by a different path.

Whoever reads exposure by address, with a history calibrated to the location, underwrites with a smaller margin of error and negotiates provisioning more effectively. Two properties a few kilometers apart, one at the foot of a slope with a history of failure and the other out of the mass's reach, fall into the same 25 km cell of a public model and get the same score. The average actuarial model cannot see that difference.

For an insurer, the cost of a single poorly-sized climate claim can exceed US$ 1 million, with an expected ROI of 25 to 60 times on the ability to anticipate.

Brazil's context amplifies the problem, and serves as a regional reference point: only 9% of the country's climate losses are covered by insurance, against 20% to 55% in developed countries, and coverage in the North and Northeast regions can fall below 2%. In 2024, 58% of climate-related claims were in the property line, precisely where structure, slope, and retaining works determine the damage.

Underwriting: physical risk by asset, at a resolution that matches the insured address, shifts pricing from a generic approximation to a specific reading. Subrogation and claims: when the insured has a record of the alert, the decision, and the action taken, the negligence discussion shifts from narrative to documented diligence, with a timestamp and a source.

A company that demonstrates active climate risk management, with a record of decisions taken ahead of each event, arrives at renewal with evidence instead of narrative. The same asset-level record organizes disclosure under the IFRS S2 standard.

The return on investment from anticipating the hazard

No system prevents landslides. What changes is the cost of dealing with them.

Mountain highway. Before i4sea, the most reliable forecasting method this mountain-terrain highway client used caught only 13% of landslides. In practice, 87% of events caught the operation by surprise, and the decision to close or reopen the stretch came only after the slope had already failed. With climate readings by slope and by stretch, rather than by regional bulletin, this client began anticipating 22 of 24 monitored landslides, a jump from 13% to 92% accuracy. Two cases still go without warning, against the 87% that used to be invisible. No climate model predicts 100% of a natural event, and it is worth distrusting anyone who promises that: what changes the game is the size of the blind spot. The question that matters is not whether risk still exists, but how much of that risk you are managing today.

The logic behind this gain has independent academic validation. An NBER working paper measured that improved hurricane forecast accuracy in the United States between 2007 and 2020 saved roughly US$ 5 billion per hurricane in damages and emergency spending, a 19% reduction in total cost. The hurricane kept happening. What changed was the quality of preparation.

According to NIST, a plan made with 2 hours of lead time costs 3 to 5 times more than the same plan made with 48 hours.

It is the same principle behind landslides: the slope will get the rain regardless; what varies is how much you pay for not having prepared.

The method: three steps to apply starting tomorrow

These steps work with a spreadsheet, an in-house system, or a vendor. None of them require hiring i4sea.

1. List the five assets that expose your operation the most, with coordinates and a threshold. Pick the five points where an unplanned shutdown hurts most: a pit slope, a cut on a mountain stretch, a track platform, a tower on a slope, or an earthworks front. For each one, record the exact coordinates, the slope's geometry, the history of prior events at that point, and the two rainfall thresholds that have already pushed that point to failure: the multi-day accumulation and the short-term intensity. That last piece of data almost always already lives in someone's head in maintenance, and was never written down. Write it down.

2. Tie every risk level to an action, an owner, and a cost. A limit with no owner never becomes a decision. For each asset, define three risk levels and tie each one to who gets alerted, which crew mobilizes, what restriction kicks in, what detour opens, how much it costs to execute that action, and how much it costs not to. "120 mm accumulated over 72 hours forecast for kilometer 223" needs to become "So-and-so positions the clearing crew the day before at 2 p.m." Done once, that table turns a decision made under pressure into a procedure.

3. Log the forecast, the decision, and the outcome, including when nothing happens. Recording events where you acted and nothing happened calibrates the threshold and prevents excessive preventive closures, just as much as recording events where you did not act and something did happen. Within six months that log becomes a calibration baseline, evidence for contracts and insurance, and an input for asset-level climate exposure reporting.

The payoff from these three steps is the same across all six sectors: the structurally lower cost of preparation replaces the cost of reaction.

Where this becomes a daily decision

A saturation threshold written into a procedure only protects the operation when the team gets the warning in time, on the channel it already uses, together with the recommended action.

The Climate AI Agent delivers the decision to the team ready-made before the event: which slope, which window, which action. It already knows the registered assets, thresholds, and history of your operation, reaches the channel your team already uses (WhatsApp, Teams, or email), and logs the source of the data behind every response.

Try the Climate AI Agent on your operation, with a free 14-day trial and no credit card required.

If you'd rather start with a diagnostic, we map the climate hazards affecting your assets and show you what the system would see today, using real data.

Request a free climate exposure diagnostic by asset.

The hours of warning your operation had, and what it did with them, is what decides the cost of the next shutdown.

Sources

Aon, via Revista Cobertura: natural disasters cost Latin America US$ 2.67 billion in Q1 2026, 20% insured.

Aon, "Climate and Catastrophe Insight 2026", via Punto por Punto: Mexico catastrophe losses over US$ 1 billion in 2025.

European Environment Agency (EEA), Economic losses from weather- and climate-related extremes in Europe: EUR 40.4 billion in economic losses in Europe in 2024 alone; less than 20% of total losses insured over 1980-2024.

Cemaden: the two landslide speeds and the 1-72 hour windows on the São Paulo coast.

Infobae, June 14, 2026: Mexico, Hidalgo, June 2026.

Dipartimento della Protezione Civile, Protezione Civile della Regione Siciliana, Il Sole 24 Ore: Niscemi, Sicily, Jan 2026.

El Desconcierto: Chile, Valdivia, July 7, 2026.

Brazil's DNIT: BR-470/RS and BR-116/RS, May 2024 event.

Molina, R. and Rudik, I., "The Social Value of Hurricane Forecasts," NBER Working Paper 32548 (2024). nber.org/papers/w32548

Source: i4sea, proprietary data (1-3 km resolution model for Latin America, calibrated with more than 10 years of climate history).

Source: real case from a mountain-terrain highway client, data provided directly by Mateus (CEO); client name withheld per his instruction.