A landslide early warning system should combine at least four layers: rainfall or another forcing, subsurface water conditions, actual slope movement, and data freshness. No single series is enough for prediction. An alarm should trigger a pre-agreed verification and action, not automatically declare that a landslide will occur.
In short
- Rainfall is the forcing, pore pressure describes the hydrologic response, and tilt or displacement shows the mechanical response of the slope.
- Rainfall thresholds transferred from another site can fail. What matters is geology, initial moisture, drainage, history and the mechanism of the specific landslide.
- A matrix of several signals gives a better basis for response than one red point, but it still requires engineering judgment and field observation.
- NO_DATA is a warning state. No movement on the screen means nothing when the measurement is stale or interrupted.
- It is worth testing the system first on scenarios: intense rain, pressure rise, movement without rain, and loss of a critical channel.
The road manager below the slope does not need another rain chart. They need decision time: send a patrol, restrict traffic, stop trains, withdraw people from the work area, or ask a geotechnical engineer for an urgent assessment. The biggest risk is that the panel shows one clear number and creates false confidence. It rained heavily, so "there will be a landslide." It did not rain, so "everything is calm." The inclinometer did not move, although the last reading came yesterday.
USGS indicates that simple intensity-duration rainfall thresholds describe landslide initiation conditions with high uncertainty because they do not capture the full ground state. Field studies combine rainfall with moisture and pore water pressure. In turn, the Polish Geological Institute uses pairs of piezometric and inclinometer boreholes because water conditions and deep movement answer two different questions. That is a sound design basis: you are not looking for one magic variable, but for a sequence of evidence.
What "early warning" means for a landslide
A landslide early warning system is a combination of measurements, evaluation rules, communication and actions that is intended to detect an adverse change early enough for a person to react. The system is not just a sensor, an app or a model. Without a post-alarm procedure, it remains only an event logger.
"Early" depends on the mechanism. A slow-moving landslide may give weeks of observable movement. A shallow debris flow after intense rain can develop quickly. A large deep-seated mass may react to rainfall with a delay, and movement can continue after the rain stops. For that reason, measurement frequency and response time are selected for the scenario, not for transmission convenience.
First describe the working model:
- where the slip surface is located or may develop;
- which layers and zones transmit or retain water;
- what process initiates movement: prolonged rain, snowmelt, toe erosion, works, drainage change, loading;
- where the first changes in pressure, tilt or displacement are expected;
- which infrastructure elements and people are exposed;
- how much time the organization needs to carry out an action.
ISO 18674-1 defines the key parameter as a physical quantity indicating the geotechnical issue under consideration. That order matters. You do not start by asking "which sensors should I buy?", but "what phenomenon and at which location do I need to observe?". A general geotechnical monitoring program is built according to the same principle.
Four evidence layers instead of one threshold
Each layer serves a different role. They should not be swapped or added together without an engineering model.
| Layer | Example data | What question it answers | Main limitation |
|---|---|---|---|
| forcing | rainfall, snowmelt, earthworks | what may have triggered the change | regional rain does not describe ground conditions |
| hydrology | pore pressure, water level, moisture | how water changes slope conditions | a point measurement may miss the flow path |
| mechanical response | tilt, displacement, deep profile | whether the ground is actually moving | local movement does not describe the whole mass |
| data quality | freshness, gaps, completeness, time | whether the other layers can be trusted | a complete series may still come from a faulty installation |
Rainfall: an important initiator, a weak standalone alarm
Intense or prolonged rain can raise pore pressure and reduce effective stress. Yet the same rainfall total can produce a different response after a dry or wet winter, on a slope with working drainage and one with blocked drainage, or in two different soils. USGS stresses that moisture and water pressure complement rainfall thresholds because they show the actual subsurface response.
Local rainfall has a different role than meteorological warning. A rain gauge on the slope measures conditions at the site. Public data provide context and regional forecast. If the system uses both, label them separately together with the update time.
Pore pressure: the link between water and stability
A piezometer measures pressure or piezometric level in a defined zone. A rise may precede movement, but its meaning depends on the position of the tip relative to the layers and the potential slip surface. No increase at one point does not rule out a local flow path elsewhere.
The most useful information is not the level itself, but the response to rainfall: start time, rate, maximum, decay time and comparison between depths. If the rain has already stopped, and pressure remains high or is still rising, the condition may require more attention than at peak rainfall intensity.
Tilt and displacement: evidence of mechanical response
A surface tilt sensor shows local rotation. A deep inclinometer helps determine the depth at which displacement is concentrated. PIG describes the summation of displacements from the bottom of the borehole upward and emphasizes that the bottom should be in stable ground below the slip surface. If the reference point itself moves, the entire profile may be misinterpreted.
Surface measurements, deep measurements and visual inspection complement one another. Tension cracks, ditch deformation, leaning trees, seepage and pavement settlement can provide information that is not captured exactly under the sensor.
Freshness: the layer that can invalidate all the others
Every alarm card should show the time of the last valid reading. If the device reports every 15 minutes by default, or more often, then a series from three hours ago is not "flat" - it is outdated. The timeout is selected according to the actual cadence, criticality and tolerated uncertainty window. The seven data SLA metrics article shows how to separate application availability from measurement availability.
Rain × pressure × movement × freshness matrix
The matrix does not replace stability calculations or thresholds approved for the asset. It organizes the first response. The states below are illustrative and cannot be used as automatic criteria for closing a road, rail line or excavation.
| Rain / forcing | Pore pressure | Slope movement | Data | Working interpretation | First action |
|---|---|---|---|---|---|
| rising | no unusual response | stable | fresh | forcing without confirmed response | observe more often, compare initial moisture |
| rising or stopped | rising consistently at several points | stable | fresh | hydrologic response is developing | increase readiness, check drainage and procedure |
| stopped | remains high | increasing tilt or displacement | fresh | hydrology and mechanics support the same scenario | urgent assessment, field observation, action by authorized person |
| no significant rain | no change | sudden jump at one point | fresh | local movement or measurement issue | preserve evidence, verify the point and adjacent series |
| any | any | no visible movement | critical gap | unknown state, not "stable" | trigger the NO_DATA procedure and a fallback measurement |
The matrix has two safeguards. First, it does not allow calm to be declared on the basis of one green indicator. Second, it does not force a single spike to be treated automatically as a landslide. In both cases, it leads to a set of evidence and action.
Response procedure: the first 15 minutes and the next steps
An alarm is the start of work. The team should know in advance who has access to the data, who may enter the field, who contacts the traffic manager and who has the authority to make the decision. If the procedure is created only after the alert, the most valuable time is lost on phone calls.
Minutes 0-5: confirm what actually arrived
Read the measurement time, freshness status, unit and location. Determine whether the exceedance concerns a physical value or a data absence. Compare the raw trace or source reading with the calculated value. Do not change the reference or threshold before securing the current view and context.
Minutes 5-15: look for spatial and process consistency
Overlay rainfall, pore pressure and movement on one shared timeline. Open adjacent points and the second measurement direction. Look for the sequence of events: forcing, hydrologic response, mechanical response. Assess whether the series behaves technically reliably and whether other critical channels are current.
After 15 minutes: execute the action from the plan
This may be a patrol in a safe zone, drainage inspection, an additional measurement, consultation with a geotechnical engineer, a restriction, or closure of the exposed section. The choice must follow a pre-approved plan. The platform can provide the alarm and evidence, but it does not take over operational responsibility for the decision.
Illustrative example. Assume that after 36 hours of rain, the level in two piezometers begins to rise with a delay of several hours. Tilt remains within the previous range. The team switches to enhanced observation mode. After the rain stops, pressure continues to rise, and two adjacent tilt points change in the same direction. This is not automatic proof of an imminent landslide, but it is a set of indicators for urgent assessment and for carrying out the action planned for that scenario.
Early warning system design checklist
- [ ] The mechanisms and hazard zones for the specific slope have been described.
- [ ] The time needed for a patrol, traffic restriction or evacuation has been defined.
- [ ] Each measurement answers a documented geotechnical question.
- [ ] Local rainfall and regional data have separate labels and roles.
- [ ] Piezometers are assigned to layers and potential flow paths.
- [ ] Movement points cover the expected activity area and a stable reference.
- [ ] The dashboard combines forcing, hydrology and movement on a shared timeline.
- [ ] NO_DATA alarms are configured for critical data.
- [ ] Thresholds have a source, unit, owner and assigned action.
- [ ] The procedure lists contacts and alternates outside working hours.
- [ ] A spike scenario, delayed post-rain response and transmission loss have been tested.
- [ ] Threshold review is scheduled after the season, works or drainage change.
What this looks like in Inclify
Inclify accepts automatic measurements from loggers via HTTP/JSON and stores them in UTC. On one dashboard, you can combine pore pressure, tilt or displacement, and temperature on two axes. Weather widgets and IMGW warnings provide external context; they do not replace a local rain gauge or the approved source for a critical decision.
Project equations convert raw values into engineering units. For each channel, OK / WARNING / ALARM states with hysteresis are available, and the NO_DATA alarm watches for interruptions in data flow. Notifications can be sent by email and SMS. Acknowledgement records the person and time, while silencing always has an expiry time.
The platform does not include an alarm for rate of change or an autonomous landslide prediction function. That process must be built from value thresholds, dashboard comparisons and human procedure. The data quality report shows cadence, completeness, gaps and freshness. This makes it possible to distinguish the absence of an observed change from the absence of a reliable observation.
Many dashboards allow the duty view to be separated from the geotechnical analysis view. Alarm configuration changes leave an audit trail, but the platform still does not replace field inspection, a geotechnical model or stability calculations. Responsibility for interpreting the combined signals and selecting the action remains with the authorized team.
Limitations: when monitoring is not enough
Point sensors observe selected locations. The slip surface may develop outside the borehole, and local flow may bypass the piezometer. The design requires geological investigation, a geotechnical model, field inspection and an assessment of whether the sensor layout actually covers the mechanism. More sensors will not fix a wrong model.
Statistical and historical thresholds are not automatically safety limits. Slope conditions may change after works, vegetation growth or drainage damage, toe cutting, or an exceptional weather sequence. Criteria must be reviewed periodically by a competent person. The platform does not replace stability calculations or the infrastructure manager's decision.
Telemetry data may be preliminary. Clock, power, transmission, reference or installation errors can mimic movement. For this reason, the procedure starts with data quality, and configuration changes leave an audit trail. The organizational model is described in detail in the article on the alarm response plan.
FAQ
Is a rain gauge enough to warn of a landslide?
No. Rainfall describes the forcing, but it does not directly show how much water reached the relevant layer or whether the slope is moving. Initial conditions, infiltration and drainage change the response. It is worth combining a rain gauge with pressure or moisture measurement and with surface or deep movement, depending on the mechanism.
What measurement frequency is appropriate?
The one that leaves time for detection, transmission, verification and action before the scenario under consideration develops. For slow movement, a less frequent reading may be enough; for a rapid response, a higher cadence is needed. The choice requires knowledge of the mechanism and the organization's response time. The platform itself accepts different intervals.
Does an increase in pore pressure mean the slope will start moving?
Not automatically. The meaning depends on the measurement location, soil properties, geometry, stresses and boundary conditions. An increase is an important indicator, especially when it has an unusual value or pattern and is accompanied by movement. A geotechnical engineer assesses it in the context of the model, slope history and other data.
How does an inclinometer differ from a surface tilt sensor?
A borehole inclinometer allows displacement to be observed with depth and helps localize the zone where movement is concentrated. A surface sensor records a local change in orientation at the installation point. The methods answer different questions and can complement one another. Neither one automatically describes the whole landslide mass without an additional model.
How should data loss be treated during intense rainfall?
As a loss of visibility during a period of elevated risk, not as stability. The NO_DATA alarm should trigger power and transmission checks, assessment of fallback measurements and the field action defined in the procedure. If a critical channel is unavailable, the decision must account for greater uncertainty and field observation limits.
Can the platform decide on its own to close a road or rail line?
It should not. It can detect an exceedance, a data gap and deliver a set of charts and a notification. The decision is made by an authorized person in line with the manager's plan, taking into account field conditions, the importance of the infrastructure and the consequences of both possible errors. Automating the message does not transfer responsibility to the platform provider.
Sources and further reading
- Polish Geological Institute - Inclinometric measurements, deep landslide monitoring - piezometer-inclinometer pairs, zero measurement and profile interpretation.
- USGS - Overview of Rainfall-Induced Landslides - limitations of simple rainfall thresholds and the role of hydrologic field monitoring.
- USGS - Real-Time Monitoring for Potential Landslides - rainfall, moisture, pore pressure, movement and near-real-time transmission.
- USGS Fact Sheet 2012-3008 - How do we monitor landslides? - selecting measurements for field conditions and the role of real-time systems.
- ISO 18674-1:2015 - Geotechnical monitoring by field instrumentation - objectives, representative parameters, plan, acceptance and reference measurements.
- FHWA NHI-14-007 - Soil Nail Walls Reference Manual - official guidance on observation and instrumentation for earth structures.
What next
Start with one exposed section. Lay out the sequence "rainfall - pressure - movement - action", identify the largest data gap, and run a NO_DATA drill during a hypothetical downpour. Book a conversation with the Inclify team if you want to connect existing piezometers or tilt sensors and build a pilot dashboard without promising automatic prediction.