The Hidden Risk in Poor Lidar Change Detection (LCD)

Why Quality and Confidence Matter in Geohazard Monitoring

In a recent article, we discussed how organizations are using repeat lidar acquisition to move beyond terrain snapshots and toward predictive monitoring. As repeat lidar becomes more common, the focus naturally shifts from identifying change to understanding what that change means.

That transition depends on confidence in the change detection results. High-resolution lidar data does not necessarily translate into high-confidence change detection, particularly when monitoring terrain movement occurring at the scale of centimeters over months or years.

In many geohazard applications, terrain movement occurs at the scale of centimeters over months or years. If the uncertainty in the analysis approaches the magnitude of the movement being measured, it becomes difficult to distinguish real terrain change from processing artifacts.

For organizations responsible for monitoring landslides, erosion, settlement, and other geohazards, that distinction has important operational consequences.

Why Confidence Matters

Low-confidence change detection creates more than technical uncertainty. It can lead to:

• chasing false positives
• missing subtle acceleration trends
• unnecessary field mobilization
• difficulty prioritizing high-risk areas

These challenges become more significant when monitoring subtle movement over large geographic areas where field verification is limited and resources must be prioritized carefully.

The objective is not simply to detect change. It is to determine whether the observed change is meaningful and actionable.

Understanding the Limit of Detection

Every lidar change detection method has a practical limit of detection.

The limit of detection represents the minimum amount of movement that can be confidently identified above the background uncertainty of the analysis. If movement falls below this threshold, it becomes increasingly difficult to determine whether observed changes represent actual terrain movement or processing noise.

This concept is particularly important for monitoring:

• slow-moving landslides
• progressive slope deformation
• settlement
• erosions
• embankment movement

Improving the limit of detection improves the ability to identify subtle movement earlier and track changes over time with greater confidence.

Why Some Change Detection Results Fall Short

Lidar change detection is conceptually straightforward:

1. Acquire lidar datasets at different points in time
2. Compare the datasets
3. Quantify the differences

In practice, the quality of the results depends heavily on both the source data and the processing workflow.

Common sources of uncertainty include:

• spatial misalignment between surveys
• GPS, georeferencing or registration error
• vegetation noise
• rasterization effects
• slope-related distortion in steep terrain

Even relatively small inconsistencies between datasets can introduce apparent change that does not reflect actual terrain movement.

As monitoring programs increasingly focus on subtle movement, these sources of uncertainty become more important.

Limitations of Traditional Surface Differencing

Many traditional workflows rely on DEM of Difference (DoD)analysis, where point clouds are converted into raster surfaces and compared over time.

DoD methods can be highly effective for identifying large-scale terrain change. However, rasterization simplifies the original point cloud geometry and can introduce uncertainty when monitoring subtle movement.

In steep terrain, even minor horizontal misalignment between datasets can produce significant vertical differences. These apparent changes may be comparable to, or larger than, the movement being measured.

For applications involving slow-moving landslides, settlement, or gradual erosion, this can reduce confidence in the results and limit the ability to identify meaningful trends.

Why Advanced 3D Methods Improve Confidence

Modern 3D point-cloud comparison methods address many of these limitations by working directly with the original lidar data rather than relying solely on rasterized surfaces.

Advanced workflows may incorporate:

• Iterative Closest Point (ICP) alignment
• Multiscale Model to Model Cloud Comparison (M3C2)
• local surface normal analysis
• uncertainty-aware comparison methods

These approaches improve the ability to:

• reduce alignment error
• preserve terrain detail
• minimize false change
• improve the limit of detection
• distinguish meaningful movement from background uncertainty

The result is a more reliable understanding of whether observed terrain change reflects actual ground movement.

From Visualization to Defensible Decisions

Lidar change detection has become an important tool for geohazard monitoring and infrastructure management. However, the value of the analysis depends on its ability to distinguish real terrain movement from processing uncertainty.

As organizations increasingly rely on repeat lidar for monitoring programs, attention is shifting from simply detecting change to understanding the confidence associated with that change.

When monitoring subtle ground movement, confidence is not a secondary consideration. It is a prerequisite for defensible engineering decisions.

‍