Editor’s Note: Lidar is helping researchers better understand landslides along California’s coastline. A four-year monitoring program combined weekly laser scanning with drone photogrammetry to detect thousands of cliff failures, showing how repeated 3D surveys can improve hazard assessment and support future early warning systems.
California asked whether we can see a coastal collapse coming….four years of weekly laser scans say we mostly can.
For anyone who lives and breathes point clouds, the part of California’s new coastal landslide research worth getting excited about involves a rather repetitive and otherwise routine activity: pointing a lidar scanner at the same crumbling cliffs, week after week, for four years.
That repeat mapping is the backbone of a March 2026 report out of the Scripps Institution of Oceanography at UC San Diego. Led by coastal geomorphologist Adam Young, the team was tasked by the state of California, through Assembly Bills 66 and 72, with figuring out whether a cliff collapse can be forecast. Turns out it mostly can, and topographic mapping is doing a lot of the heavy lifting.
The mapping setup
From 2022 to 2025, the group ran weekly scans of five cliff sections along about 12 miles of San Diego County coast, from Torrey Pines State Beach up to Encinitas. They used mobile lidar systems and drone photogrammetry, and the resolution came out in inches. Do that every week for a few years and the numbers stack up fast. The repeated surveys added up to roughly 2,500 miles of coverage along the shore.
The real payoff is change detection. By differencing point clouds from one survey to the next, the team could pinpoint exactly where the cliff lost material, when, and how much. They flagged about 4,300 erosion events this way. The workflow is semi-automated, leaning on machine-learning segmentation paired with 3D change detection, with every event then checked by hand.
From elevation change to a forecast
Stack all of this lidar data against rainfall records and patterns emerge. Big failures cluster in winter when it rains, and the report puts a preliminary number on it. After a day with more than an inch of rain, there is roughly a 70% chance of a large cliff failure somewhere in the study area within four weeks. Smaller events, ranging from 6 to 50 cubic yards, happen constantly, even in dry stretches. That statistical picture is the foundation for a regional, probabilistic warning, as distinct from the site-specific tiltmeter alerts the report also covers.
Lidar earned its keep in the forensic work too. When an estimated 200 tons of cliff dropped onto the Del Mar beach at 5 a.m. on April 21, 2024, scans taken just before and just after captured the full extent of the failure. Buried tiltmeters had flagged the accelerating ground motion days earlier, but the before-and-after point clouds are what quantified what actually came down.
There is an older archive worth noting as well. For a separate runout study, the team drew on more than 700 lidar and photogrammetric surveys collected between 2002 and 2023 across a 19-mile stretch, mapping how far debris travels once it hits the sand. The mean runout was about 19.7 feet, and 90% of cases stayed within 43 feet of the cliff base. Handy if you are the one drawing hazard zones on a map.
The report is careful to label the rainfall thresholds preliminary, and it wants the mapping pushed out to places like Palos Verdes, Big Sur, and Santa Cruz. But the take away is hard to miss. Keep scanning, keep differencing, and the elevation data itself starts to behave like an early warning system.
Source: Young, A. et al., “California Coastal Landslide Early Warning Research,” Scripps Institution of Oceanography, UC San Diego, March 2026.
Discussion Questions
For Professionals
- How could repeated lidar surveys improve hazard monitoring or asset management in your projects?
- Where do you see the greatest value for long-term change detection using repeat laser scanning?
For Students
- What challenges do you think repeated lidar mapping can solve that a single survey cannot?
- How might machine learning and point cloud analysis change the future of landslide monitoring and environmental mapping?
