After the burn: using USGS debris-flow estimates to pick which basins to survey first

A fire is out, the first storm is on the forecast, and you maintain more debris basins than you can inspect in a week. Which ones do you survey first?

The U.S. Geological Survey publishes an answer you can download for free. For recent fires it models, catchment by catchment, how likely a debris flow is and how much sediment it could deliver. That is a ranking. It is not a capacity check, and the gap between the two is where a survey earns its place.

This post walks through the USGS data for the January 2025 Eaton Fire and turns it into a survey order. It also says what the numbers cannot tell you.

What the USGS assessment gives you

For selected western fires, the USGS runs a post-fire debris-flow hazard assessment. Its program record says it uses geospatial data on "basin morphometry, burn severity, soil properties, and rainfall characteristics" to estimate the probability and volume of debris flows for a design storm.

Each fire has a downloadable data release on ScienceBase (the Eaton one is linked here). The GeoJSON and shapefile packages hold a few layers:

• Basins: catchments delineated to an outlet, one row each.
• Outlets: the outlet point for each basin.
• Segments: stream reaches, with the same fields.
• Perimeter: the burn boundary the model used.

For each design storm the data carry three results. Design storms are set by peak 15-minute rainfall intensity (I15). For Eaton they run from 16 to 40 mm/h in steps of 4, per the data specification.

Volume also comes with a 95% confidence interval (Vmin and Vmax). The Eaton metadata names the models behind these: Staley (2017) for likelihood, Gartner (2014) for volume and Cannon (2010) for the combined class.

Convert volumes to cubic yards before you put them next to a basin's design capacity. One cubic meter is 1.308 cubic yards. Every volume below is converted that way.

Worked example: the Eaton Fire

CAL FIRE's incident page lists the Eaton Fire at 14,021 acres, started January 7, 2025. The USGS output for it has 90 basin catchments. Ranked by modeled volume at I15 = 32 mm/h, they look like Figure 1. That is the value Chen, Geyman and Lamb chose for their Eaton comparison, so the two analyses use the same storm.

The volume is concentrated. The ten largest catchments hold 60% of the total modeled volume across all 90, and the top three hold 32%. Treat those shares as a guide to concentration only. Many of these catchments drain to channels, not to a maintained basin, so the total is not a forecast for your facilities.

Table 1. Top ten Eaton catchments by modeled volume, I15 = 32 mm/h. Source: USGS PWFDF Eaton v1.0, basins layer, accessed October 17, 2026. Volumes converted from cubic meters.

Three things stand out.

Burn severity is not the whole story. Segment 355 ranks third, yet only about 20% of its catchment burned at moderate or high severity. Its size and relief carry it. The Gartner volume model uses watershed relief and the area burned at moderate or high severity, so a big, steep catchment can outrank a smaller, harder-burned one. This reading is an inference from the fields in the data, not a USGS statement.

Likelihood barely separates the catchments. At 32 mm/h, 81 of the 90 catchments have a modeled likelihood of 50% or more, and 63 land in the highest hazard class. When almost everything is likely to flow, volume is what ranks.

The order does not change with the storm. The volume model scales every catchment by the same rainfall term. Ranking at 16 mm/h or 40 mm/h gives the same list; only the volumes change. You do not need to argue about which design storm to pick before you choose a survey order. You do need it later, when you compare a volume with a basin's remaining room.

The interval is wide, and that matters

The 95% interval in the Eaton output runs from 0.13 to 7.7 times the best estimate for every catchment with a nonzero estimate (89 of the 90). For segment 1111 that is roughly 21,000 to 1.2 million cubic yards around a best estimate of 159,000.

Two cautions come from the published research, not from the USGS output itself.

• Chen and colleagues surveyed 14 Eaton-area debris basins with UAV LiDAR after the February 2025 storm. They report that the USGS volume model "tends to overpredict debris volumes in basins with larger drainage areas" and performs better in smaller ones.
• The same paper notes that volumes measured right after a storm ran about 15% high because of ponded water in the basins.

So use the model for order, not for a pass or fail on one basin. The ranking tells you where to look first. It does not tell you whether a given basin will overtop.

What storm is the model using?

The design storms are not rare events. NOAA Atlas 14 gives, at 34.19 N, 118.10 W in the Eaton burn area, a 15-minute intensity of 34 mm/h for the 1-year recurrence interval and 48 mm/h for the 2-year. The Eaton design storms run from 16 to 40 mm/h.

By Atlas 14's measure these are frequent storms. At 32 mm/h, just under the 1-year intensity at this point, the model puts 81 of the 90 Eaton catchments at a 50% or higher likelihood of a debris flow. At the lowest design storm, 16 mm/h, 12 of them reach that level. In other words, a freshly burned catchment does not need an exceptional storm to produce one. This is a single point and a mean estimate, so read it as scale, not as a site value for any one basin.

From a ranking to a survey list

Here is a simple way to use the data.

1. Download the basins layer for your fire and convert V to cubic yards.
2. Join each catchment to the basin or channel feature it drains to. The USGS catchments are not your asset IDs, so this join is yours to build.
3. Sort by modeled volume at your chosen design storm. Keep the likelihood and hazard class in view.
4. Cut the list where the volume is too small to threaten the basin below it.
5. Survey the basins at the top first.

Agencies already do a version of this. Andrew Roth, environmental manager at Santa Barbara County Flood Control, describes the routine on a public video (auto-captions): "we go through a triage and prioritization process with all of those sites that we had noted from the inspections," and then they produce "literally a binder with that year's work plan." The USGS ranking is one more input to that triage.

Timing is tight. Roth says they "aim to finish this routine maintenance uh typically by December," before the rainy season (video, auto-captions). A post-fire survey list has to fit inside that window.

Why modeled volume is not capacity

A modeled volume says how much could arrive. Capacity says how much room is left. Only a survey of the basin gives the second number.

Hypothetical example. Here is a made-up basin to show how the two combine. None of these numbers come from a real facility or from the Eaton data.

• USGS-style best estimate for the storm: 15,000 cy.
• Design capacity: 20,000 cy.
• Sediment already in the basin, from a survey: 6,500 cy.

On paper, 20,000 cy of capacity covers a 15,000 cy event. With 6,500 cy already inside, the remaining room is 13,500 cy. The best estimate now exceeds it. Add the interval from the Eaton output (0.13 to 7.7 times the estimate) and the range runs from about 2,000 cy to well over 100,000 cy. One number cannot say the community below is protected.

Cleanout rules exist for this reason. LA County Public Works says debris basins are cleaned out once they are 25 percent full in unburned watersheds and 5 percent full in burned ones. A watershed counts as burned if more than 20 percent of its area burned in the previous five years. On a 20,000 cy basin, 5 percent is 1,000 cy. The made-up basin above is at 32.5 percent.

A Santa Barbara County speaker puts the goal in field terms: "approximately 1500 truckloads or 15 000 cubic yards of materials that got this Basin back to empty" (video, auto-captions; speaker identified as Walter in the captions). Back to empty is a number you can only confirm with a measured surface.

What the survey has to deliver

For a basin, the useful products are modest.

• Current surface and remaining volume. A bare-earth surface of the basin floor and side slopes, compared with the as-built or design grade.
• Stage-storage check. How much volume sits below the spillway or outlet invert, by elevation.
• Change since the last survey. The volume added since the previous clean-out or flight, with fill and cut kept separate.
• A stated accuracy. Say which standard and which checkpoints back it.

Ask for vertical accuracy the way ASPRS Positional Accuracy Standards, Edition 2 frames it: RMSEv, checked against at least 30 independent checkpoints. A statement with fewer than 30 points is a weaker claim.

Published numbers give a range for what to expect from a UAV LiDAR sensor.

• DJI specifies the Zenmuse L2 system at 4 cm vertical and 5 cm horizontal at 150 m, measured in a laboratory.
• An independent test by Tamimi and Toth (ISPRS 2024) found about 7 cm RMSE on an open road and about 21 cm under forest canopy.
• Chen and colleagues report 2 to 3 cm RMSE over bare ground at the Eaton basins, with 500 to 700 points per square meter.

A basin floor is bare ground, so the better end of that range is the realistic one. The hillsides above it are different. Burned slopes can still hold standing snags and unburned patches, so check ground-point density before trusting a slope map there (inference, not a measured result).

Accuracy turns into volume as area times elevation error. Take a made-up 1 ha basin floor. A uniform 5 cm bias is 500 m³, or about 650 cy. At 7 cm it is about 915 cy. That is roughly the size of the 5 percent trigger on a 20,000 cy basin. Real errors are not uniform and partly cancel, so treat this as a ceiling. It still shows why the checkpoint report matters when you are close to a trigger.

Who signs

If the surface and volumes feed a design, a permit or a clean-out contract, a licensee needs to be in responsible charge. In California that is a PLS or a civil engineer legally authorized to practice land surveying, per B&P Code section 8726 and the Board's Bulletin 50 on UAVs and land surveying. Twinn captures and processes the data and does not stamp. A licensed partner does.

If federal money, such as a FEMA grant, will pay for the survey, check SAM.gov for the status of any proposed sensor before you write the scope.

Limits of this approach

• The USGS output is an emergency assessment. The ScienceBase record says results "are generally representative of the conditions immediately after the fire." Vegetation recovery changes the hazard, and a later version may exist.
• The catchments are not your basins. Matching them to facilities takes local knowledge.
• The model does not know what is already in your basin. Only a survey does.
• The interval is wide, and published checks show the model can overpredict in larger catchments.
• This post uses the Eaton data because it is complete and public. Other fires differ in size, soils and rainfall, so do not read these volumes across.

Next step

Pick the most recent burn upstream of your facilities and download its USGS data release. Rank the catchments by volume, join them to your basin list, and mark the top five. Then decide whether each of those basins has a current surface, and if not, schedule one before the first storm.

Twinn Digital rents LiDAR equipment per job and is equipment-agnostic. Flights are planned with software agents and flown by a human Part 107 pilot, and the work is $2M insured. For what a debris-basin survey involves, see Twinn's debris basin page.