What the National Bridge Inventory says about California, Georgia and Florida bridges, and where drone data fits
November 19, 2026

Every public highway bridge in the United States carries a record in the National Bridge Inventory (NBI). Each record includes condition ratings that a certified inspector assigned on the last visit. That makes the NBI one of the few infrastructure datasets where you can see, structure by structure, what the inspection system already knows and what it cannot easily see.
This post reads the 2025 NBI files for California, Georgia and Florida. It shows where Poor ratings concentrate, how often bridges are inspected, and how many bridges over water have no scour evaluation on record. Then it asks a narrow question: where does drone-collected survey data supplement the inspector, and where does it not?
One position up front. Drone data does not replace the inspection. The inspector's rating is the record. Survey data can add measured geometry between ratings, and that is useful when it is scoped honestly.
What the NBI records, and how Poor is defined
The Federal Highway Administration (FHWA) publishes the inventory as delimited files by state. States submit data each spring and FHWA archives the public highway bridge data on June 15 of each year, per FHWA's table notes. The files used here cover highway bridges as of that archive date.
Three condition items matter for this post: Item 58 (deck), Item 59 (superstructure) and Item 60 (substructure, which includes abutments, piers and footings). Each is rated 0 to 9. The NBI Coding Guide calls a 7 "good condition, some minor problems," a 5 "fair condition," and a 4 "poor condition, advanced section loss, deterioration, spalling or scour."
FHWA's Good, Fair and Poor classes follow the lowest of those three ratings (Item 62 for culverts). Under 23 CFR 490.409, a bridge is Good when the lowest rating is 7, 8 or 9, Fair at 5 or 6, and Poor at 4 or below.
Two cautions about reading the classes:
- Poor is a condition class, not a closure order or a load posting. It means at least one component is rated 4 or below.
- A single weak component sets the class. A bridge with a 7 deck and a 4 substructure counts as Poor.
The counts here come straight from the 2025 files. The totals were checked against FHWA's Bridge Condition by Owner table. They match for all three states.
| State | Highway bridges | Good | Fair | Poor | Poor share |
|---|---|---|---|---|---|
| California | 25,975 | 12,239 | 12,452 | 1,284 | 4.9% |
| Georgia | 15,090 | 10,876 | 3,951 | 263 | 1.7% |
| Florida | 13,070 | 7,705 | 5,005 | 360 | 2.8% |
Source: FHWA NBI 2025, counted from the state delimited files; totals match FHWA's owner table.
California's inventory is also old. Using Item 27 (year built) and 2025 as the reference year, 65.9% of California's highway bridges are at least 50 years old. The figures are 47.4% for Georgia and 41.4% for Florida. California's median year built is 1968.
Poor ratings climb with age
Age tracks the Poor share closely in all three states. Figure 1 groups each state's bridges by age in 2025 and shows the Poor share in each group. Counts come from the NBI 2025 state files.

In California, 0.9% of bridges under 25 years old are Poor. For bridges 75 and older the share is 9.4%, or 414 of 4,414 structures. Georgia's oldest group is 5.4% (99 of 1,832) and Florida's is 7.9% (56 of 707).
Two things stand out. The oldest California band is large: 4,414 bridges are 75 or older, and 12,701 more are 50 to 74. The population heading into the age bands with higher Poor rates is the bulk of the inventory, not a tail.
Age is also not destiny. Most bridges 75 and older are still Fair or Good in every state. The ratings are what separate the structures that need attention from those that do not, which is why the quality of the ratings matters.
Owners differ: county bridges lead the Poor share
Most readers of this post work for a local agency or a consultant to one. The owner split matters for you.

In California, 6.8% of county-owned bridges are Poor (501 of 7,326), against 3.4% for the state highway agency (418 of 12,476) and 5.2% for cities (260 of 5,042). Florida's county share is 6.1% (245 of 4,042) and Georgia's is 2.8% (197 of 7,128). The pattern in the FHWA owner table is the same.
The data cannot say whether it reflects budgets, traffic, age, materials or location. What it does say is that county owners, not the state highway agencies, carry the highest Poor share in all three states.
How often bridges get inspected
The National Bridge Inspection Standards set the cycle. 23 CFR 650.311 says regular routine inspections must not exceed 24 months. Bridges with a deck, superstructure, substructure or culvert rated 3 or less, or with serious scour, must be inspected at least every 12 months. Under the rule, certain bridges meeting all of its criteria "may be inspected at intervals not to exceed 48 months," including components all rated 6 or greater.
Here is what the 2025 files record in Item 91 (inspection frequency, months):
| State | Highway bridges | 12 months or less | 13 to 24 months | 48 months |
|---|---|---|---|---|
| California | 25,975 | 236 | 19,884 | 5,855 (22.5%) |
| Georgia | 15,090 | 5 | 15,078 | 7 |
| Florida | 13,070 | 4 | 13,060 | 6 |
Source: FHWA NBI 2025, Item 91, counted from the state files. No bridge in the three files is coded between 25 and 47 months.
In California, 5,855 of 25,975 bridges (22.5%) are recorded on a 48-month interval. Georgia records 7 and Florida 6. Everything else is at 24 months or less.
So in California a fifth of the inventory can go up to four years between routine inspections. Whatever happens to the structure in that window is seen next time, if at all, as a change in a 0-to-9 rating. This is the data only, not each owner's policy. The files do not say why Georgia and Florida record almost none.
Scour: where the inventory has a gap
For bridges over water, Item 113 records scour vulnerability. The Coding Guide says a bridge coded 3 or below is scour critical, and "U" marks a bridge with an unknown foundation that "has not been evaluated for scour. Since risk cannot be determined, flag for monitoring during flood events."
Figure 3 and the table below use every bridge whose Item 113 is not "N" (not over a waterway).
| State | Bridges over water | Scour critical (0 to 3) | Unknown foundation (U) |
|---|---|---|---|
| California | 17,203 | 599 (3.5%) | 841 (4.9%) |
| Georgia | 12,619 | 82 (0.6%) | 4,470 (35.4%) |
| Florida | 9,275 | 343 (3.7%) | 1,335 (14.4%) |
Source: FHWA NBI 2025, Item 113, counted from the state files.

A "U" is not a defect. It is a missing answer. Georgia's 4,470 bridges coded U are a larger group than its 263 Poor bridges, and the Coding Guide's instruction is to monitor them during floods.
That is the gap where survey data is most plainly useful, and it is covered below.
What the inspector sees, and what survey data adds
An FHWA tech brief on small UAS for bridge inspection is a good neutral reference. It reports that in a March 2018 AASHTO survey, 35 of 44 reporting state DOTs had experience deploying UAS in some capacity. It lists the possible gains as closer access to elements "traditionally viewed through binoculars," improved inspector safety, and fewer lane closures.
It is also direct about limits: UAS "are not currently acceptable as a replacement for hands-on inspections" for fracture-critical work, though they can supplement them with images of difficult-access areas. That brief dates from 2019, so check current FHWA and state DOT guidance for your agency.
The rule sets who is responsible. 23 CFR 650.313 requires at least one qualified team leader at the bridge who "actively participates in the inspection at all times" for routine, in-depth, underwater and special inspections. A drone flight cannot satisfy that requirement.
A city official in Port Orchard, Washington said the same thing about a different inspection type. Stormwater program manager Zach Holt said that "a drone flyover does not substitute for a formal inspection for your municipal permits you still have to go out there you still have to document your field observations" (talk at 07:35). That was about stormwater permits, not bridges, but the logic carries over.
So the useful framing is supplement, not substitute. Here is where a survey-grade capture can add something, with the limits stated.
Geometry the rating does not hold
A condition rating is a judgment on a 0-to-9 scale. It does not record how far an abutment has rotated, whether a deck cross-slope still drains, or how wide a joint gap is. A georeferenced point cloud and orthomosaic can record that geometry at a date, tied to a coordinate system, within the capture's stated accuracy (see the numbers below). The FHWA brief describes the same idea: a baseline 3D model and point cloud that supports repeat flights and comparison.
A baseline for the long intervals
If 22.5% of California's bridges are on a 48-month cycle, a dated surface model of an approach, deck and channel is a record of what existed between visits. Whether it changes a decision depends on whether you re-fly and compare. A single flight is a picture. Two flights tied to the same control are a measurement.
Channel and waterway context
For the bridges coded U, a flight can capture the approach grade, channel banks, vegetation and visible debris at a stated date. The Coding Guide says scour analyses "shall be made by hydraulic/geotechnical/structural engineers," so survey data is an input to that analysis, not a replacement for it.
Access and traffic
The FHWA brief says UAS can help with "the cost and safety implications of the traffic restrictions needed for some inspections." That mostly concerns imagery. For survey work, the same flight can sometimes reduce time near live traffic. Whether it does depends on the site, the traffic control plan and the owner's rules.
What it cannot do, in numbers
Do not expect drone LiDAR to detect millimeter-scale abutment movement. The published accuracy figures do not support it.
For a capture sensor such as the DJI Zenmuse L2, DJI specifies a system accuracy of 4 cm vertical and 5 cm horizontal at 150 m. DJI measured this in its laboratory, so it is a best case. An independent test by Tamimi and Toth (ISPRS Archives, 2024) measured about 7 cm RMSE on an open road and 21 cm under forest canopy.
What follows is an inference from those figures, not a published bridge test. If two surveys each carry independent error of 4 to 7 cm, the error of their difference is about 1.4 times that, or roughly 6 to 10 cm. Movement of a centimeter, or even a few centimeters, is below what a drone LiDAR comparison can resolve with confidence. Abutment tilt and bearing movement at that scale call for fixed targets measured with a total station, or a terrestrial scanner, set up by a surveyor. Drone data is better suited to larger changes, such as a settled approach or a slumped embankment, where the change is well above the combined error.
Two more practical limits follow from the FHWA brief. Flights depend on a GNSS signal and a visual line of sight, so under-deck work with a drone can be hard. And battery life of about 20 to 30 minutes means large or complex structures need planned segments.
If you ask for a deliverable, ask for the accuracy figure in the form the ASPRS Positional Accuracy Standards, Edition 2 uses: vertical accuracy reported as RMSEv against at least 30 independent checkpoints. A vendor's sensor datasheet number is not a project accuracy.
Who can sign survey products
Twinn Digital captures and processes the data. It does not stamp. If a topographic map, cross section or set of as-built dimensions goes into a design or a permit, a licensee has to be in responsible charge.
- California: a Professional Land Surveyor, or a civil engineer legally authorized to practice land surveying. See BPELSG Bulletin 50 and Business and Professions Code 8726.
- Georgia: a Professional Land Surveyor.
- Florida: a Professional Surveyor and Mapper (PSM).
The bridge inspection rating itself is a separate professional act under the NBIS. A survey partner supplies measured geometry. The team leader still owns the rating.
Federal-aid bridge work and drone equipment
Many bridge programs use federal-aid funds. Federal funding and federal operation of certain foreign-made drones are restricted under the American Security Drone Act (Pub. L. 118-31, secs. 1821-1833). Before you scope any federally funded bridge work, check SAM.gov for the current covered-entity listings rather than assuming a manufacturer is allowed or barred, and read your funding agreement. If the work is federally funded, ask about equipment options up front, since sensors are rented per job and the choice is not locked to one brand.
Airspace and traffic planning
The FHWA brief notes that many bridges sit in controlled airspace and that the pilot must follow FAA rules, including airspace authorizations. In Port Orchard, GIS and public works inspector Darren Pedraza said about half of the city's jurisdiction falls under an airport's airspace, so altitude limits rise with distance from the airport (talk at 25:58). A bridge near an airport can need LAANC or a manual authorization before the flight. Twinn Digital's flights are planned with software agents and flown by a human FAA Part 107 remote pilot, who handles those authorizations. Traffic control follows the owner's requirements.
A practical next step
Pull your own agency's records before talking to anyone. Download your state's file from the NBI page and filter on Item 22 (owner) for your agency. Then make a short list:
- Bridges with any of Items 58, 59 or 60 at 5 or below and an inspection interval of more than 24 months.
- Bridges over water with Item 113 coded U or 6.
- Bridges where an earlier report mentions a deck-drainage, approach-settlement or abutment concern.
For each bridge on that list, write down the question a measurement could answer, such as "has the south approach settled since 2024?" Then decide the accuracy you would need to answer it and whether a drone capture can reach it.
If you are a surveyor, engineer or agency staff member planning a capture and processing job with a licensed partner stamping the survey deliverable, see Twinn Digital for surveyors and engineers.