US Bridge Condition & Scour Risk Monitor — FHWA NBI

Screen the FHWA National Bridge Inventory (NBI) — 624,193 US highway bridges as of the June 20 2025 extract — by state, overall condition (Poor / Fair / Good), and scour-critical status, and get back a normalized record per bridge with a composite 0–100 risk score, decoded NBI vocabularies, load ratings, inspection regimes and programmed-improvement costs.

Data comes straight from the official NTAD (National Transportation Atlas Database) ArcGIS service published by the US DOT — no API key, no scraping fragile HTML. 41,685 US bridges are currently rated in Poor condition; this actor finds them, ranks them, and flags why.

What changed in 1.1 (2026-08)

• posted_or_closed was wrong on 352 bridges. 1.0 computed it as "NBI item 41 is anything other than A". Code G means "new structure not yet open to traffic" — a brand-new bridge, not a load restriction. Those 352 bridges (live count 2026-08-01) were published as load-restricted. posted_or_closed is now true only for codes B, D, E, K, P, R, null when no code was published, and code G gets its own not_yet_open_to_traffic flag.
• A mid-run page failure no longer sells you a partial inventory. 1.0 broke out of the paging loop on an upstream error and emitted whatever it had already collected — an arbitrary subset presented as the complete answer to your filter, billed, run SUCCEEDED. Rows are now buffered and the run fails loudly, billing nothing.
• Truncation is visible. matching_bridge_count is the service's own count of everything that matched your filter, next to bridges_scanned and result_truncated.
• A live drift probe runs before any billable row: the layer resolves (a retired ArcGIS service answers HTTP 200 with an embedded error, not a 404), every column this actor reads still exists (an unknown field name returns zero features, not an error), the national count sits in a band, a positive canary returns bridges, a negative control returns exactly zero, and DATE_OF_INSPECT_090 still parses as MMYY — the encoding every overdue verdict depends on.
• Closed vocabularies are enforced. Every NBI condition, scour, posting and 0–9 rating code was enumerated live against the whole layer. An unrecognised code fails the run rather than being scored as though it were benign.
• The data vintage is reported (nbi_data_vintage) — the NBI is a periodically republished snapshot, not a live feed.
• 70 new fields from columns the layer has always published. The layer carries 128; 1.0 emitted 33.

What the risk score adds (the transform)

Raw NBI dumps give you 128 cryptic coded columns. This actor turns each bridge into a decision-ready row:

• risk_score (0–100) = condition weight (Poor ≫ Fair ≫ Good) × scour-critical factor (item 113 codes 0–3 boost) × inspection-overdue factor (months since last inspection vs. the designated interval) × log-scaled traffic exposure (ADT). Higher = riskier. The worst possible combination (Poor + scour failure + 2× past inspection + 100k ADT) scores 100.
• risk_flags[] — machine-readable: poor_condition, condition_unknown, scour_critical, scour_action_required, scour_unknown, inspection_overdue, no_inspection_on_record, posted_or_closed, closed, not_yet_open_to_traffic, fracture_critical, high_traffic_poor.
• Decoded vocabularies — condition letters, the full 0–9/N/T/U scour vocabulary, the 0–9 condition ratings, the 8 posting codes, owner and maintaining-agency codes, MMYY inspection dates → ISO YYYY-MM, state FIPS → postal code.

Who uses this

• Bridge engineering & inspection firms — build target lists of Poor / scour-critical / overdue / fracture-critical bridges by state for business development, with the inspection interval and last-inspection date attached.
• Infrastructure investors & muni-bond analysts — screen the bridge stock behind an issuer or geography, with programmed improvement costs and deck area.
• P&C and infrastructure insurers — exposure screening around portfolios, with load ratings, truck share, waterway adequacy and detour length.
• DOT contractors — find rehab candidates before RFPs drop.
• Journalists & researchers — "how many Poor, load-posted, high-traffic bridges are in my state?" answered in one run.

Example input

{
  "states": ["CO"],
  "conditions": ["P"],
  "scourCriticalOnly": true,
  "minADT": 1000,
  "minRiskScore": 40,
  "maxResults": 500
}

States accept postal codes ("CO"), full names ("Colorado"), or FIPS ("08"). Leave states and conditions empty to scan the national inventory (bounded by maxResults). Empty input never fails — it runs a bounded default.

Output fields (one row per bridge)

Identity & location

Field	Meaning
structure_number	NBI item 8 structure number
state	2-letter postal code
state_fips	2-digit state FIPS (the NBI's own key)
county_code	3-digit county code
facility_carried	What the bridge carries (item 7)
features_desc	What it crosses (item 6A)
location	Free-text location description (item 9)
lat / lon	Decimal-degree coordinates
source_url	A live query returning this exact record

Condition

Field	Meaning
condition / condition_code	Overall NBI condition, Good/Fair/Poor
deck_cond / deck_cond_desc	Deck rating (item 58) + decode
superstructure_cond / superstructure_cond_desc	Superstructure rating (59) + decode
substructure_cond / substructure_cond_desc	Substructure rating (60) + decode
culvert_cond / culvert_cond_desc	Culvert rating (62) + decode
channel_cond / channel_cond_desc	Channel & channel-protection rating (61) + decode — the leading scour indicator
lowest_rating	Lowest of the component ratings

Scour

Field	Meaning
scour_critical	Raw NBI item 113 code
scour_critical_desc	Full decode of the 0–9 / N / T / U vocabulary
scour_critical_flag	True for codes 0–3, FHWA's scour-critical definition
scour_action_required	True for code 4 — foundations stable, but action required to protect exposed foundations (24,218 bridges nationally)

Posting, restriction & load rating

Field	Meaning
open_closed_posted_code	Raw NBI item 41 code
open_closed_posted	Plain-English decode
posted_or_closed	True for B/D/E/K/P/R. False for G ("new structure not yet open to traffic"). Null when no code was published
posting_evaluation	Item 70 appraisal of operating rating vs legal load
operating_rating_metric_tons / operating_rating_method_code	Items 64 / 63
inventory_rating_metric_tons / inventory_rating_method_code	Items 66 / 65

Inspection regime

Field	Meaning
last_inspection_date	ISO YYYY-MM, parsed from the MMYY item 90
months_since_inspection	Whole months elapsed
inspection_frequency_months	Designated interval (item 91)
inspection_overdue	True when past the interval; null when unknown
fracture_critical	True when item 92A flags fracture-critical details
fracture_critical_code	Raw item 92A value (e.g. Y24)
fracture_inspection_freq_months	The interval embedded in item 92A
fracture_last_inspection_date	Item 93A, ISO YYYY-MM
underwater_inspection_code / underwater_inspection_required / underwater_last_inspection_date	Items 92B / 93B
special_inspection_code / special_inspection_required / special_last_inspection_date	Items 92C / 93C

NBI appraisal ratings

Field	Meaning
structural_evaluation	Item 67
deck_geometry_evaluation	Item 68
underclearance_evaluation	Item 69
waterway_evaluation	Item 71 — waterway adequacy, the overtopping signal
approach_road_evaluation	Item 72

Size, traffic & network role

Field	Meaning
structure_length_m, max_span_length_m, main_unit_spans	Items 49 / 48 / 45
deck_area_sq_m	Deck area — the denominator for cost-per-sq-m estimates
roadway_width_m, deck_width_m, traffic_lanes_on	Items 51 / 52 / 28A
min_vertical_clearance_m	Item 10. Null when NBI records the 99.99 "no restriction" sentinel
adt, adt_year	Average daily traffic and its year
percent_adt_truck	Item 109 truck share — the fatigue driver
future_adt, future_adt_year	Items 114 / 115 forecast
detour_length_km	Item 19 — network criticality if the bridge closes
functional_class_code, highway_system_code	Items 26 / 104
on_strahnet	Item 100 — Strategic Highway Network / defence route
federal_lands_code	Item 105
navigation_control_code, pier_protection_code	Items 38 / 111 — vessel-collision signals
toll_code	Item 20

Ownership, age & programmed work

Field	Meaning
owner_code / owner	Item 22 + decode
maintenance_responsibility_code / maintenance_responsibility	Item 21 + decode (can differ from the owner)
year_built	Item 27
year_reconstructed	Item 106. Null when NBI records the 0000 "never reconstructed" sentinel
is_temporary_structure	Item 103
structure_kind_code, structure_type_code	Items 43A / 43B
deck_structure_type_code, wearing_surface_type_code	Items 107 / 108A
work_proposed_code	Item 75A type of work proposed
bridge_improvement_cost_usd_thousands	Item 94
total_improvement_cost_usd_thousands	Item 96
improvement_estimate_year	Item 97. Null when NBI records the 0000 sentinel
federal_agency_code, submitted_by_code	Submitting agency

Scoring & provenance

Field	Meaning
risk_score	Composite 0–100
risk_flags	Machine-readable screening flags (see above)
nbi_source_status	ok. A row only exists when the FeatureServer answered and the full result set was retrieved
nbi_data_vintage	The FHWA extract date from the layer description (e.g. June 20 2025)
nbi_layer_name	The FeatureServer layer that answered
nbi_layer_feature_count	Total bridges in the layer at run time
matching_bridge_count	The service's own count of everything matching your filter
bridges_scanned	How many were actually retrieved and scored
result_truncated	True when matching_bridge_count exceeded what maxResults allowed
where_clause	The exact server-side filter this run applied
checked_at	ISO timestamp of the run

Inputs that populate the sparser columns

• Everything above is populated somewhere in {"states": ["CO"], "conditions": ["P"]} — the actor's own prefill, 417 rows, zero null-on-every-row columns.
• result_truncated: true, matching_bridge_count: {"conditions": ["P"], "maxResults": 50} (41,685 bridges match nationally).
• scour_action_required: true: any run touching NBI scour code 4 — 24,218 bridges nationally.
• not_yet_open_to_traffic: 352 bridges nationally carry NBI item 41 code G.
• pier_protection_code: 97,639 bridges nationally publish item 111.
• special_inspection_required / special_last_inspection_date: bridges on a special-inspection regime (item 92C begins with Y).

Use as an MCP tool

This actor is discoverable to AI agents via mcp.apify.com. Field-level output descriptions make it a clean, chainable tool: pass a state and a condition and get back structured, decoded bridge records an agent can reason over. Agents should read result_truncated and matching_bridge_count before treating a result set as complete, and must not read a null boolean as a negative.

Pricing

$8 per 1,000 results (single automatic Result/dataset-item event — one billed record per bridge). Rows are buffered and the billing gate runs before anything is pushed: if a page of the query fails part-way, the run fails and nothing is billed rather than selling you an arbitrary partial inventory.

FAQ

Does this need an API key? No — the NTAD ArcGIS service is a keyless public US DOT endpoint.

How current is the data? The NBI is a periodically republished snapshot, not a live feed. Every row carries nbi_data_vintage — the extract date FHWA stamps in the layer description (currently June 20 2025), so a new publication can never change your numbers silently.

Why did my run fail instead of returning some bridges? Because a partial inventory presented as the complete answer to your filter is worse than a failure. If a page of the query errors, or the layer fails its live drift check, or FHWA publishes a code this actor cannot score, the run fails loudly and bills nothing.

Is risk_score an official FHWA rating? No. It is an editorial composite of condition, scour, inspection overdue-ness and traffic exposure. FHWA does not publish a single 0–100 bridge risk number. Use it to rank and triage, not to make an engineering determination.

Why is posted_or_closed false on a bridge with a non-A posting code? NBI item 41 code G means "new structure not yet open to traffic" — that is not a load restriction. open_closed_posted gives you the plain-English code, and the not_yet_open_to_traffic flag tells you why.

Related actors

• National Tunnel Inventory Risk Monitor — the tunnel counterpart on the same NTAD platform.
• NFIP Flood Loss Risk Screener and FEMA NRI County Risk Profile — flood and multi-hazard context for the waterways these bridges cross.
• USACE Levee Flood Risk Screener, EPA Contaminated Site Screener — the rest of an infrastructure-diligence stack.

Each is a screening tool, not an engineering determination.