USACE Waterborne Commerce Statistics Center – U.S. Waterborne Commerce Data

It is the federal ledger of domestic waterborne commerce, kept row by row. The Waterborne Commerce Statistics Center sits inside the U.S. Army Corps of Engineers' Navigation and Civil Works Decision Support Center, and its register frames every reported cargo movement on the U.S. network: the 2023 edition enumerates 1,381 locations - 819 PORT-type entries such as Houston Port Authority, TX and 562 WATERWAY segments such as the Houston Ship Channel, with legal 'Section Included' boundary text, 52.5 miles from Galveston Harbor to the turning basin at a maintained depth of 45 feet.

Two layers make the record analyzable: a register pinning each location with a numeric id, name, PORT/WATERWAY classification and one of four coastal regions, and above it per-location cargo reports decomposing tonnage by commodity and direction across five consecutive years, with trips reports recounting the same period by vessel class and draft band - twenty-three data years, 2002 through 2024.

Within Datadory's catalog - 1,744 datasets averaging 7.81 - this record scores 9/10, with field definitions verified against live responses during research. Get a sample of this dataset cut to the ports, commodities and years you name.

Five verified observations ship with the record - two register entries, a cargo-report total row, a trips aggregate and a region:

# register entry - PORT type (region 1, Atlantic Coast, data year 2023)
locationId   : 505        name: Albany Port District, NY
locationType : PORT
description  : Section Included: ... territory on the easterly and westerly sides of the Hudson river...

# register entry - WATERWAY type (region 2)
locationId   : 2012       name: Houston Ship Channel, TX
locationType : WATERWAY    description: Galveston Harbor, TX, to Houston turning basin,
                            52.5 miles... Maintained Depth: 45 feet to Boggy Bayou

# five-year cargo report - all-commodities total row
# Houston Port Authority, TX (locationId 2031), 2019-2023 window
commodityId     : 0            # 0 = All Commodities aggregate
allDelta0..4    : 309531236 | 293833530 | 266524394 | 275940289 | 284963668
receiptsDelta0  : 94670977     # inbound tons, oldest window year
shipmentsDelta0 : 195465767    # outbound tons, oldest window year

Read the Houston row and direction matters as much as volume: 94,670,977 short tons received against 195,465,767 shipped in its oldest window year - an export-weighted gateway a single combined-total column would bury. locationId is the clean integer key (2031 Houston Port Authority, 505 Albany), so joins onto directories and position feeds never depend on name spelling.

Seventeen documented fields in two layers over one register. Identity places every row on the waterfront: locationId is the numeric register key, name carries the state suffix, locationType splits PORT from WATERWAY, regionId assigns one of four coastal regions, and description holds the legal boundary text with depths on waterways. Measurement keys cargo by commodityId and commodityName (code 0 being the all-commodities aggregate), then repeats five cells - all, receipts, shipments, intra and thru - across Delta0..4, five consecutive years per row. The trips layer reuses that spine keyed by draft band and vessel class beneath the ALL / DOM / FRN split; isTotalAggregate flags total rows. Definitions were verified against live responses during research.

Additional context on request: major-group and minor-group commodity rollups beyond the sampled codes, the full draft-band ladder behind the trips aggregates, and boundary-description text for any named waterway. Commodity labels travel inside the reports rather than as a separate lookup, so the label set confirms against delivered rows when your sample is cut.

Geography - wall-to-wall across the U.S. waterfront: four coastal regions holding 819 ports and 562 waterways in the 2023 register. Inland corridors enter through the waterway layer, which is why the Mississippi system and the Great Lakes read as first-class rows rather than aggregations.

Temporal - twenty-three data years, 2002 through 2024, every report internally spanning five consecutive years - so a single request covers the five most recent vintages for any location.

Granularity - location x commodity x direction x vessel class x draft band, annually. Five slicing dimensions at once; the nearest alternatives stop at one undifferentiated tonnage figure per port.

On Datadory's 0-10 rubric the record scores 9, against a 7.81 average across the 1,744 cataloged datasets - carried by verified documentation and depth of decomposition.

API, files, or your warehouse. Daily, weekly, or hourly.

You pick the channel and the cadence; the verified field dictionary above travels unchanged across all three.

Deliveries arrive flattened - one row per location per commodity per direction per year - so warehouse loads diff cleanly and BI tools see the same columns analysts do. Because the register schema and the report envelope are fixed regardless of which ports, commodities or vessel classes you take, switching cadence later is a settings conversation, not a re-integration project. Name the gateways and years you care about and the sample returns shaped exactly like the blocks above.

• Port-level commodity and trip panels - data scientists train on 2002-2024 history across roughly 820 ports and waterways with direction splits already columnar (data scientists use cases).
• Freight-flow modeling - receipts-versus-shipments balances at named gateways as proxies for commodity exposure (investors quants use cases).
• Client trade quantification - per-port, per-commodity, per-direction figures over five-year windows for consulting deliverables.
• Gateway selection - import gateways weighed through per-port receipts and vessel-class detail.
• Share-shift surveillance - tonnage-share drift between gateways as an early strategic signal.
• Citation-grade sourcing - attribution to the Corps back to 2002 (journalists academics use cases).

Eight personas tag this record relevant, five at the maximum relevance of 3. Data Scientists & ML Engineers get the longest port-grain panel in the slice with typed columns and stable keys. Investors & Quant Researchers get directional flow balances instead of headline totals. Market Researchers & Consultants get five-window commodity cuts ready for client decks. Developers & Data-Product Builders get one schema that generalizes from ten ports to all 819 (builders' use cases). Journalists, Academics & Students get attribution that survives peer review; Sales, Competitive Intelligence and E-commerce teams tag it at relevance 2.

This record owns the domestic tonnage fact table - the deepest decomposition in the slice. Its nearest neighbour answers a different question: BTS Principal Ports – National Transportation Atlas Database scores 10/10 and hands you a finished league-table row - 150 principal-port polygons with CY 2023 totals and a precomputed rank. This record scores 9 and hands you the parts: per-commodity, per-direction, per-vessel-class detail for 819 ports and 562 waterways back to 2002, rankings left to compute. The tradeoffs are mapped cell by cell in our head-to-head comparison. Beyond the pair, World Bank Container Port Traffic sizes 207 economies annually, NOAA CO-OPS covers real-time berth operations rather than cargo, and AIS movement feeds count hulls instead of tons.

Four honest notes up front.

First, totals arrive as parts: no rank column and no geometry - port limits are legal boundary descriptions rather than shapes, so league tables and maps are downstream work. If you need the finished scoreboard, pair this with the Principal Ports layer via the comparison above.

Second, units follow WCSC convention - short tons throughout - but the convention is implicit in program publications rather than restated beside every figure, so confirm before quoting absolute magnitudes.

Third, registers are republished per data year, so pin longitudinal analyses to one year's register when locationIds must stay stable. Fourth, commodity labels confirm against delivered rows when your sample is cut; column naming locks there too.