Goods atlas · technical reference

Methods

Definitions, aggregation, route assignment and worked reconciliations.

How to read the map

Line width represents annual throughput at that location in the selected unit: short tons or dollars. Wider lines carry more, but minimum widths, joins and caps make the numeric card more precise than a pixel comparison. A short ton is 2,000 pounds; dollar figures retain their source year and measure shipment, trade or production value rather than GDP. Zooming reveals finer geometry and additional labels without changing the underlying annual quantities. Click a line to inspect its local interval, or a place to see its reporting geography, quantities and available direction or partner detail. Follow the calculation link for the definition, and use Data to check the source vintage and coverage.

Measurement conventions

The nominal edition is a 2025 snapshot, not a single-year census. Source years span 2022–2025. A short ton is 907.18474 kg. Values are nominal current USD of their source year; benchmark-price estimates are identified separately, and no inflation adjustment is applied.

T denotes short tons, U current USD and q the selected quantity. o and d index domestic origin and destination, z a FAF zone, e a network edge and τ trade type. Hats denote atlas estimates, Σ a sum, and 1[condition] an indicator equal to one when the condition holds and zero otherwise. Additional indices and symbols are defined in the section where they are used; local meanings are not carried between metric families.

Shipment, passage, handling, customs trade and production are distinct measures. Source-estimated quantities and additional atlas assignment or valuation are identified with the calculation. See Limitations for missing data, overlap and uncertainty.

Freight regions & metros

FAF domestic origin/destination regions; imports and exports are their U.S. freight legs, not foreign endpoints

Units: annual short tons; current USD
Actual vintage: 2024 preliminary

Definitions and aggregation

q_odmcτ is shipment quantity indexed by domestic mode m and SCTG2 commodity c, in addition to the origin, destination and trade indices defined above. FAF trade types 1, 2 and 3 denote domestic, imports and exports. The preliminary 2024 columns are converted before aggregation: T = 1,000 × tons_2024 and U = 1,000,000 × current_value_2024.

O_z = Σ_d,m,c,τ q_{z d m c τ}
W_z = Σ_m,c,τ q_{z z m c τ}
S_z = O_z − W_z
R_z,K = Σ_{o≠z,m∈K,c,τ} q_{o z m c τ}, K = {1, 2, 3, 6}

O is originating freight, including within-region W; S is freight shipped outside. R is received from outside in the four retained detail modes: truck, rail, water and pipeline. The delivered zoneOut and zoneWithin arrays contain all-mode domestic/import/export tons followed by those three dollar totals. Runtime regional o and w retain these marginals. Retained od[m] rows are [origin, destination, trade type, tons, dollars, energy-class tons…]; they do not contain an all-commodity joint distribution.

For each retained mode, outside inbound and outbound exclude o = d; within includes only o = d. Partners group outside rows by the other endpoint and rank their inbound + outbound quantity in the selected unit. The leading partners plus all remaining partners reconcile to the retained four-mode exchanged quantity. Other-mode shipped/within remainders equal the respective all-mode totals minus the retained subtotals, subject to rounding. All-mode inbound and its remainder are not calculated by the served card.

Geography and rounding

FAF regions are freight statistical areas, including separately named parts of metropolitan areas and rest-of-state regions. A city label does not replace the FAF boundary or join multistate parts. The extractor sums original records in floating point, rounds OD cells to 0.1 unit and regional/national marginal cells to whole units. Independently rounded marginals therefore need not equal sums of rounded OD cells exactly.

Separate retained mode|all source tables cover all eight modes. Their internal reconstruction reconciles originating and within quantities for 132 regions (maximum absolute discrepancy 1.22 short tons; exact dollars in the tested marginals). This is source-only evidence: it does not extend the four-mode served received, partner or mode detail. Air, multiple modes/mail, other and unknown modes must not be synthesized from marginal totals.

Sources and coverage: FAF regional freight

Roads

Two-way modeled annual throughput on the clicked source interval, not the whole named route

Units: short tons; current USD
Actual vintage: FAF 2024; BEA weights 2024; geometry snapshot

County weighting and demand disaggregation

County interior points are assigned to the smallest containing FAF polygon; if none contains the point, the closest same-state polygon is used. Gazetteer points outside the largest county land polygon are replaced by that polygon’s representative point. Virginia BEA county/city combinations are attached to the county point. BEA 2024 CAGDP2 origin weights use line 91 (private goods-producing GDP) + line 34 (wholesale trade); destination weights use line 1 (all-industry GDP). GDP provides a spatial allocation weight, not a conversion from value added to shipment dollars.

For the equations below, c indexes counties, s states and ℓ GDP industry lines. g is county GDP and G its state total; a and b are normalized origin and destination shares for clusters i and j. P is an assigned path, κ its edge cost and L accumulated link quantity. Demand q_odτ sums the mode’s commodities.

ĝ_cℓ = max(0, G_sℓ − Σ_{disclosed c′∈s} g_c′ℓ) × g_c1 / Σ_{suppressed c′∈s} g_c′1
a_zi = Σ_{c∈cluster i}(g_c91 + g_c34) / Σ_c∈z(g_c91 + g_c34)
b_zj = Σ_{c∈cluster j} g_c1 / Σ_c∈z g_c1
q_oi,dj,τ = q_odτ a_oi b_dj

The first equation imputes a suppressed line-91 or line-34 weight from the state’s undisclosed remainder, in proportion to disclosed all-industry GDP. The imputation affects routing only. Counties rank by normalized origin + destination weights. Greedy spatial seeds are separated by 0.25 degree-equivalent distance in metro regions (maximum four points) and 0.55 in rest/state regions (maximum 12, or 15 when county count exceeds 100). Distance uses longitude scaled by cosine of mean latitude. Remaining counties join their closest seed. A zone without counties receives one internal point with shares 1 and 1. Exported shares round to six decimal places: conservation is exact before share rounding, approximate afterward; zero-denominator weights are not observed shipment absence.

Path assignment and link accumulation

P_oi,dj = arg min_P Σ_e∈P κ_e
L_eτ = Σ_o≠d,i,j q_odτ a_oi b_dj 1[e∈P_oi,dj]

Dijkstra produces a least-cost tree per origin point. Destination demands accumulate back along tree arcs through a sparse triangular solve. Virtual source nodes have outgoing access arcs only; virtual sink nodes have incoming arcs only, preventing transit through an allocation point. Each point connects to up to three network nodes, rejecting picks within 0.04 degree of an earlier pick. The nearby search radius is a preference: the producer permits a first pick beyond it. Access miles are 69.05 × √((Δlongitude cos latitude)² + Δlatitude²).

Truck edge cost is FHWA LENGTH × highway-class multiplier: 0.85 for classes 11/21, 0.92 for 12, 0.95 for 22, 1.25 for 19 and 1 otherwise. Class 41 and AK/HI highway geometry are excluded. Paired divided carriageways are collapsed to one retained line using proximity and opposite bearing tests; links are treated as two-way. Access candidates lie in the largest strongly connected core and on selected major classes; radius preference 0.25 degree and access cost 1.4 × miles + 0.5. Costs are generalized miles, not travel time.

Intrazonal demand is excluded from interregional routing but remains in regional totals. Alaska/Hawaii FAF regions 020, 151 and 159 have no lower-48 network assignment. Remote or disconnected demand is reported separately, not assigned a measured zero flow. Output stores [T domestic, T imports, T exports, U domestic, U imports, U exports] per source edge; the card uses its clicked local source interval. A named route, blended drawing chain and local edge need not carry the same quantity.

Sources and coverage: FAF regional freight · Highway and rail networks · County GDP routing weights · Census boundaries and internal points

Rail

Two-way modeled annual throughput on the clicked source interval, not the whole named route

Units: short tons; current USD
Actual vintage: FAF 2024; BEA weights 2024; geometry snapshot

Main-line assignment

Rail uses the county weights, point disaggregation, least-cost paths and link accumulation defined in Roads. Only the network and access costs differ.

Rail uses BTS NTAD North American Rail Network records with NET = M, valid FRFRANODE/TOFRANODE, and distinct endpoints. Cost is MILES (minimum 0.0001), with arcs in both directions. Access candidates use the largest connected main-line component; radius preference 0.45 degree and access cost 1.6 × miles + 0.5. Network mileage is not scheduled transit time; interchange penalties, service frequency and operator-specific constraints are not evidenced.

Sources and coverage: FAF regional freight · Highway and rail networks · County GDP routing weights · Census boundaries and internal points

Waterways

Passages on retained inland waterway links, upbound + downbound

Units: short tons; estimated current USD
Actual vintage: observed CY2023 → estimated 2025; FAF value rates 2024

Observed base and retained geography

USACE/BTS Link Tonnages CY2023 joins LINKNUM_1 to the National Waterway Network’s LINKNUM. Nine class pairs supply upbound/downbound short tons: COAL, PETROL, CHEM, CRMAT, MANU, FARM, MACH, WASTE and UNKWN/UNKW. Network classes I retain WTWY_TYPE {1,2,6,7,8,9,12}; G retains {6,8,9,12}. NON_US records, ocean class O, unsupported classes, lake-crossing/open-water representative routes, missing geometry and nonpositive/missing base totals are omitted. The offshore screen uses a 0.12-degree buffered land mask; suspect whole parts are refined by a >15% outside-point test. It is an approximate geometric screen, not an exact offshore length measurement.

2025 lock scaling and valuation

For river system r, commodity class c and year y, A_rc,y is the sum of LPMS annual commodity tons across locks that report both 2023 and 2025. Missing year pairs are excluded from both sums. The river-system ratio is used only if A_rc,2023 ≥ 500,000 and A_rc,2025 > 0. Otherwise the national paired-lock class ratio is used. In the executable producer, a national class with a zero baseline receives factor 1; a class absent from the national dictionary uses the overall national paired-lock ratio.

f is the selected scaling factor, p the FAF commodity USD/ton rate, and r(e) the link’s river system; ↑/↓ denote upbound/downbound.

f_rc = A_rc,2025 / A_rc,2023 (eligible river/class); otherwise national fallback
T̂_e,c,↑/↓ = T_{e,c,↑/↓,2023} f_r(e),c
p_c = U_{FAF water,c,2024} / T_{FAF water,c,2024}
Û_e = round[Σ_c(T̂_e,c,↑ + T̂_e,c,↓) p_c]

FAF’s WCSC approximation maps SCTG15→coal; 16–19→petroleum; 20–24→chemicals; 10–14,25–29,31→crude materials; 1–9→farm; 32–33→manufactures; 34–38→machinery; 41→waste; all remaining SCTG codes→manufactures. Known class rates are rounded to cents/ton in the extractor. Unknown uses total water dollars divided by total water tons. Up/down class tons are rounded independently; dollars are calculated from unrounded scaled tons and rounded once, so rounded tons × rate may not reproduce dollars exactly.

The retained summary reports national overall paired-lock factor 0.9890591223, but the retained link-weighted scaling factor is 1.0059696108: river/class composition differs. These are not contradictory national shipment estimates. The derivative contains 1,575 geometry records; multipart source links can repeat a link ID and value. Neither geometry-record count nor sum of repeated link tons is unique cargo.

Sources and coverage: FAF regional freight · Waterway link tonnages · LPMS lock scaling

Pipelines

Estimated annual throughput assigned to real line geometry; coverage incomplete

Units: estimated short tons; estimated USD
Actual vintage: FAF 2024 scaled/calibrated using EIA 2025

Demand classes and value allocation

Retained EIA geometry layers are crude 2, petroleum products 1, HGL 0 and natural gas 3. FAF mode 6 OD rows contain total tons T, value U and SCTG15–19 ton components. For an OD row let C_s be its component tons and p_s the national mode-6 dollars/ton summed across trade types. Residual T_res = max(0, T − Σ_s=15…19C_s).

k indexes demand classes; B is the class value weight before allocation, not observed class dollars.

T_crude = C₁₆; T_products = C₁₇ + C₁₈; T_gas/HGL = C₁₉ + T_res
B_crude = C₁₆p₁₆; B_products = C₁₇p₁₇ + C₁₈p₁₈
B_gas/HGL = C₁₉p₁₉ + T_resp₂₀; U_k = U B_k / Σ_k′B_k′

Coal/slurry SCTG15 tons are unassigned. The residual is priced with SCTG20’s national rate; this is a modeling proxy, not a supported chemical identification. The executable substitutes denominator 1 when all class value weights are zero, yielding zero allocated dollars rather than evidence of zero shipment value. Positive class tons only enter the class OD lists.

EIA controls and calibration

EIA monthly 2024/2025 series are summed by calendar year. Cross-PADD crude receipts PET.MCRMPP{dest}P{origin}1.M are converted from thousand barrels by ×1,000 ×0.1504 assumed short tons/barrel. Products use max(0, total petroleum MTTMPP − crude MCRMPP) ×1,000 ×0.1396. These density assumptions apply to pipeline calibration, not production-card native volumes.

a and b index origin/destination PADDs; E is the EIA petroleum receipt control after mass conversion, F the corresponding FAF class tons, and f the selected calibration factor.

f_ab,k = min(2, max(0.5, E_ab,k,2025 / F_ab,k,2024)) when E > 0
T̂_od,k = f_od,kT_od,k; Û_od,k = f_od,kU_od,k

F sums FAF class tons for a directed PADD pair. Zero/absent EIA receipt controls use the national growth ratio, not a zero factor. Intra-PADD and other pairs use national ratios: crude field production PET.MCRFPUS1.M, products inter-PADD receipts and marketed gas NG.N9050US2.M. Retained ratios are respectively 1.0269681267, 1.0441821597 and 1.0445878303. Gas/HGL uses marketed-gas growth for every pair. Executable constants and retained summaries clip to [0.5,2], using the actual executable bounds.

Graph and incremental assignment

Each class graph nodes geometric line intersections; gas and HGL share one inherited graph. Coordinates round to four decimals for node identity. Dangling endpoints within 0.05 degree of another line receive expensive undrawn connectors to a nearby line endpoint. FAF zones receive an anchor inside or within 0.75 degree of their polygon, searched within 3 degrees of the internal point. Virtual source/sink access connects up to 60 nearby nodes within 1.2 degrees and polygon distance <0.3 degree, or the single anchor fallback; access cost is 0.8 × approximate straight miles + 1.

h indexes assignment increments, α their demand shares, L accumulated tons, C the nominal class cost capacity, and k_type the line-type cost multiplier. κ is generalized edge cost; each P is its least-cost path.

κ_e,h = miles_e k_type[1 + 0.15(L_e,h/C_k)⁴]
ΔL_eτ,h = α_h Σ_od T̂_odτ1[e∈P_od,h]
α = [0.25, 0.25, 0.20, 0.15, 0.10, 0.05]

At each increment, Dijkstra assigns demand using costs based on accumulated tonnage. Type multipliers are Interstate 1, Intrastate 1.3, Gathering 3, otherwise 1. Nominal cost capacities C are 40 million crude, 30 million products and 40 million gas/HGL short tons/year; they are parameters, not measured capacities. Dollars accumulate along the same paths. Intrazonal, unanchored and disconnected demands are not drawn; connectors do not acquire observed flow records.

Sources and coverage: FAF regional freight · EIA pipeline geometry · EIA pipeline calibration controls

Air cargo airports

Airport handling on nonstop segments; international customs dollars allocated to airports

Units: short tons of freight + mail; allocated international USD
Actual vintage: 2025, all twelve T-100 months

T-100 handling

BTS T-100 domestic/international nonstop segment records retain all twelve months of 2025 and all carriers. Each segment’s short tons equal (FREIGHT + MAIL pounds)/2,000. Domestic records add tons to origin outbound and destination inbound. International records with exactly a U.S. endpoint contribute to that endpoint’s direction; U.S.–U.S. records in the international table are skipped during airport accumulation.

D and I are domestic and international directional handling tons; t is the stored direction vector (zero-based indices).

t = [D_out, D_in, I_in, I_out]
Received = t[1] + t[2]; Shipped = t[0] + t[3]
Handled = Σ_i=0…3 t[i]

The producer rounds each airport direction to whole short tons, then cards sum the vector. Domestic handling counts a domestic segment at both airports; the national domestic segment total counts it once. Neither is a unique shipment or aircraft count.

International customs-dollar allocation

Census air imports/exports are customs-port observations. A port explicitly named AIRPORT maps to the busiest international airport within 30 km; other ports consider international airports within 100 km. Retained courier exceptions add MEM to New Orleans 2002 and SDF/CVG/IND to Cleveland 4101. Distances use a 6,371-km spherical great-circle calculation. Ports lacking coordinates or candidates remain unallocated.

a indexes the airport, p the customs port and b other candidate airports. s is an allocation share and U_p,direction reported customs-port value.

s_ap,imports = I_a,in / Σ_{b∈candidates(p)}I_b,in
s_ap,exports = I_a,out / Σ_{b∈candidates(p)}I_b,out
Û_a,direction = round[Σ_p U_p,directions_ap,direction]

A direction with no T-100 tons splits equally across candidates. The runtime dollar vector u = [imports, exports] contains allocated international dollars only; domestic dollars are unavailable, not zero. Even a single-airport assignment remains a geographic allocation. uEst flags multiple-airport splits, not every modeled assignment.

Airport partners rank two-way T-100 tons, including in the dollar view; dollar partner rankings cannot be inferred. T-100 country partners and Census country-weight allocations are distinct fields. The latter apply the same port shares to customs air weight, converting kg to short tons. The retained sample filters U.S. airports to ≥50,000 handling tons or ≥$1 billion allocated value, then takes the union of top 28 by tons and top 20 by dollars (30 runtime airports in the audited payload), not every airport.

Sources and coverage: BTS T-100 airport cargo · Census customs-port trade

International vessel trade

Census customs port/district reporting geography, not individual terminal or all port cargo

Units: short tons; current import/export USD
Actual vintage: 2025 annual (December year-to-date)

Source records and direction

Census International Trade porths December 2025 year-to-date vessel weights and values supply annual foreign trade at customs reporting geography. Retained normalized fields are location_id, direction, mode, measure, commodity and value. Imports and exports remain separate. A customs-port or district anchor can represent multiple facilities; it is not a terminal meter.

p indexes customs reporting ports and d the import/export direction.

T_p,d = Σ_{vessel shipping_weight rows at p,d} kg / 907.18474
U_p,d = Σ_{vessel trade_value rows at p,d} USD
v = [round(T_imports), round(T_exports), round(U_imports), round(U_exports)]

Raw API year-to-date fields are VES_WGT_YR and VES_VAL_YR; port, country and HS2 dimensions identify reporting scope. Annual YTD values must not be summed across all twelve months. The atlas producer sums already-normalized annual records, ranks vessel tons after its eligibility screen, and retains at most 60 port markers with coordinates. Marker totals exclude domestic coastal/inland handling.

Vessel commodity detail groups retained shipping weight by commodity and direction. Country/group detail uses port × country × HS2 records; individual-country ranking excludes region-aggregate codes beginning 0. An anchor relocation preserves its reporting ID and quantity; it does not localize a district total. Port trade cannot be added to inland link throughput or FAF import/export legs: these can represent the same goods at different stages.

Sources and coverage: Census customs-port trade

Farms & crop regions

Selected county sets; separate CDFA gross production and NASS farm sales

Units: USD farm sales/gross production; native bushels; corn mass conversion
Actual vintage: NASS sales 2022; CDFA 2023 (some 2022); crop survey 2025

Sales, gross production and native crop quantities

USDA NASS 2022 Census of Agriculture county COMMODITY TOTALS – SALES, MEASURED IN $ supplies disclosed all-farm products-sold value. Let C_r be a retained county membership and S_c,2022 a disclosed county sale. The atlas sums disclosed values without inflation or growth to 2025. Withheld source values remain null; the aggregation uses zero placeholders for absent cells but these are incomplete sums, not measured zero sales.

V is county commissioner gross production value; B is corn bushels and C the selected county set.

U_r,sales = Σ_{c∈Cᵣ,disclosed} S_c,2022
U_{Central Valley} = Σ_{17 retained counties} V_c,CDFA
C_{Corn Belt} = {c: B_c,corn,2025 ≥ 25,000,000 bushels and retained county geometry exists}
T̂_corn = Σ_c∈CB_c,corn,2025 × 56 lb/bushel / 2,000 lb/short ton

Central Valley dollars use CDFA county commissioner gross production values labeled 2023, not NASS sales. The statewide CDFA extraction carries six counties’ 2022 values forward; actual row years govern interpretation. Central Valley’s retained 17 counties sum to $44,892,297,000. Corn Belt dollars are those selected counties’ 2022 all-farm sales ($74,720,406,000), not corn-only revenue. Its 2025 corn bushels and conventional mass conversion have a different commodity and vintage scope.

Named county sets supply farm-region memberships. FAF-region farm context assigns counties by interior point, not fractionally clipped farmland. Outlines are simplified/buffered county context, not cultivated acreage. The 56-pound factor is evidenced for corn only, not soybeans or all agriculture; native survey volumes are source estimates rather than directly weighed freight. NASS sales and CDFA gross production are not made comparable by sharing the dollar unit.

Sources and coverage: USDA / NASS farm sales and crops · CDFA county agricultural values · Census boundaries and internal points

Factories & industrial areas

Disclosed county/subsector shipments for named county sets; FAF-region cards use county assignments

Units: current USD value of shipments
Actual vintage: 2022

Disclosed shipments

Census 2022 Economic Census Core Business Statistics ecnbasic supplies county RCPTOT, RCPTOT_F, NAICS2022 and disclosed subsector values. RCPTOT is thousands of current dollars and converts by ×1,000. The retained county file separates all-manufacturing value_usd from subsectors_usd, preserving withheld flags/nulls.

r indexes the named industrial region, C its county set, and N its selected NAICS industries; c and n index those sets.

U_r = Σ_c∈Cᵣ Σ_{n∈Nᵣ,disclosed} 1,000 × RCPTOT_c,n,2022

When N denotes all manufacturing, the all-industry 31–33 county total is used instead of summing it together with subsectors. Gulf refining/chemicals uses NAICS324+325; the auto corridor uses 336; New Jersey–Philadelphia chemicals/pharma uses broad 325, not pharmaceutical-only 3254. The retained Silicon Valley producer uses all manufacturing for all five member counties, not a 334 electronics-only total: Santa Clara’s 334 value is withheld, but the fallback implemented for the named region is broader than that single county. Named labels describe industrial context, not a narrower measured commodity.

The production function mval sums only available county/subsector values; missing counties and null entries contribute zero to this incomplete aggregation. Top subsectors are top disclosed subsectors, not guaranteed actual leaders. Regional manufacturing context uses county-point assignment to FAF geography. County sets and parent/child records overlap; double counting and suppression prevent a complete national sum.

Sources and coverage: Census manufacturing shipments · Census boundaries and internal points

Nonfuel minerals

Whole-state mineral value displayed near selected deposits/counties

Units: current USD; native physical output only where explicitly sourced
Actual vintage: 2025 preliminary, published MCS2026

Statewide source value and local context

USGS Mineral Commodity Summaries 2026 Table 3 supplies preliminary 2025 state nonfuel mineral-production values in millions of dollars. For state s, conversion is U_s = 1,000,000 × published state value. No atlas price × quantity valuation or local allocation is established for these mineral records.

U_{named mineral area} = U_{source state}
U_{local deposit/counties} = unknown; T_local = unknown

Nevada’s $12,600,000,000 and Minnesota’s $3,160,000,000 are all-nonfuel statewide values, not Carlin gold or Iron Range iron-ore values. Other named copper/phosphate areas likewise retain their corresponding state totals. Repetition near a mineral-producing area or in a FAF-region card does not localize that total.

The Nevada boundary-context sidecar lists Elko 32007, Eureka 32011, Lander 32015 and Humboldt 32013; Minnesota lists St. Louis 27137 and Itasca 27061. These are Census 2024 1:20,000,000 cartographic county unions, not ore bodies or independently verified production boundaries. The sidecar preserves original dollars/tons and explicitly sets localValue/localTons to null. Mine-location registries provide locations, not production values. Fuel minerals are excluded, and some national mineral value is not assigned to states: selected state totals do not establish complete national coverage.

Sources and coverage: USGS nonfuel minerals · Census boundaries and internal points

Oil & gas producing regions

STEO basin/region estimates, some state/offshore totals; outlines need not match statistical geography

Units: barrels/day; cubic feet/day; estimated current USD; not short tons
Actual vintage: volumes and prices 2025

Native output and benchmark valuation

EIA STEO annual producing-region series COPRPM/BK/EF/AP/HA provide million barrels/day and NGMPPM/BK/EF/AP/HA billion cubic feet/day for Permian, Bakken, Eagle Ford, Appalachia and Haynesville. The retained extraction’s annual 2025 entry is used. Let Q_oil be million barrels/day, Q_gas billion ft³/day, P_area the selected 2025 crude first-purchase USD/barrel, H = 1,036 Btu/ft³ and P_HH = $3.52/MMBtu.

Û_oil = Q_oil × 10⁶ × 365 × P_area
Û_gas = Q_gas × 10⁹ × 365 × H / 10⁶ × P_HH
Û_region = round(Û_oil + Û_gas)

Permian/Eagle Ford use Texas crude price $64.00/barrel; Bakken North Dakota $58.91; Appalachia Pennsylvania $55.64; Haynesville Louisiana $64.95 in the retained price extraction. Henry Hub is a benchmark, not basin realized sales or wellhead revenue. The gas heat content is the retained U.S. average for consumed gas, another scope assumption.

State/offshore records use EIA crude field/marketed-gas annual volumes. Crude thousand barrels ×1,000 × state first-purchase price (PADD price fallback when withheld); gas million ft³ ×1,000 ×1.036 MMBtu/thousand ft³ ×3.52 USD/MMBtu. Native output remains barrels/day or cubic feet/day, not freight short tons. Zero short-ton placeholders denote unavailable native mass, not zero output.

Outlines use EIA play/basin context and may not coincide with the STEO statistical producing region. Child play records can have unknown value even when parent output is estimated; proportional play area is not a production allocation.

Sources and coverage: EIA producing regions and output · EIA benchmark prices and heat content

Coal regions & mines

MSHA operator quarterly mine output; PRB parent uses all Wyoming coal

Units: short tons; estimated USD for parent
Actual vintage: volume 2025; price 2024

Mine quantities and parent proxy

MSHA MinesProdQuarterly operator coal production sums four 2025 quarters by mine. Contractor counts are not added. The retained site sample takes the 40 highest-output mines, subject to ≥3 million short tons, with original site IDs and registry coordinates. Mine dollars are unavailable; the exported zero-dollar placeholder is not observed zero revenue.

q indexes the reporting quarter and P is the applicable state coal USD/short-ton price.

T_mine,2025 = Σ_{four operator quarters}T_mine,q
T_{PRB proxy} = Σ_{mines in Wyoming}T_mine,2025 = 211,680,312 short tons
Û_{PRB proxy} = T_Wyoming,2025 × P_Wyoming,2024

The Powder River Basin parent uses all Wyoming coal output, not an independently delimited basin sum. EIA Annual Coal Report 2024 average sales price is $15.23/short ton: 211,680,312 ×15.23 = $3,223,891,151.76, rounded to $3,223,891,152. Other state-derived values use state prices or rank-average fallbacks when withheld. This is 2025 quantity at a 2024 price, not reported 2025 sales.

Mine-site quantities overlap the parent proxy. Site coordinates and sedimentary-basin outlines are geographic context, not historical production footprints. All 40 retained site tons matched their source rows in the audit; that comparison is not certification of each mine location or Wyoming-to-PRB membership.

Sources and coverage: EIA benchmark prices and heat content · MSHA mine production

National statistics

Separate national freight and international trade universes; not sums of map layers

Units: annual short tons; current USD; shares
Actual vintage: FAF 2024; Census/T-1002025

Independent source universes

National totals derive from source records rather than summing visible layers or selected marker samples. For FAF trade type τ, national[τ|m] contains independently rounded 2024 short tons and current dollars aggregated across original SCTG/OD records. Census 2025 totals use normalized annual customs records. T-100 segment handling is a third universe.

q_FAF,τ = Σ_{all modes m} national[τ|m]
T_Census,d = (W_vessel,d + W_air,d) / 907.18474
U_Census,d = U_{all modes,d}
Air mass share = W_air / (W_air + W_vessel)
Air value share = U_air / U_{all modes}

Weights W are kg, values U current USD, and d imports or exports. Land-mode weights are not published in this retained Census weight universe; the mass-share denominator is therefore vessel+air, not all freight. The value-share denominator includes all trade modes. National domestic T-100 segment tons equal Σ(FREIGHT+MAIL)/2,000 with each domestic segment counted once, unlike the sum of airport inbound+outbound handling.

Layer visibility does not alter these statistical universes. Source accounting was checked against the frozen core payload for each FAF trade type/unit, Census import/export weight/value and both air-share denominators.

Sources and coverage: FAF regional freight · BTS T-100 airport cargo · Census customs-port trade

Lines, labels, zoom & loading

Cartographic display of source facts; not additional data

Units: CSS pixels; zoom; display counts, not tons
Actual vintage: current retained continuous-zoom runtime

Quantitative encodings and local facts

Rendering changes geometry/detail, not source-year quantities. In the inspected runtime, λ is the zoom-transform scale (runtime k), λ₀ its fitted home-view scale (runtime KH), and σ the projection scale. WPER = 0.085 × max(0.6, min(1.1, σ/1,500)). Quantity is normalized linearly; the exponent 0.32 applies to zoom only, not tonnage.

q_display = T/10⁶ (tons view), or U/(3 × 10⁹) (dollars view)
w_px = min(17, max(0.12, q_display × WPER × (λ/λ₀)^0.32))

The width function is a display transform, not a mass conversion: $3 billion and 1 million tons have the same baseline width by design. Ribbon handling, joined strokes, minimum widths and caps prevent direct pixel-ratio interpretation. A card reports the clicked source interval, not a blended display-chain average or whole named-route total. Network labels can use visible/named-group peaks and rounded quantities, rather than the clicked edge’s value.

In the inspected baseline, port circle size remains based on vessel tons even in dollar mode. Airport diamonds switch between handling tons and allocated international dollars with minimum/maximum sizes. Metro dots are label anchors, not quantitative circles; label priority uses domestic + export originating tons, excluding the import leg. Duplicate display names can collapse labels. Producing-area outlines describe geographic context, not quantity-proportional area; raw-material admission can use dollar thresholds even in the tons view. Native-energy tons placeholders are not measured mass.

Continuous-zoom admission, source/detail dependencies, simplification, priority and label collision rules determine visibility. Missing labels or omitted lines are not evidence of absent freight. Insets preserve source scope but use separate cartographic framing. Loading progress is preparation state; animated vehicles are not live tracked cargo. Display admission is not a national tonnage cutoff.

Sources and coverage:

Arithmetic reconciliations

Houston FAF region 486: originating versus shipped outside

The all-mode source originating shipment value is $908,120,535,876. Within-region value is $467,896,361,164.

$908,120,535,876 − $467,896,361,164 = $440,224,174,712 shipped outside

Received-from-outside value in the served card is $429,535,684,371 for truck/rail/water/pipeline only. It is neither the within-region figure nor an all-mode received total. The independent eight-mode reconstruction remains source-only evidence, not additional served detail. In tons, the displayed outside and within values derive from separately rounded source marginals and retained contract values, so sub-unit discrepancies are rounding rather than new freight.

Memphis MEM: handling and allocated value

Received = 1,263,180 + 338,895 = 1,602,075 short tons
Shipped = 1,421,732 + 257,283 = 1,679,015 short tons
Handled = 1,602,075 + 1,679,015 = 3,281,090 short tons

The vector is domestic out/in, international in/out. The $114,798,439,918 dollar figure is allocated international customs trade, not a valuation of all 3,281,090 handling tons. Applying a dollars/ton rate between these totals would mix incompatible scopes.

Historical Permian benchmark calculation

The retained annual extraction gives 6.5816049116 million barrels/day of crude and 27.554584404 billion ft³/day of gas. These digits identify the earlier audit’s reproduction inputs, not measurement precision. The raw STEO extract is now missing, so this example verifies historical input arithmetic rather than a fresh source replay.

6.5816049116 ×10⁶ ×365 ×64.00 = $153,746,290,734.976
27.554584404 ×10⁹ ×365 ×1,036 /10⁶ ×3.52 ≈ $36,676,606,723.781
round(sum) = $190,422,897,459

The crude-area price and gas benchmark have different scopes; the result is an atlas valuation estimate, not observed basin revenue.

Paired-lock factor reconciliation

Retained Mississippi farm-class locks reporting both years sum to 166,772,397 short tons in 2023 and 228,501,295 in 2025. Their factor is 228,501,295 /166,772,397 = 1.3701385787 (rounded summary 1.3701). The baseline exceeds 500,000 and the 2025 sum is positive, so the producer selects the river/class factor rather than the national farm-class fallback. These are repeated lock passages, not unique river cargo.

Coal and county-value checks

211,680,312 short tons × $15.23/short ton = $3,223,891,151.76
round = $3,223,891,152

This reproduces the Wyoming-as-PRB proxy; mine-dollar placeholders do not enter that multiplication. The retained Central Valley county sum is $44,892,297,000; the Corn Belt selected counties’ all-farm sales sum is $74,720,406,000. Nevada’s conversion is 12,600 million USD ×1,000,000 = $12,600,000,000 statewide, with no gold-only or county allocation.

Limitations

The retained production inventory contains 189 records, not a complete census of U.S. production. Selected gateways, network exclusions and suppressed source cells limit coverage. Missing, withheld, unallocated and unassigned quantities are unknown, not observed zero; compact zero placeholders do not override source scope.

Measures overlap. A shipment can appear on successive network links, at a gateway and in regional trade. Parent and child production areas can overlap. These quantities cannot be added into a national physical-economy total. Gross shipment and production values are not GDP, value added or profit.
Assignment is modeled. Truck and rail links are not traffic counts or waybills. Pipeline geometry does not validate local throughput, gas/HGL compatibility, operator dispatch or an exact shipment route. Water lock ratios proxy change on links; they are not 2025 link meters.
Geography has different roles. FAF regions, customs reporting ports, production regions and contextual outlines do not necessarily coincide. Statewide mineral values are not local deposit values. County memberships and generalized boundaries are not certified production footprints.
Some provenance is incomplete. Original waterway tonnage input rows were not recovered for full replay. The STEO extraction’s release/acquisition lineage and some geometry update dates remain unverified. Other acquisition dates are unknown unless recorded. Data identifies these gaps by dataset.
Reconciliation is not an error bound. The retained tests reproduce arithmetic and accounting. They do not establish sampling-error intervals, route-validation accuracy or local allocation uncertainty. No unsupported confidence intervals are assigned.

Fourteen retained metric families and their interpretation limits
Family	Actual vintage / units	Interpretation limit
Freight regions & metros	2024 preliminary annual short tons; current USD	FAF domestic origin/destination regions; imports and exports are their U.S. freight legs, not foreign endpointsThe four-mode card detail cannot supply all-mode inbound. Separate eight-mode retained source tables were audited internally; they are not part of the current card calculation.
Roads	FAF 2024; BEA weights 2024; geometry snapshot short tons; current USD	Two-way modeled annual throughput on the clicked source interval, not the whole named routeRoute choice, county weights and access links are atlas modeling, additional to FAF’s own estimates.
Rail	FAF 2024; BEA weights 2024; geometry snapshot short tons; current USD	Two-way modeled annual throughput on the clicked source interval, not the whole named routeRoute choice, county weights and access links are atlas modeling, additional to FAF’s own estimates.
Waterways	observed CY2023 → estimated 2025; FAF value rates 2024 short tons; estimated current USD	Passages on retained inland waterway links, upbound + downboundLock passages are repeated counts; lock ratios are proxies for nearby links, not direct 2025 meters.
Pipelines	FAF 2024 scaled/calibrated using EIA 2025 estimated short tons; estimated USD	Estimated annual throughput assigned to real line geometry; coverage incompleteCalibration uses executable factor bounds [0.5,2]; it does not force exact EIA agreement.
Air cargo airports	2025, all twelve T-100 months short tons of freight + mail; allocated international USD	Airport handling on nonstop segments; international customs dollars allocated to airportsDomestic handling counts a domestic segment at both airports. Do not sum handling to obtain unique flights or national cargo.
International vessel trade	2025 annual (December year-to-date) short tons; current import/export USD	Census customs port/district reporting geography, not individual terminal or all port cargoDomestic coastal/inland vessel activity is excluded. Customs reporting point may differ from the physical facility.
Farms & crop regions	NASS sales 2022; CDFA 2023 (some 2022); crop survey 2025 USD farm sales/gross production; native bushels; corn mass conversion	Selected county sets; separate CDFA gross production and NASS farm salesCorn mass is a conventional conversion in the retained producer, not directly weighed corn. Do not convert soybeans, other crops or native energy volumes by that factor.
Factories & industrial areas	2022 current USD value of shipments	Disclosed county/subsector shipments for named county sets; FAF-region cards use county assignmentsChemical manufacturing 325 is a broad proxy, not pharmaceutical-only 3254. County pharma is unavailable.
Nonfuel minerals	2025 preliminary, published MCS2026 current USD; native physical output only where explicitly sourced	Whole-state mineral value displayed near selected deposits/countiesNot gold-only, copper-only or iron-only value. No independent localized state-value estimate is established.
Oil & gas producing regions	volumes and prices 2025 barrels/day; cubic feet/day; estimated current USD; not short tons	STEO basin/region estimates, some state/offshore totals; outlines need not match statistical geographyHenry Hub is a benchmark, not basin realized/wellhead revenue. Volume and price scopes differ.
Coal regions & mines	volume 2025; price 2024 short tons; estimated USD for parent	MSHA operator quarterly mine output; PRB parent uses all Wyoming coalMine and parent volumes overlap. Registry coordinates are location context, not historical production boundaries.
National statistics	FAF 2024; Census/T-1002025 annual short tons; current USD; shares	Separate national freight and international trade universes; not sums of map layersNo sum of road/rail/water/pipeline link throughput, gateway handling or overlapping region production.
Lines, labels, zoom & loading	current retained continuous-zoom runtime CSS pixels; zoom; display counts, not tons	Cartographic display of source facts; not additional dataWidth caps, minimum sizes and different unit scales prevent pixel-ratio comparisons; inspect the numeric card.