Autonomy · transportation · geography

Physical AI and the Department of Transportation

Opportunities for automation across diverse geographies: what the work is, which constraint binds it, and how far along each class actually sits.

Michael Hiner, Industrial Research Fellow · The Hiner Lab
With Ronnie Gomez and Grant Markhart, Industrial Research Fellows, and Dean David Jean Charlot.

1. Scope, classes and method

This review asks what road and bridge automation is worth to a transportation agency, and it asks the question four times, in four physical settings, because the answer is not the same in each. The unit of analysis is the terrain and climate class rather than the state agency or the vehicle mode. The reason for that choice is that terrain decides what the work physically is: a crew reaching an asset across a water gap, a crew waiting for a pass to open, a crew grading an earth road and a crew rebuilding a gorge alignment are four different problems that happen to share a federal funding code.

Four classes are examined, each with a named anchor and a named counterparty. The archipelagic class is anchored in Puerto Rico and the Philippines, against the Puerto Rico Highways and Transportation Authority and the Department of Public Works and Highways. The Appalachian mountain class covers West Virginia, Tennessee, North Carolina and Virginia. The cold and avalanche class covers Alaska with the Cascade and Rocky Mountain passes, against Alaska DOT and Public Facilities, the Washington State DOT and the Colorado DOT. The arid tribal and sparse rural class covers the Navajo Nation and Hopi lands, against the Bureau of Indian Affairs road programme and the Arizona DOT.

Every quantity carries one of four verification grades. A figure is verified when it comes from a primary series, a statutory instrument or a peer-reviewed publication and was retrieved from that source rather than from a summary of it. It is reported when a named source states it and this review did not independently confirm it. It is self-published when it originates with the party running the programme it describes, including that party's own forward-looking targets, regardless of which outlet relayed it. It is modelled when it is computed here from sourced inputs, which is to say it is the authors' own arithmetic and is never presented as a measurement. Of the 156 graded figures behind this review, 88 are verified, 38 reported, 18 modelled and 12 self-published.

Each hypothesis below resolves to three things. The first is a position on the trajectory, stated in degrees out of ninety, where ninety would mean the application is routine and commercially settled. The second is the binding constraint, named from material science, actual engineering implementation, energy, computation efficiency or AI algorithms, with regulation and workforce admitted as separate classes because in this domain they frequently dominate. The third is a threshold: the specific published quantity whose movement would change the answer. A hypothesis without a threshold carrying a number is not testable and is not included here.

One further discipline governs this review. Where a quantity was sought and not found, that is recorded as a quantity this review did not locate, together with what was searched, rather than as a statement that the quantity is unpublished. Fifty-seven such records were made. Several of them bear directly on conclusions below, and Section 11 sets out the ones that matter most.

Table 1. The four terrain and climate classes, their anchors and the constraint each was found to bind on. The constraint column is a result of this review, not an assumption of it.

2. The estate, its owners, and the two markets inside it

The United States public road network is 8,865,888.676 lane-miles across the fifty states and the District of Columbia, for reference year 2024 1. Measured as centerline miles the same network is 4,208,454 2, so which denominator a coverage claim uses changes the implied result by a factor of 2.1. The Interstate System is 231,104 lane-miles of the total, 2.6 percent of the estate 1. That 2.6 percent is the only part of the network carrying mandatory annual condition data over its full extent, and it is also where the largest share of work zone fatal crashes occurs, so the sensing case and the safety case coincide there and nowhere else.

Ownership decides who can sign a purchase order. State highway agencies own 827,408 of the 4,208,454 centerline miles, counties own 1.80 million and towns, townships and municipalities own 1.33 million 2. A programme led by a state DOT therefore commands directly under 20 percent of the estate. The recurring money runs the other way: of 78.394 billion dollars in maintenance and traffic services disbursements for 2024, local governments spent 43.028 billion and state agencies 35.090 billion 12. The larger half of the maintenance market sits with roughly three thousand county and nineteen thousand municipal owners, which is also the owner class with the least survey data and the least procurement capacity.

That data gap is measurable. Table HM-60 codes lane counts for every functional system except rural minor collector and rural and urban local, which the Federal Highway Administration estimates at two lanes 1. Adding the three affected systems as published gives 4,051,719.302 rural local, 1,776,947.814 urban local and 517,390.740 rural minor collector lane-miles, which is 6.35 million of the 8.87 million total resting on an assumed lane count. Any national per-lane-mile automation figure is therefore quoted against a denominator that is partly an assumption, and the binding constraint on the national claim is inventory data quality rather than machine capability.

The maintenance pool an automation vendor addresses is also smaller than the highway budget suggests. State highway agency maintenance was 24.756 billion dollars in 2024 against 116.294 billion dollars of capital outlay, a ratio of about one to 4.7 11. The Interstate absorbed 7.082 billion of that maintenance, 28.6 percent of the state total on 2.6 percent of national lane-miles 11. Federal formula money is apportioned through named programmes, and eligibility is a real gate: the National Highway Performance Program at 30.784 billion dollars is 54.2 percent of the 56.811 billion dollar anticipated FY 2026 apportionment and is the programme tied to the condition measures, while the Highway Safety Improvement Program, the channel for work zone countermeasures, is 3.246 billion, 5.7 percent of the total 13.

Table 2. The estate divided by what is measured rather than by who drives on it. The two rows in the middle are where federal rule creates recurring demand; the third row is where the money is.

3. What compliance requires, and why it gates the sale

The strongest constraint on automated condition assessment in this domain is not sensing performance. It is the shape of the output a federal rule will accept. Bridge condition is classified by the minimum of condition rating method applied to three integer human judgements on a zero to nine scale: National Bridge Inventory items 58, 59 and 60, or item 62 for culverts. A structure is Good when the lowest of the three is 7, 8 or 9, Fair when the lowest is 5 or 6, and Poor when the lowest is 4 or below 4. A system with better physical resolution than a human inspector still fails the compliance test if it cannot emit those three integers, and the decision boundary that carries money is one integer step, from a lowest rating of 5 to a lowest rating of 4.

Pavement is gated the same way and more tightly. A section is Good only if all three of roughness, cracking percent and rutting or faulting are Good, and Poor if two or more of the three are Poor 6. A sensor suite that measures roughness excellently but cannot report cracking percent produces data that cannot form the federal measure at all. This is the specific reason a vendor accuracy claim stated on the International Roughness Index alone does not establish compliance, and it is why automated pavement assessment is a multi-sensor engineering problem rather than a single-camera one.

The same rule that constrains the output also creates the demand. For the Interstate System, data must be collected from the full extent of the mainline in one consistent lane, continuously, in nominally uniform sections of 0.10 mile, annually, and estimating conditions from samples of the full extent is not permitted 6. Against 231,104 Interstate lane-miles that is on the order of 2.3 million reportable sections a year before the non-Interstate National Highway System is added. Federal rule already forces a data volume manual methods cannot meet, which is why automated data collection vehicles are the one part of this estate where automation is the default rather than a proposal. The binding constraint there has moved downstream, to the computation efficiency of processing and quality-assuring the resulting imagery and profiles.

Where the rule creates demand it does so unevenly, and the national averages conceal it. Interstate pavement has held between 61.1 and 62.2 percent Good and between 0.8 and 0.9 percent Poor across the five reported years to 2022 5, against a federal minimum-condition threshold of 5.0 percent Poor, or 10.0 percent in Alaska 6. The national estate sits at roughly one sixth of the penalty threshold, so no national financial pressure is pushing Interstate pavement automation and any real case is state-specific. Bridges are the opposite. Good deck area on the National Highway System fell in every one of six consecutive reported years, from 43.4 to 40.4 percent, while Poor also fell from 4.5 to 4.0 percent 5. Both moving down means the estate is draining out of Good into Fair faster than Poor is being repaired, and Fair-condition deck area is precisely where preventive work has the highest return. Of 409,265,353 square metres of national deck area, 219,125,109 are Fair 3.

Table 3. The federal instruments that decide whether an automated measurement counts. Each names a quantity, and in every case the quantity is an integer or a percentage rather than a physical resolution.

4. Work zones: separating the exposure from the addressable share

Removing people from live traffic is the strongest quantified argument for maintenance automation in the federal dataset, and it is also the one most often overstated. National work zone fatalities were 863 in 2020, 963 in 2021, 903 in 2022, 905 in 2023 and 850 in 2024 8. The series has not fallen below its 2020 level in five years, and a six percent year-on-year decrease inside a five-year range of 850 to 963 does not establish a trend. The honest position is that work zone fatality exposure is flat.

The addressable share is much smaller than the headline. In 2024, 673 of the 850 work zone deaths were drivers and passengers passing through, and 169 were pedestrians and pedalcyclists 8. Worker-removal automation can address the pedestrian category and not the occupant category, and that category also contains non-worker pedestrians, so the true ceiling is below 169. A claim that maintenance robotics addresses work zone deaths as a whole overstates the addressable share by roughly a factor of five. The occupational series is separate and smaller again: between 82 and 143 worker fatalities a year at road construction sites over 2015 to 2024, of which 52.7 percent over 2022 to 2024 were workers on foot struck by a motor vehicle and a further 24.8 percent were workers in a crash 14. The two series count different populations, by roadway location and by employment status, and must not be added.

Which automation intersects the causal mechanism is a separate question from which automation removes the worker. Speeding was a factor in 34 percent of fatal work zone crashes in 2024, up from 29 percent in 2023, while total fatalities fell, and more than half of fatal work zone crashes occurred at night 8. That combination favours automated speed enforcement, automated queue warning and automated lighting or delineation over maintenance robotics, because the dominant mechanism is approaching-driver behaviour. It also locates the exposure on the night shift, which is the one where a machine's insensitivity to darkness is a genuine physical advantage rather than a marketing claim. Sequencing follows from geography as well: on the 2022 to 2024 average the Interstate carried about 47 percent of fatal work zone crashes 9 while being 2.6 percent of national lane-miles 1. Finally, proportion matters: the 850 work zone fatalities of 2024 are 2.17 percent of the 39,254 national traffic fatalities that year 10.

Table 4. The work zone safety case sized honestly. The addressable population is the fourth row, not the first, and the first row has been flat for five years.

5. The archipelagic class: the gap prices logistics, not sensing

Puerto Rico carries 19,964 miles of public road, of which the state highway agency owns 4,761 and municipalities own 15,182 15. A 76 percent municipal share means an automated inspection or maintenance programme addresses an estate whose majority owner is 78 separate municipalities rather than the territorial DOT. The managed network is 4,851.82 centerline miles and 11,394.42 lane-miles, of which 83.5 percent of centerline length is non-National Highway System local and collector road 16. Bridge condition is drifting: all bridges moved from 428 Good and 332 Poor in 2024 17 to 414 Good and 344 Poor in 2025 on a rising total, with Fair holding at about 68 percent 18.

The event that resets this estate is quantified. Hurricane Maria made landfall near Yabucoa at 135 knots and 920 millibars, with an eyewall replacement that tripled eye diameter from 9 to about 28 nautical miles before landfall, which is why the reset was island-wide 19. Damage was estimated at 90 billion dollars against a previous island record of about 5 billion, an eighteen-fold jump 19. Eighty percent of utility poles and all transmission lines were knocked down 19, and restoring power to all customers with structures deemed safe took roughly eleven months 20. Any automation depending on grid power or roadside sensing inherits that interval as an availability schedule.

Where the damage fell is the finding that matters for the class thesis. The United States Geological Survey mapped 71,431 landslides across 72 of the 78 municipalities, with zero documented on the water-separated island municipalities of Culebra and Vieques and the highest densities in the Cordillera Central interior 22. The Federal Emergency Management Agency record agrees: Utuado recorded 330 roads-and-bridges projects and Orocovis 312, against two each for Vieques and Culebra 23. Within an archipelago, the binding physical constraint on the road estate was slope and rainfall, not separation by sea. The water gap binds somewhere else, on supply and crew access.

Priced in days rather than dollars the gap is sharper still. The Port of San Juan reopened for daylight operations three days after landfall, and every airport and seaport ran at limited capacity for about seven days, with few if any materials stockpiled locally and some equipment used on the island not standard in the continental United States 21. That is the clearest published statement of what separation by water does to the cost of reaching an asset, and it applies to a spare part for an automated machine exactly as it applied to a utility pole. The resupply channel itself is ageing: four Jones Act carriers operate seventeen vessels whose containerships average 39 years against an expected useful life of about 30, and whose barges average 31 against about 27 36. That is a material and engineering constraint on the channel, and the same audit states it could not validate the shipper-reported rate examples, which is the honest limit on how far a cost premium can be quantified from it.

The permanent-work phase, not the emergency phase, is where the time goes. Across 3,036 category C projects, first obligations ran 5 in 2018, 161 in 2019 and 1,785 in 2020, with 244 projects still receiving obligations in 2026 and the latest first obligation dated 27 February 2026 23. The median project was not obligated until about two and a half years after landfall. Federal quick-release road money by contrast moved within weeks, at 72.5 million dollars for Puerto Rico 26, two orders of magnitude below the 2.98 billion dollars eventually obligated for roads and bridges 24. Automation deployed to speed damage assessment is competing against a process whose slow step sits downstream of assessment. Of 23.4 billion dollars in Public Assistance awarded as of June 2023, 1.8 billion had been spent, and about 11.3 billion required federal authorisation before the territory could expend it, with subrecipients planning to access remaining funds through 2030 and possibly beyond 25.

The Philippine anchor separates the class further. Typhoon Haiyan struck Guiuan at 168 knots on a one-minute basis and 118 knots on a ten-minute basis at the same timestamp 29, a 40 percent disagreement that any wind-load specification must declare. Yet roads and bridges together were 1.65 billion of 89.6 billion pesos of damage, about 1.8 percent, while the social sector was 61 percent 30, and 1,084,762 houses were damaged of which 489,613 were totally destroyed 30. The whole national-road reset was about 105 kilometres 31, roughly 0.3 percent of the national network. The same terrain class produced a road-dominated reset in one anchor and a housing-dominated reset in the other, which constrains any claim that the class alone fixes the binding constraint. What the water gap does price is movement: domestic inter-island freight runs at 1.20 dollars per nautical mile Manila to Davao against 0.40 for Hong Kong to Manila, a factor of 3.0 34. That multiple applies to a spare part or a replacement machine exactly as it applies to a pole. It is also not fixed by geography, because a change of crossing mode to roll-on roll-off moved commodity freight costs by 20 to 68 percent 34, a larger swing than most automation efficiency claims.

Two further quantities bound what monitoring is worth in this class. On the National Highway System, the territory's own asset management plan identifies six locations needing repair more than once across 25 years of declared emergencies 27. Predictive site-level monitoring on that network has almost no repeat-failure signal to learn from, because the reset is a whole-system event rather than an accumulation of recurring points. And the reset dominates the maintenance cycle rather than the reverse: the 2.98 billion dollars obligated for roads and bridges after Maria 24 is more than five times the highway authority's roughly 560 million dollar annual capital expenditure 28. The Philippine appropriation shows the water gap entering a maintenance budget as a standing line: alongside 8.502 billion pesos for routine maintenance of national roads and 1.788 billion for national bridges sits 1.342 billion pesos for maintenance and operations of departmental floating equipment 32, a cost with no analogue in a continental road agency. At the crossing itself the charge is published per lane metre: 283 to 366 pesos per lane metre and 1.5 to 2 hours of transit on a 24 nautical mile Batangas to Calapan leg 35, which is the marginal cost an autonomous road vehicle would still incur at an archipelagic route break.

Table 5. The archipelagic class. The two rows in bold contrast are the second and the fifth: within the island the mountain binds, and between the islands the crossing binds.

6. The Appalachian mountain class: the link budget sets the usage

The Appalachian Regional Commission region is 423 counties across 13 states, 206,000 square miles and 26.6 million residents 37. Its purpose-built corridor network is 92.1 percent built after six decades: 2,846 of 3,090 authorised miles complete, open or under construction as of 30 September 2025, with 197 miles still in location studies 42. The cost to complete the remainder was estimated at 15.154 billion dollars in 2025 dollars, up from 10.3 billion in 2021 dollars, an escalation the Commission attributes partly to national highway construction costs rising about 14 percent a year since 2021 43. Dividing the state cost-to-complete figures by miles not open to traffic gives 43.5 million dollars per mile in Tennessee, 39.2 million in Virginia, 37.3 million in West Virginia and 26.2 million in North Carolina 43, an order of magnitude above ordinary rural two-lane construction. The cheapest route to automated freight in this class is to route onto existing corridors rather than to expect new geometry to be built for it.

The class is not economically uniform, and that matters for any labour-substitution case. Appalachian North Carolina, the terrain Helene hit hardest, contains zero distressed counties, while West Virginia contains eleven distressed and twenty at risk 38. Per capita market income runs 32,783 dollars in Appalachian Virginia, 57.4 percent of the national 57,066, against 40,589 in Appalachian North Carolina 38. The wage a machine must beat is roughly four sevenths of the national figure in the part of the class where capital cost per mile is highest, which pushes the break-even point outward exactly where the terrain argument would push it inward.

Hurricane Helene supplies the stress case. Peak rainfall reached 30.78 inches at Busick in Yancey County, North Carolina, against 6.11 inches at Bluefield, West Virginia 44, a factor of five in the forcing term inside one terrain class. North Carolina recorded 107 of the 250 United States deaths, with more than a thousand of the 2,700 high-water rescues from freshwater flooding in the western mountains 44. The state assessed roughly 5,000 miles of state-maintained road as severely impacted, with 674 bridges and 712 culverts damaged and more than 7,000 private roads, bridges and culverts affected 48. Against 80,592 state-agency miles 39, 5,000 miles is 6.2 percent of the state-maintained system rendered unusable in a single event.

The legal footprint of the event shows the class boundary again. North Carolina's declaration covered 40 designated areas with both household and public assistance, Virginia's 18 and 37 areas respectively, and Tennessee's 8 and 14, while West Virginia's covered one county with household assistance and no public assistance at all 45. Obligations follow: 2.401 billion dollars of public assistance in North Carolina against 582.9 million in Tennessee, 101.7 million in Virginia and none in West Virginia 46. Federal road money moved fastest of all, with quick releases issued the day before and the day of landfall, at 250 million dollars to North Carolina against 70 million to Tennessee and 10 million to Virginia 53. North Carolina's own assessment put total damage and needs above 59.6 billion dollars 47, which is 24.8 times what federal public assistance has actually obligated, and that gap is the quantity a resilience argument turns on. Within the transportation total of 10.342 billion dollars, 2.185 billion of the 8.185 billion of road and bridge damage sits on roads no state DOT controls, 27 percent 47.

Restoration duration is measurable and long in the tail. NCDOT reported 6,900 damaged sites on 14 October 2024 50 and nearly 9,500 at eighteen months 49, so the assessment itself took about a year to converge: reporting latency, not repair latency, dominated the first phase. At eighteen months 87 percent of sites were repaired, and by July 2026 more than 99 percent of closed roads had reopened 51. The residual is where the terrain is: Interstate 40 through the Pigeon River Gorge is a two billion dollar rebuild targeted for late 2028, and permanent reconstruction on US 64, US 74A and US 176 runs to 2029 51. On the Blue Ridge Parkway, at least 57 landslides affected nearly 200 miles in North Carolina, roughly one per 3.5 miles, with more than four dozen in a single 38-mile section 52. Slope instability persists after the water recedes, so an automated route through gorge terrain faces a five-year window in which the road geometry itself is a moving target. That is a mapping and change-detection constraint rather than a driving-policy one.

The constraint that binds this class hardest is the mobile link, and it is priced. Rural combined fixed and mobile coverage at the end of 2022 was 29.4 percent in West Virginia against 52.3 percent in North Carolina, 51.0 percent in Tennessee and 42.1 percent in Virginia 54, a 23-point spread inside one class. At county resolution the spread is wider and does not track density: Ashe County, North Carolina reports 99.5 percent fixed and 21.1 percent mobile at 63.6 people per square mile, while Graham County reports 43.0 percent fixed and 28.1 percent mobile at 27.3 54. In Buchanan County, Virginia the figures are 91.0 percent fixed and 1.8 percent mobile 54, which isolates the failure precisely to the layer a moving vehicle needs. In Pocahontas County, West Virginia both layers fail together at 0.7 percent combined 54.

That constraint is moving, and this is the fastest-moving quantity located in the review. Appalachian household broadband subscription rose 11.1 percentage points between the 2014 to 2018 and 2019 to 2023 American Community Survey periods, to 86.2 percent against 89.7 percent nationally 55. Committed capital is allocated: 1.533 billion dollars to North Carolina, 1.481 billion to Virginia, 1.211 billion to West Virginia and 0.813 billion to Tennessee under the Broadband Equity, Access and Deployment programme 56. What that money buys is where terrain shows itself most clearly. Inside western North Carolina the average build cost is 4,509 dollars per location, but McDowell County runs at 18,274.73 dollars and six counties are served entirely by low-earth-orbit satellite at 600 dollars per location 58, a thirty-fold spread within one mountain region. Where the terrain wins, the state programme chose an orbital link over a trench. North Carolina's own award schedule puts service by 2030 with projects launching in mid-2026 57, which dates the earliest point at which this constraint could be materially relaxed.

Table 6. The Appalachian class. The last row is the strongest single piece of terrain evidence in this review: at the extreme, terrain does not raise the price of the trench, it replaces the trench with a satellite.

7. The cold and avalanche class: two availability markets, not one

Availability, not capability, is the governing variable in this class, and it comes in two regimes that respond to different interventions. The seasonal regime is measured in months. State Route 20 over the North Cascades closed on 18 November 2024 and opened on 22 April 2025, 155 days or 42.5 percent of the year 59,60; the following winter it closed on 4 December 2025 and reopened on 25 June 2026, 203 days or 55.6 percent, the latest opening on record 61. State Route 410 over Chinook Pass closed for 204 days in 2024 to 2025, 55.9 percent 59. Colorado 82 over Independence Pass, at a 12,095 foot summit, closes about 195 days on its typical dates, 53.4 percent 63. Two ranges and a 6,600 foot elevation difference produce the same answer, which suggests the closure fraction in this class is set by the operating decision rule rather than by elevation.

The segment over which that penalty applies is small enough to matter. The North Cascades seasonal gate closure runs from milepost 134 to milepost 171, 37 miles 62, and the Independence Pass closure covers about 37 miles between mileposts 47.2 and 84.2 63. Both are within single-charge range of current ground platforms, so on these segments the constraint is calendar access rather than platform endurance. The severity envelope on the Alaska anchor is a different matter: Thompson Pass receives nearly 46 feet of snow a year, and a single station at 6,620 feet is treated as notable enough to publicise 77, which indicates how sparse the instrumentation baseline is and makes enclosure survival a materials problem before it is a sensing one.

The seasonal constraint is written into the procurement instrument, which makes it harder to move than a weather pattern. Alaska highway construction contracts require seeding between 15 May and 15 August absent written approval, a 93-day window, 25.5 percent of the year, for the closing activity of a project 64. Asphalt may not be placed when air temperature is below 40 degrees Fahrenheit, and hot mix leveling course may not be placed when the roadway surface is colder than 40 degrees 64. Seasonal suspension of work and winter shutdown are standing contract terms 64. This names the physical variable exactly: a paving process tolerant below 40 degrees Fahrenheit would extend availability, whereas a faster or autonomous paver operating under the same threshold would not. The constraint here is material science, and an autonomous construction fleet in this class is idle by contract for part of the year, which changes the utilisation assumption under any capital-intensive automation.

The incident regime is measured in hours and is where deployed automation already sits. Colorado recorded 1,707 hours of avalanche control closures in the 2018 to 2019 season, with 1,615 explosions, 877 avalanches triggered and 51 control missions 65. Of 522 identified avalanche paths above Colorado highways, 278 are regularly monitored and controlled 65, leaving a 244-path gap. Remote control systems on three paths at Red Mountain Pass cost 744,900 dollars, roughly 248,000 dollars per controlled path 67, which sets the order of magnitude for closing that gap and is an engineering-implementation cost rather than an algorithmic one. Alaska sits earlier: two instrumented infrasound locations statewide against 200 paths reported on the Seward Highway alone, funded by a 1.1 million dollar federal grant 68. The agency's own page counts 60 common avalanche areas statewide 69, which is a different unit of account from the programme manager's 200 paths on one highway, and this review did not locate a published definition reconciling the two. Any coverage claim in this class has to state which unit it uses.

Forecasting and clearing sit on opposite sides of the automation case. Colorado's contribution to statewide highway avalanche forecasting is 1,107,265 dollars a year, 30 percent of the forecasting centre's revenue, of which 67 percent of expenditure is personnel 66. Against 522 identified paths the per-path forecasting cost is low, so automation savings in forecasting are small in absolute terms. Clearing is the opposite. A single avalanche closed the Richardson Highway at milepost 12 to 18 on 24 January 2014, with full reopening on 5 February: twelve days on the only road to Valdez, 200,000 cubic yards of snow removed from debris about 40 feet deep, with cleanup running 24 hours a day for five days 70. That is a materials-handling problem, not a sensing or decision problem. Exposure decides where an hour is worth most: Interstate 90 over Snoqualmie Pass carries about 28,000 vehicles a day of which about 5,600 are freight, under more than 450 inches of annual snowfall, against about 4,500 vehicles a day on US 2 over Stevens Pass 73.

Two findings cut against a simple cold-is-the-constraint reading. On five-year averages, winter maintenance costs 3,043.44 dollars per lane-mile in Colorado and 3,072.42 in Washington, against 1,541.37 in Alaska and 1,205.26 in Montana 72. If cold alone set the cost the ordering would reverse. Workforce density behaves the same way: 67.4 winter maintenance workers per thousand state-maintained lane-miles in Colorado and 68.2 in Washington, against 36.3 in Alaska, 28.9 in Montana and 28.1 in Wyoming 71. Colorado and Washington, not Alaska, hold the largest per-lane-mile labour pool a substituting automation could displace. A caveat belongs with those figures: Alaska's return reports labor and equipment costs summing to 26.387 million dollars in 2024 to 2025 against a stated snow and ice total of 10.035 million, with materials blank, and the same pattern appears in the two prior years 71. This review did not locate an Alaska primary source reconciling them, so the cold-cost question cannot currently be settled using that state.

Where the displaceable workforce actually is, it is already thin. The Anchorage maintenance station runs six operators on an average shift across 1,187 lane-miles of highway and 160 miles of sidewalk, after a stated 27 percent regional budget reduction and fifteen lost positions 76. This is the clearest case in the review where automation is asked to substitute for a workforce already cut rather than to slow a growing one. The equipment mix also does not transfer between members of the class: Alaska runs 8.4 road graders per thousand lane-miles against Colorado's 3.6 and Washington's 1.8, and applies 12.93 tons of abrasive per lane-mile where Colorado applies none 71. A platform validated on a plough-and-chemical regime is not validated on a grader-and-abrasive one. Finally, the benchmark any automated intervention must beat is not zero: a regulatory change, Colorado's updated Traction Law, coincided with closure hours falling 21 percent for crash closures and 35 percent for non-crash emergency responses in a single autumn quarter comparison 75.

Table 7. The cold class holds two availability markets. They are measured in different units, bind on different constraints, and only one of them currently has automation deployed against it.

8. The arid tribal and sparse rural class: the surface, then the wire

The distinguishing variable in this class is not how far apart the roads are. Navajo Nation route density is 0.580 route-miles per square mile against 0.662 for Arizona statewide and 1.187 for the United States 84, so the network is 88 percent as dense as its own state. What distinguishes it is the surface. Of 14,167.6 total roadway miles, only 23.4 percent are paved, and four owners share the network: the Bureau of Indian Affairs holds 5,994.5 miles, the Nation 4,889.9, the state 1,644.8 and counties 1,638.4 79. The federal component is starker still: 6,017 BIA-owned miles in the Navajo Region comprise 1,506 paved, 90 gravel and 4,420 earth 78. Nationally the discontinuity is an ownership boundary rather than a gradient, with 5.8 percent of tribally owned inventory miles paved against 93.7 percent of state-owned miles inside the same inventory 78. An automated vehicle crossing from a state route onto a tribal route changes problem class at the property line.

The class sits inside a larger federal estate with the same properties. The Bureau of Indian Affairs maintains 29,300 miles of road and 1,080 bridges nationally, of which approximately 16,900 miles are unimproved or earth surface, 57.7 percent, inside a total public network serving Indian Country of more than 156,000 miles 80. The Navajo Region alone holds about a fifth of BIA road miles, and it carries 66.14 million dollars of the 289.67 million dollar national BIA deferred maintenance backlog, 22.8 percent 101. Scale sets the amortisation problem: the Navajo Nation reservation and off-reservation trust land measure 24,429.534 square miles 81, and widely repeated secondary figures of 27,000 square miles run 10.5 percent above that census measurement. Hopi lands, 2,532.07 square miles 81, sit enclosed within the Navajo estate with 6,377 residents and 2,483 housing units 83, so any route network serving Hopi villages crosses Navajo jurisdiction and utilisation rather than capability governs the case there.

Paved does not mean machine-legible here either. About 10 percent of the paved system is in good or better condition, 20 percent fair, and the remaining 70 percent poor or failing 79. Lane-marking retroreflectivity, edge definition and surface friction assumptions carried over from state-highway conditions do not hold. The price ladder for changing that is published: 750 dollars per mile for blading a dirt road, 2,000 for gravel maintenance and 6,000 to 10,000 for paved maintenance 79, against 250,000 dollars per mile to gravel a road and 3 million per mile for permanent pavement 103. Total funding available over a twenty-year period is 820 million dollars against a stated need of 7.9 billion 79, a gap of 9.6 to one. If the surface does not change, the automation problem does not change, and no vehicle-side advance alters that.

Energy in this class is a delivery cost per point of presence, not a generation problem. Distributed photovoltaic capacity factor on Navajo lands is 27.9 percent, the joint highest of any tribal area in the country and 1.78 times the national residential mean of 15.7 percent 92. What binds is the wire and the meter. The utility reports 4.06 customers per mile of line against a stated urban comparison of about 500, a ratio of about 123 87, and states a cost of approximately 40,000 dollars to connect one home 86. Reliability compounds it: utility customers in Arizona lost 506.61 service-minutes in 2024 against an Arizona statewide average of 87.5, and 305.25 minutes even with major event days removed against a statewide 87.5 88. Any roadside asset depending on mains power in this class needs local storage sized to a several-hundred-minute annual outage budget rather than the tens of minutes assumed in urban designs.

Headroom on the distribution system is smaller than customer counts suggest. Connected households in this class consume 5,995 kilowatt-hours per residential customer per year against a national 10,072, 59.5 percent of the national figure 89. Adding vehicle charging to feeders sized for six megawatt-hour households is a distribution-capacity question rather than a generation one, which is the same conclusion the capacity-factor figure reaches from the opposite direction.

The trajectory on electrification is real and slow. Households lacking electricity moved from 36.7 percent in 2000 85 to an estimated 21 percent, 14,063 homes, in the 2023 report to Congress 85. The volunteer connection channel achieved more than 230 homes in 2019, 137 in 2022, 159 in 2023, 170 in 2024 and at least 200 in 2025 91, at about 3.1 homes per mile of new line 87. At roughly 170 homes a year against a residual of 10,000 to 14,063, that channel alone closes the gap on a multi-decade horizon. Two measurement cautions belong here. The 21 percent share rests on a housing denominator of 68,101 that exceeds the 2020 Census count of 57,479 units by 18.5 percent 82, and the independent check is the utility's own 43,308 metered connections 88. Separately, a primary series that does not depend on any grid estimate points the same way: of 50,333 occupied units, 32,696 heat with wood and 6,337 with electricity 90.

The reporting layer fails alongside the power layer. Of 50,333 Navajo households, 26,255 have no internet access of any kind, 52.2 percent, and 21,740 have broadband of any type 93; the Hopi figures are statistically indistinguishable at a tenth the land area 94. On Arizona federal reservations, 19.6 percent have fixed terrestrial 100/20 service and 17.3 percent have both fixed and mobile, against 65.9 percent for all tribal lands nationally 95. The metric that matters for a moving vehicle is worse and is not improving monotonically: outdoor stationary 5G-NR at 35/3 on rural federal reservations fell from 71.8 percent in 2021 to 53.8 percent in 2022 95. The Nation's own plan prices the fix at approximately 76 new towers and 1,529 miles of fibre for 267 million dollars of middle-mile work, excluding permitting and right-of-way 96, which is a connectivity build roughly one tenth the length of the road estate it would serve.

The demand side cuts both ways, and both directions belong in the record. Against automation: under the state's own medium forecast Navajo County population falls 12 percent by 2060, corridor employment falls from 669 to 588, and population around the SR 264 corridor already dropped 15 percent between 2020 and 2023 99. A per-vehicle business case here faces a shrinking denominator that no engineering advance moves. For automation: 6,946 of 50,333 households, 13.8 percent, have no vehicle available 102, 17 percent of the SR 264 corridor population has no vehicle and 48 percent lack reliable internet 97, and in the Four Corners region 583 fatalities over 2013 to 2019 included 61 percent unrestrained occupants, 43 percent with a driver at or above 0.08 blood alcohol and 36 percent involving speeding 100. Those three mechanisms are precisely what an automated driving system addresses. The per-capita safety return in this class is unusually high even though the traffic volume is unusually low: SR 264 carries 1,300 to just under 3,000 vehicles a day at speeds up to 65 miles per hour across 323 access points including 113 dirt driveways 98. That is unprotected high-speed conflict at low volume, the inverse of the dense low-speed conflict urban deployments have been tuned against. It cannot currently be demonstrated from records either, because the state agency of record states its own crash dataset is incomplete on tribal land 98.

Table 8. The arid class layer by layer. The energy rows are the ones that matter for framing: the resource is the best in the country and the delivery is the worst, so the constraint is the wire rather than the joules.

9. Does terrain partition the problem? The refuter tested

This review was designed to include a hypothesis capable of overturning its own organising choice, and that hypothesis partly survives. It states that inside the Appalachian class the automation-relevant quantities sort by ownership and connectivity rather than by slope, which if it holds means terrain is the wrong partition and the central claim fails on its own evidence. The test quantity is the share of bridges in poor condition, and it splits the class: 17.79 percent in West Virginia, 7.14 in North Carolina, 4.35 in Tennessee and 3.43 in Virginia, against a national 6.68 percent 41. Two of the four steep-terrain anchor states sit below the national figure. Only West Virginia is an outlier, at 2.66 times national. Steep terrain does not by itself predict deteriorated structures.

The alternative the hypothesis proposes does no better on the same four points. State highway agency ownership shares, computed from the published mileage, are 87.9 percent in West Virginia, 78.9 in Virginia, 74.1 in North Carolina and 14.7 in Tennessee 39. The most consolidated state has the worst bridges and the most fragmented state has nearly the best, so the ordering does not follow ownership either. Ranked against poor-bridge share the two series give a Spearman coefficient of 0.4, and rural mileage share 40 gives 0.8; on four observations a coefficient of 1.0 is required before anything can be said. Removing West Virginia leaves three states at 3.43, 4.35 and 7.14 percent, straddling the national figure. The honest reading is that on four anchor states no candidate sorting variable can be distinguished, and one outlier drives the entire picture.

There is a deeper reason the test cannot settle the question, and it is worth stating plainly because it also limits what this review claims. The terrain variable itself is unmeasured. This review searched the Federal Highway Administration Highway Statistics series, state route-characteristics documentation and design documentation for the four states and did not locate a published percent-grade or curvature distribution for any of them. The only quantified terrain statements located are a state budget office's report that western North Carolina has 25 percent more public bridges than the state average, that municipal and private roads are 48 percent of the region's total against 41 percent statewide, and that average elevation is two to three times that of the Piedmont 48. A partition cannot be confirmed or refuted against a variable nobody publishes. On the evidence needed to decide it, this question sits at about 5 of 90 degrees.

Poor-bridge share is in any case the wrong outcome to test terrain against. It is a legacy of decades of capital spending and inspection practice, not a measure of what the automation work is. Where this review does measure the work, the class separation is large and it is measured rather than inferred: 123 times on customers per mile of electrical line 87, 30 times on per-location connectivity build cost inside a single mountain region 58, 3.0 times on per-nautical-mile domestic freight 34, and a factor of about two on the fraction of the year a corridor is open 59,60,63. Terrain is therefore retained here as a partition of the physical work and of the constraint that binds it, and it is explicitly not claimed as a predictor of asset condition. Section 10 states what would settle the question: publication of a percent-grade or minimum-radius distribution for state-maintained mileage in the four anchor states, tested against those four poor-bridge shares.

Table 9. The refuter tested, including against its own alternative. Neither ownership nor rural share sorts poor-bridge condition on four states, and the terrain variable that would settle it was not located in published form. The final row is the authors' arithmetic on the four preceding rows.

10. Position and binding constraint, hypothesis by hypothesis

Eleven hypotheses were stated before the evidence was assembled, and each resolves below to a position in degrees, a binding constraint and a threshold carrying a quantity. The distribution of constraints is itself the review's most portable result. Five of the eleven bind on regulation, three on workforce, three on engineering implementation, one on material science and one on energy. Not one binds on AI algorithms, and only one binds on computation efficiency, and that one sits downstream of a compliance rule rather than upstream of a capability. On this estate, the question of whether a machine can do the work is settled more often than the question of whether the work can be bought, measured or reached.

Positions cluster low, and they cluster low for a reason that is visible in the evidence. The single application already at scale is automated pavement data collection on the Interstate, and it is at scale because 23 CFR 490.309 forbids sampling and thereby creates a data volume no manual method can meet 6. Where no rule creates the volume, adoption sits below 30 degrees almost everywhere in this review. The exception in the other direction is instructive: remote avalanche control in Colorado is at 278 of 522 paths 65 without any federal mandate, because the closure hours it removes have a price on a corridor where an hour is worth a stated two million dollars 74. Demand created by rule and demand created by corridor value are the two mechanisms visible here, and neither is a technology mechanism.

Table 10. All eleven hypotheses. Five bind on regulation and three on workforce; none binds on AI algorithms. Positions are the authors' assessments against the cited quantities, not measurements.

11. What this review did not locate

Fifty-seven quantities were searched for and not located. Recording them is part of the method, because an absent figure that is quietly replaced by an estimate becomes indistinguishable from a measurement two citations later. Four of the absences bear directly on conclusions above and are stated here with what was searched.

First, no published percent-grade or curvature distribution was located for state-maintained mileage in West Virginia, Tennessee, North Carolina or Virginia. The Highway Statistics HM series, NCDOT route-characteristics field descriptions, Tennessee design documentation and open search were used. This is the variable that would settle Section 9, and its absence is the reason that section reports a limit rather than a conclusion. Second, no federal series publishing state maintenance expenditure per lane-mile as a ratio was located; the full Highway Statistics 2024 table index was searched by title and the two component tables were retrieved and parsed 11,1. A quotient computed from them would be the authors' own arithmetic and would in any case be miscast, because the expenditure table covers only state highway agencies while the lane-mile table covers all owners. Third, no federal series recording lane-miles maintained per full-time-equivalent maintenance worker was located, which is why the workforce figures in Section 7 come from a pooled-fund programme of self-reported state returns 71 rather than from a federal count. Fourth, no documented automated-driving, connected-vehicle or teleoperation pilot on Navajo Nation or Hopi roads was located; the state DOT corridor study final report of December 2025 contains no intelligent-transportation, connected-vehicle or automation element in its recommendations 98.

Several absences are properties of the sources rather than of the subject, and they constrain what can be claimed rather than what is true. Philippine national road length here is the 2011 figure of about 31,400 kilometres 33 because the current departmental atlas could not be retrieved from its host. Puerto Rico route-kilometre and bridge counts for Vieques and Culebra separately from the main island were not located. Current county-level broadband availability was not obtainable, so the availability figures throughout carry a 31 December 2022 reference date 54,95. Alaska's winter maintenance cost components do not reconcile in the only dataset that publishes them, and no primary source reconciling them was located 71, so the question of whether cost per lane-mile rises with cold cannot be settled using the coldest state in the class. Each of these is a limit on this review, and each is stated so that a later reading can close it rather than inherit it.

Table 11. The absences that constrain the conclusions above. This review did not locate these quantities; that is a statement about this search, not about whether the figures exist.

References

1. 1 FHWA, Highway Statistics 2024, Table HM-60 'Functional System Lane-Length - 2024', sheet A, row 'U.S. Total', column 'TOTAL LANE MILES'; table dated October 17, 2025. Parsed from the published workbook, not from a page summary. https://www.fhwa.dot.gov/policyinformation/statistics/2024/xls/hm60.xlsx verified
2. 2 FHWA, Highway Statistics 2024, Table HM-10 'Public Road Length - 2024, Miles by Ownership', sheet A, row 'U.S. Total'; cross-checked against Table HM-20 'Public Road Length - 2024, Miles by Functional System', sheet A, row 'U.S. Total'. Both tables dated October 17, 2025. https://www.fhwa.dot.gov/policyinformation/statistics/2024/xls/hm10.xlsx verified
3. 3 FHWA Office of Bridges and Structures, National Bridge Inventory, 'Bridge Condition by Highway System', sheet '2025', row 'TOTALS', columns All/Good/Fair/Poor; prior-year figures from sheet '2024' of the same workbook. https://www.fhwa.dot.gov/bridge/nbi/no10/condition25.xlsx verified
4. 4 23 CFR 490.409, 'Calculation of National performance management measures for assessing bridge condition', paragraphs (b)(1)-(b)(3) and (c)(1); retrieved as XML from the eCFR versioner API for title 23 part 490 subpart D, current as of 2026-08-01. https://www.ecfr.gov/api/versioner/v1/full/2026-08-01/title-23.xml?part=490&subpart=D verified
5. 5 FHWA Transportation Performance Management, National Performance Dashboard, table 'Performance Measures: Interstate Pavement', section 'Pavement Condition on the Interstate System'. https://www.fhwa.dot.gov/tpm/reporting/national/ verified
6. 6 23 CFR 490.313, 'Calculation of National performance management measures for assessing pavement condition', paragraphs (b)(1) and (c)(1)-(c)(4); retrieved as XML from the eCFR versioner API for title 23 part 490 subpart C, current as of 2026-08-01. https://www.ecfr.gov/api/versioner/v1/full/2026-08-01/title-23.xml?part=490&subpart=C verified
7. 7 FHWA, Highway Statistics 2024, Table HM-47 'National Highway System, Miles by Measured Pavement Roughness', sheet B (Total Rural NHS) and sheet D (Total Urban NHS), row 'U.S. Total'; table dated November 2025. https://www.fhwa.dot.gov/policyinformation/statistics/2024/xls/hm47.xlsx verified
8. 8 FHWA, '2026 National Work Zone Awareness Week' fact sheet, publication FHWA-HOP-26-060, chart 'Work Zone Fatal Crashes and Fatalities'; data source stated as NHTSA FARS 2020-2023 Final Files and 2024 Annual Report File, 50 states, DC and Puerto Rico, as reported by NHTSA as of April 2026. Year-to-value mapping independently confirmed against the National Work Zone Safety Information Clearinghouse national data table. https://ops.fhwa.dot.gov/publications/HOP-26-060_FHWA_NWZAW_Factsheet.pdf verified
9. 9 FHWA, '2026 National Work Zone Awareness Week' fact sheet, publication FHWA-HOP-26-060, chart 'Distribution of Fatal Work Zone Crashes by Functional Classification and Land Use, 2022-2024 Average'; data source NHTSA FARS 2022-2023 Final Files and 2024 Annual Report File. Graded reported rather than verified because the urban/rural assignment of each pair was read from chart geometry in the PDF rather than from a labelled data table; the percentages themselves are as published. https://ops.fhwa.dot.gov/publications/HOP-26-060_FHWA_NWZAW_Factsheet.pdf reported
10. 10 NHTSA National Center for Statistics and Analysis, 'Overview of Motor Vehicle Traffic Crashes in 2024', Research Note DOT HS 813 791, April 2026, pages 1-2. Fatality counts are FARS (census); injury and crash counts are CRSS estimates. https://crashstats.nhtsa.dot.gov/Api/Public/ViewPublication/813791 verified
11. 11 FHWA, Highway Statistics 2024, Table SF-12 'State Highway Agency Capital Outlay and Maintenance - 2024', sheet SF12P4 'Total for All Areas - Classified by Functional Systems', row 'Total', maintenance columns and TOTAL column; table dated February 2026. Components verified to sum exactly to the published total. https://www.fhwa.dot.gov/policyinformation/statistics/2024/xls/sf12.xlsx verified
12. 12 FHWA, Highway Statistics 2024, Table HF-10 'Funding for Highways and Disbursements for Highways, All Units of Government - 2024', sheet HF10, rows 'Maintenance and Traffic Services' (subtotal row 59, component rows 56-58) and 'Grand Total Disbursements'; table dated March 1, 2026. https://www.fhwa.dot.gov/policyinformation/statistics/2024/xls/hf10.xlsx verified
13. 13 FHWA Notice N 4510.903, 'Fiscal Year 2026 Federal-Aid Highway Apportionments', issued June 2025, apportionment table, row 'Apportioned Total'. Graded reported rather than verified because the notice states these are the 'anticipated amounts of funds to be apportioned' under 23 U.S.C. 104 and are 'subject to change before issuing the certificate of apportionment, which is effective on October 1, 2025'; per-state rows verified to sum to the published total. https://www.fhwa.dot.gov/legsregs/directives/notices/n4510903.cfm reported
14. 14 National Work Zone Safety Information Clearinghouse (operated by the ARTBA Transportation Development Foundation in cooperation with FHWA and the Texas A&M Transportation Institute), 'Worker Fatalities and Injuries at Road Construction Sites'. The Clearinghouse cites as its origin the Bureau of Labor Statistics Census of Fatal Occupational Injuries; BLS defines a road construction site as a location of construction, maintenance and utility work on a road, street or highway, and national totals cover the 50 states and DC only. The 52.7 and 24.8 percent values were confirmed present in the page source. https://workzonesafety.org/work-zone-data/worker-fatalities-and-injuries-at-road-construction-sites/ reported
15. 15 US Federal Highway Administration, Highway Statistics 2023, Table HM-10, 'Public Road Length - 2023, Miles By Ownership', Puerto Rico row. https://www.fhwa.dot.gov/policyinformation/statistics/2023/hm10.cfm verified
16. 16 Puerto Rico Transportation Asset Management Plan 2032, Final, 29 December 2022, Table 3-1 'Comparison of Traffic, Miles, and NHS Among Small States', p. 3-2 (sources given as FHWA Statistics Tables and PRHTA roadway inventory). https://www.tam-portal.com/wp-content/uploads/sites/12/2024/04/Puerto-Rico-2022-12-29-BIL-Compliant-TAMP-2032.pdf verified
17. 17 US Federal Highway Administration, National Bridge Inventory, 'Bridge Condition by Highway System 2024', Puerto Rico rows. https://www.fhwa.dot.gov/Bridge/nbi/no10/condition24.cfm verified
18. 18 US Federal Highway Administration, National Bridge Inventory, 'Bridge Condition by Highway System 2025', Puerto Rico rows. https://www.fhwa.dot.gov/bridge/nbi/no10/condition25.cfm verified
19. 19 National Hurricane Center, Tropical Cyclone Report: Hurricane Maria (AL152017), R.J. Pasch, A.B. Penny, R. Berg, report dated 5 April 2018 (third version 14 February 2019), Table 1 'Best track for Hurricane Maria, 16-30 September 2017', p. 14, and narrative pp. 4-5. https://www.nhc.noaa.gov/data/tcr/AL152017_Maria.pdf verified
20. 20 US Government Accountability Office, GAO-20-141, 'Puerto Rico Electricity Grid Recovery: Better Information and Enhanced Coordination Is Needed to Address Challenges', 8 October 2019, p. 1 (and footnote 1), p. 36, p. 37. https://files.gao.gov/reports/GAO-20-141/index.html verified
21. 21 US Government Accountability Office, GAO-19-296, '2017 Hurricane Season: Federal Support for Electricity Grid Restoration in the U.S. Virgin Islands and Puerto Rico', March 2019, 'Logistical Challenges and Other Factors Affected Federal Support', p. 18, citing FEMA 2017 Hurricane Season After-Action Report (12 July 2018). https://www.gao.gov/assets/gao-19-296.pdf verified
22. 22 Hughes, K.S., Bayouth Garcia, D., Martinez Milian, G.O., Schulz, W.H., and Baum, R.L., 2019, 'Map of slope-failure locations in Puerto Rico after Hurricane Maria', U.S. Geological Survey data release, published 21 August 2019, doi:10.5066/P9BVMD74, item summary. https://www.sciencebase.gov/catalog/item/5d4c8b26e4b01d82ce8dfeb0 verified
23. 23 US Federal Emergency Management Agency, OpenFEMA dataset PublicAssistanceFundedProjectsDetails v2, filter disasterNumber eq 4339 and damageCategoryCode eq 'C', retrieved 8 August 2026. https://www.fema.gov/api/open/v2/PublicAssistanceFundedProjectsDetails verified
24. 24 Author's summation of the projectAmount and totalObligated fields of the 3,036 records returned by the OpenFEMA PublicAssistanceFundedProjectsDetails v2 query above, retrieved 8 August 2026. https://www.fema.gov/api/open/v2/PublicAssistanceFundedProjectsDetails modelled
25. 25 US Government Accountability Office, GAO-24-105557, 'Puerto Rico Disasters: Progress Made, but the Recovery Continues to Face Challenges', published 13 February 2024. https://www.gao.gov/products/gao-24-105557 verified
26. 26 US Federal Highway Administration press release dot17090, 'U.S. Department of Transportation Provides Additional $36 Million to Puerto Rico and US Virgin Islands for Repairs to Hurricane-Damaged Roads and Bridges', posted 27 November 2017. https://www.fhwa.dot.gov/pressroom/dot17090.cfm verified
27. 27 Puerto Rico Transportation Asset Management Plan 2032, Final, 29 December 2022, section 6.3.3 '2022 Summary', pp. 6-12 to 6-13, prepared under 23 CFR Part 667. https://www.tam-portal.com/wp-content/uploads/sites/12/2024/04/Puerto-Rico-2022-12-29-BIL-Compliant-TAMP-2032.pdf self-published
28. 28 Financial Oversight and Management Board for Puerto Rico, 'Infrastructure Budgets Focus on Reliability, Financial Stability, and Investments', 28 July 2023. https://oversightboard.pr.gov/infrastructure-budgets-focus-on-reliability-financial-stability-and-investments/ reported
29. 29 NOAA National Centers for Environmental Information, International Best Track Archive for Climate Stewardship (IBTrACS) v04r01, Western Pacific basin CSV, storm SID 2013306N07162 (HAIYAN, season 2013), fields USA_WIND, USA_PRES, TOKYO_WIND, TOKYO_PRES, LANDFALL, retrieved 8 August 2026. https://www.ncei.noaa.gov/data/international-best-track-archive-for-climate-stewardship-ibtracs/v04r01/access/csv/ibtracs.WP.list.v04r01.csv verified
30. 30 Government of the Philippines, 'Typhoon Yolanda (Haiyan) 2013 the Philippines, Post-Disaster Needs Assessment', Table 1 'Summary of Damage and Losses (Per Region)', p. 12, and 'Damage and Losses Assessment' section. https://wrd.unwomen.org/sites/default/files/2022-02/Typhoon%20Yolanda%20(Haiyan)%202013%20the%20Philippines,%20Post-Disaster%20Needs%20Assessment_0.pdf verified
31. 31 Department of Public Works and Highways, 'Typhoon Yolanda Rehabilitation and Recovery, Infrastructure Cluster', Key Achievements and 'Status by Project Category', status as of 15 September 2015, pp. 26-27, published in the Department of Budget and Management Yolanda list of releases. https://www.dbm.gov.ph/wp-content/uploads/YOLANDA/List%20of%20Releases%20(excel%20file)/DPWH.pdf self-published
32. 32 Republic of the Philippines, General Appropriations Act, FY 2026, Volume I-B, XXII Department of Public Works and Highways, A. Office of the Secretary, New Appropriations by Program and by Object of Expenditure, as printed in the Official Gazette dated 5 January 2026 and published by the Department of Budget and Management. https://www.dbm.gov.ph/wp-content/uploads/GAA/GAA2026/VolumeIB/DPWH/DPWH.pdf verified
33. 33 Asian Development Bank, 'Philippines: Transport Sector Assessment, Strategy, and Road Map', paragraphs 2 to 6, pp. 1-2. https://www.adb.org/sites/default/files/institutional-document/33700/philippines-transport-assessment_0.pdf reported
34. 34 Asian Development Bank, 'Bridges across Oceans: Initial Impact Assessment of the Philippines Nautical Highway System and Lessons for Southeast Asia', April 2010, Table 3 'Comparative Shipping Costs within Philippine Islands and the Region', p. 5, sourced to G. Llanto and E. Basilio, 2005, World Bank-PIDS Study on Ports and Shipping. https://www.adb.org/sites/default/files/publication/27510/bridges-oceans.pdf reported
35. 35 Maritime Industry Authority (MARINA), 'Philippine Nautical Highway: RORO System in the Philippines', Western Nautical Highway route tables, rates per Executive Order 909, posted May 2022. https://marina.gov.ph/wp-content/uploads/2022/05/2020-RORO-System-in-the-Philippines.pdf reported
36. 36 US Government Accountability Office, GAO-13-260, 'Puerto Rico: Characteristics of the Island's Maritime Trade and Potential Effects of Modifying the Jones Act', 20 March 2013, Table 1 and the sections 'Many Factors Determine Freight Rates' and 'Foreign Carriers Serving Puerto Rico'. https://www.gao.gov/assets/660/653182.txt verified
37. 37 Appalachian Regional Commission, "About the Appalachian Region", opening paragraph; and ARC Chartbook, "The Appalachian Region: A Data Overview from the 2019–2023 American Community Survey" (accessed 8 Aug 2026). https://www.arc.gov/about-the-appalachian-region/ verified
38. 38 Appalachian Regional Commission, "County Economic Status and Distressed Areas in Appalachia, FY 2026 Data Tables" (xlsx), sheet "County Status Count_Population", tables "Number of Counties per Economic Status" and "County Population per Economic Status", published May 2025. https://www.arc.gov/wp-content/uploads/2025/05/CountyEconomicStatusandDistressAreasFY2026DataTables.xlsx verified
39. 39 FHWA, Highway Statistics 2024, Table HM-10, "Public Road Length – 2024, Miles By Ownership", published 17 October 2025; state rows and U.S. Total row. https://www.fhwa.dot.gov/policyinformation/statistics/2024/hm10.cfm verified
40. 40 Author's arithmetic on FHWA, Highway Statistics 2024, Table HM-10 (rural TOTAL column divided by grand TOTAL column), published 17 October 2025. https://www.fhwa.dot.gov/policyinformation/statistics/2024/hm10.cfm modelled
41. 41 Author's arithmetic on FHWA, National Bridge Inventory, "Bridge Condition by Highway System 2025", All Bridges table (poor count divided by all count), dated 15 June 2025. https://www.fhwa.dot.gov/bridge/nbi/no10/condition25.cfm modelled
42. 42 Appalachian Regional Commission, "ADHS 2025 Cost-to-Complete Estimate Report", table "Status of Completion of the ADHS (Miles)", p. 5, data as of 30 September 2025 (published March 2026). https://www.arc.gov/wp-content/uploads/2026/03/ADHS-2025-Cost-to-Complete-Estimate-Report.pdf verified
43. 43 Appalachian Regional Commission, "ADHS 2025 Cost-to-Complete Estimate Report", table "Cost-to-Complete the ADHS, by State, as of September 30, 2025", pp. 6–7; costs stated in 2025 U.S. dollars; prepared with FHWA and state DOTs. https://www.arc.gov/wp-content/uploads/2026/03/ADHS-2025-Cost-to-Complete-Estimate-Report.pdf modelled
44. 44 NOAA National Hurricane Center, Tropical Cyclone Report: Hurricane Helene (AL092024), table "Total Deaths by State", p. 19; original report 19 March 2025, updated through 2025. https://www.nhc.noaa.gov/data/tcr/AL092024_Helene.pdf verified
45. 45 FEMA OpenFEMA API, dataset DisasterDeclarationsSummaries v2, records filtered on declarationTitle in {TROPICAL STORM HELENE, HURRICANE HELENE, POST-TROPICAL STORM HELENE}, retrieved 8 Aug 2026 (689 designated-area rows). https://www.fema.gov/api/open/v2/DisasterDeclarationsSummaries verified
46. 46 FEMA OpenFEMA API, dataset FemaWebDisasterSummaries v1, records for disasterNumber 4827, 4831, 4832, 4851; PA load date 8 Aug 2026, IA load date 7 Aug 2026. https://www.fema.gov/api/open/v1/FemaWebDisasterSummaries verified
47. 47 North Carolina Office of State Budget and Management, "Hurricane Helene Recovery: Revised Damage and Needs Assessment", 13 December 2024, Executive Summary pp. 5–6 and Estimates table p. 7; methodology sections state figures are extrapolated, forecast and scaled from Hurricane Florence and Tropical Storm Fred. https://www.osbm.nc.gov/hurricane-helene-dna/open modelled
48. 48 North Carolina Office of State Budget and Management, "Hurricane Helene Recovery: Revised Damage and Needs Assessment", 13 December 2024, Transportation chapter, Summary p. 63 (attributing bridge and culvert counts to NCDOT field assessment). https://www.osbm.nc.gov/hurricane-helene-dna/open reported
49. 49 North Carolina Department of Transportation, press release "This Week at NCDOT: Helene Update, NCDMV Visible Wait Times and Speed a Little. Lose a Lot", section "18-Month Helene Update", 27 March 2026. https://www.ncdot.gov/news/press-releases/Pages/2026/2026-03-27-this-week-at-ncdot-helene-update-ncdmv-wait-times-speeding.aspx reported
50. 50 North Carolina Department of Transportation, press release "NCDOT Making Steady Progress But Faces Long Road to Recovery after Hurricane Helene", 14 October 2024. https://www.ncdot.gov/news/press-releases/Pages/2024/2024-10-14-ncdot-hurricane-helene-recovery.aspx reported
51. 51 WFAE 90.7 (Charlotte NPR News), "More than 99% of roads damaged by Helene have reopened, NCDOT says", 31 July 2026, quoting NCDOT Secretary Daniel Johnson. https://www.wfae.org/2026-07-31/more-than-99-of-roads-damaged-by-helene-have-reopened-ncdot-says reported
52. 52 National Park Service, Blue Ridge Parkway, news release "Helene storm repair projects at Blue Ridge Parkway underway", 27 March 2025. https://www.nps.gov/blri/learn/news/helene-storm-repair-projects-at-blue-ridge-parkway-underway.htm verified
53. 53 FHWA, Emergency Relief Program, "Allocations/Allotments" page, tables "Recent Allocations" and "Recent Quick Releases", rows citing Hurricane Helene DR-4827-NC, DR-4832-TN, DR-4831-VA, accessed 8 Aug 2026. https://www.fhwa.dot.gov/programadmin/erelief/allocations.cfm verified
54. 54 Federal Communications Commission, 2024 Section 706 Report, FCC 24-27, adopted 14 March 2024 and released 18 March 2024, Appendix B Supplemental Figures, urban/rural state table pp. 3520 (NC), 3536 (TN), 3545 (VA), 3550 (WV); underlying Broadband Data Collection data as of 31 December 2022. https://docs.fcc.gov/public/attachments/FCC-24-27A1_Rcd.pdf verified
55. 55 Appalachian Regional Commission, The Chartbook, "Computer and Broadband Access in Appalachia", drawing on the 2019–2023 American Community Survey. https://www.arc.gov/about-the-appalachian-region/the-chartbook/computer-and-broadband-access-in-appalachia/ verified
56. 56 NTIA BroadbandUSA, "Public Resources related to BEAD Plans and Milestones / State and Territory Allocation Totals", state entries for North Carolina, Virginia, West Virginia and Tennessee, accessed 8 Aug 2026. https://ntia.gov/page/state-and-territory-allocation-totals verified
57. 57 North Carolina Department of Information Technology / Office of the Governor, press release "Governor Josh Stein Unlocks More Than $300 Million for Broadband Expansion Projects", 22 December 2025; and NCDIT "BEAD Final Proposal & Awards Approved" page. https://www.ncbroadband.gov/news/press-releases/2025/12/22/governor-josh-stein-unlocks-more-300-million-broadband-expansion-projects self-published
58. 58 Benton Institute for Broadband & Society, "The BEAD Plan for Western North Carolina: What You Need to Know", 22 September 2025, county-by-county table, analysing the North Carolina draft BEAD Final Proposal. https://www.benton.org/blog/bead-plan-western-north-carolina-what-you-need-know reported
59. 59 Washington State DOT, "Mountain pass closure and opening dates", North Cascades SR 20 historical table (series runs 1972-2026), rows for 2024 and 2025. https://wsdot.wa.gov/travel/roads-bridges/mountain-pass-closure-and-opening-dates verified
60. 60 Author's arithmetic on the two published dates in WSDOT, "Mountain pass closure and opening dates", North Cascades SR 20 table (closed Nov. 18, 2024; opened April 22, 2025). https://wsdot.wa.gov/travel/roads-bridges/mountain-pass-closure-and-opening-dates modelled
61. 61 WSDOT news release, "SR 20 North Cascades Highway closes for season Thursday, Dec. 4 at 6 p.m." (closure date, 2025); Methow Valley News, "WSDOT: Highway 20 to reopen by June 25", 21 May 2026 (opening date, record comparison). Duration is the author's arithmetic on the two dates. https://methowvalleynews.com/2026/05/21/wsdot-highway-20-to-reopen-by-june-25/ reported
62. 62 WSDOT news release, "SR 20 North Cascades Highway closes for season Thursday, Dec. 4 at 6 p.m.", milepost detail, 2025. https://wsdot.wa.gov/about/news/2025/sr-20-north-cascades-highway-closes-season-thursday-dec-4-6-pm verified
63. 63 Colorado DOT, "Independence Pass Travel Information" (typical dates, milepost range, elevation, historical closure dates 2009-2013); CDOT news release "CO 82 at Independence Pass closes for the winter season", November 2025. Duration is the author's arithmetic on the typical dates. https://www.codot.gov/projects/archived-project-sites/SH82/independence-pass/independence-pass-travel-information.html reported
64. 64 Alaska DOT&PF, Standard Specifications for Highway Construction, 2020 edition, Subsection 618-3.02 SEEDING SEASONS, p. 387 region of the text (Division 600). https://dot.alaska.gov/stwddes/dcsspecs/assets/pdf/hwyspecs/sshc2020.pdf verified
65. 65 Colorado DOT, "Avalanche Control" winter driving page, avalanche paths statement; the same figure appears on CDOT, "US 550 Red Mountain Pass Avalanche Control System" project page. https://www.codot.gov/travel/winter-driving/avalanche reported
66. 66 Colorado Avalanche Information Center, 2024/25 Annual Report, "Financial Breakdown // CAIC", FY25 CAIC Revenue and FY25 CAIC Expenditures, p. 10. https://avalanche.state.co.us/sites/default/files/2025-12/CAIC_FY25AnnualReport_Final_Web.pdf reported
67. 67 Colorado DOT, "US 550 Red Mountain Pass Avalanche Control System" project page, project cost; CDOT news, "CDOT tests new remote avalanche control equipment near Eisenhower Johnson Memorial Tunnels", November 2025. https://codot.gov/projects/us550redmtnavalanchecontrolsystem reported
68. 68 Anchorage Daily News, "Alaska DOT tests new avalanche control technology along Seward Highway", 25 March 2026, quoting Tim Glassett, statewide avalanche and artillery program manager, Alaska DOT&PF, and Stan Caldwell, director of the SMART Grants Program; same figures carried in Governing, "Alaska Tests New Tech to Predict and Prevent Avalanches". https://www.adn.com/alaska-news/anchorage/2026/03/25/alaska-dot-pioneering-new-avalanche-control-technology-along-seward-highway/ reported
69. 69 Alaska DOT&PF, Statewide Maintenance & Operations, "Snow Avalanche Programs" page, avalanche areas map description, delay statement and seasonal timeline. https://dot.alaska.gov/stwdmno/avalanche.shtml reported
70. 70 Alaska DOT&PF, "Damalanche - 2014 Avalanche on the Richardson Highway" event page, closure chronology and cleanup quantities. https://dot.alaska.gov/highways/richardson/damalanche.shtml reported
71. 71 Clear Roads pooled fund programme, "2024-2025 State Winter Maintenance Data and Statistics" workbook, sheet "2. Collected Data" column "Snow and ice total annual expenditures" and sheet "3. Calculated Stats" column "COSTS Per total lane mile"; figures are state DOT self-reported returns compiled by Clear Roads. Per-lane-mile division verified by the author against the workbook's own calculated column. https://www.clearroads.org/download/2024-2025-state-winter-maintenance-data-and-statistics/ reported
72. 72 Author's division of published five-year average totals by published lane miles, from Clear Roads, "2024-2025 State Winter Maintenance Data and Statistics" workbook, sheet "4. Average Values - Five Year", columns "Total lane miles" and "Snow and ice total annual expenditures", averaged across winter seasons 2020-21 to 2024-2025. https://www.clearroads.org/download/2024-2025-state-winter-maintenance-data-and-statistics/ modelled
73. 73 Washington State DOT, "Avalanche control" operations page, pass elevations, snowfall, traffic volumes, avalanche path count at Stevens Pass, and delay durations. https://wsdot.wa.gov/travel/operations-services/avalanche-control reported
74. 74 Colorado DOT news release, "Heavy traffic expected on I-70 mountain corridor this Presidents Day weekend, CDOT encourages drivers to prepare for winter weather", February 2025, economic impact statement and tunnel traffic count. https://www.codot.gov/news/2025/february/heavy-i70-traffic-presidents-day-weekend-prepare-for-winter-weather modelled
75. 75 Colorado DOT news release, "Mid-season check-in: Winter weather-related crashes and closures down on I-70 as updated Traction Law takes effect in Colorado", February 2020 (2019 and 2018 figures); CDOT news release, February 2025 (2024 figures). https://www.codot.gov/news/2020/february-2020/mid-season-check-in-winter-weather-related-crashes-and-closures-down-on-i-70-as-updated-traction-law-takes-effect-in-colorado reported
76. 76 Alaska DOT&PF press release PR25-0006, "DOT&PF Announces 2025 Construction Season Projects Amid Challenges", 3 February 2025 (statewide estate, construction awards); Alaska DOT&PF Central Region, "Anchorage Winter Maintenance FAQ", answers on operators on duty, miles maintained, sanding, brine and budget reduction. https://dot.alaska.gov/creg/docs/Anchorage-Winter-Maintenance-FAQ.pdf reported
77. 77 Fox Weather, "Avalanche-prone highway in Alaska gets life-saving weather station", attributing the Alaska Department of Transportation and Public Facilities as sole organisational source. https://www.foxweather.com/weather-news/thompson-pass-richardson-highway-weather-station-alaska reported
78. 78 U.S. Department of the Interior, Bureau of Indian Affairs, "BIA Road Maintenance Program," presentation to the Tribal/Interior Budget Council, March 22-23, 2016, slide "Deferred Maintenance Assessment Data Summary, Fiscal Year 2015 - BIA-OWNED ROAD MILEAGE," row N NAVAJO, columns PAVED / GRAVEL / EARTH / TOTAL. https://www.bia.gov/sites/default/files/dup/assets/as-ia/ocfo/pdf/idc1-034405.pdf verified
79. 79 Navajo Division of Transportation, 2016 Navajo Nation Long Range Transportation Plan, Final Plan, April 2016, Section 5.0 Existing Transportation System, page 5-1 and Table 5.1 "Road Ownership by Mileage" (source line: 2015 Official RIFDS Dataset). http://fortbertholdplan.org/wp-content/uploads/2016/06/LRTP-website-draft.pdf self-published
80. 80 U.S. Department of the Interior, Bureau of Indian Affairs, Budget Justifications and Performance Information, Fiscal Year 2026, page BIA-26, "Subactivity - Road Maintenance (TPA)". https://edit.doi.gov/sites/default/files/documents/2025-12/bia-2026-greenbook508.pdf verified
81. 81 U.S. Census Bureau, 2024 Gazetteer Files, National American Indian Area / Alaska Native Area / Hawaiian Home Land file (2024_Gaz_aiannh_national.txt), GEOID 2430, fields ALAND and ALAND_SQMI. https://www2.census.gov/geo/docs/maps-data/data/gazetteer/2024_Gazetteer/2024_Gaz_aiannh_national.zip verified
82. 82 U.S. Census Bureau, 2020 Census Demographic and Housing Characteristics File (DHC), Table P1 (Total Population) and Table H1 (Total Housing Units), geography 2500000US2430. https://data.census.gov/api/access/data/table?id=DECENNIALDHC2020.P1&g=2500000US2430 verified
83. 83 U.S. Census Bureau, 2020 Census Demographic and Housing Characteristics File (DHC), Tables P1 and H1, geography 2500000US1505. https://data.census.gov/api/access/data/table?id=DECENNIALDHC2020.H1&g=2500000US1505 verified
84. 84 Author's arithmetic on: NDOT 2016 LRTP Table 5.1 (14,167.6 miles); FHWA Highway Statistics 2023, Table HM-10 "Public Road Length - 2023, Miles by Ownership," dated November 19, 2024, rows Arizona and U.S. Total; U.S. Census Bureau 2024 Gazetteer county and AIANNH files for land area. https://www.fhwa.dot.gov/policyinformation/statistics/2023/xls/hm10.xlsx modelled
85. 85 U.S. Department of Energy, Office of Indian Energy Policy and Programs, "Tribal Electricity Access and Reliability: Report to Congress," August 2023 (transmitted January 10, 2024), page 51. https://www.energy.gov/sites/default/files/2024-01/EXEC-2023-000952%20-%20Tribal%20Electricity%20Access%20Reliability%20Report%20to%20Congress%20%28Final%20Draft%20-%20Clean%29-signed%20by%20S1.pdf reported
86. 86 Navajo Tribal Utility Authority testimony to the U.S. House Committee on Natural Resources, Subcommittee on Indigenous Peoples of the United States, April 21, 2021 (updated and corrected May 5, 2021), quoted verbatim in U.S. Department of Energy, "Tribal Electricity Access and Reliability: Report to Congress," August 2023, page 78. https://www.energy.gov/sites/default/files/2024-01/EXEC-2023-000952%20-%20Tribal%20Electricity%20Access%20Reliability%20Report%20to%20Congress%20%28Final%20Draft%20-%20Clean%29-signed%20by%20S1.pdf self-published
87. 87 Navajo Tribal Utility Authority, "2024 Progress Report," published via the Navajo Nation Council, pages 19-20, "For Comparison" panel. https://www.navajonationcouncil.org/wp-content/uploads/2025/01/2024_NTUA_PROGRESS_REPORT_final_web.pdf self-published
88. 88 U.S. Energy Information Administration, Form EIA-861 Annual Electric Power Industry Report, 2024 data release, file Reliability_2024.xlsx, sheets "Reliability_States" (rows for Navajo Tribal Utility Authority, Other Standard columns) and "State Totals" (row 2024 AZ, Any Standard columns). https://www.eia.gov/electricity/data/eia861/zip/f8612024.zip verified
89. 89 Author's arithmetic on U.S. Energy Information Administration, Form EIA-861 2024, Sales_Ult_Cust_2024.xlsx, sheet "States": NTUA residential sales 231,721 MWh over 38,653 customers; national sum 1,590,733,661 MWh over 157,929,484 residential customers (2,755 utility-state rows). https://www.eia.gov/electricity/data/eia861/zip/f8612024.zip modelled
90. 90 U.S. Census Bureau, American Community Survey 2019-2023 5-Year Estimates, Table B25040 "House Heating Fuel," geography 2500000US2430 (Navajo Nation Reservation and Off-Reservation Trust Land). https://data.census.gov/api/access/data/table?id=ACSDT5Y2023.B25040&g=2500000US2430 verified
91. 91 American Public Power Association, "Light Up Navajo" programme page, accessed August 2026 (programme jointly run by APPA and the Navajo Tribal Utility Authority). https://www.publicpower.org/LightUpNavajo self-published
92. 92 National Renewable Energy Laboratory analysis reproduced as Tables 7 and 8 in U.S. Department of Energy, "Tribal Electricity Access and Reliability: Report to Congress," August 2023, pages 57-58. https://www.energy.gov/sites/default/files/2024-01/EXEC-2023-000952%20-%20Tribal%20Electricity%20Access%20Reliability%20Report%20to%20Congress%20%28Final%20Draft%20-%20Clean%29-signed%20by%20S1.pdf modelled
93. 93 U.S. Census Bureau, American Community Survey 2019-2023 5-Year Estimates, Table B28002 "Presence and Types of Internet Subscriptions in Household," variables B28002_001E, _002E, _004E, _012E, _013E, geography 2500000US2430. https://data.census.gov/api/access/data/table?id=ACSDT5Y2023.B28002&g=2500000US2430 verified
94. 94 U.S. Census Bureau, American Community Survey 2019-2023 5-Year Estimates, Table B28002, geography 2500000US1505 (Hopi Reservation and Off-Reservation Trust Land). https://data.census.gov/api/access/data/table?id=ACSDT5Y2023.B28002&g=2500000US1505 verified
95. 95 Federal Communications Commission, 2024 Section 706 Report, FCC 24-27, GN Docket No. 22-270, adopted March 14 2024, released March 18 2024, Appendix B-14 "Service Availability (Millions) of Fixed Terrestrial 100/20 Mbps and Mobile 5G-NR with a Minimum Speed of 35/3 Mbps on Tribal Lands by State (December 31, 2022)"; figures derive from provider-reported Broadband Data Collection filings and FCC staff block estimates. https://docs.fcc.gov/public/attachments/FCC-24-27A1.txt reported
96. 96 Navajo Nation Broadband Office, "Empowering Navajo Communities with High-Speed Connections," presentation dated 8 May 2024, "Budget - NNBO / NN DEI" slides, posted on the Navajo Nation Broadband Office public portal (the deck carries a "not to be shared without proper authorization" banner notwithstanding its public posting). https://broadband.navajo-nsn.gov/Portals/0/homepage/Presentations/NN%20Broadband%20Plan_final%20050824.pdf self-published
97. 97 Arizona Department of Transportation, "SR 264 Corridor Planning Study" (in partnership with the Hopi Tribe), Final Report, December 19, 2025, "Socioeconomic Characteristics / Vulnerable Populations," page 14. https://azdot.gov/sites/default/files/2026-01/SR264-Corridor-Planning-Study-121925-wo_Appendices.pdf reported
98. 98 Arizona Department of Transportation, "SR 264 Corridor Planning Study," Final Report, December 19, 2025, "Access Management," "Speed Limit" and "Roadway Usage" sections, pages 12-15. https://azdot.gov/sites/default/files/2026-01/SR264-Corridor-Planning-Study-121925-wo_Appendices.pdf verified
99. 99 Arizona Department of Transportation, "SR 264 Corridor Planning Study," Final Report, December 19, 2025, "Future Conditions Analysis," pages 21-22, citing Arizona Office of Economic Opportunity forecasts and ADOT's Average Annual Daily Traffic Report. https://azdot.gov/sites/default/files/2026-01/SR264-Corridor-Planning-Study-121925-wo_Appendices.pdf modelled
100. 100 National Highway Traffic Safety Administration, Tribal Traffic Safety Initiative, "Motor Vehicle Crash Fatalities in the Four Corners Region (2013-2019)," infographic, source line "Source: FARS". https://static.nhtsa.gov/nhtsa/downloads/Events/Tribal-Traffic-Safety-Initiative/Infographics-4Corners.pdf verified
101. 101 U.S. Department of the Interior, Bureau of Indian Affairs, "BIA Road Maintenance Program," Tribal/Interior Budget Council, March 22-23, 2016, "Deferred Maintenance Assessment Data Summary, Fiscal Year 2015" and "CONSTRUCTION NEED (COST TO CONST/IMPROVE)" slides. https://www.bia.gov/sites/default/files/dup/assets/as-ia/ocfo/pdf/idc1-034405.pdf reported
102. 102 U.S. Census Bureau, American Community Survey 2019-2023 5-Year Estimates, Table B25044 "Tenure by Vehicles Available," variables B25044_003E and B25044_010E, geography 2500000US2430 (the 6,946 sum is the author's arithmetic on the two published cells). https://data.census.gov/api/access/data/table?id=ACSDT5Y2023.B25044&g=2500000US2430 verified
103. 103 Amy Linn and Alysa Landry, "Dire streets: Muddied roads and broken promises on the Navajo nation," USC Annenberg Center for Health Journalism, published April 8, 2019. https://centerforhealthjournalism.org/our-work/reporting/dire-streets-muddied-roads-and-broken-promises-navajo-nation reported

Research and review preprint, not investment advice. Figures carry the verification grade under each reference: verified (peer-reviewed or independently replicated), reported (a named source stated it, not independently confirmed), self-published (the organisation's own figure) and modelled (computed here, not measured). A modelled figure is never presented as a measurement. Corrections are welcome and will be recorded.