The DEIS in Plain English

This is the baseline the whole DEIS measures against. Because the rule bars road-building and timber harvest by default rather than case by case, it works as a blanket protection, not a site-specific one — which is exactly what rescission would end. Two things in this passage are worth holding onto. First, the document notes that despite the name, "roadless" areas already contain "thousands of miles of existing roads," and gives no mileage for that figure. Second, the 1999 presidential directive the DEIS quotes carries no citation, and two of the reference entries meant to support this history cannot be resolved against the document's own reference list — so the origin story of the rule being rescinded is not, as published, traceable.

Rescission would not by itself authorize any project. It would remove the nationwide default prohibition and shift authority to forest-level plans and local officials. The agency states elsewhere in the same document that "land management plans do not compel action or authorize projects and future decisions would be subject to local interpretation of the plan and site-specific project analysis." That sentence is offered to reassure — plans will still govern — and it simultaneously describes why plan-based protection is more discretionary, more locally variable, and easier to change than a mandatory national rule. The DEIS does not analyze that difference anywhere.

This is the master baseline. Every percentage and acreage comparison later in the document is calculated from it, which is why the filed record concentrates here more heavily than anywhere else in the DEIS. The published derivation does not close: 58.2 less 9 less 4 is 45.2, not 44.7. The 44.7-million-acre figure is contradicted by the agency's own geospatial review, which found 44.3 million acres on National Forest System land — a 0.4-million-acre difference the DEIS attributes both to post-2001 ownership change and to mapping error without apportioning it between them. The superseded 58.5-million-acre total is still stated in the present tense in the Cost Benefit Analysis supporting the rule, and the 31 percent share derived from it survives a footnote that corrects the acreage underneath it. The ten-state concentration claim carries no state-by-state acreage a reader could check. None of this requires new data: each is a discrepancy between two figures the agency has already published.

The contested history is part of why this rulemaking exists — the DEIS presents rescission as an end to decades of litigation and reversal. It does not address how that litigation came out on the merits: the courts upheld the 2001 Rule against the challenges the document lists. Two things follow from this history that the DEIS does not engage. The state-specific approach it describes as a failed experiment is the same approach a state forestry division asked the agency to reopen in this comment period, and the DEIS eliminated a state-specific alternative on a finding of equivalence that its own Chapter 2 contradicts nine pages earlier. And when the agency promulgated the 2001 Rule it found, with the concurrence of the consulting wildlife agencies, that the prohibitions were not likely to adversely affect listed species — a determination both draft Biological Assessments for this rulemaking recite in their opening pages, and which now sits unengaged against 327 findings that removing those same prohibitions is likely to adversely affect listed species.

The Tongass outcome is effectively pre-set by executive order and folded into this EIS by reference to five-year-old analysis, rather than argued through this document's own comparison of alternatives. The DEIS describes the 2023 repeal as a sequence of events and never states the substantive findings that repeal rested on — the findings this proposal reverses. An agency changing position has to acknowledge the prior policy and give reasons; presenting the reversal as chronology supplies neither. Whether five-year-old analysis can still carry the Tongass conclusions is taken up in The Tongass National Forest and in How the Analysis Was Built.

That sentence is a three-step causal chain — the rule limited management, which caused inactive management, which caused forest-health problems — and it is the central justification for the entire rulemaking. It is uncited, and the DEIS's own pages break it at each step. Table 21 reports lower wildfire ignition density inside roadless areas than on other national forest land. Table 14's insect-and-disease data are annual snapshots that cannot establish the trend the sentence asserts. The activity record in the next point shows 1.8 million acres of hazardous fuels reduction actually accomplished inside roadless areas under the rule said to have prevented it. The document also never separates management-attributable decline from climate-attributable decline, though the same sentence supplies the stated need in the Cost Benefit Analysis and therefore carries the regulatory impact analyses as well. Separately, the purpose is framed narrowly enough that it decides the alternatives: with local flexibility as the need, only alternatives granting local flexibility can meet it, which is why all three alternatives vary along a single dimension and an intermediate option was eliminated.

These are the numbers used to argue that rescission would change little in practice, and they are the least traceable figures in the document. The harvest and fuels figures come from the Forest Service Activity Tracking System, which is named in prose and appears nowhere in the reference list; the extraction is never described, and the same 500-acre average is attributed to two different land bases in the same document. The 60-mile road figure names no dataset at all, counts only permanent system roads built after 2013 — excluding the temporary roads the agency's own assessment says new construction would mostly produce — and is repeated verbatim in both Biological Assessments as though it measured road construction generally. The 1.8-million-acre fuels total does not measure rule-governed activity in the way the argument uses it, and more than half of it is attributed to wildfire use rather than to planned treatment. As for the exceptions: the agency's own Biological Assessments state that "there is no national-level tracking of the review, approval, or implementation of 2001 Roadless Rule exceptions." A 188-proposal compilation with no approval rate cannot establish that exception use was limited by the process, and the agency says it does not hold the data that could.

The DEIS presents this rulemaking as implementing a set of White House deregulatory and resource-development directives, which is a legitimate thing for an agency to say and a consequential thing for the alternatives analysis. Where an alternative is eliminated because it does not conform to an executive order's policy direction rather than because it is technically or economically infeasible or fails the purpose and need, the document has to say so and say which order and which element. The DEIS states the set of alternatives eliminated under Executive Order 14153 two different ways, at pages 37 and 315, and states the reason as a conclusion without the criterion it applied. One further wrinkle: the DEIS names its minerals analysis as the documentation of compliance with Executive Order 14154, and that analysis projects no net change in domestic leasable minerals production at all.

Announcing an intent to rescind before the environmental review is complete is the kind of sequence used to argue an agency predetermined its outcome, which cuts against NEPA's requirement that the no-action alternative get genuine consideration. The reported figures cannot be checked: no coding protocol, sampling method, or classification rule is disclosed for how a letter was distinguished from a submission, how petition signatures were attributed, or what made a letter "generally opposed." The 625,000 figure is stated by reference to "the petitions," a term with no antecedent anywhere in the document. The two places the statistics appear do not agree with each other. Most consequentially, the screen that sorted this record — including the finding that more than 90 percent of letters were "brief sentiments" rather than substantive comment — applied a definition of "substantive" that is no longer in force, and the appendix footnote reciting that definition misquotes it. Extension requests filed by state Attorneys General are unanswered on the record. The catch-all rejection basis does not say which of its four grounds applied to which concern, which is the difference between a response and a category.

By the document's own arithmetic, 30 of the 64 requests are neither held nor scheduled, and the DEIS neither states that figure nor explains what became of them. The sentence carries no source. It is also the section the compliance table designates as the record of compliance with Executive Order 13175, which requires meaningful government-to-government consultation — and a count of meetings is not a record of what was raised or what changed. Tribal consultation is taken up in full in Tribal Rights and Interests and How the Analysis Was Built; the sequencing matters here because the notifications began roughly a month after the outcome was publicly announced.

This draft is a mid-point document, and the record built now is what the agency has to answer before it can finalize. A comment that identifies a specific defect, cites the page it sits on, and names the correction it asks for is one the agency has to dispose of in writing. That is what the comments in this bundle are built to do.

The changed-conditions sentence is the load-bearing rationale for opening any of the 44.7 million acres, and it is uncited. Each of its four elements has been tested against the record and none survives intact. The drought and extreme-temperature claims are asserted for National Forest System lands generally and are not supported at the scale of the roadless areas the rule actually covers; where a source is cited, it does not reach the claim. The wildfire trend is asserted with no analysis of wildfire on the affected lands since 2001 anywhere in the document, and the fire chapter's own roadless-specific data cut against it. The insect-and-disease trend rests on Forest Health Protection data that are annual snapshots rather than a trend series — the agency cannot characterize direction of change from them. This matters beyond the DEIS: the same sentence supplies the stated need in the Cost Benefit Analysis, so the regulatory impact analyses inherit whatever it cannot support. An agency reversing a long-standing protection has to give reasons; a predicate the document's own chapters contradict is not one.

No Action does not mean nothing happens: activities already allowed under the rule's exceptions continue. Because Alternative 1 is the yardstick for the whole document, how it is characterized governs every comparison — and the two sides are not measured on the same terms. The action alternatives get quantified projections; Alternative 1 is described narratively. The objection does not depend on which alternative the Department picks: it would apply equally if the ranking ran the other way, because the defect is the absence of comparable analysis on both sides. Nor can it be deferred to project-level review, since this rulemaking is the only decision at which a nationwide prohibition is compared against its rescission. One further problem sits underneath the baseline itself: the 9.3-million-acre Tongass figure carrying the arithmetic has no stated vintage and no reconciliation against the technical corrections the DEIS says already changed the inventory.

This is the largest structural change of the three: a shift from one blanket national prohibition to reliance on dozens of individual land management plans that may restrict roads or logging in a given area but are not required to. The DEIS's framing — "more" activity being "reasonably foreseeable" — describes the margin rather than the change, since Alternative 2 removes the protective framework outright. What the agency's own Biological Assessments say about that framework is worth setting beside it: increased road construction and timber harvest are "reasonably certain to occur," and rescission "will increase the likelihood of road construction and timber harvests occurring in locations previously considered inventoried roadless areas."

"Modified" does not mean uniformly modest: how much protection is removed varies from about 60 percent of roadless acres in the Northern Region to 97 percent in the Southern Region. A reader in one part of the country sees the rule nearly intact; a reader in another sees it nearly gone. Two further problems sit in the numbers. An appendix table misallocates 3.197 million acres in Region 10, and the stated removal criteria do not account for the regional outcomes reported on the same page. And the interface acreage driving the removal is reported as a determinate figure while the agency's own sensitivity analysis establishes a measured 12 percent uncertainty band around it — an uncertainty that appears nowhere in the figures themselves.

The three named categories sum to 16.1 million acres, not 17.9 — a gap of about 1.8 million acres that appears as no line item and is never quantified. Reading the same passage a second way, by whether the categories overlap, puts the unstated remainder anywhere between 1.8 and 11.2 million acres. Those parcels are, by definition, neither roaded nor in the interface, which means the effects reasoning offered for Alternative 3 — that the land removed is already road-influenced — does not reach them. The non-federal component compounds it: the DEIS states that single quantity three different ways at 4.5, 4.585 and 4.638 million acres, and the error carries through to the acreage said to remain protected. None of this requires data the agency does not have. It is a request that the document be made consistent with itself.

Every acreage figure in the Alternative 3 effects analysis rests on the interface-and-half-mile-buffer methodology described in the alternatives chapter — and that methodology is not written into the rule text. The rule authorizes an open-ended, case-by-case process instead, with the phrase "or for other reasons" doing the work. So the boundaries actually adopted can differ materially from the ones analyzed, which is a different problem from an acreage that does not add up: it means the quantified comparison has no fixed object. As written the mechanism carries no acreage cap and no aggregate threshold triggering supplemental review, no requirement to report publicly how much land has left roadless status over time, and no cross-reference obligating Endangered Species Act consultation before a modification affecting listed-species habitat is approved — against a document that elsewhere identifies 327 listed species and 71 critical habitats in the affected area. The mechanism is also analyzed in no effects section of the DEIS at all: it appears in the alternatives chapter and in the proposed rule text, and nowhere in Chapter 3.

These are the numbers that answer "how much would actually change," and they do not reconcile with the document's own totals. The 16 percent is wrong on the DEIS's own published forested acreage: 4.8 million acres is 22.9 percent of it. The plan-direction percentages reported alongside do not sum. The 4-to-10 percent sawtimber range is a single figure covering both action alternatives, where the underlying work yields 5 to 10 percent for Alternative 2 and 4 to 9 percent for Alternative 3 — ranges that overlap across nearly their entire width, so a reader cannot tell from the published figures whether the alternatives differ on harvest at all. The supporting tables report substantially larger operable totals than the narrative quotes, and nothing in the document reconciles the two. Underneath all of it, the three thresholds that sort every forested acre into operable and inoperable — half a mile from a road, slopes under 40 percent, productivity above 20 cubic feet per acre per year — are unsourced, and the agency told its consulting wildlife agencies that the review producing them "is currently undergoing quality control and Regional/local vetting and the data presented will be updated in the final BA." The public is not told that.

These two findings sit in tension and the DEIS never reconciles them. If ignitions are already lower inside roadless areas, the case that adding roads reduces fire risk is not self-evident. The agency's own Economic Analysis states the other half plainly: "if there is an increase in public road access, the number and frequency of wildfires could rise." The full treatment of what the fire analysis does and does not compute is in Fire and Fuels Management; what matters at the alternatives level is that the wildfire rationale for choosing an action alternative rests on a benefit the document declines to quantify and a cost it concedes and does not count.

On nearly every resource the DEIS tracks, the modified alternative moves in the same direction as full rescission. It is a difference of degree — less land — not of kind. The word doing the most work in this passage is "preliminary": the determinations come from a draft assessment, the DEIS reports consultation status two different ways in adjacent pages, no Biological Opinion is published, and Alternative 3 receives no salmonid determinations at all. What that means for the species findings is taken up in Biodiversity, Wildlife, and Habitat.

This is where the range of alternatives is actually decided, and it is narrower than three options suggests. The DEIS states that of six alternatives identified during scoping, "Alternatives 1 through 5 would not meet the purpose and need for the current rulemaking, as they would not comply with Executive Order 14153" — leaving two bookends and one modified-boundary option built on a single mapping methodology. All three vary along one dimension, geographic boundaries, as the document says itself.

Each elimination has a problem the record identifies. The permanent-road-prohibition alternative was dropped for a reason two state wildlife agencies' own letters in the agency's record refute — one of them recommending, if the rule is repealed, exactly that limitation on new forest roads. The state-specific alternative was eliminated on a finding of equivalence the DEIS contradicts nine pages earlier, and a state forestry division asked in this comment period to reopen it. Twelve Southeast Alaska Tribes collectively proposed a homelands alternative, recorded in the agency's own Tribal Summary Impact Statement, and none of the four reasons given for eliminating it reaches what was proposed. The DEIS gives two different accounts of which alternatives were eliminated under the executive order, at pages 37 and 315, and states the elimination rationale as a conclusion without the criterion it applied. And an intermediate, criteria-based alternative between full rescission and full retention — the option that would test whether the purpose can be met without removing the prohibition — was never developed at all.

Set that against the sequence: the Secretary publicly announced an intent to rescind on June 23, 2025, before this analysis existed. A range this narrow, eliminated on grounds the record contradicts, is what raises the question the document does not answer — whether the three alternatives reflect a choice or a decision already made.

This list is the yardstick the DEIS itself uses later to judge how much each alternative affects these places, so it is worth carrying forward. It is broader than scenery: drinking-water sources, listed-species habitat, cultural and sacred sites, and land valuable precisely because it is undisturbed enough to serve as a scientific reference point.

Three headline totals for the same land base, none of them sourced. They are not simply alternative framings: the 58.5 figure is superseded by the document's own footnote and still appears in the present tense in the Cost Benefit Analysis supporting the rule; the 31 percent share derived from it survives the correction and cannot be derived from the 193-million-acre land base the DEIS states elsewhere; and the derivation offered for 44.7 does not close, since 58.2 less the 9 million Idaho acres and the 4 million Colorado acres is 45.2. Wilderness study area acreage inside the exclusions is stated as both 2.8 and 2.9 million. Acreage is the base unit for every downstream table on timber, fire, and habitat, so an unreconciled baseline is not a formatting complaint — no reader can reproduce the potentially affected environment from the values the document publishes.

The effect of this rulemaking is concentrated in a handful of Western states rather than spread evenly. That concentration claim is also the express premise of the Economic Analysis's geographic conclusion — and it rests on a superseded inventory baseline and on no state-by-state acreage the public can check. The 5,000-acre criterion is likewise stated without the basis that produced it, and the Colorado and Idaho exclusion acreages differ across the DEIS, the Economic Analysis, and both Biological Assessments.

One national forest accounts for roughly a fifth of all affected roadless acreage, and more than half its own land base is roadless. The 9.3-million-acre figure is reproduced verbatim across the DEIS, the Economic Analysis and both Biological Assessments, and it carries no stated vintage, no reconciliation against the technical corrections the DEIS says already changed the inventory, and no reconciliation against the 2020 Alaska analysis this rulemaking incorporates for Tongass effects. The national roadless total is reconciled to federal ownership on one page; the Tongass figure is not reconciled on the next. Given the litigation history specific to Alaska roadless management, an unsourced figure here is consequential.

This is the acknowledged discrepancy, and it is the master denominator — 56 percent of the Tongass, 66.5 percent restricted for timber, 54.5 percent restricted for roads, 2.8 percent in wilderness are all calculated from it. The DEIS names two causes and apportions the acreage between neither, so the reader cannot tell how much is genuine ownership change and how much is mapping error, or which regions it falls in. The footnote's own explanation does not scale: vertical integration errors of "tens or hundreds of acres" cannot account for 400,000. The same exclusion also defines the Endangered Species Act action area in the Draft Biological Assessment for USFWS species, so the unresolved gap reaches the section 7 analysis as well as the NEPA one. And the document asserts 44.7 million acres twice in its introductory pages while its own review found 44.3, without correcting either statement.

This subtraction is what produces the 40.1-million-acre area the DEIS analyzes, so it is load-bearing rather than incidental. The aggregate figure combines three individually unsourced numbers, and the DEIS does not say whether the three categories overlap — if they do, the net exclusion actually applied is smaller than 4 million and every impact figure expressed against the remainder shifts. The wilderness study area component is stated at two different values in the document. The same exclusion is applied verbatim in the USFWS Biological Assessment to define the action area, which means an unreconciled subtraction here bounds the species analysis too. Read as a limited-impact argument — that some of this land is safe anyway — it also does not disclose the acreage covered by no overlapping designation.

The 44.7 million acres is not the country's entire roadless inventory — roughly 13 million acres in Idaho and Colorado are governed separately and are untouched by this rulemaking. Those two rules are also the working precedent for the state-specific alternative the DEIS eliminated on a finding that state rules are equivalent to what is proposed here, a finding its own Chapter 2 contradicts. A government letter in the record states that more than 95 percent of acres under the Idaho rule carry protection similar to or stronger than the national rule, which is the comparison the elimination rationale needed to make and did not.

Plan direction is what the DEIS offers as the protection remaining after rescission, so how it is counted matters as much as how much there is. Both headline figures merge two legally different things: mandatory standards, which bind, and non-mandatory guidelines, which do not. The parenthetical "do not allow or generally do not allow" is where the merge happens, and the DEIS elsewhere concedes that plans "do not compel action or authorize projects." The plan inventory itself does not reconcile — the vintage count at page 16 cannot be squared with the stated total of 126 plans, and the figures are uncited. The agency's own Biological Assessments add two further problems: the national plan-component review producing these acreages "is currently undergoing quality control and Regional/local vetting," a caveat that appears nowhere in the DEIS; and for an identified set of species, the plans relied on for consistency predate the listings — the assessment records crayfish listed in 2016 and 2023 against plans from 1995 and 2004 that contain no protection for them. The DEIS also concedes on the same page as its vintage count that a large share of plans probably contain no direction for roadless areas at all.

The 2001 Rule was never an absolute bar — its exceptions have allowed roads, harvest, and fuels work for two decades, and that fact cuts both ways. It undercuts the claim that the rule prevented vegetation management: 1.8 million acres of hazardous fuels reduction happened under it. It also undercuts the figures themselves. The harvest and fuels numbers come from an activity-tracking system named in prose and absent from the reference list, with no query described and the same 500-acre average attributed to two different land bases in one document. The 60-mile road figure names no dataset, counts only permanent system roads built after 2013, and is qualified by "best available records" — a hedge that concedes incompleteness without saying what is missing. More than half the fuels total is wildfire use rather than planned treatment, which is a different thing from the active management the rationale credits.

This is the analytical lens that turns the setting into projected future activity, and it is the single most consequential methodology in the document — it produces the operable-acreage figure the whole harvest comparison rests on. None of its three thresholds is sourced. The half-mile road buffer, which also defines which acres Alternative 3 removes, appears with no stated basis at all, and it governs 11 million acres. The slope and productivity layers carry no source and no date. And the agency told both consulting wildlife agencies that the underlying review is still undergoing quality control and will change in the final assessment, while telling the public nothing of the kind.

More than a quarter of the system is already closed to ordinary vehicles, and the network has been shrinking for two decades because the agency could not afford what it had. Any alternative that adds mileage adds to a system the agency has spent twenty years trying to reduce for cost and ecological reasons. One figure inside this baseline does not hold: the roadless share of the network is stated as 2.5 percent in the comparison table using a 9,500-mile total that includes 100 miles of private road, and stated on the National Forest System basis at page 54, where the tables sum to 9,400.

The roads people actually drive are mostly rated fair or poor before any alternative adds a mile. The condition percentages, though, are not supported by the exhibits the sentence cites — neither the referenced figure nor the referenced table reports them — and the underlying source cannot be located by any reader. A statement about the condition of the network is the kind of fact a reader should be able to check, and here they cannot.

The agency's own numbers show a road program funded at roughly a fifth of stated need, with most of the money coming from programs expiring within about a year of this document. The DEIS does not say what replaces them, or how added mileage under the action alternatives would be paid for — only that "each new mile of road competes for limited road maintenance funding." Two of the numbers carrying that argument cannot be traced. The $39 billion replacement value is cited only to an internal database query the public cannot run, and the report behind it has no stated public access. The 4 percent multiplier that converts it into the $1.6 billion figure is attributed to no source at all: "industry standard" is an assertion, and the entire annual-need estimate rests on it. The 31,000-mile figure comes from an internal performance system, and no equivalent is reported for the roadless areas this rule actually affects.

The two figures are not comparable, and the DEIS presents them as a comparison anyway. The 2024 number excludes most of the system by mileage and is sample-based; the 2001 number is a census. Beyond scope, the comparison is in nominal dollars twenty-three years apart with no inflation adjustment anywhere in the record — in real terms the direction of change is not what the raw figures suggest. A sample-based estimate set against a census also carries sampling error the DEIS does not report. And the $6.9 billion figure itself carries no citation: the only source note nearby supports a different number entirely. Government letters in the agency's own Volume III put the backlog at other values again — $10.8 billion across a 380,000-mile system in one, and GAO's $4.85 billion for fiscal 2022 rising to $5.373 billion in fiscal 2024 in another.

These are real numbers, published by the agency, in the same document that files "agency road maintenance burden" as an unquantified cost in the Cost Benefit Analysis. Two pages after publishing per-mile costs, the analysis declines the mileage projection that would turn them into a total. Both terms are in the agency's hands; the multiplication is not performed. The cost figures also exclude bridges and major earthwork, so actual project costs run higher still.

The DEIS concedes that roads in the areas this rulemaking is about are far more expensive than its baseline figures, and then declines to state a range — leaving the case for opening those areas to be weighed without a cost for the roads themselves. "Not feasible to provide a typical cost per mile" is a defensible statement about a point estimate; it is a weaker one about a range, given the document supplies a 3-to-10-times multiplier in the very next sentence.

Roadless has never meant zero roads, and maintenance continues regardless of the outcome — the question is future construction. Both baseline figures have problems that matter for that question. The road-mileage baseline is stated as unqualified fact in the comparison table and as approximate at page 53, without the spatial rule that decides whether a segment lies inside a roadless area or the magnitude of the imprecision conceded. The DEIS and the Cost Benefit Analysis state mileages for the same geographic scope differing by about 7,500 miles, and nothing reconciles them. Most consequentially, the existing-roads data layer is defined to exclude the road categories the DEIS predicts will increase — temporary roads, and unauthorized routes the agency says it cannot count nationally — while the road-density conclusions drawn from it are stated as though it measured all roads. A government letter in the agency's own record makes the same point from the other direction: existing roads inside these lands include county, state and other ownership as well as federal.

Forest plans are the constraint the DEIS relies on once the rule is gone, and the way they are counted overstates what they do. "Allow or generally allow" and "do not allow or generally do not allow" each merge a binding standard with a discretionary guideline, so the reported restriction is not the restriction the figures imply. The 52-versus-53 percent discrepancy is minor in itself and is a visible sign that the acreage totals were not reconciled across sections.

No Action keeps construction rare and exception-driven, and the DEIS ranks it lowest on every road measure. The finding that deferred-maintenance effects would be inconsequential nationally depends on an annual construction rate rather than the 60-mile total — and the document reports that total in four places across three different time periods without ever stating the resulting rate.

This is the alternative the DEIS ranks highest for new mileage, deferred maintenance, and cost, and the analysis is expressed entirely as "more likely" and "less likely" with no projected miles, dollars, or acres. The agency did perform exactly that kind of quantified work for the Tongass alone, which is what makes the national absence a choice rather than a limitation. The record asks for a quantified, regionally stratified projection of road-mile construction and the sedimentation and fragmentation that follow, at the level of specificity already applied in Alaska. The agency's own Biological Assessment supplies both the direction and a worked example: increased road construction is "reasonably certain to occur," and a single project on the Lolo National Forest estimated approximately 19 miles of permanent and 9 miles of temporary road. Magnitude is estimable; the DEIS declines to estimate it.

Alternative 3 is a hybrid: about 13 million acres frozen under exception-only rules, the rest opened on logic tracking the proposed action. Which side of the line an acre falls on decides how much construction pressure it faces — and the line is drawn by the half-mile buffer whose basis the DEIS never states. The stated reason for removing the roaded portion is RARE II-era mapping error; the criterion actually applied is proximity to any existing road, and the difference is 11 million acres. Because the buffer sweeps in genuinely unroaded ground, the effects reasoning offered for this alternative — that the land removed is already road-influenced — does not reach the share that is not.

The ranking is consistent and it is not a quantification. No alternative gets an estimated mileage, a dollar range for the resulting backlog growth, or a projected cost. The deeper problem is that the DEIS ranks the alternatives against each other on deferred maintenance while stating in the same section that the underlying measurement cannot be made — the agency has no tracking mechanism for 82 percent of the road system. A ranking on a measure the document says it cannot compute is a conclusion without an analysis. Set that against the funding picture the same section describes, with annual road money at under a fifth of stated need and the programs covering most of the gap expiring within about a year, and the unquantified "more likely" outcomes are the whole substance of the road effects analysis.

Whichever alternative is chosen, uses already permitted keep operating. The document also concedes that its own tracking system "does not have specific tracking or reporting capabilities to identify the number or location of uses in IRA compared to other NFS lands" — so the agency cannot say how much of this activity happens inside roadless areas. That admission recurs throughout the DEIS: the agency frequently cannot report the roadless-specific share of a national figure, which is what makes the national figures hard to use as a measure of what this rule changes.

Because most inholding access already runs on existing roads, the rule's construction restrictions rarely come into play for present-day access. The stakes rise for the narrower set of cases where an inholder needs a road that does not yet exist, which is where the alternatives diverge.

For Alaska inholders the underlying right does not change with this rule. "Reasonable access" is a flexible standard satisfiable by a non-motorized route or a path across other ownership — it does not guarantee the road a landowner might want. That flexibility is in the statute, not something rescission alters.

For established mining claims, treaty rights and similar reserved rights, the agency's own account is that rescission makes no difference — they existed through the 2001 Rule and continue regardless. That narrows what rescission actually changes to newer, discretionary requests not grounded in a pre-existing right. It is worth noting the DEIS's treaty-rights analysis elsewhere in the document is limited by its own terms to "general considerations" and names no treaty, ceded territory or usual-and-accustomed area, which makes this reassurance harder to test than it appears.

Valid R.S. 2477 claims sit outside the rule's reach — asserting or perfecting one does not turn on which alternative governs. This is a legacy access category the DEIS treats as unaffected by the decision.

This is the substantive change in the topic. Under No Action a discretionary road-access request is generally turned down at screening regardless of merit; under the Proposed Action the categorical bar is gone and the request gets an individualized review the DEIS itself predicts will produce more approvals. The agency's own Biological Assessments state the same expectation in stronger terms: under the rulemaking "it is reasonably certain to occur that there would be an increase in new proposals for uses involving road construction or reconstruction." That is a firmer prediction than the DEIS's own "could result in," and it comes from the agency's biologists rather than its planners.

None of these figures carries a source, though each is in principle a directly verifiable number the agency could pull from its own databases. Taken alone the gaps are minor. The pattern is not: across the recreation, special-uses, minerals and socioeconomic sections, precise operational statistics appear without traceable sources, and a reader who wants to test the economic case for easing access has no way in. That pattern is documented in detail in How the Analysis Was Built.

These two statements are hard to hold together: a 2024 authorization of mine expansion inside an inventoried roadless area, and unbroken roadless consistency across the same decade. The DEIS does not reconcile them. Neither statement carries a Record of Decision, Federal Register citation or project document, so the scope of the authorization cannot be checked. The expansion is also omitted from Table 4 — the DEIS's own list of ongoing and reasonably foreseeable actions — which is where a cumulative effects analysis would have to pick it up, and its acreage and resource effects appear in no cumulative analysis for this rescission. Appendix 4 uses the same episode to support its finding that nothing relevant has changed since 2020, resting that finding on a comparison the record does not contain, and does not address the Biological Assessment's own Greens Creek finding at page 127.

Read narrowly, this is almost certainly true and unremarkable: adopting a rule requires no permit. Read as a description of what follows, it sits uneasily beside the rest of the document — site-specific project analysis, Endangered Species Act section 7 consultation producing "may affect, likely to adversely affect" determinations for 327 species, existing permitting under the 1872 Mining Law, and the Clean Water Act section 401 certification authority a Tribe asserted on this record. The DEIS does not say which reading it intends. Given that the same document elsewhere relies heavily on later project-level permitting as the reason effects need not be analyzed now, which reading applies is not a small ambiguity.

This distinction sets the outer limit of what the roadless rule can control. For locatable minerals the DEIS is describing a floor of access that exists regardless of roadless status. For leasable minerals the agencies have real discretion, which is where the rule's road restrictions actually bite. That is why the alternatives barely move locatable mining and meaningfully change leasable access.

Both processes give the Forest Service a real but bounded say — consent conditions and surface stipulations, not a veto over the lease itself. Once a lease exists it becomes "a contractual agreement with the government that allows reasonable access," modifiable later to add ancillary facilities. That contractual character is what makes leasable rights, once granted, hard to unwind under any alternative.

Leasing already reaches into land the 2001 Rule was meant to keep road-free, because leases predating the Rule keep their rights. The document does not state what share of the affected acreage these leased acres represent, so the reader has to judge scale without the comparison.

The 19 exceptions are offered as evidence that the rule has not blocked locatable mining. There is no denominator: without tracking of total proposals, 19 approvals could be nearly all requests or a small fraction. This is the same admission that governs the exception-use argument in the purpose and need — the Biological Assessments state flatly that "there is no national-level tracking of the review, approval, or implementation of 2001 Roadless Rule exceptions."

The agency's own data point — six of eleven major producing mines already inside roadless areas under the current rule — rebuts the idea that the rule has been a barrier to locatable mining. Where No Action does constrain is the leasable side: the no-surface-occupancy stipulation on new leases is a real limit on where roads can go.

By the DEIS's own account the locatable effect is psychological — removing a perception — not a change in what the law allows. The concrete change is dropping the default no-surface-occupancy stipulation, which makes road-building for new and renewed oil, gas and coal leases materially easier to approve.

Everything turns on which side of the retained-versus-removed line a lease falls on — and that line is drawn by the half-mile buffer whose basis the DEIS never states, under a rule provision that lets the Chief move boundaries "for other reasons." A reader who wants to know whether a specific tract keeps the no-surface-occupancy protection needs a boundary determination the document does not fix.

Existing rights are untouched by any alternative, and the real choice is how much surface-disturbing road-building is allowed for future leasing. There is a wrinkle the DEIS creates for itself here. The compliance table designates this minerals analysis as the documentation of compliance with Executive Order 14154, Unleashing American Energy — and the Economic Analysis states that "we anticipate that there would be no net change, relative to baseline conditions, in domestic leasable minerals production under the proposed rule." A section finding no net change is an odd place to locate compliance with an order directing expansion, and the DEIS does not explain the pairing.

Every one of these figures — the GDP contribution, the mine revenue, the lease acreage, the commodity value and the royalty breakdown — appears without a source. They are the economic backbone of the case for easing access, and they are also part of a wider pattern: headline figures in the affected-environment and effects chapters attributed to sources that do not contain them, or to sources absent from the document's own literature-cited list. A reader cannot check the benefit side of the ledger the DEIS is asking them to weigh.

The reference entry identifies no dataset, table or release, so the comparison cannot be reproduced. This is a defect in the accuracy of the document rather than a request for more analysis: an agency may not rest a quantitative comparison on a citation that does not identify the data behind it. A government letter in the agency's own record shows what a checkable version looks like — a state reporting outdoor recreation at $13.9 billion in value added and 130,000 direct jobs, sourced to a named state report.

The quoted passage cannot be verified against the 2020 document from what this record contains, so neither its accuracy nor its original context — including any caveats about mitigation or site conditions — can be confirmed. The deeper problem is structural: an earlier document's expectation that mining would expand is treated as sufficient justification for a later approval, with no site-specific analysis of what the expansion actually did to habitat, water quality or roadless character. That is the same move Appendix 4 makes to carry the whole 2020 Tongass analysis forward, and it is taken up in The Tongass National Forest.

That quantified detail on benefits is not matched by comparable analysis of the costs the benefits depend on. The asymmetry is visible in the Cost Benefit Analysis itself, which enumerates forgone recreation benefits at a minimum of $6.1 million annually and forgone passive use values at $5.3 to $11.5 million, files agency road maintenance burden as an unquantified cost while publishing per-mile road costs two pages earlier, and leaves invasive-species establishment and treatment at zero despite conceding elsewhere that activity under the action alternatives could increase spread. The net present value the agency ultimately reports spans from a $92 million loss to a $199 million gain — a range the DEIS itself never states.

The forest at issue is western and conifer-dominated, which is also where the DEIS's own tables put wildfire, drought and insect risk. A large share of it — pinyon/juniper — is explicitly non-commercial, so "more acres opened to management" does not translate directly into more timber cut. The forested total itself is the denominator for every harvest percentage in this section, and the DEIS reports it three different ways: one figure from satellite land cover, another from forest inventory, a third in the Biological Assessment, differing by about 4.5 million acres, with no statement of which the effects analysis uses.

Old growth and mature stands together are two-thirds to three-quarters of the forested roadless landscape, and they carry the most stored carbon and structural habitat value. How much of them falls inside the acreage opened to harvest is one of the most consequential numbers in the document. The agency's own mature and old-growth inventory is listed in the reference section and no finding is drawn from it anywhere — the reader cannot tell whether it was considered and found not to change anything, or simply not applied.

Two statements in that sentence are contradicted by the source cited for all of it, and both are correctable from that source. The regional figure is wrong and the sentence does not reconcile with itself: the five removal regions total 75.0 MMT, not 63.5, and the DEIS's own arithmetic confirms it — 63.5 less the 31.5 it reports is 32.0, which cannot produce the 43.5 net figure in the same sentence, while 43.5 plus 31.5 is 75.0. The published 63.5 is that total with the Eastern Region's 11.5 MMT of removal left out. Second, the DEIS credits the sink to "increasing forest biomass," and the source says the opposite about this particular sink: "nearly all the annual net carbon sink (-38.1 MMT CO2 Eq.) is due to transfers from the live biomass carbon pool to dead wood carbon pool in the Western United States." Roughly 88 percent of what the DEIS presents as growing forest is, in its own source, carbon moving out of living trees into dead wood. That difference is material, because dead wood in the fire-prone West is a transient pool subject to re-release through combustion and decay — and the DEIS concedes on the same page that carbon loss from fire and insects "can be significant, long lasting, and expected in any of the alternatives." The sign convention is not the issue: parentheses and minus signs in the source denote removal, so "sequestered 43.5 MMT" states that correctly.

The document's own numbers show roadless-area risk at or below risk on managed land outside designated areas — 34.2 percent in Northern Region roadless areas against 50.4 percent outside them. That comparison cuts against the argument that limited management inside roadless areas caused forest-health decline. The figures behind this discussion rest on citations a reader cannot follow: paywalled articles, and in one case a citation resolving to nothing in the document's own reference list. This is not a request that the agency buy public access to copyrighted work — it is that an analysis stating its own reliability should rest on sources a reader can check, and here the only statement of that reliability is itself unverifiable.

The 500-acre average is the baseline the entire comparison rests on — what No Action continues, and the yardstick for any increase. It comes from an activity-tracking system named in prose and absent from the reference list, with no query described, and the same figure is attributed to two different land bases in this document and in both Biological Assessments. The DEIS also uses the 2001 Rule's own design language — that stewardship cutting was "expected to be infrequent" — as evidence the rule failed, then carries that characterization into the effects analysis as a predictive finding supported nowhere. A rule working as written is not evidence it did not work.

National harvest is set to rise regardless of what happens to the Roadless Rule, and the DEIS says "there could be more opportunities within the analysis area due to the agency's emphasis on increased timber harvest" — meaning part of that national increase is expected to land inside former roadless areas. This is where the no-action baseline contradicts itself: at page 225 the DEIS says roadless harvest does not rise under no action, citing for that the section at page 79 which says some of the projected 25 percent increase lands there. Which is right determines how large an increment the rescission gets credited with. The Economic Analysis adds a third statement — that "the proposed rule does not include changes to the projected timber sale quantity guided by existing land management plan direction" — and nothing nets the three together.

No Action is the only alternative where roadless harvest is decoupled from the national push described above — and that decoupling is exactly what the p. 79 discussion contradicts. Everything downstream for this alternative assumes natural processes remain the dominant force. Because Alternative 1 is the baseline, the asymmetry matters: the action alternatives get quantified projections and this one gets a narrative, so the comparison the rulemaking turns on is not made on common terms.

This is the largest projected increase of the three, and the numbers carrying it do not hold. The 16 percent is wrong on the DEIS's own published forested acreage: 4.8 million acres is 22.9 percent of it. The narrative's operable acreage does not match the document's own supporting tables, which report 13.4 million acres for the same category — a gap of nearly 9 million acres reconciled nowhere. The three thresholds producing the operable estimate are unsourced, and the agency told both consulting wildlife agencies that the underlying review "is currently undergoing quality control and Regional/local vetting and the data presented will be updated in the final BA," a caveat the public never sees. The DEIS frames these as upper-bound comparison estimates rather than predictions, which is fair — but an upper bound computed on a denominator the document contradicts is not a bound a reader can use.

A real increase over No Action, geographically narrower than Alternative 2 — and carrying the same table-versus-narrative gap, 4.3 million acres in the text against 10.0 million in the appendix. The Alternative 3 acreage is separately mislabelled in the vegetation conclusion, and its percentage cannot be derived from any forested acreage the DEIS publishes. Most consequentially, the two alternatives' published ranges — 5 to 10 percent and 4 to 9 percent — overlap across nearly their entire width, and the Comparison of Alternatives collapses them into a single "4 to 10 percent" figure no analysis in the record actually derives. On the disclosed information a reader cannot tell whether the alternatives differ on harvest at all, which is the comparison the document exists to support.

This is the agency's own accounting of how much old growth sits inside the footprint where harvest becomes more likely, and its own acknowledgment that roughly half is not merchantable. What is missing is the effect of the roads needed to reach the rest. The DEIS cites a general finding that fragmentation can reduce biodiversity by 13 to 75 percent and never applies that range to the forest types and areas at issue — while its own Appendix 5 computes a full quantitative fragmentation baseline for exactly this land, and the comparative effects conclusion never uses it.

Expanded access is presented as a forest-health tool without any projection of how much mortality it would prevent — the risk is quantified region by region and the benefit is qualitative. That asymmetry runs through the whole rationale: the DEIS's own Tables 14 and 21 contradict the causal chain the rulemaking rests on, showing roadless-area insect risk at or below managed-land risk and roadless ignition density below roaded-land density. A claim that restricted management caused forest-health decline needs to survive the document's own tables, and it does not.

The DEIS holds both terms of this calculation and never multiplies them. It quantifies the carbon stock. It quantifies the projected harvest increase. It then states that the effect "would be small" without converting the harvest into tons of carbon or a net greenhouse gas comparison across alternatives. That is a magnitude finding resting on a comparison the document does not perform, and the fix is one of two things: quantify the change in stock and net flux under each alternative on the same basis as the volume projection, or correct the finding to state that the magnitude was not determined. The document's only response to commenters raising the same point is a cross-reference back to this section.

The comparison is: essentially no change under No Action, the largest opening under the Proposed Action, a narrower middle under the Modified Rule. Whether that means more logging on the ground depends on plan direction, markets and budgets the DEIS says it cannot predict. Plan direction is doing more work here than the document acknowledges: it is offered as the constraint that survives rescission, its acreage figures merge binding standards with non-binding guidelines, its vintage count does not reconcile to the stated total of 126 plans, and the DEIS concedes on the same page that a large share of plans probably contain no direction for roadless areas at all.

Three problems cut across the section. The baseline acreage is inconsistent — 44.7, 58.5 and 58.2 million acres appear for the same land base without reconciliation, and the superseded 58.5 figure is still stated in the present tense in the Cost Benefit Analysis. The forested denominator is stated three ways. And the treatment of costs against benefits is one-sided: harvest volumes, acreages and percentage increases are stated with precision while fragmentation, carbon loss and cumulative habitat effects are described qualitatively, even where the document has already computed the underlying metrics. The caution that plan restrictions "could change over time through future amendments" is worth reading carefully, since plan direction is simultaneously the reassurance offered that protection survives rescission.

This framing makes wildfire the primary lens for judging the alternatives, which means the fire data has to carry the rationale. It largely does not, and the reasons are specific rather than atmospheric: the ignition gradient the DEIS measures is never applied to the alternatives, the containment rates are never applied to the road classes the rescission would produce, and the emissions comparison the document states as a finding is one it says it did not perform.

These are the most consequential numbers in the document, because they are the baseline any claimed fire benefit is measured against — and they point the opposite way from the rationale. Roadless areas already ignite at less than a third the density of roaded land, and the document states the mechanism itself: human-caused ignitions increase with proximity to roads. Two errors in the figures need correction. The high and moderate severity values are transposed between the text and the table, understating high-severity fire in exactly the land base at issue. And the two burned-area tables report on incommensurate bases — one measuring acres burned by fires originating within a designation, the other an intersection-based figure — while the text draws a comparison across them as though they were the same measure.

The document's own conclusion is narrower than the rulemaking built on it. Hazard is not risk to people — the index "does not incorporate the presence of communities, infrastructure, or other highly valued resources and assets" — and most high-hazard roadless acreage is not near a community. So the community-protection case is strongest on a minority of the acres, and the rule is nationwide. One figure here needs correction: the reported hazard percentage is computed over a denominator that includes 12.5 million acres of Alaska roadless area the hazard layer does not measure. And the interface acreage that defines which land Alternative 3 removes is reported as determinate while the agency's own sensitivity analysis puts a measured 12 percent uncertainty band around it.

This is the agency conceding the premise of its own rationale. Fuels work is already legal under the current rule; 1.8 million acres of it happened; and a study the DEIS itself cites found the rule was not a meaningful constraint. The document never reconciles that with its claim that the rule "contributed to the lack of active management." The Economic Analysis says the same thing again — hazardous fuels treatments "can occur within inventoried roadless areas if they do not include road construction or tree cutting, or if exceptions are granted." A government letter from a state with roadless areas makes the point from the field: under the current rule those areas "already allow road construction and reconstruction in specific situations, including fire."

This is the most exposed claim in the document, because it is the central justification and it is contradicted rather than merely uncited. The drought and temperature premises are asserted at a scale the record does not reach. The wildfire trend is asserted with no analysis of wildfire on the affected lands since 2001, against a fire chapter reporting lower ignition density inside roadless areas. The insect-and-disease trend rests on annual snapshots that cannot establish direction of change, against a risk table showing roadless-area risk at or below managed-land risk. The middle link — that the rule limited management — is defeated by the 1.8 million acres of fuels work accomplished under it and by the agency's own admission that it does not track exception use at all. And the whole chain never addresses the competing explanation the record supplies: climate-driven change, fire-suppression history, and drought operating independently of who was allowed to log.

No Action is the baseline the fire benefits of the other alternatives are measured against, and it is described qualitatively while the action alternatives get acreage. The claim that exception use "would continue to be infrequent" is a prediction resting on the same exception record the agency says it does not track.

This is where the fire analysis stops short. The DEIS concedes the ignition increase and discards it on an access assumption it never quantifies. It has everything needed to do otherwise: Table 21's ignition densities by designation, and a road-distance relationship showing human-caused ignitions rising with proximity to roads. Applying those to the road mileage the alternative would produce is the calculation the document assembled its own data to support and then did not perform. The agency's Economic Analysis states the direction plainly — "if there is an increase in public road access, the number and frequency of wildfires could rise" — and leaves the resulting cost at zero in the cost column. Government letters in the agency's own record supply the magnitude the DEIS omits: over 53 percent more human-caused ignitions in the first 200 metres from a road; 88 percent of wildfires human-caused with most starting near roads; more than 80 percent of wildfires occurring within 800 metres of a roadway; ignition density highest within 50 metres of roads.

This is the more geographically targeted version — less interior flexibility, tighter focus near communities, which is a coherent trade. It depends entirely on the interface layer defining which acres are removed, and the DEIS never identifies that layer or states the basis for the half-mile road buffer paired with it. A government letter in the record puts the interface overlap at about 23 percent — roughly 10.2 million acres — a figure the document's own numbers have to be reconciled against.

This is the strongest evidence in the document for the road-access case, and the DEIS does not carry it to its conclusion. Containment performance depends sharply on road standard: high-standard roads help substantially, low-maintenance roads barely at all. The rescission would principally produce temporary and low-standard roads — the agency's own assessment says timber harvest "would likely generate the highest need for permanent and temporary roads" and that new permanent roads are relatively unlikely. Applying Table 26's rates to the road classes the alternatives would actually generate is the connection that determines whether the wildfire benefit exists, and it is never made. The initial-attack comparison has a separate problem: success is defined by outcome, and the DEIS concedes that less-than-full-suppression strategies are used in roadless areas, which makes the designations non-comparable on that measure.

These are not irreconcilable in principle — roads could raise ignitions and still improve suppression, and the net could go either way. The problem is that the DEIS asserts the net without computing it. At page 113 it concludes "the net impact of new roads on fire occurrence is likely small or insignificant, given the potential for improved response times and increased initial attack success," and the record supports only the response side of that netting. The ignition increment is conceded and never quantified anywhere in the document. A net finding with one term missing is not supported at page 113, in the fire analysis at pages 93–113, or in any source the DEIS cites.

The tools that have done most of the fuels work in roadless areas — managed wildfire and prescribed fire — do not depend on road access. The tools the Proposed Action would expand have played a small role. That is directly relevant to the added value of road-building for fuels reduction specifically. The fuel deficit is also not unique to roadless areas, which undercuts any claim that roadless status caused the buildup. One figure here needs reconciling: the wildfire-use accomplishment total cannot be squared with the document's own burned-acreage tables, and the post-fire assessment criteria that would explain the difference are never stated.

The sourcing gap runs through the whole section — the harvest and fuels baseline, the ignition-density comparisons, and the road-containment statistics all appear without a traceable dataset. The emissions admission is the sharper problem. The Air section states an affirmative comparative finding that the action alternatives reduce wildfire emissions relative to no action, and three pages earlier states that the net analysis supporting it was not performed. The agency holds the inputs — treatment acres and harvest volumes by alternative — so either the comparison gets made, or the finding comes out of Table 39, the Air conclusion, and the summary comparison of alternatives.

Because a standing executive order commits agencies to avoid worsening the problem, this analysis bears on whether the decision complies with a directive already in force. The agency's own Biological Assessment states the mechanism in terms the DEIS does not: "roads create favorable conditions for invasive plants by providing light gaps, dispersal corridors, and reduced competition," and vehicles and road maintenance move them. It also reports that of the 134 listed plant taxa in its analysis, 81 — 60 percent — are threatened by invasive or non-native species. That figure appears nowhere in the DEIS.

The DEIS ties its own two central topics together here — ground disturbance and roads are documented vectors. The follow-through is what is missing: the document never quantifies how much new road construction would add to that risk, though it quantifies road construction potential elsewhere in some detail.

This is the baseline the whole section rests on, and it measures the wrong thing. Mapped extent is a record of what has been surveyed, not what is present — and roadless areas are less surveyed precisely because they lack roads. A government letter in the agency's own record makes the point directly: roadless areas "often have less non-native invasive plants due to the lack of roads and other pathways generally associated with their spread and distribution." So a low mapped percentage in roadless areas partly reflects the absence of the very vector the rescission would introduce. The DEIS acknowledges the undercount in one sentence and then reasons throughout the section as though the baseline were complete. Neither the 309,000-acre figure nor the 438-species count is sourced, and no survey coverage or data vintage is disclosed. On the arithmetic: 309,000 acres against roughly 40.1 million is about 0.77 percent, but 270,787 against Alternative 3's roughly 27.1 million is about 1.0 percent — so "less than 1 percent," applied to both, does not hold for Alternative 3.

Doing nothing does not mean no risk — spread continues, driven by vectors already present. The document ties this partly to resources for monitoring and treatment, a variable it does not quantify for any alternative.

The DEIS acknowledges the mechanism and describes its magnitude only qualitatively — "may be minimal," "could increase" — with no acreage projection, spread-rate estimate or cost figure. That last omission is concrete rather than abstract: the Cost Benefit Analysis's quantified cost column lists forgone recreation benefits and forgone passive use values and carries nothing at all for invasive-species establishment or treatment, even though the DEIS concedes here that the action alternatives could increase spread.

The DEIS treats the two action alternatives as functionally the same here, which is defensible — risk follows where and how intensely ground is disturbed, not nominal acreage. It also means the smaller footprint of Alternative 3 buys little on this resource.

The bottom line rests on two things the section does not establish: a baseline the document itself says is incomplete, and an assumption that future mitigation will work, with no statement of what those measures achieve in practice. The fuels-reduction offset is a real argument and it is stated without a magnitude on either side — neither the invasive acres added by disturbance nor the invasive acres avoided by reduced wildfire.

This is presented as a commenter claim and left there — no literature named, no statement of whether the agency verified or disputes it, and no dedicated assessment of road density or traffic against invasive occurrence anywhere in the main analysis. The agency's own Biological Assessment supplies exactly the mechanism the claim describes, so the material to engage it is in the record. Attributing a substantive point to commenters and neither adopting nor rejecting it is how a reasonably foreseeable indirect effect goes unanalyzed.

The same asymmetry appears across the document: economic benefits in precise dollars, ecological costs in qualitative terms. Here it has a specific consequence. The Cost Benefit Analysis explains that it leaves costs unquantified "because the proposed rule does not prescribe site-specific activities," and that reasoning is what keeps a conceded effect at zero in the cost accounting while the benefit side carries dollar figures to two significant digits.

Reducing substantive opposition to a percentage, with no topic-by-topic account of the agency's reasoning for proceeding, is the pattern documented in How the Analysis Was Built: the notice-of-intent record was screened under a superseded definition of "substantive," the appendix groups comments into themes and assigns a disposition to none of them, and the four-part rejection list does not say which ground applied to which concern.

Roads, not logging alone, are the persistent disturbance — most effects fade, road effects compound if maintenance lapses. The document discloses a large and growing maintenance funding gap in its roads section and never connects it to the environmental risk of building new roads under the action alternatives. The agency's own Biological Assessment states the ranking plainly: "the causal effect of road building for peak/baseflows creates the largest effects to peak/baseflows of any other activity in this consultation." This is also where the DEIS's treatment of permanent resource commitments sits: it states that no alternative causes irreversible or irretrievable commitments, on the same page it finds the underlying road and harvest activity reasonably foreseeable — and the reason it gives is defeated by its own definition and precedent on that page.

Effects range from short to long term with severity, and the document notes even temporary roads and skid trails "are likely to reduce productivity for several decades." Because roadless areas are relatively undisturbed, new compaction here is a change from an intact baseline rather than an addition to degraded ground — which is precisely why "several decades" of reduced productivity is the kind of commitment the irreversible-and-irretrievable analysis exists to disclose.

The DEIS draws the conclusion itself: roadless areas in the Intermountain, Southwestern, Southern and Eastern regions "are at higher risk for impairment due to additional activities" because more of their soil is already Fair to Poor. The regions with least room for additional damage are the ones the document flags as most exposed.

The 100-to-300-times figure is uncited — no study, author or dataset for one of the most dramatic statistics in the analysis. The practices credited with holding sediment to acceptable levels are also never specified: the section does not say which apply to soil, water or air individually, or name the governing manual. That matters because the agency's own Biological Assessments discount exactly this reassurance. They state that "the nature and timing of future projects, and inclusion of minimization measures are unknown which means the effects of sediment and turbidity magnitude is uncertain," and say only that there is "reason to believe these minimization measures will be included in any potential future project" — belief, not the mandatory practice the comparison of alternatives credits at page 32. One assessment also finds that magnesium chloride road treatment causes direct fish mortality and food-web effects, an impact no practice in the comparison addresses.

This figure is sourced, unlike the sediment multiplier — a distinction worth noting when weighing which numbers here are traceable. The analytical gap is different: the DEIS treats landslides and debris flows as a soil-resource metric and never asks who is downslope. In the same chapter it maps communities exposed to wildfire. A government letter in the agency's own record reports the consequence directly — a dramatic increase in landslides, with one autumn storm producing 55 slides that damaged waterways and highways — and another warns that roads and logging on steep ground would worsen damage from storms like Hurricane Helene. Naming a downslope receptor is not new analysis; it is the question the chapter asks about fire and not about slope failure.

Protection against widespread soil damage rests on two things: practices the Biological Assessments say may or may not be applied, and low operable percentages limiting where activity could occur — not a numeric cap on acres disturbed or sediment delivered. The operable percentages are themselves in question, since the operability thresholds are unsourced and the underlying review is unvetted by the agency's own account. And the DEIS never converts foreseeable additional harvest and road-building into a quantified projection of resulting sedimentation or fragmentation, though it performs that granularity for the Tongass alone.

The nutrient and temperature effects are framed as temporary and self-correcting. The Pacific Northwest data point complicates the framing that western water quality is generally better — on impairment it is the western outlier, which the document flags and does not carry into a different regional risk ranking. One authority is missing from the framework entirely: Clean Water Act section 401 state and tribal certification, which a Tribe asserted on this record over "any federal permit or license that may result in a discharge into our waters." That authority is a direct constraint on the project-level flexibility the rescission promises, and the DEIS does not mention it.

No single project is likely to change flooding or yield on its own. The harvest-versus-roads distinction is the meaningful one for permanent road construction: roads change drainage pathways, so their hydrologic effects recover more slowly even though harvest disturbs more area. The agency's Biological Assessment puts the same point in ranking terms — road building creates the largest peak and base flow effects of any activity in the consultation.

Tens of millions of people draw water from these watersheds, and the document frames the background climate trend as potentially calamitous before any rule change — a demanding baseline for adding effects to. What the analysis does with all this is the problem. Table 35 inventories intakes and populations served by region and names no watershed. The effect statement for the proposed action is a conditional clause with no consequent — the analysis runs six pages and states no effect under any alternative. Alaska is excluded. Section 401 certification authority is absent. Government letters in the agency's own record supply what the table omits: 350 major watersheds in roadless areas serving millions; a named 36-square-mile watershed supplying over 35,000 residents; two million North Carolinians whose drinking water traces to a roadless area; a utility opposing federal action that risks water quality for 400,000 users in one metropolitan area, drawing 60 percent of daily supply from one range.

This is the plainest statement of the water trade-off: No Action closest to natural, the Proposed Action opening the most acreage and the most people's supplies, the Modified Rule the same pattern over fewer intakes. The reassurance carrying it — that practices reduce impacts to Clean Water Act levels — is the one the agency's own assessments say may or may not be applied to any given future project.

This is the thinnest part of the analysis and it carries a great deal. "Minimal" is undefined — no threshold, metric or comparison — and the finding rests on a forest plan the DEIS elsewhere calls indeterminate. The DEIS also materially alters the 2020 finding it relies on, in a way that removes the grounds this rulemaking most affects, and states its fish conclusion two ways at pages 243 and 318 that do not agree. The Tongass aquatic conclusion rests on a forest-wide average road density that cannot support it, and omits fish passage entirely — road-stream crossings, the mechanism a Tribal government reports observing directly as "forest road failure, fish passage obstructions and sediment delivery." Against that thin treatment, the NMFS Biological Assessment reaches a likely-to-adversely-affect determination for every listed salmonid species in the action area, records a determination for Oregon Coast coho and its critical habitat that the DEIS analyzes nowhere, and reproduces the agency's own critical-habitat designation finding that tributary habitat quality is "excellent in wilderness and roadless areas" — the baseline the adverse-effect finding never states.

Air quality is generally good across national forests, and the Pacific Southwest differs in kind: nearly two-fifths of its roadless acreage sits in an area already failing the federal ozone standard. The document singles it out as the region where new activity matters most for air.

By the document's own table, the single largest air factor is rated worse under No Action than under either rescission alternative, which is what makes the air comparison come out favorably for the action alternatives. And the document concedes in this same discussion that a quantitative comparison of wildfire emissions across alternatives "is not possible at this time." So Table 39's decisive row is a qualitative judgment presented in a comparative table, and the affirmative finding drawn from it — that the action alternatives reduce wildfire emissions relative to no action — rests on an analysis the DEIS states three pages earlier was not performed. The agency holds the inputs: treatment acres and harvest volumes by alternative. Either the comparison gets made, or the finding comes out of Table 39, the Air conclusion and the summary comparison of alternatives.

The conclusion is conditional by design: effects are foreseeable in general terms and deferred to project decisions, with existing legal requirements named as the safeguard rather than any limit set here. Three things undercut that structure. The plan direction offered as protection merges binding standards with non-binding guidelines and, by the agency's concession, may not exist for a large share of roadless areas. The practices offered as mitigation are ones the Biological Assessments say may or may not be applied. And the deferral itself is the mechanism by which the irreversible-commitments requirement is discharged — on the same page the underlying activities are found reasonably foreseeable, and for effects no later project decision will revisit, since this rulemaking is the only point at which a nationwide prohibition is compared against its removal.

These figures anchor everything downstream, and none can be traced to its source. The species count cannot be reconciled to the Biological Assessment it comes from. The 79-plus-19 critical habitat count cannot be reproduced from the incorporated assessments, which matters because the count bounds the set of habitats whose exposure the effects analysis addresses — an unreproducible count means the analysis is bounded by a set no one can verify, and the same data is what the agency carried into consultation. The "thousands" of sensitive species is an uncited sentence, and the section it refers the reader to does not supply the baseline. The assessment itself explains part of why: its databases "do not differentiate between DPSes of vertebrate species listed under the ESA, which causes potential inflation of exposure," and reptile representation in those databases is poor. What the DEIS does not carry forward is the assessment's own enumeration — 134 plant taxa, 74 endangered and 60 threatened, 38 with designated critical habitat — which is the kind of baseline a viability conclusion needs.

The DEIS supplies the science and does not apply it. It cites the global 13-to-75-percent range without locating the affected areas within it or identifying which habitat types and species would bear the effect — while its own Appendix 5 computes a full quantitative fragmentation baseline for exactly this land: mean patch area, largest patch index, shape index, nearest-neighbour distance, by region and forest type. Those metrics never reach the comparative effects conclusion. Two further problems sit in the appendix itself: the mean patch area values do not reconcile between the regional and forest-type tables, and the basis on which patches were delineated is never stated, so the metrics cannot be reproduced. The assessments supply the effects side the DEIS leaves general: fragmentation is "explicitly identified as a threat to 14 plant species in this analysis and likely is a concern for many more," increased road density affects salmonids "typically through habitat fragmentation and sedimentation," and for wolverine the federal listing record turns on "the fragmented nature and distribution of wolverine habitat."

No Action is the option least likely to disturb biodiversity near-term, because it keeps roads and logging out. The DEIS states a counterweight — reduced ability to intervene against conifer encroachment, insect outbreaks and severe fire, so that long-term "potential adverse effects to biodiversity from lack of management would be more likely" — and qualifies it itself, noting those effects are "less certain to occur." That qualification is doing real work: the certain effects run one way and the speculative ones the other.

The DEIS is explicit that effects cut both ways — "a road- or timber harvest-related beneficial effect for one species, may, in fact, represent an adverse effect for another." What is missing is magnitude on the cost side. Two examples the agency itself produced and the DEIS did not use: the Biological Assessment states that roads "can have a negative effect on bird abundance up to one kilometer away, on average," and it computes that of 137 wolverine detections in inventoried roadless areas, 69 fall in operable areas. Those are exposure figures, and the DEIS's species summary — which the document says summarizes the assessments — is a presence-and-absence table that omits every one of them.

Concentrating change at the edges rather than the core is a real difference, and the document links it to lower risk given that "the size of a protected area has been positively related to biodiversity." Two things qualify it. The 17.9-million-acre figure does not reconcile: its named categories sum to 16.1 million, leaving 1.8 million acres unaccounted for, and read another way the unstated remainder runs to 11.2 million. Those residual parcels are by definition neither roaded nor in the interface, so the edge-effects reasoning offered for this alternative does not reach them. And the rule text does not fix the boundary the acreage assumes — it lets the Chief modify boundaries "for other reasons."

This places the real decisions downstream, at the project level and in future plan revisions, which is accurate as far as it goes — rescission authorizes no project. It also means the comparative conclusions rest on what is expected to happen later under standards this document does not set. The fragmentation metrics that would have made the comparison quantitative were computed in the appendix and left there.

This is the most consequential open question in the document, and the DEIS reports it two ways in adjacent pages: page 167 says consultation is anticipated, page 168 says it "is ongoing and will be completed prior to a final decision." No Biological Opinion is published. Alternative 3 receives no salmonid determinations at all. The single reliability statement offered — that gaps were resolved conservatively — is contradicted by the assessments themselves, which record error running in both directions: occurrence data that are "opportunistic and do not represent the results of exhaustive structured species surveys nor do they include negative occurrence data," databases that inflate exposure by not distinguishing population segments, and exposure "assumed" for whole taxonomic groups for want of data. The USFWS assessment omits its status-of-the-species section entirely as "not prudent" given scope, while the companion NMFS assessment supplies exactly that content for the same action. And the exposure screen has no road-effect zone, against the assessment's own finding that roads affect bird abundance up to a kilometre out. There is also a reversal no one has explained: when the agency promulgated the 2001 Rule it found, with the Services concurring, that these same prohibitions were not likely to adversely affect listed species — a determination both assessments recite in their opening pages.

The conclusion is reassuring on its face and rests on a baseline that is not there. The affected-environment sentence establishing what sensitive species are present carries no citation, and the section it points to does not supply the enumeration. The scoring method — how exposure and effect were assigned, what made a likelihood high, moderate or low — is summarized briefly without the thresholds or landscape metrics that would let anyone check it, and the assessment concedes exposure was assumed for many species for want of data.

The DEIS states two things about this that cannot both be true. At page 169 it says existing protections afforded by land management plans were reviewed and carried into the analysis. At page 291 it says "the relationship of how forests manage habitat conditions for spotted owl or grizzly bear are not reflected in this analysis." Those are the two species the biodiversity discussion uses most as examples. The inconsistency reaches the exposure analysis behind determinations for 327 species and 71 critical habitats, and resolving it requires the agency to state which plan direction was and was not carried in. The Biological Assessment adds a related problem the DEIS never mentions: for an identified set of species the plans relied on for consistency predate the listings — crayfish listed in 2016 and 2023 against plans from 1995 and 2004, and a plant whose 1995 forest plan "does not contain species-specific protections... because the plant was listed" afterward.

The provision contains no cross-reference to the Endangered Species Act — no requirement that a modification affecting designated critical habitat or occupied habitat for a listed species go through section 7 consultation first — against a document reporting 327 species and 71 critical habitats with adverse-effect determinations tied to roadless status. It carries no acreage cap, no aggregate threshold triggering supplemental review, and no public tally of how much land has left roadless status over time. Without those, cumulative habitat effects from repeated modifications accumulate with no point at which further review is required. The assessment states the consequence for one species in categorical terms: "by definition core habitat for grizzly bears is lost when a road is constructed, unless it is mitigated by the creation of new secure habitat elsewhere," and it concedes that the contribution of roadless-area loss to that outcome was not analyzed.

The mechanism is stated and the baseline measures survey effort rather than extent, for the reasons set out in Non-Native Invasive Plant Species. The figure the assessment supplies and the DEIS omits is the one that connects this to biodiversity: 81 of 134 listed plant taxa — 60 percent — are threatened by invasive or non-native species.

The purpose statement announces a balancing of competing values and never performs it against the conservation evidence in the agency's own record — the connectivity, refuge and watershed values this same document describes roadless areas as providing. A purpose built around the costs of the status quo, without an equally developed account of what large connected habitat supplies, narrows the range of alternatives before the tradeoff is weighed. That is not an abstract concern here: it is the stated ground on which the one alternative answering the wildlife findings — retaining boundaries and the permanent-road prohibition while allowing temporary roads and harvest — was eliminated, for a reason two state wildlife agencies' own letters in the record contradict.

This is a large inventory of quiet, undeveloped land that a substantial share of national forest visits depends on, and because 90 percent is already in the least-developed classes, any shift toward roads and motorized access has more room to change its character than it would elsewhere. The visitation figures carrying that point are unsourced. The DEIS states that its visitor monitoring program "does not track separately the number of recreation visits to IRAs," and in the very next sentence reports the 125-million figure with no citation, no footnote and no dataset named — while the sentences on either side are sourced. The Economic Analysis states the same limitation more plainly: "the number of recreation visits to IRAs is unknown and, thus, the recreation visitor spending in local communities associated with IRA recreation cannot be directly quantified."

These are the yardsticks for every comparison that follows, and the agency flags the underlying map as an approximation. The decision rules, buffer distances and data sources behind it are not documented, so the acreage figures that follow are reasonable estimates rather than counts — which is worth holding in mind when the same document presents Alternative 3's acreage as determinate.

No Action preserves the current character with the least change, and the DEIS never quantifies what that is worth — no acreage of unfragmented setting, no recreation-value metric, no dollar figure attributable to keeping the rule. The action alternatives get acreage tables and operability percentages; the baseline gets adjectives. This is the same asymmetry that runs through the comparison of alternatives, and it cannot be deferred to project review: no future project document will compare a nationwide prohibition against its rescission.

The exposure is concentrated in the semi-primitive non-motorized class — the second most protected, and nearly half of all affected land. The DEIS frames part of this as opportunity while documenting that roughly half the non-primitive settings are more likely to change. The operable-area figures carrying the estimate rest on the unsourced operability thresholds and on a review the agency told its consulting Services is not yet vetted.

The same kind of change over less land. The asterisk is that these acreages assume a fixed boundary — half a mile from roads, or within the interface — and the rule text for Alternative 3 sets no distance or interface criteria at all, describing a discretionary, case-by-case process decided by the Chief. The acreage of settings actually affected could be more or less than the tables project, and the document supplies no cap.

The document establishes that these areas meet a high bar precisely because they have been largely free of roads and logging — which is the causal claim the alternatives comparison then tests.

The document distinguishes effects that fade in a few years from a permanent road dropping an area from high or very high integrity to moderate or low. The mitigation it relies on — site-specific scenery design criteria — is applied project by project after this decision, not committed to here. Note the tension with the soils discussion, which states that even temporary roads "are likely to reduce productivity for several decades": visual recovery and ecological recovery are being described on very different timescales for the same disturbance.

The DEIS states the trade-off in its own terms: easier access for some users, less of the quiet remote setting other users specifically seek, and more conflict between them under Alternative 2. It does not net the two, and no quantity is attached to either side — which leaves the comparison resting on the adjectives.

By the DEIS's own figures Alternative 2 puts roughly 1,584 miles of operable-or-complex designated trail at higher risk against about 390 under Alternative 3 — a substantial difference in how much of the congressionally designated network could see closures, construction noise or long-term visual change. The trail baseline itself does not reconcile: the category mileages and percentages at page 186 do not sum to any single stated total, and no data system or pull date is given for any of them, while every other resource inventory in the same chapter identifies both.

Large new campgrounds inside roadless areas are unlikely under any alternative; the realistic change is smaller infrastructure following new roads. The recreation-infrastructure baseline has an internal contradiction in the agency's own footnote, and the $1.28 billion figure appears twice on a single page and nowhere else in the record, with no derivation in the Economic Analysis and no disclosed snapshot date for the underlying database. Deferred maintenance is an estimate, not a count, and none of the method behind it is stated.

Businesses built on unroaded terrain could lose operating area under the Proposed Action while the same alternative creates road-based opportunity elsewhere. The DEIS presents both and nets neither. The permit and revenue figures are unsourced, like the rest of this section's operational statistics.

This is one of the document's more direct statements that the Proposed Action could reach formally designated Wilderness, not just the areas being rezoned — because those areas currently absorb recreation pressure and hold intensive activity away from wilderness boundaries.

The relative ordering is consistent throughout the document and is probably right. The figures describing the scale are the problem, and the wilderness passage is the clearest case. The DEIS states that "the National Wilderness Preservation System includes approximately 111.9 million acres; of these, approximately 36.7 million acres are on National Forest System lands," and neither figure carries any citation — no inventory, no database, no date. The same subsection's visitation figures are likewise unsourced, and the derived percentages cannot all be computed from the acreage the record states elsewhere. This is not a citation-practice complaint: the wilderness discussion is where the DEIS establishes that designated wilderness remains available as a comparable recreation setting if roadless protections are removed, so the substitution finding rests entirely on figures the document does not source. The separate "over 4 million acres" of existing wilderness, study area and wild river protection — used to suggest a cushion exists regardless of this rulemaking — is likewise uncited, its components do not sum to the stated total, and whether the three categories overlap is never stated, which matters because that subtraction is what produces the 40.1-million-acre area the whole document analyzes.

The category is intentionally wide, so ground disturbance almost anywhere in a roadless area has some potential to touch a cultural resource. The agency's own statement that it cannot enumerate what counts underscores how much assessment is pushed to the project level — and makes the absence of any baseline in this document harder to justify, not easier.

This is the test that determines whether a site receives Section 106 protection. Traditional cultural places and sacred sites can qualify without a building or artifact, because their significance is ongoing living use rather than age.

Where the line falls between new disturbance and additional disturbance in an already-disturbed place is a judgment made project by project, and this document neither standardizes it nor discloses how it would be applied.

This defers the entire substantive analysis. The document provides no inventory of known historic properties in the affected areas, no ethnographic data on Tribal use, no estimate of how much acreage is likely to contain significant sites, and no assessment of the risk that ground disturbance destroys resources never recorded in the first place. That last category is the one deferral cannot reach: a survey at project stage finds what is findable, and the sites destroyed are the ones nobody knew were there. Deferring all of it without baseline data now does not discharge either Section 106 or NEPA's disclosure requirement, and the agency already holds the data a programmatic baseline would need — region-level site density is exactly what its own records contain.

The scale and age of what is at stake is large and much of it predates written records, so damage is irreversible and unrecorded losses may not be recognized as losses at all. The DEIS does not translate any of this into an acreage-level estimate of where significant sites are likely to be.

Because exceptions stay infrequent, this alternative ranks least-impact of the three. The prediction that exception use remains infrequent rests on an exception record the agency states elsewhere it does not track nationally.

The conclusion ranks impacts "more frequent and broader in scale under alternative 2," with Alternative 3 affecting "fewer cultural resources... because additional activities would be permitted on less land area." That ranking is drawn purely from land area, and land area cannot support it: site density is not uniform, and the areas Alternative 3 removes — near roads, near communities — are in many landscapes exactly where habitation sites concentrate. The comparison needs site-density data, which the agency holds and the document does not use.

The DEIS identifies roads, not harvest generally, as the activity most likely to cross the legal adverse-effect line — and as a harm that continues after construction through public access. Only Tribes with ancestral ties can identify some of these effects, which places the weight on consultation actually occurring and being substantive. That is the same consultation record the document reports as a tally with 30 of 64 requests unaccounted for.

The Section 106 process is described correctly as a matter of law and left abstract as a matter of fact: no existing programmatic agreements named, no consultation protocols described, no baseline surveys of where sites are. Those are what determine whether avoidance works in practice. Flagging for avoidance requires knowing where to flag.

Thirty of the sixty-four are unaccounted for — not held, not scheduled, and not explained as declined, withdrawn or pending — and the sentence reporting the tally carries no source. That gap sits in the only section the DEIS designates as its record of compliance with Executive Order 13175. The Tribes' own accounts on this record are more specific than the tally: one Tribal government records that at its September 3, 2025 quarterly meeting with four national forests "there was no mention of this" rulemaking; another states that Tribes participating as cooperating agencies in the earlier Alaska rulemakings had their "sovereign input... repeatedly ignored or marginalized." And the agency's own Tribal Summary Impact Statement records that the central procedural concern is the government "proceeding without adequate government-to-government" consultation. A count of meetings does not say what was raised, what changed, or why — and disclosing that disposition requires no protected information at all.

Reciting the list does not commit the document to analyzing how rescission affects each value. Combined with the absent baseline and the area-based ranking, the pattern is a value identified and not analyzed. There is a further wrinkle specific to this resource: the DEIS's assurance that later Section 106 review will hold down impacts rests on an assumption it never states — that Section 106 procedures will still be in place and unchanged. The agency's own Tribal Summary Impact Statement records the opposite concern, that "future regulatory protections (like Section 106) may be vulnerable to streamlining efforts." An analysis resting on future process should say what happens if that process narrows.

Tribal rights here are not a stakeholder interest to be weighed like any other; they are binding obligations existing independent of what the agency decides about roads. Each Tribe's rights differ by its own treaties and court orders, which is why the DEIS stresses site-specific consultation. That is also what makes the analysis's own limit consequential: the effects discussion is bounded by its own terms to "general considerations" and names no treaty, no ceded territory, and no usual-and-accustomed area anywhere. The Tribes on this record name them precisely — one stating it "reserved its right to take fish at its usual and accustomed fishing places in perpetuity," another that roadless areas "remain some of the last places Tribal members can exercise their treaty rights to hunt, fish, and gather," a third identifying sacred sites, ancient trail networks and usual-and-accustomed areas by name. The compliance table rests the agency's treaty-rights showing on a section that identifies none of them.

The DEIS documents that the Tribal input it received was consistently opposed to the rescission it evaluates. It does not identify which lawsuits Tribes cited or how that history bore on the choice among alternatives. The agency's own Tribal Summary Impact Statement adds what this summary omits: that "the majority sentiment among Tribal governments consulted is opposition," that Tribes "emphasized that they are sovereign governments, not mere stakeholders" and "requested consultation based on mutual concurrence," and that twelve Southeast Alaska Tribes "collectively proposed a Tribal alternative for the protection of traditional homelands" — an alternative the DEIS eliminated for four reasons, none of which reaches what was actually proposed.

A qualitative framework means the comparison rests on description rather than measurement — no acres of sacred sites near proposed roads, no miles of treaty-reserved fishing streams inside newly opened areas. The document concedes the quantitative information exists.

An existing channel for Tribal involvement, independent of the roadless rule. The DEIS does not say whether rescission expands or contracts it — which is a live question given that Tribes identified co-stewardship as one of their six themes.

Protection depends on disclosure, and disclosure carries risk — which is why the confidentiality provisions the DEIS cites elsewhere exist. Sites Tribes have not disclosed may not receive the protection in practice. A Tribal government on this record states the problem directly: ancestral village sites, travel routes and areas of deep significance "are not always documented." That is also why an area-based ranking of cultural effects cannot work: the resources least documented are the ones most at risk.

The DEIS establishes the category and never states what entered the analysis through it. Whether Indigenous Knowledge shaped the alternatives, the boundary criteria or anything else is the question the consultation record is supposed to answer, and the record is a tally.

All three keep future projects subject to their own consultation; the difference is how much construction and logging becomes possible. Whether project-level consultation is a sufficient answer depends on how the consultation on this rulemaking has gone, which is the subject of the last points in this section.

The DEIS states these risks candidly rather than minimizing them. It identifies no specific trail, site or area, leaving that to future consultation — which is a defensible allocation only if the consultation record supports it.

These are established mechanisms, not speculation, and the DEIS ties them to treaty-reserved rights rather than general environmental quality. What the document does elsewhere is inconsistent with this passage: it treats road access as a benefit to deer, while the Department's own consultation record credits the road prohibition with Alaska deer and moose recovery. A Tribal community on this record states the consequence in the plainest available terms — it qualifies as a federally designated food desert, with the nearest grocery store in another town.

The DEIS conditions adequate protection on two things it does not supply — robust monitoring protocols and integration of Indigenous Knowledge — and leaves both to project decisions this document does not make.

The ranking depends on future consultation working as described, and the adequacy of consultation on this rulemaking is what the next points examine.

No source is cited. The arithmetic leaves 30 requests — nearly half — with no stated disposition. The sequence compounds it: the Secretary publicly announced intent to rescind on June 23, 2025, roughly a month before Tribal notification began. Tribes on this record describe what the notification consisted of: an email to Tribal leaders on July 22, 2025, with one Tribe stating that "email notice does not constitute" consultation; a comment period of fourteen business days, which a Tribal government states "does not allow for meaningful participation during this early and critical stage"; and another noting that the reduced notice period "on top of the failure to offer consultation – is wholly inadequate for meaningful Tribal engagement." The agency's own guidance sets a substantially longer minimum. And the DEIS reports meetings without reporting what was learned or what changed, which is the showing Executive Order 13175 and the case law both look to.

Read within this document, "the DEIS and proposed rule" means this one — and this one's Notice of Intent published in August 2025. Hearings held in 2019 cannot have been conducted for it. The sentence needs correction. Underneath the date problem, the document takes two positions on Section 810 that cannot both stand: at page 246 it states no evaluation is required because this is a programmatic rulemaking, and at page 319 that the 2020 evaluations "remain valid and adequate" under Section 810. It also discloses no hearing locations, communities involved, resources evaluated or findings reached. Two government letters in the agency's own record state the requirement covers the Chugach as well as the Tongass, and the Tribal Summary records Alaska Native entities demanding "assurance that the ANILCA Section 810 subsistence evaluation is performed and disclosed prior to a Record of Decision." The evaluation the DEIS incorporates covers the Tongass only, which leaves the Chugach rescission with none at all — roughly five million acres of Alaska roadless area unanalyzed for subsistence.

Roughly 57,500 acres of currently inventoried roadless area could leave that status independent of this rulemaking — a foreseeable, overlapping change disclosed in a footnote and folded into no cumulative effects analysis. What makes it sharper is where the footnote sits: on the same page where Appendix 4 concludes that circumstances relevant to the effects analysis "have not changed." Treating the rulemaking as programmatic does not excuse the omission absent a specific commitment to analyze it later, and none is given.

Two problems with leaning on the older document. It is not in the DEIS's own bibliography for pinpoint citation, so the public cannot check what is being relied on. And "fully disclosed" is not accurate for every resource: the fire and fuels effects for the Tongass are not disclosed there. Seven years will have passed between the underlying subsistence hearings and any decision here, and the DEIS itself discloses intervening developments — harvest-volume directives, the pending conveyance legislation, a mine expansion, a principal harvester's exit from logging — without analyzing any of them against that baseline. The Section 810 determination being carried forward was also made for a rule scoped to the Tongass, and this action is nationwide.

This is the closest thing in the document to a map of who feels the effects, and it is populated with a single variable. The agency built exactly the county-level frame that an analysis of community-level effects requires, filled it with population and rurality, and then reasoned about county character from data it did not collect — no economic-dependence typology, no poverty rate, no persistent-poverty status, all of which come from a federal classification series the DEIS already cites elsewhere.

This is the agency identifying which communities would feel a policy change most, and it stops there. Table 47's results are not carried forward into the alternative-by-alternative effects discussion, so the document identifies 158 forest-dependent counties and 21 economically dependent ones and never reports what happens to them under each alternative. The counts themselves carry no citation. The DEIS's own compliance catalogue also identifies no authority under which the distribution of effects across communities is analyzed — while its effects text concedes those effects fall unevenly.

This decline is the backdrop for expanding harvest opportunity, and the headline figure is not supported by the data printed in the same paragraph. The claim needs correction to a figure the record supports, identifying what is measured — volume, value or acreage — the source, and the endpoint years. There is also a competing explanation the document does not address: independent research finds timber sales on at least the Tongass have run at a deficit, with administrative costs exceeding revenue, which would make part of the historic decline a matter of sale economics rather than regulatory limits. The one growing category is worth noting for a different reason: restoration and thinning work is growing, and it is the category least dependent on new road access.

These establish the scale of existing extractive activity, and essentially every 2024-specific figure here is unsourced. That matters because these are the numbers framing what more roadless access could add — and because the same pattern recurs: headline figures in the affected-environment and effects chapters attributed to sources that do not contain them, or absent from the document's own literature-cited list.

Recreation is by far the largest economic interest tied to these lands, by the DEIS's own comparison — and the citation chain under it does not hold. The FY2023 and FY2024 figures are attributed to a source published years before the data they support, across at least six separate statistics. The mining-and-utilities comparison rests on a reference entry identifying no dataset, table or release, so it cannot be reproduced. And the visitation figures themselves sit next to the DEIS's own statement that visits to roadless areas are not tracked separately. These numbers anchor the recreation side of the comparison, and none of them can be checked as published.

This is the agency's own estimate of the economic value riding on the areas being opened, and it dwarfs the projected timber and mineral gains discussed below, which run in single-digit millions per year. Three defects sit in the derivation. The $8.5 billion base is uncited. The 31 percent allocation is applied to produce figures that do not reproduce — the published magnitudes work out to 30 percent, not 31. And the source key carrying every roadless-attributed spending figure has no entry in the document's Literature Cited section at all. The Economic Analysis states the underlying limitation plainly: "the number of recreation visits to IRAs is unknown and, thus, the recreation visitor spending in local communities associated with IRA recreation cannot be directly quantified." An acreage-proportionality assumption is standing in for that, and the document never states it.

This is the reference point for both action alternatives, and it contradicts the document's own harvest discussion: at page 79 the DEIS states that some portion of the projected national increase would land inside the potentially affected areas. Which statement governs determines how large an increment the rescission gets credited with — the entire economic case for the action alternatives is the difference between these two accounts.

The DEIS's own figures show the timber gain is small next to the recreation value tied to the same land. What the document does not report is the net. The Cost Benefit Analysis does: a net present value spanning from a $92 million loss to a $199 million gain at 3 percent over 15 years. The DEIS states no net figure at all. The favorable end of that range also drops a cost the agency's own footnote names — a $6.1 million annual recreation floor — while counting a timber benefit the same passage labels "hypothetical." That $6.1 million figure is itself derived as an upper limit and then called a minimum. Two further omissions on the cost side: the analysis publishes per-mile road costs and files agency road maintenance burden as unquantified, and it leaves invasive-species establishment and treatment at zero after conceding activity could increase spread. The harvest projection carrying the benefit side is stated five separate ways across the record in three different ranges, on a volume model never disclosed.

A scaled-down version on paper. The quantified effects assume the half-mile and interface boundary is fixed, and the rule text sets no distance or interface criteria at all — it lets the Chief modify boundaries "for other reasons." The acreage and dollar effects modeled here may not match what the adopted rule produces. And the two alternatives' harvest ranges overlap across nearly their entire width, which means the economic difference the comparison reports may not be a difference the disclosed information can establish.

This is the agency tempering its own wildfire rationale: the practical treatment gain from new roads is limited by cost and budget, not only by the rule's prohibitions. The Economic Analysis says the other half — that "while wildfire risk reduction in IRAs is not quantified in this analysis, any reduction in risk provides meaningful benefits" — so the benefit is unquantified on one side while the conceded ignition cost sits at zero on the other. Both documents acknowledge that increased public road access could raise wildfire number and frequency; neither carries it into the accounting.

These are described qualitatively while timber, mineral and recreation figures get dollars — the same asymmetry that runs through the document. It is worth noting that the Cost Benefit Analysis does price one of them: forgone passive use values at $5.3 to $11.5 million annually. So the agency has demonstrated it can quantify existence value when it chooses to, which makes the qualitative treatment here a choice rather than a limitation.

The DEIS frames a trade-off between groups and never asks which groups. It does not identify whether the burdens fall on low-income, minority, Tribal or subsistence communities, and it runs no distributional screening of the divergent impacts its own socioeconomic discussion describes. The county frame needed to run that analysis is already built — 397 counties, 158 forest-dependent, 21 economically dependent — and populated with one variable. Two specific gaps follow. The DEIS identifies Southeast Alaska as where its own measure of community dependence binds hardest and carries that finding into no analysis of who bears the subsistence consequences. And the Regulatory Flexibility certification divides a loss the analysis itself calls local across every firm in the country, publishing two denominators three orders of magnitude apart — which is how a $9 million regional loss becomes an insignificant per-firm impact. One more thing sits in the county-revenue rationale: the DEIS ties timber harvest to county tax revenue, and the agency's own analysis states that "timber revenue after FY 1999 does not impact the calculation of payments to states and counties from the Secure Rural Schools Act."

The special treatment is tied to size and a two-decade rulemaking history, not chosen arbitrarily. That history also makes the Tongass the forest with the deepest litigation record on roadless management, so the extra attention carries correspondingly higher stakes if the analysis under it does not hold.

The detailed numbers below are not new work; they summarize a six-year-old document. The DEIS never discloses the vintage of the field data behind those findings — fish habitat surveys, forest-health monitoring, subsistence-harvest data — and never assesses whether the harvest-volume directives issued after 2020 change the alternatives comparison being carried forward. "Fully disclosed" is also not accurate across the board: the Tongass fire and fuels effects are not disclosed in the borrowed document. And the 2023 repeal now being reversed is presented as a sequence of events, with the substantive findings it rested on never stated.

This treats Alternatives 2 and 3 as producing identical Tongass outcomes, which sits awkwardly beside the rest of the document, where Alternative 3 retains prohibitions on about 13 million acres elsewhere. Whether Alternative 3's Tongass provisions are truly equivalent to the 2020 full exemption, or merely similar, is not explained. The equivalence matters because it is doing the same work as the equivalence finding used to eliminate the state-specific alternative — treating two regimes as interchangeable without demonstrating it.

The central estimate is small and the legal ceiling is not. Removal opens all 9.3 million acres to the maximum, and the document itself notes "considerable uncertainty concerning the future development of Southeast Alaska's road system" — so the modest projection is not a bound. The road figures underneath it inherit the national problems: a construction baseline that counts only permanent system roads built after 2013, and a road layer defined to exclude the temporary roads the agency's own assessment says harvest would mostly generate.

Volume barely moves and the pool of old growth legally available grows by roughly three-quarters. That gives managers far more locations to draw from, including remote ones, while the headline volume looks unchanged. There is also a live question about whether the volume figure can be relied on at all: the Tongass timber effects conclusion and the adequacy finding supporting it both depend on a market demand analysis the agency says is unavailable.

The document says fragmentation and wildlife effects would be greater than under no action — not zero, just not dramatically so by its own estimate. Appendix 4 concedes an old-growth change while the DEIS elsewhere finds harvest composition unaffected, and the Tongass analysis carried forward rests on one or the other. Which one is not stated.

A roughly 20 percent increase in average road density, framed as leaving watersheds mostly functioning, without identifying which subwatersheds fall in the remaining tenth. The fish conclusion is where this section is thinnest. It rests on a forest-wide average road density that cannot support a habitat conclusion, states the finding two ways at pages 243 and 318 that do not agree, materially alters the 2020 finding it relies on in a way that removes the grounds this rulemaking most affects, and omits fish passage entirely — road-stream crossings being the mechanism most directly at issue when road density rises. Appendix 4 disposes of Tongass fish habitat in a single tiered sentence resting on a forest plan the DEIS elsewhere calls indeterminate. Against that, the NMFS Biological Assessment finds every listed salmonid in the action area likely to be adversely affected.

The narrow footprint rests on which species count as present in ESA terms, and the supporting record is contested. The DEIS's account of consultation for the NMFS-managed critical habitats is internally contradictory, and the assessment it relies on concedes it cannot resolve the salmon population units at issue — its databases "do not differentiate between DPSes of vertebrate species listed under the ESA, which causes potential inflation of exposure." The Gulf of Alaska Chinook candidacy and not-warranted determination are both stated without a Federal Register citation. And the NMFS assessment describes the NEPA document it supports as an environmental assessment, which is a correction the record needs.

Removal concentrates a meaningfully larger share of harvest — nearly a fifth — into the most remote settings, even as the primitive-acreage share stays under 1 percent either way. Note the asymmetry in the road-closure point: roads are counted as closed for purposes of limiting recreation gain, and the same roads remain on the landscape for purposes of fragmentation, sediment and ignition.

This is the most concrete place-named detail anywhere in the DEIS — specific waterways the agency itself identifies as most exposed. Nothing comparable appears for scenery in any nationwide section, which is exactly the contrast the last point in this chapter takes up.

The DEIS records a Tribal preference for keeping the current rule, in its own voice, driven by competition for culturally important trees. That is a finding the document makes and then does not carry into the alternatives comparison. The surrounding consultation record is a tally of meetings with 30 of 64 requests unaccounted for, and the agency's own Tribal Summary records twelve Southeast Alaska Tribes proposing a homelands alternative that was eliminated for reasons not reaching what they proposed.

The DEIS takes two positions that cannot both stand: no evaluation is required at page 246, and the 2020 evaluations "remain valid and adequate" under Section 810 at page 319. It also states elsewhere that "ANILCA subsistence hearings were conducted for the DEIS and proposed rule, consistent with Section 810 in 2019" — hearings that predate this proposal's Notice of Intent by six years and cannot have been conducted for it. Two government letters on this record state the requirement reaches the Chugach as well as the Tongass, and the evaluation being incorporated covers the Tongass only, leaving roughly five million acres of Alaska roadless area with no subsistence evaluation at all. Food security appears in Volume I only in the summary of comments about it, while a Tribal community on this record states it qualifies as a federally designated food desert.

The wage figures barely move, and they are the only quantified economic effects the DEIS discloses for the Tongass — 2016 estimates, presented undated. Every circumstance the re-evaluation appendix itself identifies postdates them, including a principal regional harvester's exit from logging in 2021, which the DEIS discloses elsewhere without reconciling it against this baseline. Appendix 4 reconciles the harvest volume driving these figures and does not reconcile the figures. Against that, the subsistence economy the document calls foundational gets no dollar or resource-quantity estimate at all.

This is the contrast that makes the national analysis harder to defend. The agency performs granular quantitative work — acres, road miles, viewsheds by name, jobs, harvest by age class over a hundred-year horizon — for one forest, and describes effects "in a general, programmatic, and qualitative way" everywhere else, for a rescission reaching 44.7 million acres. The distinction is not that the Tongass was analyzed and the rest could not be: it is that a pre-existing dataset happened to be available to borrow here. NEPA's requirement for a detailed statement of reasonably foreseeable effects applies to effects within the agency's control everywhere, and the record asks for the same work nationally — a quantified, regionally stratified projection of road-mile construction and the sedimentation and fragmentation that follow, at the level of specificity already applied at pages 238 to 239.

These carry no stated vintage and no reconciliation against the technical corrections the DEIS says already changed the roadless inventory, and they cannot be reconciled with the 2020 analysis this rulemaking incorporates for Tongass effects. The national acreage is reconciled to federal ownership on one page; the Tongass figure is not on the next. Appendix 4 compounds it: it asserts that "roadless management on the Tongass National Forest has been consistent with the 2001 Roadless Rule for over a decade" while the same appendix recounts the 2020 full exemption, its 2023 repeal, and a November 2024 decision authorizing mine expansion inside an inventoried roadless area. It answers the Alaska Mental Health Trust land exchange by incorporating a determination the DEIS does not publish. And it cites, for its own re-evaluation authority, a regulation that does not contain the paragraphs cited — 7 CFR 1.9 is the Department's official-records and administrative-appeals provision, with no paragraph (e)(7) or (e)(8).

None of the reported figures can be checked. No coding protocol, sampling method or classification rule is disclosed — how a letter was distinguished from a submission, how petition signatures were attributed, what made a letter "generally opposed," how mixed and form letters were handled. The 625,000 figure refers to "the petitions," a term with no antecedent in the document. The two places the statistics appear do not agree with each other. And the screen that produced them, including the finding that more than 90 percent of letters were "brief sentiments" rather than substantive comment, applied a definition of "substantive" no longer in force — with the appendix footnote reciting that definition misquoting it. Correcting that is not cosmetic: the definition determines which comments the agency owes a response to, so re-screening the record under the standard actually in force is what the correction requires. Requests for a longer period, public meetings and cooperating-agency status appear in the summary with no showing they were granted, and extension requests filed by state Attorneys General are unanswered on the record. For comparison, the rule now being rescinded was adopted on a 60-day comment period with over 187 public meetings.

The range widened in response to comment and was then narrowed back. Alternatives 1 through 5 of the 2020 Alaska FEIS were eliminated because "these alternatives do not meet the purpose and need for the current proposed rulemaking, as they would not comply with the policy direction in Executive Order 14153" — a statement of policy non-conformance rather than a finding of technical or economic infeasibility or failure to meet the purpose and need, which is what the statute asks for. The DEIS also states the eliminated set two different ways at pages 37 and 315, and gives the rationale as a conclusion without the criterion applied. What remains is two bookends plus a boundary-modification option built on one mapping methodology. The specific eliminations each have a problem: the permanent-road-prohibition alternative was dropped for a reason two state wildlife agencies' own letters refute; the state-specific alternative on an equivalence finding Chapter 2 contradicts nine pages earlier; the twelve-Tribe homelands alternative for four reasons none of which reaches what was proposed. No intermediate, criteria-based alternative between full rescission and full retention was ever developed.

The statement omits forest-health and ecological-integrity values and is then used to eliminate every alternative that does not maximize flexibility — which is a purpose narrow enough to select its own answer. The causal claim under it is uncited and contradicted: no acreage, no fire or insect statistics, no comparison to non-roadless areas, against a document whose own Table 21 reports lower ignition density inside roadless areas and whose own record shows 1.8 million acres of fuels work accomplished under the rule said to have prevented it. The exception-use half of the claim rests on a 188-proposal compilation reporting no approval rate, and the agency's own Biological Assessments state there is "no national-level tracking of the review, approval, or implementation of 2001 Roadless Rule exceptions." The 2001 Rule's existing fire-risk exception already authorizes the treatment the purpose identifies, which means an alternative retaining the rule while streamlining that exception process was available and never analyzed.

The arithmetic leaves 30 requests with no stated disposition, in the only section designated as the Executive Order 13175 compliance record, in a sentence carrying no source. The record contains more specific accounts than the tally: a Tribal government recording that this rulemaking went unmentioned at its September 2025 quarterly meeting with four national forests; Tribes stating that email notice does not constitute consultation; a fourteen-business-day window against the agency's own longer minimum; Tribes who served as cooperating agencies in the earlier Alaska rulemakings stating their input was "repeatedly ignored or marginalized." The agency's own Tribal Summary records the central concern as proceeding "without adequate government-to-government" consultation, and records that Tribes "requested consultation based on mutual concurrence." A count of meetings does not show what was raised or what changed, and disclosing the disposition of a concern requires no protected information.

Read within this document, "the DEIS and proposed rule" means this one — whose Notice of Intent published in August 2025. Hearings in 2019 cannot have been conducted for it, and the sentence needs correction. Beyond the date, it asserts compliance with a statute carrying specific mandatory content and states none of it: no hearing locations, no communities notified, no resources evaluated, no findings. The DEIS also takes two incompatible positions on whether Section 810 applies at all. And the evaluation being relied on covers the Tongass only, while two government letters on this record state the requirement reaches the Chugach — leaving that forest with no subsistence evaluation for this action.

This is tiering, and the question is whether the reuse was checked or asserted. The DEIS does not disclose the vintage of the field data carried forward — fish habitat, forest-health monitoring, subsistence harvest — even as it separately discloses recent defoliator outbreaks causing "mortality and decline of forested stands" before concluding those changed circumstances "have been found to result in no substantial changes to the analysis and the analysis remains adequate." That conclusion is asserted, and the DEIS's own insect-and-disease baseline contains no Alaska data from which to reach it. Post-2020 timber-volume directives are never assessed against the alternatives comparison being reused. The Tongass timber conclusion depends on a demand analysis the agency says is unavailable. The appendix's continuity claim is uncited and sits beside its own account of the 2020 exemption, the 2023 repeal and a 2024 mine-expansion authorization. It answers the Alaska Mental Health Trust land exchange by incorporating a determination the DEIS does not publish. And it cites for its authority a regulation whose paragraphs do not exist as cited.

Three gaps. Footnote 74 discloses pending ANCSA legislation affecting up to 115,000 acres, roughly half inside the roadless inventory — a foreseeable overlapping change disclosed on the same page Appendix 4 concludes circumstances have not changed, and folded into no cumulative analysis. The Greens Creek North Extension is omitted from Table 4, the document's own list of ongoing and reasonably foreseeable actions. And the Alternative 3 boundary-modification mechanism carries no acreage cap, no aggregate reporting requirement and no cross-reference obligating Endangered Species Act consultation, which is how many individually modest changes accumulate without a point at which review is triggered. Two entries in the table also point somewhere other than they claim: the minerals analysis designated as documenting compliance with the energy executive order projects no net change in production, and the treaty-rights section designated as the compliance showing names no treaty.

The document reports consultation status two ways on adjacent pages — anticipated at 167, ongoing at 168 — publishes no Biological Opinion, gives Alternative 3 no salmonid determinations, and never states when consultation would conclude relative to a decision. A likely-to-adversely-affect finding triggers formal consultation, and a preliminary screening is not a substitute for a completed Opinion. The single reliability statement offered is contradicted by the assessments, which record error in both directions: occurrence data "opportunistic" and lacking negative records, databases inflating exposure by not distinguishing population segments, exposure assumed for entire taxonomic groups. The exposure screen has no road-effect zone despite the assessment's own finding that roads depress bird abundance up to a kilometre out. And the DEIS states two incompatible things about whether plan habitat direction entered the analysis — reviewed at page 169, not reflected at page 291, for the two species the biodiversity discussion uses most.

No screening identifies which communities bear the divergent impacts the document itself describes — and the frame to run it is already built and left unpopulated: 397 counties, 158 forest-dependent, 21 economically dependent, classified by population and rurality alone. The compliance catalogue identifies no authority under which distribution of effects is analyzed at all. Table 47's results are never carried into the alternative-by-alternative discussion. The DEIS identifies Southeast Alaska as where community dependence binds hardest and analyzes nobody's subsistence consequences there. And the small-entity certification divides a loss the analysis calls local across every firm in the country. Whether the prior executive order on environmental justice stands or not, the hard-look duty to analyze effects the agency's own findings put at issue does not depend on it.

The sequence is in the document's own chronology, in three consecutive sentences: announcement, then notification of Tribes in July, then the comment period in August. A public commitment to an outcome made before the process meant to weigh it is the kind of record used to argue predetermination, and it bears directly on whether the no-action alternative received genuine consideration. The document does not address the sequence anywhere. It is also the point at which the purpose-and-need gate becomes consequential rather than technical: alternatives were eliminated for failing a purpose announced before the analysis existed.

The reference apparatus does not support the document as published, in three distinct ways. Some figures are attributed to sources that say something different: the carbon passage misstates its only cited source in two respects, and the wilderness acreage figures at page 195 carry no citation at all while the substitution finding rests on them. Some in-text citations resolve to nothing — at least 66 distinct author-year citations appear in the text with no matching entry in the 399-entry reference list, and the effects analysis for Alternative 2 rests at two points on one of them. And a large share of what is listed cannot be obtained: 157 of 399 entries — 39 percent — are paywalled, carry no locator of any kind, or point to URLs that fail. Four of those carry the socioeconomic analysis. Beyond that, many of the document's load-bearing statistics carry no citation at all: the acreage-scope figures in the Introduction, the ignition-density comparisons, the road-containment percentages, the recreation baseline, the mineral and energy dollar figures. This is a different kind of problem from disagreement over judgment: it goes to whether the record can be traced and checked, which is what a hard look is supposed to make possible.