NEPA Decarbonization Technology Analysis: Deliverable 4
Review Timelines for Categorical Exclusions, Environmental Assessments, and Environmental Impact Statements
Executive summary
NoteKey findings
- Review duration climbs from days to months to years. Median review duration is 20 days for CE, 117 days (~4 months) for EA, and 1,021 days (~2.8 years) for EIS — the expected complexity ladder, now quantified across the full database (complete timelines; Figure 5).
- Timeline completeness varies by process, and the gaps are structural. Complete timelines (both an initiation and a decision date, in valid order at day- or month-precision) reach 50% for CE, 56% for EA, and 32% for EIS (Figure 3). CE and EA are initiation-limited (no Notice-of-Intent requirement, often no recorded start); EIS is decision-limited — Records of Decision are usually separate documents outside the corpus, so EIS decision coverage is just 42.3%. LLM adjudication of the ambiguous cases meaningfully lifted complete coverage for all three processes.
- Post-FRA, EIS documents got shorter while EAs held steady. Across all 3,683 EA/EIS projects (all energy types, regulatory pages from the full text), mean EIS length fell from 318 to 269 pages after the Fiscal Responsibility Act, while EA length stayed near 33–39 pages — though the post-FRA window is still small (Section 3).
- The longest reviews are prime case-study candidates. Genuinely protracted, evidence-verified reviews — SunZia Southwest Transmission (14.6 yr) and Cushman Hydroelectric (18.1 yr, hand-verified) — are surfaced for CATF investigation and separated from extraction errors (an “initiation” anchored to a historical citation), which are excluded; two other long candidates (Energia Sierra Juarez, Grain Belt Express) remain pending date verification (Section 2.4).
- Process change, not office-by-office learning, drove the CE speed-up. CE reviews got much faster across the 2010s, but the decline hit every field office at once regardless of caseload: within any year an office’s accumulated experience is uncorrelated with its speed, and low-volume offices converged to high-volume speed without doing comparable reps. A process change reaching all offices at once — batch processing, procedural streamlining — not office-by-office learning, is the likely driver; a wholly separate DOE register-office corpus reproduces the null independently (Section 5).
- CEQ rule changes don’t produce clean before/after breaks. Segmenting reviews by the CEQ rule in effect at decision (1978 rules / 2020 rewrite / 2022 Phase 1 / the provisional post-2024 window) mostly re-slices the same secular trends visible year by year; the post-2024 samples are thin, coverage-confounded, and read as provisional (Section 4).
- Geothermal NEPA is two different worlds, and office-level comparison is not feasible. The
Geothermaltag pools BLM western resource reviews (NV/CA/UT) with a much larger body of DOE grant-era research and heat-pump CEs filed nationwide; only a handful of reviews sit in any one BLM field office, so the geothermal analysis is a descriptive office inventory plus a state-cohort map, not an office ranking (Section 6).
This report delivers:
Timelines for categorical exclusions, environmental assessments, and environmental impact statements — including segmentation by regulatory period (pre/post-FRA), energy category, and decision year — plus a timeline-outlier deliverable for case-study investigation and a document-length / FRA page-limit analysis refreshed on Phase 2 decision dates.
Methodology
This deliverable extends the Phase 1 timeline analysis — which examined review timelines for decarb projects only — to the entire NEPATEC 2.0 database, which includes decarb, fossil, and other project types. It does so with a rebuilt extraction pipeline that adds authoritative agency dates and two machine-learning models.
How timeline dates are extracted
This section improves on the Phase 1 timeline method. For every project we identify two milestones — the initiation (when the review begins: an application received, a Notice of Intent, or scoping) and the decision (when it concludes: a signature, a Record of Decision, or a Finding of No Significant Impact). The pipeline finds these in five high-level steps:
1 · Extract candidate dates
Pull every date from every NEPA document in NEPATEC 2.0, capturing the surrounding sentence, document location, and metadata so each date's meaning is clear
↓
2 · Classify each date's role
A machine-learning classifier (SetFit) labels every candidate as an initiation, a decision, or neither
↓
3 · Rank the candidates
A second ranking model (LightGBM) orders the initiation and decision candidates by how likely each is the operative date for the project
↓
4 · Anchor on authoritative dates
Where the BLM or DOE NEPA register records an official project-start or decision date, that authoritative date is used in place of a parsed one
↓
5 · Resolve the hardest cases with an LLM
For projects where the operative date is still ambiguous, a large language model (Claude Haiku) selects the best remaining candidate
↓
Complete timeline
A project's timeline is complete when both an initiation and a decision date are found
How Phase 2 analysis improves upon Phase 1
- Full-database coverage — Phase 1 covered decarbonization technology projects only (20,725); Phase 2 covers the entire NEPATEC 2.0 database (61,881 projects).
- API registry dates — Phase 2 incorporates authoritative dates from BLM and DOE NEPA registers as an authoritative source of official project-start and decision dates. We pull in roughly 37,000 project-matched register dates, which is particularly helpful for CE and EA reviews.
- Cleaner retrieval and granularity handling — the pipeline uses year-month decision dates (e.g. a Final-EIS cover month) where no better dates are available. (A section-aware retrieval tier retrieves decision/initiation sections directly; a 2026-07-23 re-run of the section index and retrieval chain activated it, adding 21,289 section-tier candidates to the current dataset.)
- Improved classifier and a ranking model — Phase 1 used a single date classifier (
DeBERTa); Phase 2 uses an improved classifier (SetFit) and adds a dedicated ranking model (LightGBM) that decides which candidate is the operative initiation/decision, sharply improving which candidates get sent to the LLM.
Date provenance
Every extracted date traces to one of two sources: the BLM/DOE NEPA registers (official agency dates pulled via the registers’ metadata API) or document text (parsed from the NEPA documents themselves, including LLM-adjudicated picks and Final-EIS publication dates). Figure 1 shows the split by process and endpoint — a transparency check on where the timeline rests. Registers anchor a meaningful share of CE dates; EA and EIS rely overwhelmingly on document text.
Timeline analysis
Timeline completeness
Figure 2 shows the share of projects with complete timelines by review process; Figure 3 breaks down which component is missing. The patterns differ by process:
- CE (50% complete): decision coverage is high (90.3%) but initiation coverage is low (59.7%) — most CEs are single-signature determinations with no recorded start.
- EA (56% complete): the strongest of the three, but still initiation-limited — EAs have no Notice-of-Intent requirement and often skip scoping.
- EIS (32% complete): initiations are usually documented (an NOI is required), but decision coverage is only 42.3% because Records of Decision are frequently separate documents not in the corpus.
Comparison with Phase 1. Phase 1 measured completeness for decarb projects only and reported CE ~30% · EA ~62% · EIS ~48%. On the full database, Phase 2 reaches CE 50% · EA 56% · EIS 32%. CE is higher — the much larger universe is dominated by routine, register-dated determinations — but EA and EIS land below their Phase 1 decarb-only rates, for two reasons this analysis surfaced:
- The aggregate is diluted by non-decarb projects. Phase 1’s denominator was decarb energy; Phase 2’s is everything. Decarbonization EIS alone completes at 39.6% (closer to Phase 1’s 48%) — it is the heterogeneous “Other” bucket (27.6% complete) that pulls the EIS aggregate down to 32%.
- The remaining gaps are structural, not fixable by better parsing. EIS completeness is capped by missing Records of Decision (a separate document, often outside the corpus — decision coverage only 42.3%); CE and EA completeness is capped by missing initiations, because neither requires a Notice of Intent and a start date is frequently never recorded. A minority of EIS “projects” are also comment letters or document fragments with no milestone date to extract.
Coverage also varies by energy category (Figure 4): decarbonization EIS projects complete at 39.6%, well above the EIS aggregate of 32%, because decarb EISs are better documented than the heterogeneous “Other” bucket that drags the aggregate down.
Duration analysis
Durations use all complete timelines (both dates present, including those resolved with a proxy date such as a Final-EIS publication date in place of a missing Record of Decision). Month-only dates are imputed to the mid-month 15th.
On that basis, durations track the expected complexity ladder from CE to EA to EIS — a median of about 20 days for CE, 117 days (~4 months) for EA, and 1,021 days (~2.8 years) for EIS (n = 27,275 / 1,736 / 1,329). Figure 5 summarises the spread; Figure 7 shows the full distributions; Figure 8 plots individual project spans.
Duration by review type
Splitting duration by energy category (Figure 6) exposes an artifact: fossil EAs run far shorter than decarbonization or “other” EAs, a median of ~40 days, vs ~310 days for Decarb and ~209 days for Other. This is not because fossil EAs are genuinely faster: 84% of them take their start date from the BLM register rather than the document. This is a broad convention across routine BLM oil-and-gas EAs — drilling permits, lease sales, and well-abandonments alike (well-abandonment is 53% of complete fossil EAs, not the whole story) — where the recorded “project start” is a late administrative entry made within days of the decision. Because fossil is the largest EA group (758 of the 1,736 complete EAs), those artificially short spans pull the overall EA median down. Restricting fossil EAs to the document-anchored starts lifts their median to ~5 months — much closer to the Other and Decarb EAs and the more defensible review-length estimate — but that view survives for only a fraction of them.1
The full duration distributions (Figure 7) show the same complexity ladder as a shape, not just a median: CE durations pile up near zero (most are days-to-weeks determinations), EA spreads out to a year or two, and EIS is broad and heavy-tailed, with a meaningful mass of reviews running 3–10+ years. That long EIS tail is real (large, contested reviews spanning multiple administrations) and is examined directly in the outlier case studies below.
Figure 8 plots each complete review as a horizontal bar from its initiation to its decision date, sorted by length within each process so the spread in durations is visible one project at a time rather than only in aggregate. CE spans collapse to short marks, EA bars range from a few months to a couple of years, and the EIS panel is dominated by multi-year bars — the same CE-to-EA-to-EIS complexity ladder seen in the histograms, now resolved to individual reviews.
Duration for solar projects
Phase 1 delivered a solar-only version of the duration analysis; this subsection replicates it on the Phase 2 timeline. For projects tagged Renewable Energy Production – Solar, complete timelines show the same complexity ladder — a median of about 0.7 months for CE (n = 811), ~12 months for EA (n = 58), and ~22 months (1.9 years) for EIS (n = 71). Against the all-decarbonization reference (Figure 9, dashed ticks), solar EIS reviews conclude about a year faster — ~22 vs ~34 months — while solar EAs run on par with decarbonization EAs generally (~12 vs ~11 months) and CE determinations are indistinguishable (~1 month in both). This replicates the Phase 1 finding for EIS — Phase 1’s solar figure reported ~22 months (n = 66) against Phase 2’s ~22 here (n = 71) — but softens Phase 1’s EA contrast (~13 months solar there) to parity. Two caveats: the solar EA and EIS cells are small (n = 58 and 71), and solar CEs are register-heavy like CEs generally, so the register-anchoring caveats above apply equally here.
Duration by technology
Within each energy group, review length also varies by the specific technology — the cleaned tech_group tag that defines the decarbonization-vs-fossil split. Figure 10 and Figure 11 break duration down by technology for EA and EIS reviews only: CE determinations are uniformly ~1 month across every technology and add no contrast, and the most CE-dominated technologies (biomass and CCS on the decarbonization side, rural energy on the fossil side) have too few substantive reviews to chart at all — itself telling, since their NEPA footprint is almost entirely categorical exclusions.
Among decarbonization technologies, the ordering tracks physical footprint and contestedness: hydropower (~37 months / 3.1 years) and transmission (~28 months) run longest — large, linear, multi-jurisdiction projects — followed by nuclear (~20), wind (~18), and solar (~19) in the middle band, with geothermal near ~14 months, “other clean” near ~13 months, and energy storage the quickest at ~10 months. Among fossil technologies, coal (~36 months / 3 years) and offshore oil & gas (~26 months) dominate the long end, while land-based oil & gas is an order of magnitude faster (~2 months, n = 849) — the same register-anchored well-abandonment and drilling-permit flood that compresses the fossil-EA median noted earlier. The two panels are separate figures on their own horizontal scales; the technologies with the longest tails (hydropower, transmission, coal) are the same large-infrastructure reviews that populate the outlier case studies below.
Projects by decision year
Figure 12 shows project volume by decision year (all projects), faceted by review process. CE volumes ramp from the late-2000s (ARRA-era funding) with a second rise into the 2020s (BIL/IRA); EIS decisions lag CE and EA by several years, reflecting longer review cycles. Figure 13 restricts to the Department of Energy — the single largest agency in the dataset — where the ARRA/BIL/IRA grant-and-loan cycles are even more pronounced.
Timeline outliers: case-study candidates
Per the deliverable request, timeline outliers are surfaced for CATF staff to investigate via case studies (including whether NEPA itself was a cause of delay). Outliers are init→decision spans longer than 5,000 days (~13.7 years), produced reproducibly by code/deliverable04/10_outliers.R.
The distinction that matters is real review versus extraction error. A multi-year span can be a genuinely protracted review or an artifact where the “initiation” was anchored to a historical citation (a facility’s construction date, a reserve’s founding, a prior plan, a regulatory-compliance milestone) rather than the NEPA start. The two are separable only by reading the evidence text, which the outlier table preserves.
Analysis caveats
- EIS outliers are mostly real. EIS median duration is already ~3 years; a 14–17-year EIS sits in the legitimate tail (transmission, restoration, resource-management plans genuinely run that long).
- CE outliers are mostly errors — CEs are fast determinations (median 20 days), so a multi-year CE almost always reflects a mis-extracted initiation. CE outliers are therefore excluded from the client list.
- Initiation dates before ~1990 are a strong error signal. Of the four EIS outliers with 1980s-or-earlier initiations, three are confirmed errors (the “initiation” is a park founding, a prior lock’s construction-completion date, or a RCRA compliance milestone).
The decarb EA/EIS candidates below are the highest-value case studies. SunZia (ROW application Sep 2008 to Record of Decision Apr 2023) is the clearest long-permitting case — its dates, along with Cushman Hydroelectric’s 18.1-year FERC relicensing span, were verified by hand against the documents’ own procedural text. Energia Sierra Juarez and Grain Belt Express look similar but remain pending verification: the extracted Energia span bridges the original EIS and its 2023 supplement, and Grain Belt’s endpoints are a state CPCN filing and an FEIS publication date rather than federal NEPA milestones. Confirmed errors are excluded, for example the Palisades Nuclear Restart “19-year” span, which glues a 2005 license-renewal application to a 2024 restart Record of Decision (two different NEPA actions).
| Project | Process | Energy | Lead agency | Years | Initiation | Decision | project_id |
|---|---|---|---|---|---|---|---|
| Interstate Highway 35 Capital Express Central Project From United States Highway 290 East to United States Highway 290 West/State Highway 71 Travis County, Texas | EIS | Other | 29.8 | 1993-03-15 | 2023-01-05 | d4cb63aaadf006f332c6015042498d08 | |
| Closure of Nonradioactive Dangerous Waste Landfill (NRDWL) and Solid Waste Landfill (SWL) | EA | Other | Department of Energy | 24.5 | 1985-11-15 | 2010-05-01 | da4d59f465693f67613fb04a3be56a30 |
| Lower Cache Creek, Yolo County, Woodland and Vicinity, California Flood Risk Management | EIS | Other | 23.6 | 1996-05-06 | 2019-12-27 | 7da2bac1ab510ecd349877382a4569ae | |
| 2021 Land Management Plan Helena – Lewis and Clark National Forest | EIS | Other | Forest Service | 20.0 | 2001-10-26 | 2021-10-15 | ca3d703bbef5b90c42e4900a179cf696 |
| Rawlins Resource Management Plan | EIS | Other | 17.7 | 2002-02-25 | 2019-11-13 | 252300407e5f5b7ae87c30984faa46da | |
| Construction and Operation of The Molecular Foundry | EA | Other | Department of Energy | 16.3 | 1986-12-01 | 2003-03-06 | 8237e9b86021dba1b116398c74644f23 |
| West Mojave Route Network Project | EIS | Other | Bureau of Land Management | 16.3 | 2003-06-07 | 2019-10-02 | 48e2e33fc1555b1ef01a7c5bbc3e6c4f |
| West Mojave Route Network Project | EIS | Other | Bureau of Land Management | 16.3 | 2003-06-07 | 2019-10-02 | 34a8fa6c4bb7081831ba3e843af15db3 |
| Clearwater National Forest Travel Planning | EIS | Other | Forest Service | 16.1 | 2007-11-28 | 2024-01-15 | 9757c3901c0240237625ff424be4cdac |
| Clearwater National Forest Travel Planning | EIS | Other | Forest Service | 15.7 | 2007-11-28 | 2023-08-15 | 572a8342af3551b4c1dd71d910ea5645 |
| Pima County Multi-Species Conservation Plan | EIS | Other | 15.4 | 2000-09-07 | 2016-01-15 | 95c319c8d504a7e1436abf6de96f0a7c | |
| Placer County Conservation Program | EIS | Decarb | United States Fish and Wildlife Service | 15.2 | 2005-03-15 | 2020-05-16 | 9e65f5e1614fae11f2d17be8d96fff63 |
| Proposed Roan Plateau Planning Area Resource Management Plan Amendment | EIS | Fossil | Bureau of Land Management | 14.8 | 2000-11-16 | 2015-09-22 | 90edf860d9396c4d21b0c28165054c4f |
| Malheur National Forest Site-Specific Invasive Plants Treatment Project | EIS | Other | Forest Service | 14.7 | 2000-06-26 | 2015-03-15 | 15b199816215e2ab7b2a01534eaaaeca |
| Sterling Highway Milepost 45-60 Project | EIS | Other | Federal Highway Administration | 14.6 | 2003-07-15 | 2018-03-04 | 4a660923bc39c4d901fa0e43743556c3 |
| Resource Management Plan for the Ring of Fire Planning Area | EIS | Other | Bureau of Land Management | 14.6 | 2009-06-26 | 2024-02-05 | f0519ad10cf558982119d91fc5070055 |
| Las Vegas Disposal Boundary | EIS | Fossil | Bureau of Land Management | 14.6 | 1990-05-08 | 2004-12-17 | 1ae98cccdd70178ad07bb9139acff4f1 |
| Energia Sierra Juarez U.S. Transmission Line Project | EIS | Decarb | Department of Energy | 14.6 | 2009-02-25 | 2023-10-06 | 3b7e393de77613cc10641d9a352796ae |
Document length & FRA page limits
This section builds directly on Phase 1 Deliverable 5 (“Document Length Over Time and Fiscal Responsibility Act Impact”), which analyzed document length for decarb EA/EIS projects. Here we reproduce that analysis on the entire Phase 2 EA/EIS corpus, all energy types, refreshed on Phase 2 LLM-adjudicated decision dates (inclusion requires only a decision date — the time axis and FRA classification depend only on it). Each subsection below links to its Phase 1 counterpart for comparison.
NoteMethod & scope
- Regulatory pages = body word count ÷ 500, excluding embedded appendices and low-content pages, per 40 C.F.R. § 1508.1(bb) — the measure that matches the FRA limits (EA ≤ 75 pages; EIS ≤ 150, or 300 if extraordinarily complex).2
- These counts are computed from the actual page text by
fra/01_extract_pages.py(a DuckDB pass overphase2/data/processed/{ea,eis}/pages.parquet— 6.1M EIS pages — detecting the appendix boundary and low-content pages), producing regulatory pages for 5,032 EA/EIS projects (2,765 EA, 2,267 EIS) across all energy types — not just decarbonization. - The pre/post-FRA analysis below uses the 3,683 projects that also have a decision date (EA 2,239 / EIS 1,444).
- FRA date = enactment (June 3, 2023), matching Phase 1 D5 and the timeline-duration period split above.
Document length over time
Figure 14 tracks regulatory page length by decision date across all EA/EIS. Two patterns stand out. EA length sits low and essentially flat across the decade — the rolling mean hugs the low tens of pages, with only occasional long outliers — consistent with EAs being short by design. EIS length is far higher and far more dispersed: individual EISs run from tens of pages to well over a thousand, and the rolling mean hovers in the low-to-mid hundreds with no strong secular trend before the FRA. The post-FRA window (right of the red line) is denser — more recent decisions — and is where the EIS rolling mean steps down.
Compared with the decarb-only Phase 1 version, two differences follow from the wider Phase 2 scope. First, the all-energy EA cloud sits well below Phase 1’s: Phase 1’s decarb EAs averaged ~62 regulatory pages, whereas the full corpus averages ~33, because the large pool of short fossil and well-abandonment EAs (median ~7 pages) pulls the all-energy mean down. Second, the EIS picture is the same in shape — high, heavy-tailed, and dominated by project-to-project variation rather than a clear time trend — confirming Phase 1’s caution that average trends obscure substantial spread, now borne out on a roughly 5× larger sample.
Pre vs post-FRA
Across all EA/EIS, EIS regulatory length fell after FRA (mean 318 → 269 pages; median 261 → 212), while EA length was essentially flat (mean 33 → 39). The post-FRA window is short, so these are provisional (n = 218 EA, 97 EIS). The direction matches the decarb-only Phase 1 result closely (Phase 1 D5 — Pre vs Post FRA and its distribution view): Phase 1’s EISs fell from a mean of 368 → 270 regulatory pages, and the all-energy corpus here falls 318 → 269 — landing at essentially the same ~270-page post-FRA EIS mean despite a very different sample. EA stayed flat in both (Phase 1 62 → 57; here 33 → 39), reflecting that EAs were already well under the 75-page limit and had little to compress.
Breaking the same comparison out by energy category (Figure 16) shows the post-FRA EIS decline is broad — it appears in decarbonization and fossil EIS, not just one segment — while EA length is essentially flat across all categories. That consistency makes the EIS drop more credible as a genuine post-FRA effect rather than a composition artifact.
The full distributions (Figure 17) confirm the shift is not an outlier effect: the EIS distribution moves down post-FRA (lower median, thinner upper tail) while the EA distribution barely moves. Splitting that distribution by energy category — shown separately for decarbonization (Figure 18), fossil (Figure 19), and other (Figure 20) projects so each panel is legible — shows the EIS contraction holds in both decarbonization and fossil EIS, while EA length stays flat across all three.
Descriptive statistics
Table 2 reports the full descriptive statistics — mean, median, standard deviation, and interquartile range — by process and FRA period across all energy types (the all-energy counterpart to Phase 1 D5’s descriptive-statistics table).
| FRA Period | N |
Regulatory Pages (body word count ÷ 500)
|
||||
|---|---|---|---|---|---|---|
| Mean | Median | SD | P25 | P75 | ||
| EA | ||||||
| Pre-FRA | 2,021 | 33 | 19 | 44 | 7 | 41 |
| Post-FRA | 218 | 39 | 24 | 42 | 7 | 63 |
| EIS | ||||||
| Pre-FRA | 1,347 | 318 | 261 | 268 | 128 | 440 |
| Post-FRA | 97 | 269 | 212 | 250 | 130 | 394 |
| Post-FRA projects are those with a decision date on or after June 3, 2023. Regulatory pages exclude embedded appendix pages and low-content pages (maps, figures, blanks); documents whose filename already omits appendices use their page count directly, and documents without extractable text are excluded. | ||||||
FRA page-limit compliance
The Fiscal Responsibility Act sets presumptive page limits — 75 pages for an EA and 150 pages for an EIS (300 if the agency documents extraordinary complexity). Figure 21 scores the post-FRA EA/EIS projects with a decision date against those limits (regulatory pages).
- EAs comply at a high rate. 83% of post-FRA EAs (n = 218) fall within the 75-page limit; the 17% that exceed it generally do so by a small margin.
- EISs are more mixed, but most are within reach of the complexity ceiling. Only 32% of post-FRA EISs (n = 97) land within the baseline 150-page limit, but a further 33% sit between 150 and 300 pages — i.e. within the extraordinary-complexity ceiling — so 65% are within the 300-page threshold overall. The remaining 35% exceed even 300 pages, consistent with the long right tail of complex transmission, restoration, and resource-management EISs seen in the duration analysis.
Two caveats temper this. First, the post-FRA window is short (218 EA / 97 EIS projects with both a regulatory page count and a decision date), so these rates are provisional and will firm up as more post-FRA decisions land. Second, the FRA limits are presumptive, not absolute — agencies may exceed them with senior-official approval or a documented extraordinary-complexity finding — so “exceeds limit” flags a document for review, not an automatic compliance failure.
Review timelines by CEQ regulatory regime
NEPA’s procedural rules were rewritten three times in four years. The natural regulatory regimes — keyed to each rule’s effective date — are summarized below. (The FRA, June 3, 2023, is a statutory amendment to NEPA, not a CEQ rule; it is already the pre/post-FRA breakpoint used in the duration and page-length analyses.)
| Regime | Effective | What changed |
|---|---|---|
| 1978 regulations | 1978–2020 | Original CEQ regulations; essentially unchanged for ~42 years. |
| 2020 Final Rule (Trump) | Sep 14, 2020 | Major rewrite: presumptive 2-yr EIS / 1-yr EA limits, page limits, narrowed effects/cumulative-impacts definitions. |
| 2022 Phase 1 Rule (Biden) | May 20, 2022 | Restored 1978 purpose-and-need and cumulative/indirect effects. |
| 2024 Phase 2 Rule | Jul 1, 2024 | Implemented the FRA's statutory NEPA amendments (codified page/time limits, CE sharing, lead-agency roles). |
| 2025 rescission | Apr 11, 2025 | CEQ interim final rule (90 FR 10610) removed 40 C.F.R. §§ 1500–1508; agencies revert to their own NEPA procedures. |
Figure 22 lays out that sequence on a single axis — three regulatory rewrites plus the 2025 rescission inside five years, with the FRA (a statutory change, not a CEQ rule) marked separately.
Everything in this section rests on the same complete-timeline frame as the rest of the deliverable — the 27,275 CE, 1,736 EA, and 1,329 EIS reviews that carry both an initiation and a decision date in day- or month-resolution and valid order (50%, 56%, and 32% of all CE / EA / EIS projects; identical to the complete-timeline definition in Figure 2, so these are the same reviews the headline completeness figures count). The 1978 bucket is by far the largest not because coverage is fuller there but because it absorbs every decision from the start of the corpus through September 13, 2020 — roughly four decades, with CE activity concentrated in the 2010s. Of the 32,463 CEs decided before the 2020 rule, 17,223 (53%) have both dates — CE decision coverage runs 90% corpus-wide (registers plus determination-form dates) against 60% initiation coverage, so completeness is roughly half, and half of a very long, activity-dense window is still ~17,000 reviews.
This segmentation is deliberately independent of the two breakpoint sets used above: it complements the reg_period funding eras (ARRA / BIL / IRA) and the FRA statutory split by cutting on the CEQ regulatory rule effective dates instead. Because a review is a span, we assign each one to the rule in effect at its decision date — so every figure reads as “reviews decided under a regime,” not “reviews governed by” it. The 2024 Phase 2 Rule and its 2025 rescission are folded into one provisional 2024+ bucket: the rescission window falls far below the n = 30 display floor for every process, the standalone Phase 2 window falls below it for CE, and a single merged bucket keeps the provisional post-2024 period readable rather than a mix of displayed and suppressed cells (counts in the caveats below).
The regime medians move very little for the two smaller-review processes, and what movement there is tracks sample composition rather than any rule:
- CE reviews stay near-instant under every pre-2024 regime — a median of 0.6, 0.7, and 0.8 months under the 1978, 2020, and 2022 regimes. The
2024+bucket reads 37 months, but on only 46 reviews — an artifact of which CE reviews happen to carry a complete, recently-dated timeline, not a regime effect. - EA medians sit in a narrow 2.2–4-month band across the three pre-2024 regimes, rising to 8.9 months in the thin
2024+bucket (n = 55). - EIS medians are the longest and the noisiest — 34 months (1978), 29 (2020), 39 (2022), 27 (
2024+, n = 50). The apparent dip under the 2020 regime is a pooling artifact, not evidence that the Trump rule sped EISs up. - The 1978 bar pools ~42 years and understates recent durations. Restricting the 1978 rule to its final stretch (2015–2020 decisions) lifts the EIS median from 34 to 40 months — so the pooled 1978 comparator in Figure 23 is the wrong reference for a before/after-rule read. Use the year-by-year trend in Figure 24 for that intuition instead.
Figure 24 places the CEQ rule dates on the annual median-duration line, alongside the FRA statutory marker. Because durations have drifted secularly (CE especially), the year-by-year view is the honest way to see whether anything changed around a rule date — the regime bars above cannot separate a rule effect from the underlying trend.
Analysis caveats
- Decided-under, not governed-by. A multi-year EIS decided in 2022 was scoped and largely conducted under earlier rules; the decision-date assignment is a labeling convenience, not a claim about which regulations shaped the review.
- The post-2024 samples are thin and confounded. Only 46 CE, 55 EA, and 50 EIS reviews fall in the
2024+bucket, and they are entangled with the same coverage ramp that limits the post-FRA split — recently-decided reviews are over-represented by whichever timelines happen to be complete. - The 2025 rescission window is folded into
2024+. The 2025 rescission window (18 CE, 19 EA, 9 EIS) falls far below the n = 30 display floor for every process; the standalone 2024 Phase 2 window (28 CE, 36 EA, 41 EIS) falls below it for CE, though its EA and EIS counts clear the floor. They are analyzed only as the merged2024+bucket in Figure 23 so the provisional post-2024 period reads as one consistent cell rather than a mix of displayed and suppressed windows. - For a before/after-rule read, use Figure 24, not Figure 23. The pooled 1978 regime bar spans four decades of secular decline; the year-by-year trend is the only view that can separate a rule date from the underlying drift.
- Associational, not causal. Nothing here isolates a regulatory effect from composition, agency mix, or calendar time.
Field-office experience: BLM and DOE
The deliverable asks whether lead-agency offices process NEPA reviews faster as they accumulate experience, a “learning-curve” effect. We test it on two structurally independent corpora: BLM projects map to the field office that handled them (from case metadata), while DOE projects map to the administering or grant office that processed them (through the CX register). Both arms measure the same thing — whether an office’s review duration falls as its own cumulative caseload grows — on document-anchored, complete CE timelines over each agency’s full window. All conclusions here are associational, not causal.
The two arms rest on very different sample sizes (Figure 25). Of 26,016 BLM-led projects, 16,249 parse to a field office and only 31 clear the ≥ 30 document-anchored-CE bar the regression requires; of 32,305 DOE-led projects, 11,707 link to a register office and only 12 clear it. A handful of high-volume offices carries each regression, so the design leans on the calendar control rather than any single office’s trajectory.
Process change, not office experience, drove the CE speed-up. CE reviews got much faster across the window, but the decline hit every office at once rather than the busiest offices pulling ahead. Within any given decision year, an office’s accumulated caseload is essentially uncorrelated with its speed — the per-year Spearman correlation between cumulative reviews and duration ranges -0.30 to +0.22, centered on zero — and the quieter half of offices converges to the same speed as the busier half without ever accumulating a comparable caseload (Figure 26). That is the signature of a system-wide change, not office-by-office learning.
The office fixed-effects regression confirms it formally. The raw gradient looks like a real speed-up, -6.8% per doubling of an office’s caseload (95% CI -13.2% to +0.1%), but that is the corpus-wide decline in disguise. Once each decision year is given its own baseline, the apparent speed-up is gone: for BLM the estimate in fact reverses sign to +11.0% (slower, not faster, per doubling; 95% CI -2.9% to +26.8%, not statistically distinguishable from zero). The DOE arm flattens to genuinely near-zero on its own full window, an apparent -6.2% raw gradient that the calendar control erases to +3.2% (95% CI -4.7% to +11.7%) across 12 administering offices in a wholly separate corpus.
What cannot be separated is which systemic change did it: register-anchored batch processing, procedural streamlining, or agency-wide practice all reach every office simultaneously and are observationally equivalent. The claim is associational, not causal.
The convergence in Figure 26 is the clearest picture. Offices are ranked by their total document-anchored CE caseload; the busier half is the 16 BLM offices at or above the median, each handling at least 46 reviews, and the quieter half the 15 below. Each line is the median review duration among that half’s decisions in each two-year bin (only bins where the half has at least 10 decisions are shown). The quieter half — which never builds a comparable caseload — still converges to the busier half’s review duration; if reps were what made offices fast, that gap would persist, but instead it closes.
How to read Table 3: each row is a test of the learning-curve idea for one agency. We line up every office’s reviews in the order it completed them and ask whether reviews later in an office’s own sequence — when the office had more accumulated experience — ran shorter. The effect is expressed per doubling of experience: an estimate of −5% would mean that each time an office’s completed caseload doubles (its 30th review → 60th → 120th), its reviews run about 5% shorter. The first estimate asks the question directly, and shows the apparent learning gradient (negative — later reviews look faster). The second asks it again after allowing each calendar year its own baseline speed, so that system-wide changes that made all offices faster in the same years — new procedures, batch processing — are not mistaken for office experience. That control eliminates the effect in both agencies: offices with more accumulated experience were not systematically faster than offices with less.
| Agency | Reviews / Offices | Experience effect — no year control | Experience effect — with year control |
|---|---|---|---|
| BLM field offices | 1,757 / 31 | -6.8% (CI -13.2 to +0.1) | +11.0% (CI -2.9 to +26.8) |
| DOE administering offices | 4,821 / 12 | -6.2% (CI -9.9 to -2.4) | +3.2% (CI -4.7 to +11.7) |
| CE, document-anchored, over each agency's full window (no calendar filter). Register-anchored sensitivity fits and the EA infeasibility note (no BLM office clears ≥ 30 document-anchored EA durations) are retained in d4_fieldoffice_model.csv. | |||
Analysis caveats
- Register-anchoring artifact. Durations rest partly on register/proxy start dates that are late administrative entries; the analysis uses document-anchored initiation dates to limit this, and reports the register-anchored fits only as a flagged sensitivity view.
- Retained pre-2012 artifact rows. A few early BLM rows are known extraction artifacts — a 9-row batched cluster (office AZ-A010) all sharing one initiation → decision span, and historical-citation “initiations” (e.g. a 7,227-day span anchored to a 1985 citation). Under the symmetric no-filter design both arms keep their full windows, so these rows are retained rather than one-sidedly dropped; they inflate the raw (uncontrolled) BLM estimate only and do not affect the year-controlled result (they show up as the tall early point of the quieter half in Figure 26).
- ~38% of BLM projects are unparsed (37.5%). EISs account for most of the gap — only 7.4% parse (large/programmatic EISs are filed under titles, not DOI-BLM case numbers) — versus 63.9% of CE and 66.8% of EA.
- Project mix within an office. Short well-abandonment CEs dominate some offices;
project_energy_typeis a covariate, but a “faster” office could reflect a shift toward simpler actions rather than any change in processing. - DOE administering offices are not BLM field offices. They come from the CX register’s
officefield (grant/program offices such as NETL and Golden), not site-specific case handlers; the DOE arm is a structural replication, not a like-for-like comparison.
Geothermal review timelines by BLM field office
The Geothermal tag spans two worlds that should never be pooled, and the by-office slice the deliverable asked for is too thin to compare. Of 873 geothermal projects only 109 are BLM-led, only 61 carry a parseable field-office code, and only 39 of those have a complete timeline — spread across 21 offices whose busiest holds just 9 complete reviews, well below any floor for a duration comparison. The analysis is therefore a descriptive office inventory plus a cohort-level state map, and it focuses on CE reviews (the substantive EA/EIS record is a handful of projects — see the caveats).
The two worlds are BLM western-resource geothermal — leases, exploration, and power plants concentrated in NV, CA, and UT, median CE 34 days — versus the much larger body of 743 Department of Energy projects: ARRA-era heat-pump and research categorical exclusions filed nationwide, including states with no geothermal resource base, median CE just 8 days with a 2010–2012 decision surge. Pooling a resource-development review with a research-grant CE would be meaningless, so every number here is read within a cohort, never across. (Office-code recovery is a dead end: only 2 of the 48 unmatched BLM projects carry any office-like code.)
The office inventory
Figure 27 sizes the problem in three stacked panels. The top panel is the funnel — of the 873 geothermal projects, 764 are the DOE & other tier, 48 are BLM-led with no parseable office code, and only 61 reach the field-office inventory. The middle panel is the literal by-office answer for those 61: every BLM field office is a short bar — the busiest, NV-C010, holds 18 projects — and most are categorical exclusions. The bottom panel is newly feasible: the DOE-tier projects carry no field-office code but link to the CX register’s office field, and 456 of the 764 do so — dominated by the Golden Field Office (311 projects), then the National Energy Technology Laboratory and EERE headquarters. These are administering and grant-program offices, not BLM-style field offices — two inventories of two different kinds of unit.
The two worlds on a map
Figure 28 plots the complete CE reviews by state and cohort. The reading is immediate: the BLM bubbles (navy) sit only in the western resource states, while the DOE & other bubbles (light blue) blanket the country — because they are grant CEs tied to funding, not geothermal geology. Bubble size is the median duration in months, not project count, so the map also shows that duration is not a regional constant: even within the DOE tier, states range from near-instant heat-pump CEs to long research reviews.
Analysis caveats
- Two worlds, one tag. “Geothermal” pools BLM western resource development with DOE grant-era heat-pump and research CEs; the split median (34 days for BLM vs 8 days for everything else) is a composition artifact, not a process finding. Every number here should be read within a cohort, never across.
- Office-level is descriptive only. With a busiest office of 9 complete reviews (2 of them document-anchored), no office clearing the ≥ 10-review floor, and ~44.0% of BLM geothermal projects unparsed (48 of 109, office recovery largely infeasible), the inventory cannot support a duration comparison, a learning-curve test, or any office ranking.
- Register artifact, and DOE offices are not field offices. Durations rest partly on register/proxy start dates (the document-anchored counts isolate the artifact-free reviews, and they are even thinner); separately, the 456 DOE-tier projects with a named office link to administering / grant-program offices in the CX register (the Golden Field Office, NETL, EERE-HQ), not to BLM-style field offices — the DOE-office panel is an inventory of a different kind of unit, not a BLM-vs-DOE office comparison.
- Small-n everywhere, associational. The map’s per-state medians rest on small samples, and only 22 complete geothermal EA/EIS timelines exist corpus-wide (their individual durations remain in
d4_geothermal_timeline_points.csv); nothing here isolates a causal effect of office, agency, or geography.
Reproduction
Run from the repository root in the nepa conda environment:
# Timeline pipeline (candidates -> classify -> select -> LLM adjudicate)
conda run -n nepa python phase2/code/deliverable04/run_pipeline.py
ANTHROPIC_API_KEY=$(security find-generic-password -s nepa-anthropic -w) \
conda run -n nepa python phase2/code/deliverable04/06_adjudicate_llm.py \
--mode candidate_adjudication --process CE EA EIS --model claude-haiku-4-5-20251001 --workers 24
# Analysis + figures
Rscript phase2/code/deliverable04/08_create_figures.R # coverage, durations, FRA, energy
Rscript phase2/code/deliverable04/08_create_figures_solar.R # solar duration analysis
Rscript phase2/code/deliverable04/08_create_figures_technology.R # duration by technology
Rscript phase2/code/deliverable04/10_outliers.R # timeline outliers (case-study candidates)
# FRA document-length analysis (regulatory pages from the full page text, all EA/EIS)
conda run -n nepa python phase2/code/deliverable04/fra/01_extract_pages.py --run
Rscript phase2/code/deliverable04/fra/02_create_figures.R # document length / FRA page limits
# CEQ regulatory-regime durations (requires 08_create_figures.R outputs above)
conda run -n nepa python phase2/code/deliverable04/ceq_regime/01_build_tables.py
Rscript phase2/code/deliverable04/ceq_regime/02_create_figures.R
# Field-office experience (BLM + DOE arms)
conda run -n nepa python phase2/code/deliverable04/field_office/01_parse_offices.py --run
conda run -n nepa python phase2/code/deliverable04/field_office/01b_build_doe_offices.py --run
Rscript phase2/code/deliverable04/field_office/02_create_figures.R
# Geothermal timelines by BLM field office (tiered inventory + state map)
conda run -n nepa python phase2/code/deliverable04/geothermal/01_build_tables.py
Rscript phase2/code/deliverable04/geothermal/02_create_figures.ROutputs are written to phase2/output/deliverable04/ (figures/, diagnostics/).
Draft generated 2026-07-31 | NEPA Decarbonization Technology Analysis — Phase 2, Deliverable 4
Footnotes
The EA panel’s magenta Fossil (doc-anchored) row recomputes the fossil-EA duration using only document-stated initiation dates (
initiation_source_type ≠ register), setting aside the BLM register’s administrative “project start” entries that compress the register-anchored median to ~40 days. Its ~5-month median is the more defensible estimate of true review length, but it survives for only 124 of the 758 complete fossil EAs — the register is the sole start-date signal for the other 634. Both rows are shown so the coverage-versus-plausibility trade-off is explicit; CE and EIS are unaffected by this issue and keep their three standard energy rows.↩︎We use regulatory pages rather than raw PDF page counts because raw counts substantially overstate the length the statute limits — they include embedded appendices and sparse pages (covers, maps, dividers). Mean raw pages run roughly 2× regulatory pages for EAs and higher for EISs, so a raw count cannot be compared to the 75/150-page limits.↩︎