The four active DT logs document four discrete decision tracks. Read separately, each describes a specific gap: a missing capacity mechanism, a departing generation buffer, a hydrological monitoring deficit, a missing additionality requirement. This memo reads them together — as a single compound failure mode. The argument is that Finland and the Nordic system are losing all three independent elasticity reservoirs simultaneously, through mechanisms that are structurally independent but functionally convergent.
The DT logs map onto WP-015's three structural shifts plus one amplifying mechanism:
| DT log | Layer | WP-015 shift | Mechanism |
|---|---|---|---|
| DT-004 | Demand (Input) | Shift I — demand rigidification | Data centre 24/7 load with near-zero price elasticity. Growing baseline rigidity. 3.4 GW pipeline without binding flexibility obligation. |
| DT-002 | Domestic supply buffer | Shift II — buffer erosion | CHP decommissioning removes the only thermally-correlated dispatchable buffer. ~500 MWe retired 2024–2025, coal ban 2029, no replacement. |
| DT-003 | External buffer | Shift II extension | Norwegian reservoir system — the Nordic region's multi-week energy battery. Drought correlation with Finland increasing. No coordinated stress protocol. |
| DT-001 | Market allocation | Shift III — market residualisation | No capacity mechanism. PPAs pre-commit capacity away from spot. Market operates on residual volume and loses ability to signal scarcity when it matters most. |
The cascade logic follows directly:
Despite tracking different decision domains, all four DT logs converge on a single structural claim: the problem is not a capacity shortage — it is a coordination and temporal failure. The system has or could acquire sufficient capacity in aggregate, but the mechanisms that would deploy that capacity during a multi-day winter stress event are simultaneously degrading.
| DT log | Specific gap | Common thread |
|---|---|---|
| DT-001 | No capacity mechanism → no instrument for multi-day adequacy | The system cannot convert stress into an adequate response in time — not because capacity is absent in aggregate, but because the mechanisms that would deploy it have been removed, pre-committed, or never built. |
| DT-002 | CHP removal → no thermally-correlated domestic buffer | |
| DT-003 | No Nordic hydro stress protocol → external buffer unmanaged | |
| DT-004 | No additionality requirement → rigid load grows without compensation |
The four logs are internally consistent but reveal three structural tensions when read in combination.
| DT log | SE1 role |
|---|---|
| DT-003 | External buffer — Nordic hydro reserve that Finland can import from |
| DT-002 | Replacement production — substitute for decommissioned CHP via Aurora Line |
| DT-004 | Competitor — industrialisation (Hybrit, Stegra, data centres) absorbs SE1 capacity |
The one causal link that remains unproven is the connection between DT-003 (hydrological buffer collapse) and DT-001 (market residualisation). They currently appear as parallel stressors rather than a feedback loop.
The potential feedback mechanism would run: reservoir depletion → reduced Norwegian export capacity → reduced Nordic spot market depth → accelerated PPA formation as actors seek to secure supply → further spot market residualisation → DT-001 conditions worsen. If this loop is real, WP-015 shifts from a structural risk model to an endogenous instability model. This is the empirical question that separates "structural fragility" from "self-reinforcing collapse."
Under what conditions does elasticity collapse materialise vs. remain latent?
| Condition | Collapse avoided | Collapse materialises |
|---|---|---|
| DT-004 load growth | Data centres sign additionality PPAs → system capacity grows with load | Non-additionality PPAs → residual market shrinks |
| DT-002 CHP exit | Replacement dispatchable capacity built before 2029 coal ban | No replacement → thermally-correlated buffer gone by 2029 |
| DT-003 hydro stress | Nordic joint protocol activated → coordinated reservoir management | No protocol → SE1 switches to competitor role silently |
| DT-001 market depth | Capacity mechanism in place before 2027 → market has floor | Mechanism delayed to 2031+ → three winters without instrument |
| All four simultaneously | Any two of four conditions avoided → partial elasticity retained | All four conditions materialise simultaneously → WP-015 jump-regime |
Sections 1–6 and Appendix A were built on a definition of elasticity inherited from the 2021–2025 system: elasticity as dynamic response to stress. The empirical test in A.4 partially falsified the coupling hypothesis under that definition. But the critique this revision responds to identifies a more fundamental problem: the 2021–2025 reference period describes a different type of system than the one Finland is entering.
The classical Nordic balance model assumed: slowly growing load, CHP as structural presence, Norwegian hydro as passive buffer, spot market as dominant allocation mechanism, PPAs as marginal phenomenon, SE1–FI flows as symmetric. All of these assumptions are in structural transition simultaneously. The consequence is not that SM-002's empirical tests were wrong — they measured what they were designed to measure. The consequence is that the question was wrong.
| 2021–2025 system | Emerging system | |
|---|---|---|
| Allocation mechanism | Spot market responds to stress in real time — price signal activates reserve capacity | PPA layer pre-allocates capacity before stress — spot operates on residual volume only |
| Load structure | Weather-correlated demand — peaks in cold weather, lower in summer | Rigid 24/7 baseload (data centres, hydrogen, electrified heat) + weather peaks on top |
| CHP function | Thermally-correlated dispatchable buffer — output peaks when demand peaks | Exiting — its loss is not a capacity deficit but a loss of built-in response timing |
| Hydro role | Passive external buffer — available on request via price signal | Less predictable — drought correlation increasing, SE1 industrial draw competing |
| Crisis signal | Appears first in production shortfall — visible in EPP, SP, import-gap | Appears first in access asymmetry — some actors have contracted supply, others face residual spot price |
| Elasticity definition | Capacity that can be activated in response to stress | Capacity that is not yet pre-committed at the moment of stress |
In the pre-allocated system, the relevant state variable is not installed capacity or even dispatchable capacity — it is Freely Allocatable Capacity (FAC): the share of total capacity that is not already committed via long-term contracts at any given moment of stress.
This is the M variable problem made concrete: M (market depth) is not a separate variable — it is the ratio FAC / total_capacity. As FAC shrinks, the spot market operates on a smaller fraction of the system. The price it discovers is the price of scarcity in the residual, not the price of the system. This is WP-015 §4's market residualisation rendered as a measurable quantity — if FAC were observable.
In the pre-allocated system, a stress event does not necessarily appear as an EPP spike or a Fingrid shortage status alarm. It appears first as price divergence between contracted and spot supply. Rauramo (Fortum, winter 2025–2026) confirmed the mechanism: annual production sold at under 4 snt/kWh in fixed contracts while spot price approached 20+ snt/kWh during cold peaks. The physical electrons were produced. They were not available to spot buyers at any price that reflected the stress — because they were already committed.
This has three implications for SM-002's monitoring framework:
EPP underestimates structural stress in the pre-allocated system. EPP measures the gap between consumption and domestic production at the system level. It does not measure the gap between freely-allocatable production and consumption. When PPA penetration is high, EPP can read Normal while residential and industrial spot buyers face Elevated or BP-like prices.
The DT-004 Priority Inversion Risk is not a worst-case scenario — it is the baseline operating mode of the pre-allocated system under stress. Data centres have contracted supply. Households buy from the residual. The inversion is not a failure — it is the intended outcome of the contracting structure.
WEM §12's market depth indicator (M) is the most important instrument in the pre-allocated system — and it is the one that does not yet exist. FAC is not published by Fingrid, NordPool, or any public Nordic data source. The absence of this data is itself diagnostic: the system is transitioning away from a market structure that made FAC irrelevant (when everything was spot-priced, FAC = total capacity) toward one where FAC is the critical variable and no one is measuring it.
The structural findings of §§1–6 and Appendix A remain valid with this redefinition:
Es structural decline is confirmed — CHP MW are exiting and not being replaced. In FAC terms, this means fewer MW are available for free allocation even before PPA pre-commitment is considered.
Ex conditional coupling is confirmed — SE1 can supply Finland under normal conditions but switches to competitor role under simultaneous Nordic industrial stress. In FAC terms, SE1's freely-allocatable export capacity to Finland is shrinking as Hybrit, Stegra, and Nordic data centres claim larger shares of SE1 production through PPAs.
The SE1 regime switch function (A.3) is the correct formal structure — but its state boundaries should now be defined in FAC terms: SE1 is in Buffer state when FAC_SE1→FI is high, Competitor state when FAC_SE1→FI approaches zero because SE1 industrial PPAs have absorbed available production.
External critique (April 2026) identified a structural weakness in §§1–6: the "simultaneous degradation" assumption is asserted through narrative symmetry rather than demonstrated through a formal model. SM-002 as written is coherent as a stress narrative but underspecified as a causal model. This appendix attempts to collapse the four-layer architecture into a minimal state-space representation that is falsifiable and forces each elasticity term to be declared explicitly.
The system is described by five measurable state variables:
| Variable | Description | Measurement proxy | Historical counterexample |
|---|---|---|---|
| Ed | Demand elasticity — price sensitivity of total load, including data centre baseline | DIESL 2025 short-run price elasticity estimates; DR activation volumes | 2022-Q3: industrial demand response reduced Finnish peak load ~400 MW during price spike |
| Es | Domestic supply elasticity — dispatchable capacity activatable within 0–4h | CHP MW available; hydro regulation capacity; BESS (DS 398/399); WEM FS layer | Winter 2023: OL3 commercial operation offset ~1600 MWe of dispatchable gap; CHP deficit masked |
| Ex | External import elasticity — SE1 capacity realisable when Finland needs maximum import | TRR (WEM §12 DS 60/87); hydro_RF; SE1–FI price spread; Fenno-Skan + Aurora Line utilisation | Winter 2021: sustained high SE1 import (800–1200 MW) during extended cold period without SE1 stress |
| M | Market depth — spot market's share of total dispatchable capacity (non-pre-committed) | PPA penetration rate; forward vs. spot volume ratio; residual spot capacity at stress hours | No clean counterexample identified — this is the weakest empirical layer in the model |
| T | Institutional response time — time from observed stress signal to activated policy response | DT-001 timeline (2023 working group → 2031 proposal); Toolbox 2023 → no adoption 2026 | 2022 European energy crisis: several EU member states activated emergency measures within weeks — but Finland did not introduce new instruments |
Version 1.1 used a multiplicative form C(t) = [Ed · Es · Ex · M] / T. Second-round critique identified two structural errors in this specification:
Error 1 — T is endogenous, not exogenous. Institutional response time is not a fixed parameter. It is a function of stress level: severe, visible crises accelerate political response (2022 European energy emergency produced binding policy within weeks in several countries). Treating T as constant overestimates collapse probability in acute-crisis scenarios and underestimates it in slow-onset scenarios where no single event triggers a political response threshold.
Error 2 — multiplicative form assumes independence. The product Ed · Es · Ex · M implies all four terms are statistically independent. They are not. Es and Ex are partially substitutable (domestic hydro and SE1 import serve similar buffering functions). More critically, the SM-002 thesis is specifically about correlated degradation under stress — but the multiplicative form tests components separately, not their joint behaviour. A component-by-component falsification test cannot validate or invalidate a coupling claim.
The revised specification separates the structural condition from the coupling condition:
This reformulation separates two claims that v1.1 conflated: (1) that individual elasticity terms are declining, which is partially evidenced by DT-001–DT-004, and (2) that they decline together under stress, which is the coupling hypothesis and remains empirically untested.
SE1 has been identified as analytically under-identified — it operates as a "role-shifting actor" in the DT logs (buffer in DT-003, import source in DT-002, competitor in DT-004) without a formal regime-switch rule. The revised model gives SE1 an explicit state function:
| SE1 state | Condition | Role for Finland | Observable signal |
|---|---|---|---|
| Buffer | SE1 hydro > 65% median AND SE1 industrial load < 8 GW AND SE1–FI price spread > 0 | Export available; Ex near maximum | TRR > 0.6; DS 60 positive (→FI) |
| Neutral | SE1 hydro 45–65% median OR SE1 industrial load 8–12 GW | Export limited; Ex moderate | TRR 0.3–0.6; price spread near zero |
| Competitor | SE1 hydro < 45% median OR SE1 industrial load > 12 GW (Hybrit + Stegra + DC baseline) | SE1 net importer; Ex near zero or negative | TRR < 0.3; DS 60 near zero or reversed; SE1 price > FI price |
The SE1 state is thus a function of three observable variables: NVE hydro level, SE1 industrial draw, and price spread. All three are available in real time through the WEM §12 transmission layer (DS 60, DS 87, DS 30) and the NVE proxy (hydro_RF). The regime transition is now observable, not inferred.
The coupling condition was tested using reconstructed monthly cold-peak data for 2020–2025, drawn from Fingrid statistics, Energiateollisuus annual reports, and NVE Magasinstatistikk. The test covers 14 cold-stress months and 5 non-stress summer months. Full hourly Fingrid API data (DS 124, 87, 60) would produce more precise results — this is a first-order test using available aggregate data.
Es proxy: (CHP + hydro) / consumption · Ex proxy: SE1 import / NTC capacity (1600 MW)
| Test | Result | Interpretation |
|---|---|---|
| ρ(Es, Ex) during cold peaks (n=14) | +0.464 (p=0.095) | Positive correlation — Es and Ex move in the same direction. This suggests partial substitution, not simultaneous collapse. Coupling hypothesis not supported in normal winters. |
| Es trend 2020–2025 (cold peaks only) | r = −0.89 (p=0.001, excl. 2022) | Strong structural decline confirmed — CHP decommissioning visible. Es fell from 0.37 (2020) to 0.30 (2025). This part of SM-002 is empirically supported. |
| Ex trend 2020–2025 (cold peaks only) | r = −0.07 (p=0.81) | No systematic decline — Ex varies widely year to year. SE1 import has not structurally degraded in the 2020–2025 window. |
| 2022 anomaly: Ex during energy crisis | Mean Ex = 0.281 vs. 0.582 in other years | 2022 shows Ex CAN collapse synchronously with stress — but this was driven by an external shock (gas price crisis), not structural SE1 industrialisation. |
Revised SM-002 thesis: Es is structurally declining (CHP exit, confirmed). Ex is not structurally declining but is vulnerable to synchronous collapse during systemic shocks. The coupling condition is not a permanent feature of the system — it is a conditional risk that activates under exceptional circumstances. This distinction matters for policy: the primary addressable risk is Es (domestic dispatchable capacity), not the coupling mechanism itself.
| Test | Data required | Status |
|---|---|---|
| PPA formation rate vs. NVE reservoir level 2020–2026 | Nordic PPA announcements vs. NVE weekly filling | Not yet compiled — tests DT-003 ↔ DT-001 causal link |
| T(stress) function: Finnish policy response time in 2022 | Policy decision timeline vs. EPP/NVE stress signals | Preliminary: Finland did not activate new instruments in 2022 — T remained long |
| Hourly ρ(Es, Ex) during 72h+ cold-calm events | Fingrid DS 124, 87, 60 hourly 2019–2026 | Requires Fingrid API access — monthly aggregate used as proxy above |
Adaptive substitution: Real power systems exhibit layered substitution — price spikes, fuel switching, demand response scaling, emergency imports, last-resort dispatch. The model treats institutional latency as long inside the 2027–2030 window. The endogenous T(t) formulation partially corrects this, but the substitution dynamics between Es and Ex are still modelled as independent rather than as partial substitutes.
High-variance reliability vs. clean regime transition: If outcomes depend on idiosyncratic single-component failures (Stegra furnace breakdown, OL3 outage, Aurora Line maintenance window), the system is in a high-variance reliability space rather than a deterministic regime-transition space. These are different claims. The Monte Carlo structure of WP-014's OGAS2 framework is better suited to high-variance reliability analysis than SM-002's structural model.
M remains the weakest empirical layer: No historical counterexample exists for market residualisation producing allocation failure at Nordic scale. The DT-003 ↔ DT-001 causal link is the model's most consequential unproven assumption and the highest-priority empirical test.
The SE1 regime switch function (A.3) is partly driven by a structural factor not visible in the 2021–2025 reference data: the Nordic nuclear age cliff. Sweden's six remaining reactors (Ringhals 3+4, Forsmark 1+2+3, Oskarshamn 3) were all commissioned 1980–1985 and approach 60-year operational limits 2040–2045. Vattenfall's SMR programme targets first operation circa 2035, creating a decade-long capacity gap during which SE1's structural export surplus progressively erodes. Combined with SE1 industrial draw (Hybrit, Stegra, data centres), SE1's regime distribution shifts toward Competitor each year — independent of Norwegian hydrology.
Finland's own nuclear concentration risk compounds this: OL3 at 1,600 MW represents 12–20% of Finnish winter demand as a single unit. Each 50-day maintenance outage is a direct Es shock measurable in WEM Firm Share. The original OL3+OL4 configuration would have halved this exposure; OL4's 2015 cancellation made the concentration permanent. Loviisa's VVER-440 fuel supplier transition (TVEL → Westinghouse, 2024–2030) adds further operational uncertainty through the 2027–2032 window. These are forward-looking structural changes not captured in the A.4 empirical test; the A.3 state thresholds should be reviewed as SE1 industrial load data from Hybrit and Stegra becomes available post-2026. See SM-003 §7.1 for full analysis.