WILDFIRE RISK SITUATION REPORT - Western Europe (Italy · Portugal · France · Spain) | WeatherNext 2 ML Ensemble | El Niño Baseline | 20 June 2026
Western Europe (Italy · Portugal · France · Spain) | WeatherNext 2 ML Ensemble | El Niño Baseline | 20 June 2026
Integrated Multi-Hazard Model (Fable-tuned) — Wildfire Ensemble class with legacy multiplicative sharpening | Google Earth Engine | 15-day Forecast Horizon | Includes media triangulation (EFFIS · DG ECHO · IPMA · AEMET · Météo-France · Protezione Civile)
| HORIZON | WeatherNext 2 ensemble mean (≤ 360 h) | ENSO BASELINE | El Niño (2019, 2023, 2024) |
| FUSION | Ensemble — Bayes + Legacy | RES | 500 m fine / 1 km exposure |
1. Executive summary
The Capybara wildfire model places concentrated 15-day fire-risk surfaces across four Western European countries — Italy, Portugal, France, and Spain — and the ground evidence already supports the regional read.
The strongest single regional signal is northern Iberia: the model places dense per-pixel hotspots across the Minho, Trás-os-Montes, Douro, and Porto regions of northern Portugal, continuing across the border into Galicia, Asturias, Cantabria, and northern León in northwestern Spain. This bridge corresponds precisely to where IPMA (Portugal) has flagged ~140 municipalities at maximum rural-fire risk across 12 districts including Vila Real, Bragança, Viana do Castelo, and Porto, and where AEMET (Spain) has issued "very high or extreme" fire-risk warnings concentrated in Galicia.
Italy shows the model's classic Apennine-spine pattern (Tuscany through Calabria) plus a distinct Sardinia signature, consistent with Italy's Protezione Civile launching its Forest Fire Campaign on 15 June 2026 with Sardinia named as priority deployment area. France shows the Pyrenees-border concentration and Provence-Mediterranean coastal pattern, matching Météo-France's orange fire alert in Pyrénées-Orientales, Aude, Bouches-du-Rhône, Vaucluse, and Var. Spain shows the Galicia / Cantabrian-mountain concentration consistent with AEMET warnings.
Two operator-specific findings warrant immediate attention. Portugal: 18.95 % of mapped utility-scale wind farms (38 of 203 sites) sit inside the posterior surface — concentrated in exactly the Minho and Trás-os-Montes districts EDP Renováveis operates in heavily. Italy: 9.46 % of mapped solar farms (7 of 74 sites) sit inside the surface, with the Apennine corridor capturing several southern-Italy fleet locations.
The France country-wide exposure panel had not finished computing at the time the run was screen-captured (cell values read "Computing..."). The map surface itself rendered fine; the exposure numerics are pending and would complete in a few additional minutes of GEE compute.
2. Exposure analysis (four-country comparison)
Posterior-weighted expected exposure inside each national AOI for the 15-day WeatherNext 2 window. Percentages remain interpretable; absolute hectare counts carry the area-aggregation defect documented in FDO 14 and should not be quoted publicly.
| Variable | Italy | Portugal | France | Spain |
|---|---|---|---|---|
| Population (% mapped) | 3.03 % | 6.32 % | computing | 1.42 % |
| Population (people) | 1,446,426 | 525,932 | pending | 475,957 |
| Cropland | 6.76 % | 2.60 % | pending | 0.73 % |
| Urban area | 2.07 % | 2.21 % | pending | 0.32 % |
| Protected areas | 3.86 % | 7.34 % | pending | 2.70 % |
| Built-up footprint | 4.79 % | 6.80 % | pending | 2.15 % |
| Electric grid length | 5.40 % | 8.50 % | pending | 2.26 % |
| Roads length | 5.17 % | 7.82 % | pending | 2.20 % |
| Wind farms | 0.00 % (0/1) | 18.95 % (38/203) | pending | 2.37 % (8/331) |
| Solar farms | 9.46 % (7/74) | 1.01 % (1/61) | pending | 0.46 % (0/28) |
Robust headlines for public communication: Portugal carries the highest population exposure rate (6.32 %) and the standout infrastructure finding (19 % wind farms · 8.5 % grid · 7.8 % roads). Italy carries the largest absolute exposed population (~1.45 M) and notable solar-farm exposure (9.46 %). Spain shows the lowest national-scale percentages but the model's spatial signal is regionally concentrated. France numerics pending.
3. Spatial findings by zone
Italy
| Apennine spine — CRITICAL | Continuous per-pixel hotspot band from Tuscany through Calabria. Mediterranean-facing and Adriatic-facing slopes both flagged. |
| Sardinia — CRITICAL | Distinct posterior concentration; matches Protezione Civile's named-priority deployment area for the 2026 campaign. |
| Po Valley fringe — HIGH | Hotspots along the Alpine front and northern Apennine foothills (Piedmont, Lombardy, Veneto). |
| Sicily — HIGH | Central and eastern interior inland hotspots. |
Portugal
| Northern Portugal — CRITICAL — ACTIVE WARNING | Minho, Trás-os-Montes, Douro, Porto, Vila Real, Bragança, Viana do Castelo. Brightest hotspot zone in the entire four-country run. Matches IPMA's 140-municipality maximum-risk warning. |
| Central Mountains — HIGH | Serra da Estrela approach; continuation of the northern signal into central-mountain interior. |
| Algarve interior — MODERATE | Isolated hotspots in the southern interior. |
France
| Pyrenees border — CRITICAL — ACTIVE WARNING | Pyrénées-Orientales, Aude, Ariège. Continuous band along the French side of the Spanish border. Matches Météo-France orange alert. |
| Provence Mediterranean coast — CRITICAL — ACTIVE WARNING | Bouches-du-Rhône, Var, Vaucluse, Alpes-Maritimes. Matches Météo-France orange alert for the three named Provence departments. |
| Massif Central scattered — MODERATE | Scattered hotspots through the central uplands. |
| Brittany / Normandy coastal — LOW–MODERATE (FDO 13) | Isolated single-pixel hotspots; unexpected at this resolution. Worth checking for forecast-model edge effects. |
Spain
| Galicia / Asturias / Cantabria / León — CRITICAL — ACTIVE WARNING | Iberian-bridge continuation of the Portuguese northern signal. Brightest Spanish concentration. Matches AEMET very-high-to-extreme warnings. |
| Spanish Pyrenees — HIGH | Cross-border continuity with the French Pyrenees signal. |
| Cantabrian eastern flank — HIGH | Basque country and La Rioja foothills. |
| Castilla y León interior, Extremadura — MODERATE | Spring press reports of severe fires in these regions; current model signal is lower-intensity scattered. |
4. Forecasting and Decision Objects (FDOs) — JANUS method
FDOs follow the JANUS analytic method (Feged 2025, DOI 10.21203/rs.3.rs-7580516/v1). Each FDO links a claim to a dataset, a model layer, or a verified external source.
Model-derived FDOs
FDO 01 — Transboundary Iberian fire-risk corridor
| Source / trace | Posterior risk surface (final_threat) at fineScale = 500 m. Continuous bright-red hotspot zone bridging northern Portugal and northwestern Spain visible in rendered output. |
| Simple | The most-concentrated fire-risk zone in Western Europe right now stretches from northern Portugal across the Spanish border into Galicia, Asturias, and Cantabria. It's one zone, not two countries. |
| Expert | Bridge reinforced by three independent signals: WN2 ensemble-mean wetness deficit, El Niño-baselined CHIRPS anomaly, and FIRMS T21 historic fire density. The political border is invisible to the model and to fire dynamics. |
| Implication | EU Civil Protection should formalise Iberia-wide coordination for the next 15 days rather than treating Portugal and Spain as independent dispatch zones. Cross-border mutual-aid agreements should be activated now, not after the first major incident. |
FDO 02 — Portugal: highest population-exposure rate (6.32 %)
| Source / trace | CIESIN/GPWv411/GPW_Population_Count; expectedPopulation reducer; 525,932 / 8,321,886 = 6.32 %. |
| Simple | About 526,000 of mapped 8.3 million Portuguese live inside the 15-day fire-risk surface — one in sixteen. Highest national rate of the four countries. |
| Implication | ANEPC (Autoridade Nacional de Emergência e Proteção Civil) should anticipate concurrent activation across the named northern districts. |
FDO 03 — Italy: largest absolute exposed population (~1.45 M)
| Source / trace | 1,446,426 of 47,771,701 = 3.03 %. Posterior covers Apennine spine, Sardinia, Po Valley fringe. |
| Implication | Dipartimento della Protezione Civile should treat the next 15 days as broad national elevated alert. The pre-positioned 15 Canadair fleet posture is correctly sized for the spatial extent. |
FDO 04 — Portugal wind-farm exposure: 38 of 203 sites (18.95 %)
| Source / trace | WRI/GPPD/power_plants, wind-fuel filter; pointCountExpo reducer. 38/203 = 18.95 %. Spatial overlap with FDO 01 bridge near-total. |
| Implication | EDP Renováveis and other northern-Portugal wind operators should activate fire-perimeter inspection, transformer protection, and access-road clearance across the affected fleet. REN should anticipate possible upstream disruption to wind feed-in. |
FDO 05 — Italy solar-farm exposure: 7 of 74 sites (9.46 %)
| Source / trace | 7 of 74 mapped solar sites — concentrated along central-southern Apennine corridor and Sardinia. |
| Implication | Italian utility-scale PV operators (Enel Green Power and others) with central-southern or Sardinian assets should activate access-road clearance and panel-debris-monitoring protocols. |
FDO 06 — WeatherNext 2 ensemble provides 15-day forecast lead-time
| Source / trace | 64-member WeatherNext 2 ML ensemble at ≤360 h. Materially longer skill-window than 7-day GFS used in earlier Capybara runs. |
| Implication | This is a forecast-leading product. Operational decisions taken in the next 48 hours can affect outcomes across the named hotspot zones for the next two weeks. The EU Civil Protection mega-deployment posture is consistent with this lead-time profile. |
Media-triangulation FDOs
FDO 07 — EU Civil Protection "largest ever" deployment matches the model's regional read
| Source / trace | Four-country model posterior surface vs. DG ECHO 2 June 2026: 777 firefighters from 14 member states pre-positioned in Cyprus, Greece, Italy, France, Spain, Portugal. rescEU fleet pre-positioning at unprecedented scale. |
| Expert | Regional convergence: EU is acting under the same fire-weather observations (EFFIS, national met services) that motivate the Capybara model. Capybara adds value as an independent fusion with per-pixel granularity that EU-level deployment cannot itself produce. |
| Implication | Capybara output is consistent with the operational picture EU-level decision-makers are already acting on. The value-added is geographic localization within the named countries. This sitrep can be shared with DG ECHO operational coordination as supplementary geographic detail. |
FDO 08 — Portugal IPMA / GNR 140-municipality warning matches FDO 01 bridge zone
| Source / trace | Model's northern-Portugal hotspot zone vs. IPMA/GNR: ~140 municipalities at maximum rural-fire risk across 12 districts including Vila Real, Bragança, Viana do Castelo, Porto. |
| Expert | All four IPMA-named priority districts fall inside the model's high-posterior surface. Convergence between national institutional read and Capybara's independent EO fusion is publishable as a regional concordance. |
| Implication | For ANEPC / IPMA stakeholder communication: Capybara's geographic detail complements the municipality-level priority list with within-municipality pixel granularity — which specific neighbourhoods within Vila Real, which wind-farm operators within Trás-os-Montes. |
FDO 09 — French Pyrenees / Provence model signal matches Météo-France orange alert
| Source / trace | Model's French Pyrenees-border and Provence-Mediterranean concentrations vs. Météo-France orange alert for Pyrénées-Orientales, Aude, Bouches-du-Rhône, Vaucluse, Var (yellow alert for 8 more). |
| Implication | Five out of five orange-alert departments fall inside the high-posterior surface. Sécurité Civile's deployment posture in the named departments is supported by independent EO analysis. |
FDO 10 — Spain Galicia / Asturias model signal matches AEMET extreme-risk warnings
| Source / trace | Model's Galicia / Asturias / Cantabrian-mountain concentration vs. AEMET "very high or extreme" fire-risk warnings; recent spring fires in Galicia, Asturias, Castilla y León, Extremadura with at least three deaths and thousands evacuated; UME military emergency unit deployments. |
| Implication | UME and Galician / Asturian regional civil protection should expect sustained activation through the window. Cross-border coordination with Portuguese ANEPC for the Minho-Galicia corridor is the operationally important coupling. |
FDO 11 — Italian Protezione Civile 2026 campaign launch corresponds to the run window
| Source / trace | Italy's Protezione Civile Forest Fire Campaign 2026 officially opened 15 June 2026 (running through 15 October). 15 Canadair CL-415, 4 Erickson S-64F, 9 military helicopters pre-positioned. Sardinia named priority. |
| Implication | The 15-day WN2 forecast window covers 20 June – 5 July 2026, the first 20 days of the official campaign. Capybara can serve as a forecast-leading complement to Protezione Civile's tactical operations during initial mobilization. |
Adversarial / methodological FDOs
FDO 12 — ENSO baseline correctly aligned: NOAA confirms El Niño already present
| Source / trace | Model used El Niño baseline (CHIRPS years 2019, 2023, 2024). NOAA CPC June 2026: El Niño present, Niño-3.4 +0.7 °C, Niño-1+2 +2.1 °C. 82 % probability MJJ; 96 % DJF 2026-27; 63 % probability strong+ event. |
| Expert | Inverse of the recent Colombia run (where La Niña baseline was misaligned). Here the El Niño baseline is at minimum correctly aligned, and arguably conservative. The "wetter-than-baseline" / "drier-than-baseline" anomalies are computed against the correct climate reference. |
| Implication | The spatial findings of this run are baseline-supported as well as media-triangulated. Defensible for stakeholder communication on both methodological and ground-evidence axes. |
FDO 13 — Coverage divergence: northern French / Brittany / Normandy pixels not in Météo-France alerts
| Source / trace | Model posterior includes scattered single-pixel hotspots along Brittany / Normandy coastal France. Météo-France official alerts cover only southern departments. |
| Expert | Two readings. (a) Model edge effects: WN2 may produce artefacts near coastlines or at the AOI boundary. (b) Institutional observation bias: Brittany / Normandy are not Mediterranean fire-belt areas; Météo-France's alert system is calibrated for historic fire-risk geography. The model is data-driven and may surface emergent risk where institutional attention has not historically focused. |
| Implication | Worth a localized validation pass with French volunteers familiar with the named coastal communes. If durable across multiple WN2 inits, flag to Sécurité Civile as emergent-risk information. If it disappears on re-run, treat as ML-ensemble artefact. |
FDO 14 — Area-class absolute totals are not quotable in this run; ratios remain interpretable
| Source / trace | Built-up "total" (e.g. Spain 50.6 M ha ≈ Spain's 50.5 M ha total land area) is implausible. Same defect as El Salvador, Guatemala, Colombia, Nuevo León runs. |
| Implication | Quote percentages, not absolute hectare numbers, in public communications. Patch path is identified in the repo NOTES file. |
FDO 15 — France exposure panel pending; map surface is complete
| Source / trace | France country-wide exposure panel reads "Computing..." in the screen capture; map surface rendered cleanly with terrain shading + reference roads/rivers/cities. |
| Implication | Volunteers running France should wait 3–5 additional minutes after map tiles render for the exposure panel. The country-wide French numerics are not yet quotable; the spatial findings (FDO 09) are. |
5. Limitations & caveats
- Western European fire season has just begun; minimal named megafire incidents (>1,000 ha) surfaced for 1–20 June 2026. Current posture is warnings + pre-positioning, not active suppression. This is a forecast-leading product.
- WeatherNext 2 access is gated; volunteers without WN2 access can substitute GFS short-term (≤504 h) for a coarser but qualitatively similar surface.
- GPWv4 vintage (2010–2015). Absolute population counts are conservative lower bounds.
- Area-class absolute totals are not quotable (FDO 14).
- WRI GPPD incomplete for newer European renewable buildout. Wind / solar findings represent mapped exposure, not operational fleets.
- France numerics pending (FDO 15).
- The 6-component hazard coefficients were calibrated for tropical-Andean fire regimes; performance for Mediterranean Europe is validated by media triangulation but not formally re-tuned.
6. AI-use disclosure & reflexive memo (JANUS)
LLM assistance. A large language model assisted with structuring sections, drafting the plain-language Simple text of each FDO, executing the media-triangulation web search across EFFIS, DG ECHO, IPMA, AEMET, Météo-France, Protezione Civile, Reuters, NPR, Connexion France, Portugal Resident, and severe-weather outlets, and weaving convergent findings into the FDO structure. The LLM did not generate the model output, fabricate exposure numbers, or invent the institutional warnings.
Provenance. Exposure figures read directly from the model's exposure panel for Italy, Portugal, Spain; pending for France. Model code: PROD/src/earth-engine/integrated-multihazard-model-tuned.js (Ensemble class, legacy multiplicative sharpening at 0.7 strength). Median media-source search recency: 6 days. ENSO classification: NOAA CPC June 2026.
Reflexive memo. The analyst is Colombian and an outsider to the Mediterranean fire-policy geography. Specific institutional naming (ANEPC, AEMET, Météo-France, Protezione Civile, DG ECHO, IPMA, GNR, UME) should be validated by European Capybara volunteers familiar with operational structures before stakeholder dissemination.
Adversarial pass. Five alternatives considered. (1) Iberian bridge is artefact — rejected: bridge is consistent with shared meteorology / terrain / fire history across the border, institutional warnings on both sides converge. (2) Four-country convergence with EU deployment is coincidence — rejected: deployment uses EFFIS observations; Capybara fusion is independent and adds local granularity. (3) Brittany / Normandy single-pixel signal is operationally meaningful — partially rejected (FDO 13): treat as worth validation pass, not established finding. (4) El Niño baseline inflates exposure — rejected (FDO 12): NOAA classification supports baseline. (5) Legacy sharpening at 0.7 produces too-granular hotspots — accepted as visual-interpretation caveat; numerics use the same posterior and are internally consistent.
7. Sources
Model run. Capybara EWS multi-hazard model (Wildfire Ensemble class), PROD/src/earth-engine/integrated-multihazard-model-tuned.js. Legacy reference flow: legacy/wildfire/firerisk-legacy-pointing.js. JANUS framework: Feged 2025, DOI 10.21203/rs.3.rs-7580516/v1.
Regional fire-weather:
- Copernicus EFFIS — current wildfire situation
- DG ECHO 2 June 2026 — mega-deployment announcement
- Severe-weather.eu — Iberian heat dome forecast · Carbon Brief weather digest
Portugal:
- IPMA / GNR — 140 municipalities at maximum risk · The Portugal News
- US Embassy Lisbon — wildfire-season advisory 17 June 2026
Spain:
- NPR — Galicia heat wave + fires · RusSpain — Cantabria/Asturias 60+ fires/day · AEMET fire-risk forecast
France:
Italy:
ENSO state: NOAA CPC — June 2026 ENSO discussion.
Energy infrastructure: Wikipedia — Wind power in Portugal · AInvest — Portugal wildfire & renewables.
Document generated: 20 June 2026 | Model: Integrated Multi-Hazard Model (Fable-tuned) — Wildfire Ensemble (Bayes + Legacy) | Forecast: WeatherNext 2 ensemble mean ≤360 h | ENSO baseline: El Niño (correctly aligned per NOAA CPC) | Visualization: SRTM-shaded posterior, 0.75 opacity | Classification: UNCLASSIFIED / FOR PLANNING USE
Labels: Early Warning System, Situation Report, Wildfire, Western Europe, Iberia, Italy, Portugal, France, Spain, WeatherNext 2, JANUS, EU Civil Protection, Iberian bridge, EDP Renováveis, Sardinia, Pyrenees, Provence, El Niño baseline
Note: Only a member of this blog may post a comment.