FLOOD RISK SITUATION REPORT Nuevo León, Mexico | GFS Short-Term | Bayesian Multi-Hazard Model | June 15, 2026
FLOOD RISK SITUATION REPORT
Nuevo León, Mexico | GFS Short-Term | Bayesian Multi-Hazard Model | June 15, 2026
Integrated Multi-Hazard Model (Fable-tuned, Bayesian flood class) — Google Earth Engine | 7-day Forecast Horizon | Sub-region run (Nuevo León admin-1) | Hazard-dominant prior balance | Includes media triangulation (Protección Civil NL / CONAGUA / SADM / SDP / Milenio / Infobae)
| HORIZON | GFS short-term (≤168 h) | ENSO BASELINE | El Niño (2019, 2023, 2024) |
| FUSION | Bayesian — hazard_dominant | RES | 500 m fine / 1 km exposure — standard profile |
1. Executive summary
The Bayesian flood model places an extraordinarily dense posterior surface across Nuevo León for the 7-day GFS window: 93.25 percent of the model's mapped population — approximately 3.16 million of 3.39 million people — falls inside the posterior exposure surface, alongside 93.95 percent of mapped urban area and 100 percent of mapped wind farms (1 of 1 site). This is the highest population-exposure reading produced by any Capybara run to date, and reflects a Nuevo León whose population is overwhelmingly concentrated in the Monterrey metropolitan area — itself sitting in the dominant high-posterior zone alongside the Sierra Madre Oriental orographic intercept that funnels runoff directly into the metro.
Two operational findings warrant immediate attention. Reservoir saturation: Presa La Boca at 96.3 percent and Presa Cerro Prieto at 94.7 percent as of 11–14 June 2026 (SADM/MVS bulletins). At those levels, incoming runoff translates directly to spill or release risk; both reservoirs supply Monterrey, and their downstream channels run through dense urban infrastructure. The single mapped utility-scale wind farm — 100 percent exposure — sits inside the high-posterior band; the operator should anticipate access disruption.
The visualization in this run uses the new terrain-shaded rendering: SRTM hillshade is multiplied into the posterior palette and low values are masked transparent, so the Sierra Madre Oriental ridge structure reads clearly through the color overlay. The shadowed eastern flank of the Sierra (Galeana → Aramberri → Doctor Arroyo) is part of the posterior, which is meteorologically consistent (orographic intercept on the Gulf-of-Mexico-facing slope) though it does not yet have ground-report corroboration for this window — see FDO 11.
Methodological note worth surfacing: the El Niño baseline used by this run is approximately aligned with the current ENSO state, where NOAA CPC has us in neutral conditions with 62 percent probability of El Niño emergence by August. The corridor identification is therefore baseline-supported as well as triangulated by ground reports. Area-class absolute hectare totals carry the same documented defect as prior Capybara reports — percentages remain interpretable, absolute hectare counts should not be publicly quoted (FDO 13).
2. Exposure analysis
Posterior-weighted expected exposure inside the Nuevo León AOI for the 7-day GFS window under the El Niño-baselined Bayesian fusion, hazard_dominant balance, sub-region run at standard resolution. ⚠ marks rows where absolute totals are affected by the area-aggregation defect (FDO 13); percentages remain valid.
| Variable | % of total | Expected (absolute) | Total (denominator) |
|---|---|---|---|
| Population (expected exposed) | 93.25% | 3,159,103 ppl | 3,387,600 ppl |
| Cropland area | 17.94% | 12,416 ha ⚠ | 69,211 ha ⚠ |
| Urban area | 93.95% | 57,253 ha ⚠ | 60,939 ha ⚠ |
| Oil palm area | N/A — not present in AOI | ||
| Protected areas | 60.57% | 293,417 ha ⚠ | 484,444 ha ⚠ |
| Built-up footprint | 29.06% | 1,889,339 ha ⚠ | 6,500,648 ha ⚠ |
| Electric grid length | 42.10% | 2,196 km ⚠ | 5,215 km |
| Roads length | 31.45% | 5,152 km | 16,382 km |
| Oil & gas sites | N/A — sparse-point artefact | ||
| Wind farms | 100.00% | 1 site | 1 site |
| Solar farms | N/A — sparse-point artefact | ||
Robust headlines for public communication: Population 93.25 % (≈ 3.16 M people). Urban area 93.95 %. Electric grid 42.10 %. Wind-farm exposure 100 % (single mapped site). These are percentages volunteers and partners can quote directly.
3. Spatial findings by zone
| Monterrey AMM — CRITICAL — ACTIVE | Monterrey, San Pedro Garza García, San Nicolás, Guadalupe, Apodaca, General Escobedo, García, Santa Catarina. The densest posterior zone in the AOI. Currently flooding (FDO 07). The Santa Catarina river runs through the metro; Sierra Madre runoff feeds it directly. |
| Sierra Madre Oriental flank — CRITICAL — UNVERIFIED | Galeana → Aramberri → Doctor Arroyo → General Zaragoza. Continuous orographic intercept; clearly visible in the terrain-shaded posterior surface. No fresh June 2026 ground reports surfaced (FDO 11 — possible observation bias). |
| Northern Nuevo León border — HIGH — WEAK CORROBORATION | Anáhuac, Lampazos de Naranjo, Bustamante. Posterior signal present; CONAGUA / ABC alert bulletins name these municipalities; no impact reports yet. |
| Southern Nuevo León — MODERATE | Linares → Mier y Noriega → Doctor Arroyo. Posterior present at lower intensity; agricultural land use; less urban-flood relevance. |
| Reservoir downstream channels — OPERATOR-SPECIFIC | La Boca → Río La Silla; Cerro Prieto → downstream; El Cuchillo → Río San Juan. The coupling between saturated reservoirs and urban downstream is the most actionable finding (FDO 09). |
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 — 3.16 million people inside the posterior surface (93.25 % of mapped population)
| Source / trace | CIESIN/GPWv411/GPW_Population_Count.first() at populationScale = 1000 m, posterior-weighted via expectedPopulation reducer. |
| Simple | About 3.16 million of the model's mapped 3.39 million Nuevo Leonenses live inside the 7-day flood-posterior surface. Nearly everyone in the state. |
| Expert | Three factors compound. (a) Population concentration: ~5 of 5.8 M residents live in the Monterrey AMM, entirely inside the posterior. (b) Hazard_dominant mode weights the forecast term heavily; GFS is wet across the state for this window. (c) El Niño baseline ≈ current state, so the wetter-than-baseline anomaly is real. GPWv4 vintage means the percentage applied to current ~5.8 M state population implies ~5.4 M actually exposed. |
| Implication | Protección Civil Nuevo León should treat the next 7 days as a state-wide elevated alert. AMM-municipality civil protection (Monterrey, SP Garza García, SN, Guadalupe, Apodaca, GE, García, Santa Catarina) should activate concurrently. |
| Limits | GPWv4 vintage. Percentage robust; absolute count is a lower bound. |
FDO 02 — 93.95 % of mapped urban area inside the posterior surface
| Source / trace | MCD12Q1 LC class 13 (urban), fraction-preserving aggregation at exposureScale = 1000 m. |
| Expert | 93.95 %. The AMM accounts for the overwhelming majority of mapped urban land in the state; the AMM is inside the posterior. The 6 % outside are smaller cabeceras in northern and southern municipalities. |
| Implication | Urban civil-protection in the AMM should activate flood-response protocols. SADM downstream channels (Santa Catarina, La Silla, Pesquería, San Juan) warrant pre-positioning. Hospitals (Hospital Universitario, Metropolitano, OCA, IMSS Centro Médico Nacional del Noreste) should prepare for displaced and injured arrivals from peri-urban informal settlements. |
FDO 03 — Single mapped wind farm 100 % inside the surface
| Source / trace | WRI/GPPD/power_plants filtered to wind fuel types; posterior sampled via pointCountExpo. Expected count = 1.00 of 1. |
| Expert | 100 % of a single mapped site. WRI GPPD is incomplete for Mexico's recent renewable buildout (Tres Mesas, Reynosa-adjacent fleets not represented). The 100 % is exposure of mapped infrastructure, not the operational national fleet. |
| Implication | The operator of the mapped site should activate access-road clearance, foundation-runoff inspection, and substation-protection protocols for the window. CFE / CENACE should anticipate transmission-side disruption. |
FDO 04 — Sub-region selector path validated end-to-end
| Source / trace | Run parameters confirm profile=standard with AOI restricted to Nuevo León via GAUL admin-1. State boundary is visible as a hard cutoff against Tamaulipas and Coahuila in the rendered posterior. |
| Expert | The tuned multi-hazard model's subRegionSelect lets users of large countries run one admin-1 region at full standard resolution rather than the whole country at coarse resolution. This run validates the path: Mexico as country, Nuevo León as sub-region, standard profile retained, full exposure panel returned, terrain-shaded visualization rendered. |
| Implication | Generalizable: Brazil → Amazonas, USA → Texas, China → Sichuan should expect identical UX. Useful proof-of-concept for volunteer demonstrations. |
FDO 05 — Terrain-shaded posterior visualization works as designed
| Source / trace | The output map renders the posterior via the new terrainShade helper (SRTM 90 m hillshade × data palette, low values masked transparent, 0.75 default opacity). Sierra Madre Oriental ridge structure is clearly visible through the purple posterior in the rendered image. |
| Implication | The hillshade-blended visualization preserves the spatial information in the posterior while making the underlying terrain context legible. Especially useful here because the Sierra Madre flank is part of the explanation for the eastern-flank signal — flat color would have hidden it. |
FDO 06 — Reservoir saturation as the operational coupling
| Source / trace | The model's posterior surface covers the watersheds of Presa La Boca (Río La Silla), Presa Cerro Prieto (Río Pilón / Río San Juan tributary), and Presa El Cuchillo (Río San Juan). |
| Expert | The model identifies upstream watersheds whose runoff feeds the saturated dams (FDO 09 provides the dam-level ground evidence). Posterior + dam state produce a coupled urban-flood vulnerability: incoming runoff cannot be absorbed at current levels; either the dams spill or downstream channels run high. |
| Implication | The single most actionable finding of the run. SADM and CONAGUA's Cuenca del Río Bravo regional office should coordinate spill-management. Downstream-channel municipalities (Cadereyta, China, Doctor Coss) need 24-h notice for any controlled release. |
Media-triangulation FDOs
FDO 07 — REAL-TIME CONCORDANCE: Monterrey flooding on the same day this report was generated
| Source / trace | Model's Monterrey AMM corridor (FDO 01) vs. 15 June 2026 ground reports: intense thunderstorms produced flash flooding, vehicle entrapments, and road collapses across the AMM. Protección Civil Nuevo León rescued at least seven people from submerged vehicles in Monterrey; a man swept by current was extracted by emergency teams. |
| Simple | The flood is happening right now. The model and the news are reporting the same event with the same geography on the same day, from opposite directions. |
| Expert | The closest model-to-ground concordance any Capybara run has produced — report and ground event are simultaneous. The 93 % AMM population exposure (FDO 01) is matched in real time by emergency-services activations. |
| Implication | Capybara's flood model identified the priority zone before the ground reports rolled in, using purely Earth-observation data. This concordance is publishable as the inaugural "real-time JANUS triangulation" case in the Capybara portal. |
| Limits | Causal-validity caveat: the model run is dated to the same day, so this is a co-occurrence rather than a forecast that preceded the event by several days. A scheduled-run regime (e.g. 06:00 UTC daily) would produce true forecast-leading triangulation. |
FDO 08 — CONAGUA bulletin confirms the forecast hazard layer
| Source / trace | CONAGUA bulletin for 16 June 2026: "lluvias muy fuertes con puntuales intensas (75–150 mm) en el este de Nuevo León," with hail, lightning, landslide, and flood risk noted. |
| Expert | CONAGUA's 75–150 mm forecast is consistent with the GFS field driving the model's hazard layer. Independent agency, same forecast direction. The CONAGUA aviso names eastern Nuevo León — the model's high-posterior band. |
| Implication | Convergence between CONAGUA's national-meteorological-service forecast and Capybara's posterior surface — independent data, independent fusion — makes the spatial findings defensible for stakeholder communication. |
FDO 09 — Reservoir saturation operational finding (Presa La Boca 96.3 %, Cerro Prieto 94.7 %)
| Source / trace | SADM live monitoring; MVS Noticias bulletins, 11 and 14 June 2026: Presa La Boca at 96.3 %, Presa Cerro Prieto at 94.7 %, Presa El Cuchillo at 61.1 %. |
| Simple | Monterrey's two biggest water reservoirs are nearly full. There is almost no buffer for the next week's runoff. |
| Expert | La Boca and Cerro Prieto effectively full — the additional storage each can absorb before spill/release is a few centimeters of pool height. With CONAGUA forecasting 75–150 mm over eastern Nuevo León and the model's high-posterior surface covering both watersheds, the next 7 days carry serious dam-management risk. The Santa Catarina river — already historically prone to catastrophic overflow events (Alex 2010, Alberto 2024) — would carry that runoff through dense urban Monterrey if dam releases coincide with continued precipitation. |
| Implication | The single most actionable finding for AMM water operations. SADM, CONAGUA Cuenca del Río Bravo, and AMM municipal civil-protection commissions should coordinate spill-management protocols. Downstream-channel municipalities (Cadereyta, China, Doctor Coss) need notice before any controlled release. |
FDO 10 — Synoptic forcing: mid-level cyclonic circulation, no named system
| Source / trace | Atlantic 2026 season opened 1 June; NOAA outlook below-normal (8–14 named systems). No active Atlantic tropical cyclone affecting NE Mexico as of 15 June. Driver: mid-level cyclonic circulation over NE Mexico + Gulf-of-Mexico moisture; possible "Arthur" being monitored (10 % development probability). |
| Implication | Important framing: the high-posterior surface does not require a named tropical cyclone. A 75–150 mm event over a saturated watershed driven by extratropical-tropical interaction is meteorologically sufficient. If the season's first named system does develop and tracks into the western Gulf, the posterior would intensify substantially. Stakeholders should not gate response on tropical-cyclone formation; the threshold is already crossed. |
Adversarial / methodological FDOs
FDO 11 — Coverage gap: Sierra Madre Oriental eastern flank flagged by model, no June 2026 ground reports
| Source / trace | Model's continuous high-posterior across Galeana → Aramberri → Doctor Arroyo → General Zaragoza, clearly visible in the terrain-shaded posterior. Media search returned only an April 2026 hail event (different hazard, different season); no June 2026 flood / landslide / precipitation impact reports. |
| Expert | Two readings. (a) Observation bias: rural mountain municipalities with low media-coverage density and limited stringer presence. The prior tempering work in the tuned model is the correction for this asymmetry. Absence of news does not mean absence of risk. (b) Forecast over-call: orographic-intercept interpretation correct in mechanism but actual rainfall and runoff may be smaller than posterior implies. |
| Implication | Volunteers with ground contacts in the sierra municipalities should check on them — Capybara's value-added would be flagging risk where the news cycle has not yet looked. Protección Civil Nuevo León's regional cells are the channels. |
FDO 12 — Methodological strength: El Niño baseline is approximately aligned with current ENSO state
| Source / trace | Model run used El Niño baseline (CHIRPS years 2019, 2023, 2024). NOAA Climate Prediction Center, June 2026: ENSO-neutral, 62 % probability El Niño emerges JJA 2026 and persists through end-2026. |
| Expert | Inverse of the Colombia-run caveat (La Niña baseline against neutral-to-El-Niño state, mismatch). Here the El Niño baseline closely aligns with the emerging state — the "wetter-than-baseline" anomaly is read against a roughly correct expectation. Corridor identification is therefore baseline-supported as well as triangulated. |
| Implication | ENSO_YEAR_TABLE should be updated to add 2026 as neutral_to_elnino. Maintenance task, not a defect. |
FDO 13 — Area-class absolute totals are not quotable in this run; ratios remain interpretable
| Source / trace | Cropland total 69 k ha (SADER lists ~150 k ha cultivated). Built-up "total" 6.5 M ha ≈ entire state area, implausible. |
| Implication | Quote percentages, not absolutes. Same defect documented across El Salvador, Guatemala, and Colombia runs. Patch path identified; logged as repository maintenance. |
5. Limitations & caveats
- The model run is dated to the same day as the active event (15 June 2026). This is co-occurrence rather than predictive forecast lead-time. Scheduled daily runs would convert this into true 24–168 h forecast triangulation.
- GPWv4 vintage (2010–2015). Absolute population counts are a lower bound; current Nuevo León population is ~5.8 M.
- GFS native resolution is 0.25° (~25 km). Sub-municipal precipitation variation in the AMM is not resolved.
- The 6-component hazard coefficients were calibrated for Colombia; Mexico-specific runoff regimes empirically validated by FDO 07 but not formally re-tuned.
- Reservoir state is not internally modelled. Model identifies upstream watersheds; SADM's live monitoring is the authoritative source for dam state.
- WRI GPPD power-plant database is incomplete for Mexico's renewable buildout.
- Area-class absolute totals are not quotable in this run (FDO 13).
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 Protección Civil Nuevo León, CONAGUA, SMN, SADM, El Norte, Milenio, Infobae, SDP Noticias, Récord, and Reuters Mexico, and weaving the convergent findings into the FDO structure. The LLM did not generate the model output, fabricate exposure numbers, or invent the ground-report events.
Provenance. Exposure figures read directly from the model's bottom-left exposure panel. Model code: PROD/src/earth-engine/integrated-multihazard-model-tuned.js. Median media-source search recency: 1 day. ENSO classification: NOAA CPC June 2026 discussion. Reservoir-state source: SADM live monitoring + MVS Noticias bulletins (11–14 June 2026).
Reflexive memo. The analyst is Colombian, not Mexican, and is an outsider to Nuevo León's specific governance structures. Local Capybara volunteers with Monterrey ground familiarity should validate institutional naming (Protección Civil NL, SADM, CONAGUA Cuenca del Río Bravo, CFE/CENACE) before stakeholder dissemination.
Adversarial pass. Five alternative explanations considered. (1) The 93 % population reading is a model defect — rejected: AMM concentration is real geography. (2) Monterrey concordance is coincidence — rejected: model and ground report the same geography from independent data within hours. (3) Sierra Madre flank is forecast over-call — partially accepted as caveat. (4) 100 % wind-farm finding is portfolio-scale meaningful — rejected: dataset incomplete. (5) El Niño baseline produces inflated exposure percentages — partially accepted: percentages robust as ratios, magnitudes slightly inflated against strict-neutral baseline.
7. Sources
Model run. Capybara EWS multi-hazard model (Bayesian flood class), PROD/src/earth-engine/integrated-multihazard-model-tuned.js. JANUS analytic framework: Feged 2025, DOI 10.21203/rs.3.rs-7580516/v1.
Ground-report triangulation (June 2026):
- Monterrey flash flooding, 7 rescues, 15 June 2026 — SDP Noticias · Infobae 16 June · Milenio 15 June
- Prior-week baseline 8 May 2026 — Infobae · Récord
- Santa Catarina river historical context (TS Alberto 2024) — SDP Noticias
- CONAGUA aviso 16 June — N+ · Milenio
- Atlantic hurricane outlook 2026 — Unotv NOAA · Azteca Chiapas — Arthur monitoring
- Reservoir levels — MVS 11 June · MVS 14 June · SADM live monitoring
- Hurricane Alex 2010 historical context — Frontiers in Environmental Science
ENSO state: NOAA Climate Prediction Center, June 2026 ENSO discussion — CPC ENSO Discussion.
Document generated: June 15, 2026 | Model: Integrated Multi-Hazard Model (Fable-tuned) — Flood (Bayesian) | Forecast: GFS short-term ≤168 h | Sub-region run: Nuevo León (standard profile) | ENSO baseline: El Niño (aligned with current state per NOAA CPC) | Prior balance: hazard_dominant | Visualization: SRTM-shaded posterior, 0.75 opacity | Classification: UNCLASSIFIED / FOR PLANNING USE
Labels: Early Warning System, Situation Report, Mexico, Nuevo León, Monterrey, Floods, Bayesian fusion, JANUS, real-time triangulation, Sierra Madre Oriental, dam saturation, La Boca, Cerro Prieto, terrain-shaded visualization
Note: Only a member of this blog may post a comment.