Capybara Center for Risk Analysis · situation report · under JANUS

Flood Exposure Situation Report · Puebla, Mexico

Posterior-weighted expected exposure, Bayesian flood model, triangulated against Civil Protection reporting

HazardFlood, short-horizon exposure

AreaEstado de Puebla

Run2026-07-06 22:16 UTC

HorizonGFS short-term, 168 h (~7 days)

Provenance. Legacy Capybara Bayesian flood model (Google Earth Engine), model=bayesian, prior_balance=balanced. Config: GFS 168h · fineScale=500 · exposureScale=1000 · profile=standard · pop=WorldPop100m. ENSO baseline (as run): el_nino analog years 2019, 2023, 2024 (matches the current NOAA El Nino advisory). Code as FDO: derivedFrom: capybara-legacy [PID pending].

Legend flood risk: low → high (grey, light blue, blue, deep blue, dark purple = highest)

The model flags extensive flood exposure across the wildland and valley terrain of Puebla, concentrated in the Sierra Norte and Sierra Nororiental in the state's north and in the Mixteca and Sierra Negra in the south, with the highest-intensity (dark purple) cells in the northern sierra around Huauchinango, Xicotepec and Venustiano Carranza. Independent Civil Protection reporting for the same window corroborates this geography: Puebla City suffered a severe flood on 28 June 2026 and Tropical Storm Boris struck fifteen northern-sierra municipalities on 10 June. The run is now conditioned on the El Nino baseline, consistent with the current forecast state, and the figures are scenario estimates of exposure. The corroboration establishes consistency, not predictive skill.

One caveat to read first

Absolute population counts are under review and should not be used as issued. The in-domain population total (61,218) is implausibly low for the analyzed area, and is inconsistent with the built-up total (3.36M ha). This is a known aggregation issue with the new 100 m population layer: population is a per-pixel count that must be summed, not averaged, when resampled to the 1 km exposure grid, so it is being undercounted by roughly two orders of magnitude. The percentages below remain valid because the scaling cancels; the absolute people-exposed figures are omitted from the prose pending a fix.

ENSO baseline (resolved). An earlier version of this run used La Nina analog years; it has been re-run on the El Nino baseline (2019, 2023, 2024), which matches NOAA's June 2026 El Nino advisory. Switching from the La Nina to the El Nino baseline lowered the modeled exposure modestly (for example population share from 51% to 40%, urban from 63% to 54%), consistent with El Nino's drier, less cyclone-active signal for eastern Mexico. The figures below are the El Nino run.

§AWhere the exposure concentrates

Two broad clusters dominate. In the north, a large, contiguous dark-purple cluster covers the Sierra Norte and Sierra Nororiental around Huauchinango, Xicotepec, Venustiano Carranza and Metlaltoyuca, extending toward the Veracruz border near Poza Rica. This is the highest-intensity zone in the run and coincides with the most disaster-prone terrain in the state. In the south, flagged cells spread across the Mixteca and Sierra Negra toward Tehuacan and the Oaxaca border, with a further band through the central Serdan valley below La Malinche. The Puebla City metropolitan valley shows moderate (blue) exposure. Read together, the flagged surface traces the state's mountainous, high-rainfall river country, where steep terrain, saturated soils and dense drainage combine.

The model's seasonal diagnostic reinforces the timing: modeled area at risk in the baseline El Nino years rises through a clear mid-year peak across roughly June to September, the core of the Mexican rainy season, and falls away toward winter. That is exactly the window in which the June 2026 events below occurred, so the seasonal shape is consistent with when Puebla actually floods. The observed-water series carries the standard caveat that it lags (JRC monthly history ends about 2021), so the current-year observed line is not a live signal.

§BScenario exposure by asset

Flood exposure fractions are high across the board, which is expected: unlike wildfire, floods concentrate in the valleys and channels where people and infrastructure also concentrate, so a large share of the in-domain assets falls in the flagged surface. Figures are rounded and framed as planning estimates under the stated (La Niña) configuration.

Asset	Share exposed	Expected (in-domain)
Urban area	54.1%	~31,700 ha
Cropland	32.6%	~165,700 ha
Roads	30.6%	~3,870 km
Electric grid	30.5%	~2,139 km
Built-up footprint	30.3%	~1,019,500 ha
Protected areas	17.4%	~60,000 ha
Population (share only)	40.1%	absolute count under review

The standouts are urban area (54%) and the lifeline networks: about a third of the in-domain electric grid and road length fall in the flagged surface. For a flood scenario, that infrastructure share is the operationally important number, since inundated roads and substations drive both the emergency response burden and the recovery cost.

§CTriangulation against Civil Protection and press

Corroboration · current window (June–July 2026)

On 28 June 2026, Puebla City recorded one of its worst floods of the century: more than five hours of rain, over 64 mm in under two hours, drainage collapse, sixteen flood points, water up to 70 cm, preventive evacuations and one death, with the historic center (over the buried San Francisco river) among the worst hit (Quadratin; La Jornada; Puebla Civil Protection PC 044/2026). Earlier, on 10 June 2026, Tropical Storm Boris affected fifteen municipalities, mainly in the Sierra Norte, Nororiental and Negra, causing floods and landslides in Eloxochitlan, Zoquitlan, Chichiquila and Huauchinango (Quadratin), and heavy rain damaged 84 homes in Ciudad Serdan (La Jornada de Oriente). State authorities report maintaining special vigilance over the Sierra Norte, Sierra Nororiental and Mixteca. These are precisely the regions the model flags most strongly.

Historical context

The northern cluster sits over the terrain of Mexico's deadliest modern flood-landslide disaster: in October 1999, extraordinary rains over the Sierra Norte de Puebla killed around 350 people, including roughly 129 in the Teziutlan landslide, and affected 92 municipalities. The model's concentration in this zone is consistent with a well-documented structural susceptibility, not a novel claim.

{ "@id": "nanopub:tri_MX_puebla_flood_2026-07",
  "assertion": { "subject": "model high-risk flood surface, Sierra Norte/Nororiental de Puebla",
    "predicate": "corroboratedBy",
    "object": "TS Boris (10 Jun) + Puebla City flood (28 Jun 2026) + PC vigilance" },
  "provenance": { framing: "consistency, not predictive skill",
    ensoBaselineAsRun: "la_nina", ensoActualState: "El Nino advisory Jun 2026",
    sources: "puebla.gob.mx; jornada.com.mx; quadratin 2026-06/07",
    derivedFrom: "capybara-legacy [PID pending]" },
  "pubInfo": { "identifier": PID pending, "license": "CC-BY-4.0" } }

§DRecommended use

1. Treat this as operational readiness guidance for the rainy season, not a prediction of where water will stand.
2. Prioritize the northern sierra cluster (Huauchinango, Xicotepec, Venustiano Carranza, Teziutlan corridor) for pre-positioning and shelter readiness, given both the model and the historical record.
3. Cross the flagged surface with the road and grid networks, since a third of each is exposed and lifeline disruption drives the response.
4. Retain the earlier La Nina run as a wet-case sensitivity bound alongside this El Nino most-likely scenario.
5. Resolve the population-layer aggregation before any people-exposed figure is quoted externally.
6. Freeze and date this report at issuance so its scenarios can be evaluated against the season.

§EAssumptions & how to read the figures

The model returns a probability-weighted exposure field, not a point prediction. Percentages are the expected share of each in-domain asset within the flagged surface under the stated configuration. They are planning estimates, not counts of confirmed events, and have not been validated against realized outcomes. The absolute population counts in this run are affected by a known aggregation issue (see caveats) and are omitted pending a fix.

§FSources

• Coordinacion Estatal de Proteccion Civil, Puebla. Comunicado PC 044/2026 (flooding, zona metropolitana). puebla.gob.mx
• La Jornada, Puebla City storm and flooding, 28 June 2026. jornada.com.mx
• Quadratin Puebla, flood history and 28 June 2026 event. quadratin.com.mx
• Quadratin Puebla, Tropical Storm Boris, 15 municipalities, 10 June 2026. quadratin.com.mx
• La Jornada de Oriente, tropical-cyclone season vigilance (Sierra Norte, Nororiental, Mixteca). lajornadadeoriente.com.mx
• NOAA / CPC, El Niño Advisory and ENSO discussion, June 2026. noaa.gov
• Gobierno de Puebla, Programa Especial de Proteccion Civil, Temporada de Lluvias y Ciclones 2026. proteccioncivil.puebla.gob.mx

Capybara Center for Risk Analysis · Tecnologico de Monterrey · under the JANUS framework (Feged-Rivadeneira, 2025; archived release doi.org/10.5281/zenodo.21170613).
Legacy flood model run. Probabilistic exposure estimates, not deterministic forecasts of inundation or damage. Triangulation establishes consistency, not predictive skill.