SevereEnded

Disruption index anomaly in Portugal

🇵🇹 Portugal

Onset: 2026-07 · First detected: 2026-07-20T05:34:46.723615+00:00 · Last observed: 2026-08-10T00:01:05.691434+00:00 · Ended: 2026-08-10T05:55:04.473847+00:00

Generated at: 2026-08-26T00:00:54.752022+00:00 · The radar runs every 6 hours; data may lag by up to 6 hours.

Plain-language brief

What happened, who is affected, what similar past episodes looked like and what to do — every number links to its source.

What happened

ObservedIn 2026-07, the combined disruption picture for Portugal moved well outside its usual range — rated severe — the highest of our three alert levels.[Composite Disruption Index (monthly methodology)]
ObservedThe reading behind this alert: index at 69.3 (the full technical detail is in the sections below).[Composite Disruption Index (monthly methodology)]

Who is affected

ObservedHotels, travel managers, OTAs and tour operators with guests in Portugal during the event window are the most directly affected, together with travellers holding bookings there.[Composite Disruption Index (monthly methodology)]

What similar past episodes looked like

Historical associationIn similar past high-disruption months, overnight stays in Portugal averaged 53.1% below normal calm-month levels — a historical association from our published research panel, not a prediction.[estimates.historical_correlation]
Historical associationLooking at the 3 most similar months in our 2014-2026 research panel, overnight stays one month later differed from a year earlier by a median of +11.4% — again a historical association, not a prediction of this event.[estimates.analogues]

What to do now

RecommendationFor now, keep an eye on the situation in Portugal; no change to bookings is needed yet.[detection.observed]
RecommendationLet affected Portugal bookings move to new dates without penalty while the disruption lasts.[detection.observed][estimates.historical_correlation]

Impact

index	69.31
components	{"aviation":0,"disaster":89.43,"health":0,"rail":0,"weather":0}

Historical monthly context — not current conditionsAs of: 2026-07 · Disruption index: 0.88 · 0 months older than this event

How this was detected

Rule: monthly_index_anomaly

Method: deterministic-rule

Thresholds

index_watch	25
index_severe	50
z_min	2
z_floor_index	10

Observed

month	2026-07
disruption_index	69.31
components	{"aviation":0,"disaster":89.43,"health":0,"rail":0,"weather":0}

Baseline

n_points	137
window_hours	—
mean	33.81
std	21.73

Insufficient baseline — absolute threshold only.

Sources

Composite Disruption Index (monthly methodology)snapshotcollected: 2026-08-24

Corroborating evidence

Corroboration not applicable: live snapshots describe current conditions, not the period of this event.

Estimates

Estimates are labeled model outputs derived from published historical artifacts — they are context, not forecasts.

Model output — historical correlation estimatehistorical-correlation

r_nights	-0.35
r_arrivals	-0.40
best_lag_months	2
best_lag_r	-0.40
baseline_deviation_pct	-53.08
n_months	137

Caveats:

correlation is association, not causation
baseline_deviation_pct is the observed average nights deviation during historical high-disruption months (index > 50) vs calm months
coefficients are dominated by the 2020-2021 pandemic period and may overstate the impact of smaller disruptions
computed on monthly national aggregates; sub-national and short-lived events are diluted

Observed history — analogue episodes, not a predictionanalogue-episodesindex_equiv = 60

Episode	score	index	h0	h1	h2	h3
🇮🇹 Italy · 2023-05	1.627	76.89	16.71	1.95	1.78	-0.24
🇩🇪 Germany · 2022-05	1.588	80.11	250.14	66.78	18.41	3.96
🇫🇷 France · 2025-05	1.581	79	-2.34	11.45	5.56	1.18

median_h1	11.45
median_h2	5.56
median_h3	1.18

Caveats:

analogues are historically similar months, not a forecast of this event
similarity is computed on a 10-feature panel embedding (tourism volumes, disruption/stringency indices, seasonality); it ignores event semantics
the index_equiv mapping from severity to disruption-index points is a disclosed modelling assumption, not an observation
observed_nights_yoy_pct mixes the event effect with everything else that happened in those months

Recommendations

Monitor the situation in Portugal and re-check this report before confirming new bookings.
based on: detection.observed
Enable flexible rebooking for trips to Portugal departing within {days} days.
based on: detection.observed, estimates.historical_correlation

For AI agents

Machine-readable JSON for this page:

/data/api/v1/events/ev-monthly-anomaly-pt-20260701-62a9e0.json

Add as an MCP server (from the repo root):

claude mcp add travel-trends -- .venv/bin/python -m etl.mcp_server

API documentation→