SevereEnded

Weather warning spike in Latvia

🇱🇻 Latvia

Onset: 2026-07-07 · First detected: 2026-07-07T18:00:53.173423+00:00 · Last observed: 2026-07-07T18:00:53.173423+00:00 · Ended: 2026-07-08T06:00:48.021482+00:00

Generated at: 2026-07-12T00:00:51.613701+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

ObservedOn 2026-07-07, our automated monitoring flagged an unusual spike in official weather warnings for Latvia — rated severe — the highest of our three alert levels.[Meteoalarm EUMETNET]
ObservedThe reading behind this alert: active official weather warnings at 210 (the full technical detail is in the sections below).[Meteoalarm EUMETNET]

Who is affected

ObservedHotels, travel managers, OTAs and tour operators with guests in Latvia during the event window are the most directly affected, together with travellers holding bookings there.[Meteoalarm EUMETNET]

What similar past episodes looked like

Historical associationIn similar past high-disruption months, overnight stays in Latvia averaged 71% 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 +70.8% — 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 Latvia; no change to bookings is needed yet.[detection.observed]
RecommendationContact guests and travellers with upcoming stays in Latvia before they contact you, and point them to the official sources linked in this report.[detection.observed]

Impact

warning_count	210
max_severity	Extreme

Historical monthly context — not current conditionsAs of: 2026-05 · Disruption index: 0 · 2 months older than this event

How this was detected

Rule: weather_warning_spike

Method: deterministic-rule

Thresholds

abs_min_warnings	150
severe_count	400
z_min	2
z_floor_warnings	25
min_baseline_points	12
min_baseline_span_hours	48

Observed

warning_count	210
max_severity	Extreme

Baseline

n_points	32
window_hours	523.63
mean	97.75
std	67.14

Sources

Meteoalarm EUMETNET snapshotcollected: 2026-07-07T18:00:53.173423+00:00

Corroborating evidence

Independent live sources scanned for disruption-significant entries about this country. Only strict, disclosed rules count — routine entries are never counted.

minbuza-nl-travel1 matching entriesSnapshot
Rule: count the event country's Dutch MinBuZa advisory only when its colour is orange or red; green/yellow advisories are routine
- Letland — orange

Sources scanned: auswaertiges-amt-api-de, emsc-seismic, france-diplomatie-travel, gdacs-live, mae-ro-travel, minbuza-nl-travel, uk-fcdo-advisories, us-state-advisories, viaggia-sicuri-it

Estimates

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

Model output — historical correlation estimatehistorical-correlation

r_nights	-0.46
r_arrivals	-0.46
best_lag_months	0
best_lag_r	-0.46
baseline_deviation_pct	-71.01
n_months	136

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
🇫🇷 France · 2023-04	1.525	83.11	7.20	9.43	-0.13	-2.01
🇵🇱 Poland · 2022-04	1.522	83.48	325.47	132.20	39.64	6.04
🇨🇾 Cyprus · 2015-04	1.505	82.35	—	—	—	—

median_h1	70.81
median_h2	19.75
median_h3	2.02

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 Latvia and re-check this report before confirming new bookings.
based on: detection.observed
Proactively inform guests with upcoming stays in Latvia about possible disruptions.
based on: detection.observed
Enable flexible rebooking for trips to Latvia departing within {days} days.
based on: detection.observed, estimates.historical_correlation
Shift promoted inventory away from Latvia toward unaffected destinations until {month}.
based on: detection.observed

For AI agents

Machine-readable JSON for this page:

/data/api/v1/events/ev-weather-spike-lv-20260707-6b749d.json

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

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

API documentation→