EDSEnvironmental Decision System

Observational field note

When the Pressure Fell

An EDS observational field note on successive Atlantic storm passages recorded in Espinho during early 2026.

20 January to 10 March 2026 · local station record · UTC

Supported sequences Peak 6 h fall · hPa
Nils / Oriana-12.911
Kristin-12.288
Leonardo / Marta-10.568
Measured local sea-level-pressure changes. Event windows were named only after telemetry-first screening.
99.256%Occupied minute-bin coverage
71,464Occupied minute bins of 72,000 expected
60 sMedian acquisition interval
3Supported official sequences

Observation window

A continuous local record through a crowded winter.

The study uses the Espinho EDS station record from 20 January through 10 March 2026. A preserved BSON snapshot was restored in a network-isolated, read-only extraction process. The operational database was not contacted.

Minute-bin coverage is the continuity measure. The extraction contains 71,571 telemetry rows, including 107 additional readings inside already occupied minute bins. Those rows are retained and are not presented as extra minutes.

71,571Extracted telemetry rows
71,464 / 72,000Occupied / expected minute bins
9Gaps longer than five minutes

Name-blind screening

The station trace came before the storm name.

Candidate pressure episodes were identified from the station record without using named-event dates. Only after those candidates were fixed were they compared with official IPMA, ANEPC and APA material.

This order reduces the risk of selecting a pressure feature simply because a known storm name makes it look important.

  1. 01 · Aggregate

    Build valid hourly means

    Hours require at least 50 valid sea-level-pressure samples.

  2. 02 · Screen

    Apply frozen pressure thresholds

    Mark falls of at least 6 hPa in 3 hours, 10 hPa in 6 hours or 16 hPa in 12 hours, with an approved low-pressure alternative.

  3. 03 · Align

    Compare with official windows

    Add a name only where an authoritative source supports the overlapping period.

Method boundary: these thresholds are a reproducible event-screening contract. They are not a meteorological storm classification, a forecast model or a national warning threshold.

Event sequence

Five windows. Three supported sequences. Two necessary negatives.

Every approved comparison window remains visible. Supported means a frozen-threshold station episode overlaps an independently documented official sequence. It does not mean EDS measured regional storm intensity.

  1. 01
    Ingrid21 to 24 January
    -8.855 hPaPeak 6 h fall
    Noticeable, below threshold
  2. 02
    Kristin27 to 30 January
    -12.288 hPaPeak 6 h fall
    Supported
  3. 03
    Leonardo / Marta3 to 8 February
    -10.568 hPaPeak 6 h fall
    Supported
  4. 04
    Nils / Oriana10 to 14 February
    -12.911 hPaPeak 6 h fall
    Supported
  5. 05
    March adverse-weather period5 to 7 March, unnamed
    -2.866 hPaPeak 6 h fall
    No major local signature

Supported events

Three different pressure stories.

The aligned figures show raw five-minute traces and hourly means across pressure, relative humidity, air and dew point, and dew-point spread. Each official window is a comparison frame, not an exact claim about local arrival or departure.

Kristin · supported

The cleanest precisely timed example.

The station recorded a -12.288 hPa peak six-hour fall. IPMA independently places Kristin's explosive cyclogenesis between 21:00 UTC on 27 January and 03:00 UTC on 28 January.

Aligned EDS station traces for Storm Kristin from 27 to 30 January 2026, including sea-level pressure, relative humidity, air and dew point, and dew-point spread Enlarge figure
Kristin comparison window. Raw five-minute trace plus hourly mean. Local EDS observations in UTC.

Leonardo / Marta · supported

A sequence, not one isolated pulse.

The 3 to 8 February window contains multiple frozen-threshold episodes and a -10.568 hPa peak six-hour fall. The broad label preserves the official sequence instead of assigning every local pulse to Leonardo alone.

Aligned EDS station traces for the Leonardo and Marta sequence from 3 to 8 February 2026, including sea-level pressure, relative humidity, air and dew point, and dew-point spread Enlarge figure
Leonardo / Marta comparison window. Broad attribution follows the IPMA February bulletin and ANEPC material.

Nils / Oriana · supported

The strongest local transition.

This window records a -12.911 hPa peak six-hour fall and the largest 12-hour fall among all five defined windows.

Aligned EDS station traces for the Nils and Oriana frontal sequence from 10 to 14 February 2026, including sea-level pressure, relative humidity, air and dew point, and dew-point spread Enlarge figure
Nils / Oriana comparison window. The value is local to the Espinho station and is not a national severity measure.

Strongest local transition

A twenty-hectopascal fall in twelve hours.

The Nils / Oriana window contains the largest 12-hour fall among the five approved comparison windows. The magnitude is compelling because its scope is precise: one local EDS station, one frozen analysis method, one defined window.

Peak 12 h pressure fall-20.223hPa · local station record

Strong evidence does not need a larger claim.

This value records a rapid change in hourly mean sea-level pressure at Espinho. It does not measure national storm intensity, wind, rainfall, waves, flooding or damage.

Five normalized EDS station pressure traces shown in cyan against nearby non-overlapping control windows shown in grey Enlarge figure
Hourly sea-level-pressure change from each window start. Controls were selected without inspecting whether their weather looked benign.
Bar chart comparing peak six-hour pressure fall in five event windows with the median of nearby non-overlapping control windows Enlarge figure
Event window in cyan and control median in grey. These comparisons are descriptive, not a forecast-skill score.

Negative evidence

The method also had to say no.

Keeping qualified and negative cases visible prevents the story from becoming a highlight reel. A national warning period and a named depression do not guarantee a major pressure transition at one local station.

Ingrid · below threshold

Noticeable, but not promoted.

-8.855 hPa · 6 h

Ingrid's fall is more negative than every selected control, but it misses the frozen six-hour threshold. The note preserves that result as substantial local movement, not as a supported discovery episode.

5 to 7 March · no major signature

A warning period can remain locally quiet.

-2.866 hPa · 6 h

The official adverse-weather period produced no major local pressure signature. That result is compatible with regional warning context and demonstrates why local observation and national guidance answer different questions.

Claims and limits

Observation is not prediction.

This field note tests whether a preserved local record captures repeatable environmental transitions around independently documented storm sequences. It does not test whether EDS forecast them.

Observed

What the station recorded

  • Minute-level local continuity.
  • Pressure, humidity, temperature and dew-point behavior.
  • Strong and weak local transitions.

Supported

What the comparison shows

  • Three station episodes overlap official sequences.
  • Event-versus-control differences are reproducible under the frozen method.
  • Ingrid and March remain qualified or negative.

Not established

What this note cannot claim

  • Forecast accuracy or warning lead time.
  • Historical advisory or action performance.
  • Wind, rainfall, waves, impacts or national severity.

Data, method and references

The public record is smaller than the research workspace.

These downloads contain approved public-safe metrics, comparisons, integrity summaries and methods. They exclude raw BSON, complete raw telemetry, identifiers, internal logs and private infrastructure data.

Every published file is recorded with SHA-256 identity and provenance in the download manifest.

Event candidate matrix

Five approved windows, station metrics, controls and threshold outcomes.

CSV · public-safe derivativeDownload CSV

Station integrity summary

Daily row counts, minute-bin coverage, cadence and gap measures.

CSV · public-safe derivativeDownload CSV

Approved candidate metrics

Machine-readable core metrics and event results used by this note.

JSON · approved Phase 0 outputDownload JSON

Normalized event and control comparison

Public-safe vector figure for all five approved windows.

SVG · approved Phase 0 figureDownload SVG

Public methodology

Frozen discovery, control-selection and historical-output rules.

Markdown · approved Phase 0 methodDownload method

Official reference timeline

Approved IPMA, ANEPC and APA event-context records.

CSV · official-source inventoryDownload CSV

Privacy and claim statement

Public boundaries for privacy, timing and scientific language.

Markdown · approved Phase 0 reviewDownload statement

Download manifest

SHA-256 identity, source path and provenance for every published artifact.

JSON · publication recordDownload manifest

Official references

Study figure