Skip to content
AlertWatch World

Methodology

How AlertWatch World processes and presents weather alerts

AlertWatch World combines weather alert information from several official sources into a common public map and dashboard.

Those sources do not all use the same fields, warning categories, geographic structures, timestamps, severity systems, or update practices. AlertWatch therefore performs a series of normalization, validation, classification, and publication steps before records are displayed.

This page explains those steps and the limitations that matter when interpreting the site.

AlertWatch World does not independently verify whether a weather hazard is occurring on the ground. The issuing authority remains the authoritative source for its warning.

On this page

1. Collection

AlertWatch currently receives weather alert information through four main production pipelines:

These pipelines are processed by AlertWatch’s automated update system before the resulting records are published for the website.

Visitors’ browsers do not directly query the NWS, ECCC, or WMO alert APIs. The automated AlertWatch workflow retrieves and prepares the information, and the website then downloads AlertWatch’s published data files.

2. Normalization

Each source adapter converts available source fields into a common AlertWatch record structure.

Depending on the source, normalized fields can include:

Not every field exists for every source or warning. Missing information is not automatically inferred merely to make records look uniform.

3. Severity

AlertWatch uses five common display levels:

Extreme · Severe · Moderate · Minor · Unknown

These categories allow different source systems to appear in a common dashboard, but the mapping is source-specific.

NWS

NWS records retain the source-provided severity where available, with an Unknown fallback where the required severity information is unavailable.

WMO/SWIC

WMO severity values are mapped into AlertWatch’s Extreme, Severe, Moderate, Minor, and Unknown display levels.

ECCC land alerts

Canadian land alerts are interpreted from the source’s warning characteristics and colour/category information according to AlertWatch’s ECCC normalization rules.

These rules do not mean that every source uses identical definitions of severity.

Canadian Marine

Canadian Marine warnings are displayed as Unknown severity because the selected source information does not provide an equivalent severity field that AlertWatch considers defensible for the common scale.

Unknown does not mean harmless or low risk. It means AlertWatch does not have a sufficiently comparable severity value for that record.

4. Geography

Alert geometry is handled differently depending on the issuing source.

NWS

Where an NWS alert includes usable source geometry, AlertWatch can use it directly. Where appropriate, affected forecast, coastal, or marine zone references can be resolved against official NWS geographic reference data.

ECCC land

Canadian land alerts use geographic information supplied through the ECCC data source.

WMO/SWIC

Multiple geometries associated with the same CAP warning identity can be combined for display. AlertWatch applies geographic simplification to WMO geometry to reduce publication size and browser rendering cost.

Points and polygons

Point geometry is displayed as a small map marker.

Polygon geometry represents the affected area supplied or derived from the source information. A polygon should not be interpreted as a surveyed or exact physical boundary of the hazard itself.

5. Canadian Marine warnings

Canadian Marine information is processed through a dedicated pipeline rather than being treated as ordinary land alerts.

AlertWatch applies checks for:

Ordinary forecasts are not published as alerts.

Where a warning names a subdivision that AlertWatch cannot resolve reliably, the record can be withheld rather than incorrectly assigning the warning to the entire parent marine area.

Canadian Marine severity remains Unknown for the reason described above.

6. Waterspout classification

AlertWatch includes a Waterspout filter, but a record is not classified as waterspout-related merely because the word “waterspout” appears somewhere in the text.

Classification uses reviewed, source-specific evidence and intentionally conservative rules.

For NWS records, selected structured and textual indicators can qualify a warning.

For WMO records, classification uses constrained English-language threat rules associated with the warning’s context.

A simple incidental mention is insufficient.

Some Canadian marine products may involve waterspout terminology without being assigned the Waterspout classification until AlertWatch has sufficient reviewed evidence for that product type.

Therefore, the Waterspout filter should be understood as a curated classification of qualifying AlertWatch records, not as a complete inventory of all waterspout observations or occurrences worldwide.

7. Duplicate records and grouping

AlertWatch suppresses exact duplicate alert identifiers during publication.

Some source-specific lifecycle information can also be used to handle replaced or cancelled records. For example, the NWS pipeline can recognize explicit update or cancellation references to older records.

This should not be confused with universal event deduplication.

Separate warnings can legitimately remain separate when they apply to different:

The Major Alerts presentation can visually group similar records so that the dashboard is easier to read, but the underlying alert records and geometries remain distinct.

8. Countries Reporting

The Most Alerted Countries ranking uses normalized country labels from records in the current filtered view.

For this ranking, AlertWatch normalizes matters such as case, Unicode representation, and unnecessary whitespace. It does not attempt to redraw political geography or merge territories solely for ranking purposes.

The Countries Reporting summary is calculated separately, counting distinct country codes or, when a code is unavailable, country names in the current filtered view.

Country alert totals represent published AlertWatch records.

They do not measure:

Different warning systems divide hazards into records differently. One authority may issue one broad warning while another may issue multiple warnings for smaller areas.

For that reason, country counts are not a ranking of danger or safety.

9. Latest Alerts

The Latest Alerts panel uses the most defensible available source timestamp for each source type.

Examples include:

If a usable source time is unavailable, AlertWatch can display “Time unavailable” rather than manufacturing a timestamp.

“Latest” refers to this selected publication/update timing. It does not mean AlertWatch independently confirmed that the hazard began most recently.

10. Publication freshness

AlertWatch tracks the generation time of its published datasets.

The freshness indicator describes the age of the AlertWatch publication, not the age of every warning contained within it.

The current display begins indicating aging data after two hours and identifies the dataset as stale after three hours.

A fresh publication can still contain source records with older issue times.

Likewise, a stale publication means AlertWatch has not successfully replaced that publication recently enough; it does not establish whether individual hazards have ended.

11. Open browser tabs

AlertWatch does not currently perform continuous background downloads of the complete alert dataset while a Weather page remains open.

Some clocks or relative-time labels can continue updating in the browser, but those changes do not mean a new alert publication has been downloaded.

Visitors who keep AlertWatch open for long periods should reload the page when current information is important.

12. Source failures and last-good data

AlertWatch’s update system includes safeguards intended to prevent a temporary source failure from replacing a usable publication with incomplete data.

Worldwide / WMO failure

If the WMO portion of an update fails, AlertWatch can retain the previous complete Worldwide publication, including its original publication timestamp, rather than presenting the failed update as new Worldwide data.

This means the newer U.S./Canada direct-source dataset can sometimes coexist with an older Worldwide publication.

Canadian Marine failure

Marine processing can preserve the exact previous valid Marine records in certain failure situations.

These fallback mechanisms improve continuity, but retained records must not be mistaken for newly downloaded information. Publication freshness remains important.

13. Worldwide coverage

The default Weather view combines:

“Worldwide” describes the scope of this combined publication. It does not guarantee complete coverage of every jurisdiction or warning.

WMO notes that warnings can be missing because feeds are unavailable, alerts do not comply with required formats, or technical issues prevent display. Sea-area coverage may also be incomplete.

WMO/SWIC coverage and information notes

14. WMO record lifecycle and expiry limitations

International warning systems vary in how lifecycle and expiry information is supplied and interpreted.

AlertWatch currently uses WMO’s effective-warning data as part of its international publication, but it does not apply one universal expiry-removal rule to every underlying WMO record.

As a result, some published WMO records may contain expiry timestamps that precede the time at which the AlertWatch dataset was generated, or may lack an expiry time altogether.

An expired timestamp by itself does not allow AlertWatch to determine with certainty that the underlying hazard ended because lifecycle practices vary among issuing authorities and feeds.

For this reason, AlertWatch describes these as:

published or currently listed alert records

rather than claiming they have all been independently verified as ongoing hazards.

Terms such as “live,” “active,” or “right now” on the interface should therefore be understood in the context of the source publication, not as independent confirmation of current conditions.

15. Global Alert Activity

The Global Alert Activity chart records the number of records in successful, validated Worldwide publications over time.

It currently maintains a rolling seven-day history, with at most one stored observation for each UTC hour. If more than one successful publication occurs during an hour, a later observation can replace an earlier observation for that hour.

Missing observations are left as gaps rather than being converted to zero.

The chart is independent of:

It represents the number of published alert records in AlertWatch’s Worldwide dataset, not:

16. Historical retention

The Global Alert Activity file maintains a bounded rolling history for the public chart.

That does not mean older generated data can never exist elsewhere, such as repository version history or development records.

The public chart should therefore be understood as a rolling visualization, not as the complete historical archive of every warning AlertWatch has ever processed.

17. Why AlertWatch totals can differ from other websites

Alert totals can differ between AlertWatch, an issuing authority’s own website, WMO/SWIC, or another third-party service for legitimate reasons.

Differences can result from:

A difference in totals does not by itself establish that either service is incorrect.

For urgent decisions, consult the original issuing authority.

18. Validation

Before a new publication is accepted, AlertWatch performs automated checks intended to catch structural or data-quality problems.

These checks support publication consistency and help prevent obviously incomplete data from replacing valid data.

They do not independently verify:

The issuing authority remains responsible for its warning.

19. Map limitations

AlertWatch’s map is a visualization of warning information.

Displayed areas can represent:

These should not be interpreted as exact hazard footprints.

Weather phenomena can extend beyond a polygon, affect only part of a displayed area, or change after the AlertWatch publication was generated.

Map boundaries and place names are provided for geographic context and do not imply an AlertWatch position regarding legal or political boundaries.

20. What AlertWatch does not do

AlertWatch World does not:

AlertWatch’s purpose is to make available official warning information easier to discover, compare, and understand.

21. Corrections and methodology changes

AlertWatch’s processing rules may be refined as source systems change, new edge cases are identified, or more reliable methods become available.

Changes may affect:

Material changes to the public methodology should be reflected on this page.

The underlying issuing authorities remain responsible for their original warning information.

22. Using AlertWatch safely

For general awareness, AlertWatch can provide a broad view of official weather-alert activity.

For decisions involving personal safety, travel, evacuation, marine activity, severe weather preparation, or emergency response, consult and follow the instructions of the applicable official authority.

The absence of a displayed alert does not prove that an area is safe or that AlertWatch World has complete coverage.