NOAA Weather Alert Monitor

What does NOAA Weather Alert Monitor do?

NOAA Weather Alert Monitor queries the official National Weather Service API (api.weather.gov) and turns raw alerts into scored, ranked decisions. Most weather-alert tools hand you the CAP feed and leave you to read it; this one scores every alert on the standard severity × urgency × certainty matrix, maps the NWS response field to an operational recommendedAction (take-shelter, evacuate, avoid-area, monitor), flags which alerts actually need attention, and returns a run-level summary of what's happening across your area — so an automation or dashboard can branch on a number and an enum instead of parsing prose.

It operates in two distinct modes. Active mode (the default) fetches all currently live warnings, watches, and advisories in real time. Historical mode searches recent past alerts within a specified date range — useful for post-event analysis, insurance risk assessment, and climate research. Set a watchlist name to turn on cross-run monitoring: each run is compared to the last, flagging new, escalated, downgraded, cancelled, and resolved alerts for a location. Each alert record carries the full CAP payload — headline, safety instructions, affected areas, SAME and UGC geocodes, issuing NWS office, timing fields — plus the decision layer on top. The NWS API is free, open, and requires no API key.

Why use NOAA Weather Alert Monitor on Apify?

• No API key or authentication required -- the NWS API is completely free and open. This actor handles all communication, required headers, pagination, and data transformation automatically.
• Structured, normalized output -- raw NOAA GeoJSON responses with deeply nested properties are transformed into clean, flat JSON records ready for processing, filtering, and export.
• Automatic cursor-based pagination -- the actor follows NWS pagination links to retrieve complete result sets, handling up to 5,000 alerts across up to 20 API pages per run.
• Scheduled monitoring -- run on a recurring Apify schedule (every 15 minutes, hourly, daily) to build a continuous weather alert feed with zero infrastructure.
• Cloud execution with integrations -- no servers or cron jobs needed. Built-in REST API access, webhook triggers, and integrations to Slack, Google Sheets, Zapier, and more.

Key features

• Deterministic priority scoring -- every alert gets a 0-100 alertPriority from the standard CAP matrix (severity 50%, urgency 30%, certainty 20%) plus a priorityLevel band (critical / high / moderate / low / informational). No LLM, no randomness — the same alert always scores the same, with a priorityFactors breakdown showing how.
• Operational recommended action -- recommendedAction maps the NWS response field to a routing enum (take-shelter, evacuate, avoid-area, prepare, monitor, all-clear, no-action) and actionRequired is a one-field automation gate.
• Timing intelligence -- timing.status (active / upcoming / expired), minutes-until-onset, minutes-until-expiry, and an expiringSoon flag, so you can tell what's happening now from what starts in six hours.
• Cross-run monitoring -- set a watchlistName and each run is diffed against the last: changeFlag (NEW / UPDATED / ESCALATED / DOWNGRADED / CANCELLED / UNCHANGED), priority delta, and a resolved-alert list when warnings clear.
• Run-level summary record -- one recordType: "summary" record with the headline, critical count, action-required count, the highest-priority alert, breakdowns by level/severity/event/category, and the next alert to expire (also pinned to the SUMMARY key-value record).
• Dual-mode operation -- active alerts for real-time monitoring, or historical search with date ranges for research and trend analysis.
• Three-dimensional NWS filtering -- filter by severity, urgency, and certainty at the API, plus a post-scoring minPriority worklist filter and an outputProfile (minimal / standard / full).
• State and territory filtering -- any US state or territory by two-letter code (CA, TX, FL, PR, GU, etc.) or empty for nationwide.
• SAME and UGC geocodes -- FIPS county codes (SAME) and NWS zone codes (UGC) for county-level geographic targeting.
• Up to 5,000 alerts per run -- configurable maxResults cap with automatic cursor pagination.

How to use NOAA Weather Alert Monitor

1. Go to the NOAA Weather Alert Monitor page on Apify and click Start.
2. Choose your mode: leave Active Alerts Only checked for live monitoring, or uncheck it for historical search.
3. Enter a State/Territory Code (e.g., TX for Texas) or leave empty for all US alerts.
4. Optionally set Severity, Urgency, Certainty, or Event Type filters.
5. For historical mode, set Start Date and End Date in YYYY-MM-DD format.
6. Set Max Results (default 100, up to 5,000).
7. Click Start and wait for results (typically 5--30 seconds).
8. Download results from the Dataset tab in JSON, CSV, Excel, or XML format.
9. Optionally, set up a recurring schedule (e.g., every 15 minutes) for continuous monitoring.
10. Access results programmatically via the Apify API or connect integrations (Slack, webhooks, Google Sheets, Zapier).

Input parameters

Parameter	Type	Default	Description
activeOnly	Boolean	true	When checked, returns only currently active alerts via /alerts/active. Uncheck for historical search with date range filters via /alerts.
area	String	"CA" (prefill)	Two-letter US state or territory code (e.g., CA, TX, FL, NY, PR). Automatically uppercased. Leave empty for all US alerts.
severity	Select	--	Filter by severity: Extreme, Severe, Moderate, Minor, or Unknown.
urgency	Select	--	Filter by urgency: Immediate, Expected, Future, Past, or Unknown.
certainty	Select	--	Filter by certainty: Observed, Likely, Possible, Unlikely, or Unknown.
eventType	String	--	Specific NWS event type (e.g., Tornado Warning, Flash Flood Warning, Winter Storm Watch). Must match NWS naming exactly.
startDate	String	--	Start date in YYYY-MM-DD format. Only used when activeOnly is unchecked (historical mode). Converted to T00:00:00Z.
endDate	String	--	End date in YYYY-MM-DD format. Only used when activeOnly is unchecked (historical mode). Converted to T23:59:59Z.
maxResults	Integer	100	Maximum number of alerts to return (1--5,000).
minPriority	Integer	--	Only output alerts scoring at or above this alertPriority (0--100). The run summary still counts every alert; suppressed alerts are logged.
watchlistName	String	--	Set a name to turn on cross-run monitoring for a location/query. Each run is compared to the last under this name. Leave empty for a one-off run.
outputProfile	Select	standard	minimal (decision fields only), standard (decisions + core alert fields), or full (every field).

Input examples

Active severe alerts in Texas:

{
    "activeOnly": true,
    "area": "TX",
    "severity": "Severe",
    "maxResults": 50
}

Historical tornado warnings with date range:

{
    "activeOnly": false,
    "area": "OK",
    "eventType": "Tornado Warning",
    "startDate": "2025-04-01",
    "endDate": "2025-04-30",
    "maxResults": 500
}

Extreme urgency with observed certainty:

{
    "activeOnly": true,
    "area": "FL",
    "severity": "Extreme",
    "urgency": "Immediate",
    "certainty": "Observed",
    "maxResults": 100
}

All active alerts nationwide (no filters):

{
    "activeOnly": true,
    "maxResults": 1000
}

Monitor Florida for high-priority alerts (cross-run change detection):

{
    "activeOnly": true,
    "area": "FL",
    "minPriority": 60,
    "watchlistName": "florida-ops",
    "outputProfile": "minimal"
}

Run this on a schedule under the same watchlistName and each run flags what is NEW, ESCALATED, or RESOLVED since the last one — only emitting alerts scoring 60+.

Tips for best results

• Active mode is best for monitoring -- it returns only currently live alerts and responds faster since the NWS active endpoint is optimized for real-time queries.
• Date filters are ignored in active mode -- startDate and endDate only apply when activeOnly is unchecked. This is by design of the NWS API.
• Event type strings must match exactly -- use official NWS event names like Tornado Warning, Flood Watch, or Winter Storm Warning. Partial matches and case variations do not work.
• State codes are case-insensitive -- the actor automatically uppercases the area parameter, so ca, Ca, and CA all work identically.
• Combine all three filter dimensions for the most targeted monitoring. Setting severity to Extreme, urgency to Immediate, and certainty to Observed captures only confirmed, life-threatening events in progress.
• Schedule runs every 15--30 minutes for continuous monitoring. Connect a webhook or Slack integration to receive notifications when new severe alerts appear.

Output example

Each alert is output as a flat JSON object combining the raw CAP fields with the decision layer. Here is a realistic tornado warning (standard profile):

{
    "recordType": "alert",
    "schemaVersion": "2.0.0",
    "alertPriority": 95,
    "priorityLevel": "critical",
    "recommendedAction": "take-shelter",
    "actionRequired": true,
    "messageStatus": "new",
    "isCancellation": false,
    "timing": {
        "status": "active",
        "minutesUntilOnset": null,
        "minutesUntilExpiry": 37,
        "durationHours": 0.6,
        "expiringSoon": true
    },
    "priorityFactors": {
        "severity": { "value": "Extreme", "contribution": 50 },
        "urgency": { "value": "Immediate", "contribution": 30 },
        "certainty": { "value": "Observed", "contribution": 20 }
    },
    "whyItMatters": "Tornado Warning (Extreme/Immediate) for Cleveland, OK. In effect for ~37 min more.",
    "summary": "CRITICAL: Tornado Warning for Cleveland, OK. Recommended: take-shelter.",
    "alertId": "urn:oid:2.49.0.1.840.0.2025.04.15.14.23.0045",
    "event": "Tornado Warning",
    "headline": "Tornado Warning issued April 15 at 2:23PM CDT by NWS Norman OK",
    "severity": "Extreme",
    "urgency": "Immediate",
    "certainty": "Observed",
    "category": "Met",
    "status": "Actual",
    "response": "Shelter",
    "areaDesc": "Cleveland, OK; McClain, OK",
    "senderName": "NWS Norman OK",
    "sent": "2025-04-15T14:23:00-05:00",
    "effective": "2025-04-15T14:23:00-05:00",
    "onset": "2025-04-15T14:23:00-05:00",
    "expires": "2025-04-15T15:00:00-05:00",
    "ends": "2025-04-15T15:00:00-05:00",
    "sameCodes": "040027|040087",
    "ugcCodes": "OKC025|OKC087",
    "nwsHeadline": "TORNADO WARNING IN EFFECT UNTIL 300 PM CDT",
    "changeFlag": null,
    "extractedAt": "2025-04-15T19:24:12.345Z"
}

When watchlistName is set, changeFlag, previousPriority, priorityDelta, and runsSeen are populated instead of null. Every run also emits one recordType: "summary" record (and, in watchlist mode, a recordType: "watchlist-summary" record).

Output fields reference

Decision layer (added by this actor on top of the raw CAP fields):

Field	Type	Description
recordType	String	alert, summary, watchlist-summary, no-results, or error.
alertPriority	Number	0--100 deterministic CAP priority (severity 50% + urgency 30% + certainty 20%). Sort key.
priorityLevel	String	Band: critical, high, moderate, low, informational.
recommendedAction	String	take-shelter, evacuate, avoid-area, prepare, execute-instructions, assess, monitor, all-clear, no-action.
actionRequired	Boolean	True when critical/high, in effect, and not cancelled. Automation gate.
messageStatus	String	new, update, cancellation, acknowledgement, error (from CAP messageType).
isCancellation	Boolean	True when the message withdraws a prior alert.
timing	Object	{ status, minutesUntilOnset, minutesUntilExpiry, durationHours, expiringSoon }.
priorityFactors	Object	Per-axis contribution to the score (severity, urgency, certainty).
priorityReason	String[]	Plain-English reasons behind the score (dropped in standard/minimal profiles).
whyItMatters	String	One-line plain-English context (event, severity/urgency, area, timing).
summary	String	LLM-friendly one-line decision summary.
changeFlag	String/null	Watchlist mode: NEW, UPDATED, ESCALATED, DOWNGRADED, CANCELLED, UNCHANGED. Null when off.
previousPriority	Number/null	Priority on the prior run (watchlist mode).
priorityDelta	Number/null	Change in priority since the last run (watchlist mode).
runsSeen	Number/null	Runs this alert has appeared in (watchlist mode).

Raw CAP fields (from the NWS alert):

Field	Type	Description
alertId	String	Unique NWS alert identifier in URN OID format.
event	String	NWS event type name (e.g., Tornado Warning, Flood Watch).
headline	String/null	Human-readable headline with event type, date/time, and issuing office.
severity	String	Severity level: Extreme, Severe, Moderate, Minor, or Unknown.
urgency	String	Urgency level: Immediate, Expected, Future, Past, or Unknown.
certainty	String	Certainty level: Observed, Likely, Possible, Unlikely, or Unknown.
category	String	CAP alert category (e.g., Met for meteorological, Geo for geophysical).
status	String	Alert status: Actual, Exercise, System, Test, or Draft.
messageType	String	Message type: Alert, Update, or Cancel.
response	String	Recommended action: Shelter, Evacuate, Prepare, Monitor, AllClear, or None.
areaDesc	String	Human-readable description of affected areas (counties, zones, cities).
senderName	String	Name of the issuing NWS forecast office (e.g., NWS Norman OK).
sender	String	Email address of the sending NWS office.
sent	String	ISO 8601 timestamp when the alert was sent.
effective	String	ISO 8601 timestamp when the alert becomes effective.
onset	String/null	ISO 8601 timestamp of expected event onset. Null if not specified.
expires	String	ISO 8601 timestamp when the alert expires.
ends	String/null	ISO 8601 timestamp when the event is expected to end. Null if not specified.
description	String	Full NWS hazard description with event details, source, and expected impact.
instruction	String/null	Official safety instructions and recommended protective actions. Null if not provided.
sameCodes	String	SAME (Specific Area Message Encoding) FIPS county codes for EAS broadcast targeting, pipe-separated.
ugcCodes	String	UGC (Universal Geographic Code) NWS zone/county identifiers, pipe-separated.
nwsHeadline	String/null	Condensed NWS headline extracted from nested alert parameters. Null if not present.
extractedAt	String	ISO 8601 timestamp when this record was extracted by the actor.

How it works

Input Config
    |
    v
Select Endpoint ──> activeOnly=true?  ──yes──> /alerts/active
    |                                   ──no──> /alerts (+ date params)
    v
Build Query Params (area, severity, urgency, certainty, event)
    |
    v
Fetch Page ──> User-Agent + Accept: application/geo+json
    |
    v
Parse GeoJSON Features ──> pagination.next exists? ──yes──> Follow cursor URL
    |                       (cap: 20 pages)          ──no──> Continue
    v
Transform Each Feature ──> Flatten properties, extract geocodes + NWSheadline
    |
    v
Push to Apify Dataset + Log Summary (event types, severity counts, sources)

Dual-endpoint architecture

The NWS API exposes two separate alert endpoints. The actor selects between them based on the activeOnly flag:

• /alerts/active (default) -- returns only currently live alerts. This endpoint is faster and returns smaller payloads because expired alerts are excluded server-side.
• /alerts -- returns both active and historical alerts. This endpoint accepts start and end date query parameters for time-range searches.

When activeOnly is true (or unset), the actor uses the active endpoint. Setting activeOnly to false switches to the full alerts endpoint with date filter support.

Date filter restriction

The startDate and endDate parameters are only appended to the request URL when activeOnly is explicitly false. This matches the NWS API design: the /alerts/active endpoint does not accept date parameters since all returned alerts are, by definition, currently active. Date values are formatted with time suffixes -- T00:00:00Z for start and T23:59:59Z for end -- to cover complete days in UTC.

Cursor-based pagination

Unlike page-number pagination, the NWS API uses cursor-based pagination. Each API response may include a pagination.next URL pointing to the next batch of results. The actor follows these URLs sequentially until one of three conditions is met: no next URL is present, the maxResults cap is reached, or a safety limit of 20 pages is hit. This approach is more reliable than offset-based pagination because it handles data that changes between requests.

NWS headline extraction

NWS alerts contain a nested parameters object with additional metadata. The NWSheadline field (accessed at parameters.NWSheadline[0]) provides a condensed, uppercase headline that is distinct from the main headline field. The actor extracts this into the nwsHeadline output field. This headline is typically shorter and more direct, making it useful for notification subject lines and dashboard displays.

SAME and UGC geocode extraction

Each alert includes a geocode object with two code systems. SAME codes (Specific Area Message Encoding) are six-digit FIPS county codes used by the Emergency Alert System for broadcast targeting on TV, radio, and wireless devices. UGC codes (Universal Geographic Code) are NWS-specific zone and county identifiers used in forecast operations. The actor extracts both arrays from the nested geocode object, defaulting to empty arrays when codes are absent.

User-Agent requirement

The NWS API requires a valid User-Agent header on all requests. Requests without this header are rejected with a 403 error. The actor sends apify-noaa-weather-alerts/1.0 ([email protected]) to comply with this requirement, along with an Accept: application/geo+json header to receive GeoJSON-formatted responses.

Decision layer

After each alert is fetched, the actor scores it deterministically — no LLM, no randomness — from the CAP fields the NWS already publishes:

• alertPriority (0--100) = severity (50%) + urgency (30%) + certainty (20%), each from a fixed weight table, with the per-axis contribution exposed in priorityFactors. priorityLevel bands it into critical / high / moderate / low / informational.
• recommendedAction maps the CAP response field (Shelter, Evacuate, Avoid, Prepare, Monitor, AllClear) to a stable routing enum; actionRequired is true only for in-effect critical/high alerts that have not been cancelled.
• timing parses onset/effective and expires/ends against the current time into a status (active / upcoming / expired), minutes-until, and an expiringSoon flag.

Alerts are streamed highest-priority first.

Run summary record

After processing all alerts, the actor emits one recordType: "summary" record (also pinned to the SUMMARY key-value record) containing:

• headline -- a one-line read of the run (e.g. "2 critical alerts for TX — Tornado Warning highest priority").
• criticalCount / actionRequiredCount and breakdowns byPriorityLevel, bySeverity, byEvent (top 15), byCategory.
• topAlert -- the highest-priority alert, and nextToExpire -- the active alert closest to expiry.

How much does it cost to run?

The NWS API is completely free with no usage fees. The only cost is Apify platform compute time. Since the actor makes lightweight HTTP requests with no browser rendering, runs are very fast and inexpensive.

Scenario	Estimated alerts	Estimated time	Estimated cost
Active alerts for one state	5--100	~5 seconds	~$0.001
Active alerts nationwide	200--2,000	~15 seconds	~$0.005
Historical search with date range	500--5,000	~30 seconds	~$0.01

Costs are based on Apify platform usage at standard rates with 256 MB memory. Actual costs depend on alert volume and pagination depth. Scheduled monitoring (e.g., every 15 minutes, 24/7) costs approximately $0.10/day.

Limitations and responsible use

• US-only coverage -- the NWS API covers the 50 US states, DC, and US territories (PR, GU, VI, AS, MP). International weather alerts are not available.
• Historical depth is limited -- the NWS API retains historical alerts for a limited window (typically days to weeks). It is not a long-term archive.
• Event type exact match -- the eventType filter requires the exact NWS event name string. Typos, partial names, or case variations return no results.
• Pagination safety limit -- the actor follows up to 20 pages of cursor-based pagination. Extremely broad queries exceeding this are truncated.
• No geometry data -- the actor extracts alert properties but does not include GeoJSON polygon coordinates. Use the NWS API directly if you need geographic boundaries.
• Update handling -- an updated alert (messageType: Update) carries messageStatus: "update"; in watchlist mode it is flagged changeFlag: "UPDATED". Within a single run the actor does not merge an update onto its original — both records appear.
• Not a substitute for official warnings -- this is a data and decision-support tool for monitoring and research. The recommendedAction is operational routing derived from the NWS response field, not life-safety advice. For life-safety decisions, always rely on official NWS alerts at weather.gov.
• API availability -- the NWS API may experience degraded performance during severe weather events when demand is highest. The actor retries transient failures with backoff and, if the API stays unavailable, emits a typed error record and finishes successfully rather than crashing the run.
• Date precision -- startDate and endDate are converted to full ISO 8601 timestamps (T00:00:00Z and T23:59:59Z). Sub-day time precision is not supported through the actor input.

NWS alert severity levels

Level	Description
Extreme	Extraordinary threat to life or property
Severe	Significant threat to life or property
Moderate	Possible threat to life or property
Minor	Minimal or no known threat to life or property
Unknown	Severity is unknown at the time of issuance

NWS alert urgency levels

Level	Description
Immediate	Responsive action should be taken immediately
Expected	Responsive action should be taken within the next hour
Future	Responsive action should be taken in the near future
Past	Responsive action is no longer required
Unknown	Urgency is not determined or not applicable

NWS alert certainty levels

Level	Description
Observed	Event has been observed or confirmed (e.g., tornado sighted)
Likely	Probability greater than 50% the event will occur
Possible	Probability between 25% and 50% the event will occur
Unlikely	Probability less than 25% the event will occur
Unknown	Certainty is not determined or not applicable

API integration examples

Use the Apify API to run the actor programmatically and retrieve results in your applications, pipelines, or automated workflows.

Python:

from apify_client import ApifyClient

client = ApifyClient("YOUR_API_TOKEN")

run = client.actor("ryanclinton/noaa-weather-alerts").call(run_input={
    "activeOnly": True,
    "area": "TX",
    "severity": "Severe",
    "maxResults": 50,
})

for item in client.dataset(run["defaultDatasetId"]).iterate_items():
    print(f"{item['event']}: {item['headline']}")
    print(f"  Severity: {item['severity']} | Urgency: {item['urgency']}")
    print(f"  Areas: {item['areaDesc']}")
    print()

JavaScript:

import { ApifyClient } from "apify-client";

const client = new ApifyClient({ token: "YOUR_API_TOKEN" });

const run = await client.actor("ryanclinton/noaa-weather-alerts").call({
    activeOnly: true,
    area: "FL",
    severity: "Extreme",
    urgency: "Immediate",
    maxResults: 100,
});

const { items } = await client.dataset(run.defaultDatasetId).listItems();
for (const alert of items) {
    console.log(`${alert.event}: ${alert.headline}`);
    console.log(`  Instruction: ${alert.instruction}`);
}

cURL:

# Start the actor run
curl -X POST "https://api.apify.com/v2/acts/ryanclinton~noaa-weather-alerts/runs?token=YOUR_API_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{
    "activeOnly": false,
    "area": "CA",
    "eventType": "Flood Warning",
    "startDate": "2025-01-01",
    "endDate": "2025-01-31",
    "maxResults": 200
  }'

# Retrieve results (replace DATASET_ID from the run response)
curl "https://api.apify.com/v2/datasets/DATASET_ID/items?token=YOUR_API_TOKEN&format=json"

FAQ

What types of weather alerts does this actor return? Every alert type issued by the National Weather Service, including Tornado Warnings, Severe Thunderstorm Warnings, Flash Flood Warnings, Flood Watches, Winter Storm Warnings, Blizzard Warnings, Ice Storm Warnings, Hurricane Warnings, Tropical Storm Warnings, Heat Advisories, Excessive Heat Warnings, High Wind Warnings, Wind Advisories, Fire Weather Watches, Red Flag Warnings, Dense Fog Advisories, Freeze Warnings, Frost Advisories, Tsunami Warnings, and dozens more. Any alert published through the NOAA Weather API is captured.

What US states and territories are supported? All 50 US states, Washington DC, and US territories including Puerto Rico (PR), Guam (GU), US Virgin Islands (VI), American Samoa (AS), and Northern Mariana Islands (MP). Use standard two-letter postal codes.

Can I search for multiple states at once? The NWS API area parameter accepts a single state code per request. To monitor multiple states, leave the area field empty to search all US alerts nationwide, or run the actor separately for each state.

Why do my date filters have no effect? Date filters (startDate and endDate) only work in historical mode. Make sure activeOnly is set to false. When activeOnly is true (the default), the actor uses the /alerts/active endpoint which does not accept date parameters.

How often are NWS alerts updated? The NWS updates alerts in near real-time. New alerts are typically available within seconds of issuance. For continuous monitoring, schedule the actor to run every 15--30 minutes.

What are SAME codes and UGC codes? SAME codes are six-digit FIPS county identifiers used by the Emergency Alert System (EAS) for broadcast targeting on TV, radio, and wireless devices. UGC codes are NWS-specific zone and county codes used in forecast products. Both are useful for programmatic geographic matching and integration with emergency notification systems.

How do I set up continuous weather monitoring? On the Apify platform, create a scheduled task that runs this actor at your desired interval (e.g., every 15 minutes). Configure a webhook or integration (Slack, email, Zapier, Google Sheets) to receive or store the output dataset after each run. This provides automated, continuous weather alert monitoring without any server infrastructure.

Use in Dify

Drop this actor into Dify workflows via the Apify plugin's Run Actor node. Each alert returns scored, classified, and recommended as structured JSON — critical / high / moderate / low plus the recommendedAction and actionRequired your downstream node branches on. A generic weather-alert scraper pointed at the same NWS feed returns raw CAP JSON; this returns decisions.

• Actor ID: ryanclinton/noaa-weather-alerts
• Sample input (monitor a state for high-priority alerts, only surface what changed):

{
    "activeOnly": true,
    "area": "FL",
    "minPriority": 60,
    "watchlistName": "florida-ops",
    "outputProfile": "minimal"
}

A Dify if/else node routes cleanly on the stable enums — no prose parsing:

• recommendedAction == "take-shelter" or "evacuate" → page the on-call / fire the emergency SMS branch
• actionRequired == true → escalate to a human; false → log only
• priorityLevel == "critical" → high-urgency channel; "moderate" / "low" → digest channel
• changeFlag == "NEW" or "ESCALATED" (watchlist mode) → alert; "UNCHANGED" → suppress, so a scheduled run only fires when something actually changed

Opt-in modes Dify workflows can lean on: set watchlistName to turn each scheduled run into a change feed (new / escalated / resolved alerts), and minPriority to keep only alerts worth acting on. The recommendedAction enum and the run-level summary record's headline are usable verbatim in a notification — no LLM rewriting needed.

Related actors

Actor	Description	Link
FEMA Disaster Declaration Search	Search FEMA disaster declarations by state, type, and date range	ryanclinton/fema-disaster-search
GDACS Disaster Alerts	Monitor global disaster alerts from earthquakes, floods, cyclones, and volcanoes	ryanclinton/gdacs-disaster-alerts
UK Environment Agency Flood Warnings	Search active flood warnings and alerts across England	ryanclinton/uk-flood-warnings
Weather Forecast Search	Get detailed weather forecasts for any global location	ryanclinton/weather-forecast-search
USGS Earthquake Search	Search recent and historical earthquake data from USGS	ryanclinton/usgs-earthquake-search
Website Change Monitor	Monitor any weather-related web page for content changes	ryanclinton/website-change-monitor