Weather and emergency
A live warning polygon, turned into population at risk, critical facilities, infrastructure and social vulnerability.
01 / The problem
In the first ten minutes of a warning, nobody has time to run a query, and that is exactly when somebody has to tell a county how many people are in the path and which buildings hold the ones who cannot leave.
02 / The decisions it serves
Not a report. A decision somebody has to sign.
Each of these is a question your team already asks, and currently answers with a spreadsheet, a phone call and a fortnight.
Who is in the path
Population inside the actual polygon, not inside a circle drawn near it.
Which facilities matter most
Ranked by who cannot get themselves out. Nursing homes and schools first, always.
Which communities are least able to absorb it
Vulnerability by tract, so the response goes where the need is rather than where the noise is.
What to tell four different rooms
The duty officer, operations, the elected officials and the public all need the same analysis in different words.
03 / It is willing to say no
It does not hand you numbers. It gives you a verdict.
An assistant that only ever agrees with you is a mirror, not an analyst. This one holds a line, names the thing that kills the deal, and will tell you to walk away from a site you already like.
The thresholds it holds you to
-
It reads the actual warning polygon, not a circle drawn around a point. The geography of a tornado warning is the whole point of it.
-
Census tracts scoring above 0.75 on social vulnerability are flagged as high vulnerability.
-
Facilities are ranked by who cannot get themselves out: nursing homes and long-term care first, then schools, then hospitals, then fire and EMS, then government buildings.
-
Three modes, and the language changes with them. A warning says people need to shelter now. A watch says prepare. Post-storm says who was in the impact zone.
-
Renters carry higher displacement risk than owners, and it always says so.
What it will not let you ignore
-
Every number comes from a tool result. It will not invent a population figure under time pressure, which is exactly when a fabricated number does the most damage.
-
If a data layer is missing, it says the gap out loud rather than failing quietly and letting you assume coverage you do not have.
-
If several polygons are on the map, it confirms which one before running. It never assumes.
04 / Why not just ask an AI
A fluent answer and a defensible one are not the same thing.
Every model will answer a location question now, and most of the answers sound right. Ask twice and you get two of them. Ask where the number came from and the room goes quiet.
-
It reads the actual warning polygon. The geography of a tornado warning is the entire point of it, and approximating it with a radius is not a small error.
-
Every number comes from a tool result. It will not invent a population figure under time pressure, which is precisely when a fabricated number does the most damage.
-
If a layer is missing it says the gap out loud, rather than failing quietly and letting you assume a coverage you do not have.
05 / What it runs on, and what you get
The data underneath
Licensed and authoritative, and named in the output, so the number keeps its source when it travels.
- Baron Weather
- Esri GeoEnrichment
- CDC Social Vulnerability Index
- Your critical-facility layers
What you walk out with
An impact summary for the duty officer, a formatted situation report for operations, a plain-language briefing for elected officials, and a public messaging draft. Same analysis, four rooms.
The method it used and the vintage of every number travel with it, so it still holds up when somebody asks you why in six months.
How an answer travels06 / It pairs with
Nobody makes one decision in isolation. Most teams run this alongside two or three of these.
Run Weather and emergency on something you already decided. See if it agrees with you.
The honest test of a method is whether it reaches the conclusion your best analyst already reached, and tells you plainly when it does not.
Twenty minutes, and you can argue with the weights.