Delta Impact Scale and Delta GEOINT Readiness
Two labels sit on every event in Off-Nadir Delta. The Delta Impact Scale says how much physical harm the reports establish. Delta GEOINT Readiness says whether satellite imagery can check the event. This page states exactly how each is worked out.
Both are Off-Nadir Delta's own. No public body issues, reviews or endorses them. The thresholds inside the Delta Impact Scale are public and are cited one by one below; how they are arranged into levels is ours.
Delta Impact Scale
An event gets the highest level whose threshold a reported figure crosses. Nothing is weighted or added up: one figure crossing one threshold is enough, and a second figure at the same level changes nothing.
| Level | The event reaches it when | Where the threshold comes from |
|---|---|---|
| 0 | A statement, a procedure or a plan. | Our own criterion. |
| 1 | Something physical happened; the reports give no deaths and no damaged structure. | Our own criterion. |
| 2 | 1 or more killed, or a structure or facility damaged. | UCDP Georeferenced Event Dataset (an event is recorded at 1 direct death); Copernicus EMS damage grading (the class “Damaged”). |
| 3 | 10 or more killed; 100 or more injured, displaced or evacuated; or a structure or facility destroyed. | EM-DAT entry criteria (10 deaths, or 100 people affected, injured or homeless); Copernicus EMS damage grading (the class “Destroyed”). |
| 4 | 100 or more killed, or 80,000 or more displaced. | USGS PAGER (Orange: 100–999 estimated fatalities); GDACS alert levels (Orange: more than 100 dead or 80,000 displaced). |
| 5 | 1,000 or more killed, or 800,000 or more displaced. | USGS PAGER (Red: 1,000+ estimated fatalities); GDACS alert levels (Red: more than 1,000 dead or 800,000 displaced). |
- No figure, no level. When the reports give no casualty count and no damaged structure, and it is not known whether anything physical happened, the event is shown as “Not established”. We do not fill in a default.
- Damage counts only for structures. “Destroyed” or “damaged” raises the level only when the reports name a building, a facility or a vessel — not a car or a window.
- Order inside a level is the reported number itself — deaths, then injuries. No weights are involved.
- The figures are read from the reports by an AI model and can be wrong or missing. The level is only as good as that reading; the event detail shows each figure with the report it came from.
Why the level is taken this way. The public scales themselves work like this. USGS PAGER issues one alert from estimated fatalities and one from estimated losses, with “the higher of the two setting the overall alert level”. EM-DAT records a disaster when it meets “at least one” of its entry criteria. GDACS raises a flood alert at a number of dead or a number of displaced. We did not use a weighted formula, because the weights would be ours to choose.
What is ours, stated plainly. Dividing the ladder into six levels is our choice. So is applying fatality thresholds written for earthquakes and floods to attacks, explosions and accidents — we found no public scale that grades a single attack or explosion by severity. Levels 0 and 1 rest on no public threshold.
Delta GEOINT Readiness
The class is looked up from three recorded facts. There is no score and no weighting, and severity plays no part: how serious an event is and whether a satellite can check it are different questions.
- Does the event leave a physical mark a satellite can read?
- How precisely is it located — to a facility or street block, to a town, or more coarsely?
- Is freely available 10 m class imagery enough, or does it need sub-metre imagery?
| Class | Meaning |
|---|---|
| A | Checkable now with free imagery: it leaves a readable mark, the site is known, and 10 m class imagery is enough. |
| B | Checkable with commercial imagery: it leaves a readable mark and the site is known, but it needs sub-metre imagery. |
| C | Checkable once the place is narrowed down: it leaves a readable mark, but it is located only to a town or city. |
| D | Not imageable yet: it leaves a readable mark, but the location is too coarse to point a sensor at. |
| E | Not checkable from orbit: no mark a satellite can read. |
What is public and what is ours. What an image of a given quality lets an analyst do is described by a public scale, the National Imagery Interpretability Rating Scale (NIIRS): the Civil NIIRS Reference Guide, prepared by the IRARS Committee, rates images from 0 to 9 by the tasks they support. Where the imaging recommendation on an event carries a NIIRS level, it is shown with the recommendation. The table above — which combination of facts gives which class — is ours. Whether an event leaves a readable mark is judged by an AI model from the reports.
We use “GEOINT” as defined in U.S. law (10 U.S.C. § 467) and in the National Geospatial-Intelligence Agency's GEOINT Basic Doctrine, Publication 1.0. That is a definition of the discipline. It does not set out a way to score events, and this class is not one of NGA's.
The AI readings shown beside them
The event detail also shows “Severity (AI reading)” and “GEOINT score (AI reading)”, each from 1 to 10. Those are an AI model's reading of the reports against the written scales on the methodology page. They can be higher than what the reported figures establish; the Delta Impact Scale is what the figures establish. Lists, filters and map symbols still use the AI readings.