How DPS Improves on the Saffir-Simpson Scale

The Saffir-Simpson Hurricane Wind Scale was published in 1971. It rates a storm on a single variable: peak one-minute sustained wind. That made sense in 1971. It does not make sense in 2026, when satellite altimetry, scatterometer winds, ocean heat content, and integrated kinetic energy are all measured in real time. The Destructive Power Score (DPS) is a 0–100 hurricane rating that uses all of them.

Why the Category scale falls short

Saffir-Simpson groups every hurricane into five categories based on peak wind alone. Two storms with identical peak winds can have radically different destructive footprints, and the scale cannot tell them apart.

Hurricane Ike (2008) made U.S. landfall as a Category 2. It killed 195 people and caused $38 billion in damage — more than several Category 4 and 5 storms in the same era. Ike's destructive footprint was not driven by wind speed; it was driven by an unusually large wind field that produced a Category 4–equivalent storm surge across the entire upper Texas coast. Saffir-Simpson called it a 2 and moved on.

The opposite problem also occurs. A small, intense Category 5 over open ocean — say a recurving Western Pacific super-typhoon that never crosses land — rates the same on Saffir-Simpson as a Category 5 that grinds across populated coastline for 36 hours. One is a curiosity; the other is a catastrophe. The wind scale flattens the distinction.

What DPS measures

DPS is a composite 0–100 score computed from five physically meaningful components, each derived from public NOAA, NHC, JTWC, and IBTrACS data. None of the components are subjective.

The components are combined into a normalized 0–100 score with basin-specific adjustments — Atlantic, Eastern Pacific, Western Pacific, North Indian, South Indian, and South Pacific storms each have their own calibration so the scale behaves consistently regardless of where the storm formed. A score of 50 means roughly the same thing in the Atlantic as in the Western Pacific. A score of 95 in either basin is among the most destructive storms ever recorded.

A note on Stage 5 — the presentation layer

Stage 5 is honestly the most editorially-loaded part of the pipeline, and it's worth being explicit about it.

The output of Stages 1–4 is a pre-compression "raw" value that can range from ~30 (weak tropical storms) into the hundreds for the most extreme storms in the historical record — Katrina, for example, has a Stage-4 output around 127. The structurally meaningful number for analytical use is this pre-compression value; it's what the formula actually computes from the underlying physics.

The displayed 0–99 score is then produced by mapping that raw value through the square-root compression curve. That mapping is a presentation choice, calibrated so that the most destructive storms in modern history (Katrina, Maria, Haiyan) cluster near 90, with the 99 ceiling unreachable except by storms beyond any current historical reference. A linear pass-through, a logistic curve, or a different (T, S) pair would all produce the same ordering of storms but different displayed spacings between them.

This matters because the difference between an 87 and a 92 is editorial, not structural. The underlying hazard ranking is robust to the compression choice; the specific displayed value reflects a calibration intended to make the public scale intuitive (Devastating / Historic bands match how readers describe these storms) rather than to encode a precise quantitative interpretation. Anyone using DPS for downstream quantitative work — risk pricing, portfolio scoring, regulatory filings — should reference the pre-compression value alongside or instead of the displayed score.

How DPS maps to outcomes

The 0–100 scale is calibrated against historical outcomes — damage in inflation-adjusted dollars, fatalities, and post-storm survey reports — across roughly 200 storms from 2015 onward. The ranges below are not strict bins; they are typical outcome buckets.

Worked example: Ike vs. Charley

Hurricane Charley (2004) struck southwest Florida as a strong Category 4, peak winds of 130 kt. Hurricane Ike (2008) struck Galveston as a Category 2, peak winds of 95 kt. Saffir-Simpson rates Charley as the more dangerous storm. The damage record disagrees: Ike caused roughly four times the inflation-adjusted economic loss.

The reason Saffir-Simpson misses it is visible in two numbers Saffir-Simpson does not look at:

• Wind field radius. Ike's R34 (tropical-storm-force radius) was ~275 nm. Charley's was ~85 nm. Ike was more than three times wider.
• Integrated kinetic energy. Ike's peak IKE was ~150 TJ. Charley's was ~12 TJ. Ike carried more than ten times the total kinetic energy.

DPS scores Charley around 71 and Ike around 84. That ordering matches the damage record. The categorical ordering does not.

How well does DPS predict real damage?

A score is only as good as what it predicts. We test DPS against independent ground truth — the federal disaster response — across the 35 U.S. storms in our dataset that carry FEMA records, and compare it to the metrics a forecaster would otherwise reach for: peak wind, integrated kinetic energy, and minimum central pressure. Two measures: the AUC (how cleanly each metric separates the storms that drew a FEMA major-disaster declaration — 30 of 35) and the rank correlation (how well each orders storms by counties declared). Higher is better on both.

DPS leads on both — because it weighs storm size, duration, and surge geography, not just peak intensity. The scatter plots every storm by its DPS against the breadth of its federal disaster footprint:

The dots furthest from the line are the honest edge cases, and they reflect a property of the outcome metric, not a flaw in DPS. Above the line sit weak-but-wide systems — Isaias, Debby — whose multi-state footprint (and the county-by-county federal response it triggered) far outran their modest intensity. Below it sit compact, ferocious storms — Harvey, Milton — that concentrated historic rainfall and wind on relatively few counties. "Counties declared" rewards geographic spread; DPS measures destructive power; the two diverge exactly where you would expect.

This is a deliberately honest test, not a victory lap: the sample is small and U.S.-only, the FEMA-declaration outcome is itself imperfect, and DPS still misses some exposure- and rainfall-driven damage. But on the question it is built to answer — which storm carries more destructive power — it out-predicts every conventional single-number metric.

Validating the surge model against tide gauges

DPS's surge term is a SLOSH-surrogate model — a fast parametric estimate of open-coast surge potential, not a measurement. Where we can, we check it against reality. NOAA CO-OPS tide gauges record actual water levels through a storm, and the true storm surge is the residual — observed water level minus the predicted astronomical tide — which strips out the normal rise and fall of the tide and isolates the storm's own contribution.

For storms with gauge coverage we compare the model's surge potential when the storm was nearest a gauge against the highest residual that gauge recorded. One caveat dominates the comparison: a fixed gauge rarely sits at the exact point of peak surge, and many sit inside sheltered harbors, so an observed value below the modeled open-coast potential is normal and expected — not a model error. The check is most useful in the other direction, flagging the rare case where the observed surge exceeds the model. These observed gauge values are also available per storm on the map, via the opt-in “Observed surge” layer.

What DPS does not do

DPS is a destructive-potential score, not a forecast. It is computed from observed or forecast track and intensity data; it does not generate the track itself. NHC, JTWC, and the various ensemble models produce the track; DPS interprets it.

DPS is also not a replacement for the official watch/warning system. The National Hurricane Center's products are the operational source of record for evacuation decisions. DPS is meant for situational awareness, historical comparison, and research — telling you that the Category 2 bearing down on your coast is in fact more dangerous than the Category 4 that grazed the same coast a decade ago.

Destructive potential is not the same as realized damage

DPS measures a storm's destructive power — wind, surge potential, storm size, and duration of coastal exposure. It is deliberately not a damage estimate. The dollars a storm actually costs also depend on three things DPS does not fold into the core power score: exposure (how much population and property sit in the path), secondary hazards (tornadoes, freshwater/inland flooding), and vulnerability (building codes, tree cover, grid resilience). Exposure and rainfall geography are carried by the companion Impact-Area (IAS) and Economic-Risk (ERS) scores; tornado and grid-outage damage are not yet modeled at all.

The clearest example is Hurricane Isaias (2020). It made landfall in North Carolina as a Category 1 and scores a DPS in the low 30s — a genuinely modest storm by power. Yet it drew major-disaster declarations across roughly 100 counties in eight states. The damage came from a seaboard tornado outbreak, freshwater flooding, and days of tree-driven power outages across the densely populated Northeast — none of which is "storm power." A low DPS for Isaias is the correct reading of its strength; the gap to its damage footprint is exactly the part DPS leaves to exposure and vulnerability, and a reminder that no single number captures every dimension of a storm's threat.

Open data, reproducible methodology

StormDPS is intended as a transparent alternative to closed proprietary scales. The underlying historical scores for ~200 storms are published on the Data page as CSV and JSON. The full computational methodology — including the basin-specific coefficients, IKE integration, and SLOSH-surrogate surge model — is open and citable.