SWRainAlertData
How NOAA Radar Rainfall Data Works, and Where It Fails
Stage IV is the quality-controlled record that arrives late. MRMS is the near-real-time estimate that arrives first. Compliance monitoring needs both, for different reasons.
“Synced with NOAA” appears on every product in this category and means almost nothing on its own. NOAA publishes several precipitation products with materially different properties. Which one a tool reads — and what it does when that product cannot see your site — is the substance behind the claim.
How a radar rainfall estimate is made
A weather radar does not measure rainfall. It measures reflectivity — how much energy is scattered back by whatever is in the beam. Converting reflectivity to a rain rate uses an empirical relationship (a Z–R relation) that varies with the type of precipitation. Convective summer downpours and winter stratiform rain do not scatter identically.
The estimate is then corrected against ground gauges, gridded, and accumulated over time. Both NOAA products run some version of this chain. They differ in how the correction is done and who checks it.
Stage IV
Stage IV is a multi-sensor analysis: radar combined with rain gauge reports, assembled regionally and manually quality-controlled by NOAA’s River Forecast Centers before national distribution.
- Strength: human review. It is the more defensible record of what fell, and it has a long, consistent archive suitable for retrospective analysis.
- Cost: latency. Quality control takes time, so it is not the product that tells you a threshold was crossed twenty minutes ago.
MRMS
The Multi-Radar/Multi-Sensor system, developed at NOAA’s National Severe Storms Laboratory, mosaics the national radar network at high resolution with automated gauge bias correction, updating on a fast cycle.
- Strength: speed and resolution. It is the product that supports alerting inside a 24-hour inspection window.
- Cost: no human in the loop. Automated correction handles the ordinary cases well and the unusual ones less well.
The part almost nobody handles correctly
Both products have to represent the absence of an observation, and they do it with sentinel values — reserved numbers that mean something other than a depth. MRMS uses -3 for no radar coverage and -1 for missing data.
These are negative numbers sitting in a field of rainfall depths. Any code that treats the grid as plain numeric data — clamping negatives to zero, or summing without filtering — converts “we could not see this site” into “no rain fell at this site.”
This is the single most important correctness rule in rainfall-based compliance monitoring. A site the radar could not observe must never look identical to a site where it did not rain. One of those needed an inspection. In SWRainAlert, sentinels become no reading — shown as a gap on the chart, a blank cell in the CSV, and an explicit “we cannot see this site right now” on the screen.
Why we run both, and never average
The two products disagree regularly, because they are built differently. A real example from a Georgia site: Stage IV reported 0.47″; MRMS reported 0.62″. The permit threshold is 0.50″.
One source says no inspection is required. The other says one is required within 24 hours. Running a single source means that decision was made for you by a data-ingestion choice nobody documented.
The tempting fix is to average them: 0.545″, which crosses the threshold, so the outcome looks right. It is the wrong move. 0.545″ is a number neither source reported, and in an audit you would be defending a figure with no provenance. Store both, alert on the higher, and record the divergence as a fact about the storm.
Practical reading guide
| Question | Product | Why |
|---|---|---|
| Did a threshold just get crossed? | MRMS | Latency low enough to act inside the window |
| What is the defensible record of the storm? | Stage IV | Human quality control |
| Did the sources agree? | Both | Disagreement is information, not noise |
| Was this site observed at all? | Sentinels | A gap must read as a gap |
Choosing which product answers which question
Common questions
What is the difference between NOAA Stage IV and MRMS?
Stage IV is a multi-sensor precipitation analysis quality-controlled by NOAA's River Forecast Centers, which makes it more defensible but slower to arrive. MRMS is the Multi-Radar/Multi-Sensor product from the National Severe Storms Laboratory, offering higher resolution and near-real-time updates with automated rather than human gauge bias correction.
How accurate is radar rainfall data?
Radar estimates rainfall rather than measuring it, and accuracy varies with distance from the radar, terrain, and precipitation type. Gauge bias correction improves it substantially. For compliance purposes the practical answer is to use radar to detect that a threshold was likely crossed, and to keep the permit's named system of record, which may be an on-site gauge, for the record itself.
What does it mean when rainfall data is missing?
NOAA products encode missing data and no-radar-coverage as sentinel values rather than as zero. Missing data means the site was not observed, which is different from an observation of no rain. Any system reporting 0.00 inches for an unobserved site is misrepresenting its data.
A note on what this is. General guidance, not legal advice. Your permit, your SWPPP and your state’s requirements govern, and they differ. SWRainAlert is a monitoring and recordkeeping aid, not a compliance guarantee — where a permit names an on-site rain gauge as the system of record, the gauge is the record.