Products built from many radars
- View
One grid built from the whole network, updating every couple of minutes, and four derived products that are estimates rather than observations.
In this lesson you will
- Describe what MRMS is and its update cadence.
- Interpret MESH and state its known bias.
- Interpret azimuthal shear and rotation tracks.
- State what these products do and do not tell a spotter.
One grid instead of many radars
Every product so far has been from one radar, with that radar's cone of silence, its blockages and its range limits. MRMS, the multi-radar multi-sensor system, merges the whole network plus Canadian radars, gauges, satellite and model data onto a single national grid and updates it about every two minutes.
The gaps in one radar are not the gaps in the next. A storm in one radar's cone of silence is at useful range from another; a sector blocked by terrain from one direction is clear from another. Merging fills holes that no single site can fill, which is why the derived products below are built on MRMS rather than on individual radars.
MESH, and the bias you must carry with it
Module 5 introduced the maximum estimated size of hail as a radar proxy. Here is the operational picture, and the caveat is the important half.
- It is an estimate derived from the reflectivity profile above the freezing level. It is not a measurement of a stone and nobody claims it is.
- It underestimates the largest hail, and this is documented rather than suspected. Skill falls away badly for the very biggest stones, partly because hail bigger than the radar wavelength scatters in ways the relationship does not capture.
- It has known failure modes: highly tilted storms in strong deep-layer shear, left-moving supercells, and storms with a very large bounded weak echo region.
- Around 29 to 30 mm is about the most skillful threshold for separating severe from sub-severe hail, which is a different and more achievable task than sizing a stone.
If MESH could size hail accurately, a measured hail report would add little. It cannot, it is biased low exactly at the sizes that matter most, and the bias is worst for the giants. A spotter with a ruler and a coin is measuring the thing the radar is estimating. Module 5 lesson 2 is the procedure.
Azimuthal shear and rotation tracks
- Azimuthal shear
- A calculated measure of how sharply velocity changes across the beam, computed in near-surface and mid-level layers. It is a way of finding rotation automatically rather than by eye.
- Rotation tracks
- Azimuthal shear accumulated over time and painted as a swath on a map. Extremely useful after the fact: it shows where rotation went and how it varied, and because it is built on the merged grid it partly compensates for cones of silence, beam broadening and terrain blockage.
Azimuthal shear is not a tornado observation. It is a diagnostic computed from velocity, and every limitation of velocity in lesson 6 is inherited by it, including beam height. And it is artificially inflated very close to the radar, where the geometry exaggerates it, which is a known artifact rather than a real signal.
Probabilistic products
You may hear ProbSevere mentioned on a net. It is a model that fuses radar, satellite, lightning and model data into a probability that a storm will produce severe weather in the near future, now in its third version. It is a forecaster tool and this course prints no thresholds from it.
A 70% probability is not a statement that something will happen. It is a statement that in a large number of similar situations it happens about seven times in ten. A probability that does not verify was not wrong, and a spotter who treats these numbers as predictions will conclude the tools are unreliable when they are behaving exactly as designed.
What this platform does with all of it
This is the same principle modules 5, 6 and 7 each arrived at, and this lesson is where the machinery behind it becomes visible. Reports submitted here are checked for plausibility, not exactness, against these products.
- A convective reflectivity gate, so a severe report during obvious convection is corroborated.
- A hail estimate gate, using a MESH threshold around the severe discrimination figure above, because that is the task MESH is actually good at.
- A rotation gate, using azimuthal shear, for reports of rotation.
A report that disagrees with these products is flagged for a human. It is never automatically discarded. Every product in this lesson is an estimate with a documented bias, and several of them are biased in the direction that would reject correct reports. Module 6 lesson 7 made the same argument for flooding.
Check yourself
A spotter measures 3.5 inch hail. MRMS MESH over that location peaked around 2 inches. What is the right conclusion?
- The report is plausible. MESH is a documented underestimate for the largest hail
- The report is probably wrong by about 1.5 inches and should be corrected downward
- The report should be rejected, as it disagrees with the radar-derived value
- MESH and the report cannot be compared, since MESH does not estimate hail size
Hint
This lesson names the exact size range where MESH performs worst.
Answer
The report is plausible. MESH is a documented underestimate for the largest hail
MESH underestimates large hail, and skill collapses for the very biggest stones, so a measured report well above the MESH value is exactly the expected disagreement rather than evidence against the report. This is the case the flag-for-review rule exists to protect: auto-rejecting it would delete the most significant report of the day.
Sources
- Smith et al. 2016, Bull. Amer. Meteor. Soc. 97(9), 1617–1630: multi-radar multi-sensor severe weather and aviation products
- Murillo & Homeyer 2019, Wea. Forecasting 34(5), 1363–1384: the development of a practical MESH-based hail metric
- NWS Warning Decision Training Division, radar and warning guidance
- NSSL, multi-radar multi-sensor system
- Cintineo et al. 2024, Wea. Forecasting 39(12): ProbSevere version 3, improved exploitation of data fusion