Tornadoes on radar

Radar does not see tornadoes. It sees rotation, and sometimes it sees what a tornado has picked up. Those are different observations with different confidence levels, and conflating them is the most common error in reading a warning-scale storm.

There are three signatures worth knowing, and they form a ladder of increasing certainty.

1. The hook echo — the weakest of the three

The famous one, and the least diagnostic. A hook is a curl of reflectivity on the rear-right flank of a supercell, wrapping counter-clockwise. It forms because precipitation is being drawn around the mesocyclone — the hook is the rain being wrapped, not the tornado.

Many hooks never produce a tornado, and many tornadoes occur with no visible hook, particularly in rain-wrapped storms and linear systems. A hook says “this storm has organized rotation dragging precipitation around it.” That is worth your attention and it is not a detection. Treat it as the thing that makes you look at velocity.

2. The velocity couplet — the real detection

Look for strong inbound immediately adjacent to strong outbound, with a tight boundary between them. That is rotation on an axis perpendicular to the beam.

• Tightness beats magnitude. A 60-knot difference across two gates is far more concerning than the same difference spread over several miles. The first is a small violent circulation; the second is broad, ordinary storm-scale rotation.
• Height above ground. The beam rises with range. A couplet 15,000 feet up in a distant storm is a midlevel mesocyclone, not a tornado. The same couplet at 1,500 feet is another situation entirely.
• Depth. Rotation continuous through several elevation cuts is a deep, organized mesocyclone. Rotation at one level may be noise, aliasing, or a brief spin-up.
• Persistence. A couplet surviving several volume scans means something. One that appears for a single scan often does not.

The aliasing trap: velocity wraps past a radar’s maximum unambiguous value, so a very fast inbound can be reported as a fast outbound. It mimics an extremely tight couplet. The tell is a sharp noisy boundary with no supporting structure in reflectivity and no coherence across tilts.

3. The debris signature — confirmation, after the fact

This is the only radar signature that constitutes evidence a tornado is on the ground right now, and it comes from dual-polarization. A tornado debris signature requires three things co-located:

• High reflectivity where the circulation is — often a distinct blob, hence “debris ball”
• Correlation coefficient dropping below roughly 0.80 — the diagnostic half
• ZDR near zero — debris tumbles randomly rather than falling in a preferred orientation

The CC drop is the physics. Correlation coefficient measures how uniform targets are within a sample volume. Rain is very uniform and sits near 0.98. Lofted debris — lumber, insulation, soil, vegetation — is wildly non-uniform, and CC collapses. No ordinary precipitation does this.

Reading them together

No single field decides. The confident read is a stack.

Each field answers a question the others cannot. That is the argument for a tool that shows all of them from the same volume rather than one derived product at a time — you are cross-examining a single moment from several angles.

What this page will not tell you

Whether to take shelter. That is not a radar-reading question. Warnings come from the National Weather Service, which has training, context and ground truth no radar image carries. If you are in a warned area, act on the warning.

Radar is a tool for understanding storms, not a replacement for the warning system. Anyone telling you otherwise is selling you something.

Next: the supercell, the storm type most of these signatures live inside.