Rain, Snow, and Fog: How Weather Affects AV Safety

Here's a number that surprises everyone: 3,550 AV/ADAS incidents — the vast majority — occurred in clear weather. Only 432 happened in rain. Just 19 in snow. If you expected autonomous vehicles to crash most in storms, the data says otherwise. But the explanation isn't what you think.

Why Clear Weather Dominates

The answer is exposure. Americans drive the vast majority of their miles in clear weather. According to FHWA data, roughly 75% of all vehicle-miles-traveled occur on dry roads under clear skies. If 75% of driving happens in clear weather, you'd expect about 75% of crashes in clear weather — and the AV data shows exactly that pattern (3,550 of ~4,700 weather-reported incidents, or ~75%).

This means AVs aren't disproportionately crashing in good weather. They're crashing in proportion to when they drive.

Rain: A Real Risk Factor

The 432 rain incidents deserve closer attention. Rain accounts for roughly 10% of reported-weather incidents but only about 5–6% of total driving miles. This suggests rain may be a genuine risk multiplier for AV systems — which makes intuitive sense. Water on camera lenses, spray from other vehicles, reduced road friction, and altered braking distances all stress automated systems.

For camera-based systems like Tesla, rain is particularly challenging. Water droplets on cameras create distortion and blur. While Tesla uses heated cameras and software-based rain compensation, the physics of water on a lens are hard to fully overcome.

Snow: Rare but Revealing

Only 19 incidents occurred in snow — less than 0.5% of the total. This isn't because AVs handle snow well. It's because most AV operations avoid snowy conditions entirely. Waymo operates primarily in Phoenix and San Francisco (minimal snow). Tesla's Autopilot is known to disengage or limit functionality in heavy snow. The 19 snow incidents likely come from ADAS vehicles (mostly Teslas) in northern states where drivers push through winter conditions.

Fog and Other Conditions

Fog, sleet, and other low-visibility conditions account for small numbers but are worth monitoring. Fog is particularly dangerous because it degrades both camera and lidar performance. Lidar beams scatter off fog particles, creating noise. Cameras lose contrast and depth perception. Radar remains the most fog-resistant sensor — which raises questions about Tesla's decision to remove it.

The Rate vs. Count Problem

Looking only at counts is misleading. What we really need is a crash rate per mile driven in each weather condition. If AVs drive 100 million miles in clear weather and 5 million in rain, the 432 rain crashes represent a much higher per-mile rate than the 3,550 clear-weather crashes. Unfortunately, total AV miles by weather condition aren't publicly available — a significant data gap.

What This Means