Weather forecasts are 95% accurate, and nobody believes them
A meteorologist says forecasts are about 95% accurate in the three-to-five-day window, but the public only remembers the time it got rained on; climate is not future tense, it's present continuous.
The video won't play here. Listen to the audio instead:
The argument · tap a timestamp to hear it
The forecast isn't wrong; human memory is biased
The public broadly believes weather forecasts are often wrong, but the number Shepard gives is roughly 95% accuracy in the three-to-five-day window. The problem isn't the model, it's memory: people remember the cookout that got rained on, the kid's ballgame that got rained on, and forget all the days the forecast nailed it. He uses the kicker analogy — perfect all season, misses one in the Super Bowl, and that's the only one anyone remembers. What this changes: forecast reliability is far higher than its public reputation, and that reputation is shaped by statistical bias.
— Dr. Marshall ShepherdBomb cyclone isn't a new word, it's new exposure
Many people assume bomb cyclone, polar vortex and derecho are signals that the climate has changed. Shepard rejects that outright: these terms have existed in meteorology for decades, they've just recently come into heavy use in mass media. So you can't use the popularity of a term as an anchor for climate change. What is actually changing is intensity — heavier downpours, more frequent and more severe heat waves, rapid hurricane intensification, higher sea levels pushing more water in with the same storm, bigger storm-surge losses.
— Dr. Marshall ShepherdThe reason for optimism is that we know what to do
Shepard says he doesn't want to stop at bad news, because the good news is this isn't a problem that still needs head-scratching research — we know what to do: cut carbon emissions, and we must adapt, because some changes have already crossed irreversible points. He cites the 2021 Pacific Northwest heat wave: most homes there had no air conditioning, Canada recorded 120 degrees Fahrenheit, it was called a once-in-a-millennium event, but climate change made it 150 times more likely, causing nearly $9 billion in losses and 1,400 heat deaths. The adaptation strategy is to retrofit those houses with air conditioning.
— Dr. Marshall ShepherdWarming makes blizzards more intense, not less
The counterintuitive mechanism: hurricanes, downpours and blizzards all start with water vapor. The Clausius-Clapeyron equation says a warmer atmosphere holds more water vapor. For every 1 degree Fahrenheit of warming, water vapor increases about 4% (7% in Celsius). A century ago the Northeast US averaged 22 degrees Fahrenheit in winter; now it's up to 26, and some years 30 — that's a 32% increase in water vapor over 22, and 30 is still below the 32-degree freezing point. So a warmer planet means more water in the atmosphere, and blizzards can hit harder.
— Dr. Marshall ShepherdThe hurricane cone is probability, not a path
During Hurricane Ian in 2022, some people evacuated from one part of the cone to another — a total misreading. The cone means the storm center has a 67% chance of falling somewhere inside it, not that the storm will travel down the center line of the cone. Shepard says the public broadly struggles with probability, uncertainty, two or more processes happening at once, and anything that can't be simplified to their level — so they simplify complex information down to their level, and the cost is bad decisions.
— Dr. Marshall ShepherdA 20% chance of rain does not mean it won't rain
Shepard tells the story of getting rained on while tubing with his son: the woman next to them complained that it was only a 20% chance of rain and meteorologists are always wrong. But that day wasn't 0% — they just happened to be in that 20% area. Statistically, 80% of the area stayed dry, and they landed in the 20% that got wet. One way to compute POP is confidence of rain × fraction of area affected: 60% confidence over 60% of an area gives a POP of 36%. So "20% chance" doesn't mean "it won't rain" — it's describing a regional probability distribution.
— Dr. Marshall ShepherdSatellites cover more ground, balloons resolve more detail
On the trade-off between weather balloons and satellites: satellites give more consistent observations over a wider area, while balloons only have data at the launch point and launch time, sometimes only twice a day. But satellite resolution is far coarser. Shepard uses a phone camera analogy: more pixels make a better photo, and the finer the resolution of an observation or model, the more fine-scale weather it can capture. A thermometer or barometer at your location has far better resolution than the blurry one-square-mile reading a satellite gives you.
— Dr. Marshall ShepherdIn their own words · checked verbatim
weather is your mood and climate is your personality
Dr. Marshall Shepherd12:24
Almost all of the time, about 95% of the time within three to five days. It's just that it's human nature for people to remember the bad ones.
Dr. Marshall Shepherd14:27
Climate change is now we've got to stop using future tense. We're living it right now
Dr. Marshall Shepherd22:42
there's a 67% chance that the center of that storm can be anywhere in that hurricane cone
Dr. Marshall Shepherd41:31
it wasn't a 0% chance of rain. In fact, we were probably in the 20% region that day.
Dr. Marshall Shepherd1:00:12
it's fluid dynamics, but it's accessible to many in the public. So they think it's much more.
Dr. Marshall Shepherd1:09:35
Figures
| Weather forecast accuracy, three to five days | About 95% | 14:27 |
| Water vapor increase per 1 degree Fahrenheit of warming | About 4% (7% in Celsius) | 29:09 |
| Canada's record high during the 2021 Pacific Northwest heat wave | 120 degrees Fahrenheit | 23:47 |
| How much more likely climate change made the 2021 heat wave | 150 times | 23:47 |
| Economic losses from the 2021 heat wave | Nearly $9 billion | 23:47 |
| Probability the storm center falls inside the hurricane cone | 67% | 41:31 |
| Hurricane track forecast error | 50 to 100 nautical miles | 57:03 |
| Share of Earth's water that is usable | 0.3% | 26:00 |
Glossary
- Clausius-Clapeyron equation
- The physical relationship describing how a warmer atmosphere holds more water vapor.
- wet bulb globe temperature
- A heat-stress index combining temperature, humidity, wind speed, solar radiation and cloud cover.
- bombogenesis
- When colliding cold and warm air masses cause a low-pressure system's pressure to drop rapidly — a bomb cyclone.
- atmospheric river
- A long, narrow band of moisture that can carry as much water as the Amazon River.
- ENSO
- The warm-cold cycle of equatorial Pacific sea-surface temperatures — El Niño and La Niña.
- AMO
- A warm-cold pattern in Atlantic sea-surface temperatures that flips roughly every 70 years.
How to listen
Engineers and product people interested in weather, climate communication and probability communication; anyone doing risk communication, early-warning systems, agriculture or energy decisions.
The first ten minutes of host monologue and guest résumé can be skipped.