The term wet bias refers to weather forecasters deliberately reporting a higher probability of rain than their predictive models show. The Weather Channel has been empirically shown, and has also admitted, to having a wet bias in the case of low probability of precipitation (for instance, a 5% probability may be reported as a 20% probability) but not at higher probabilities (a 60% probability will likely be reported accurately). Blogger Dan Allan noted that the channel is also biased at the upper end (a probability of 90% or higher will be rounded up to 100%). Local weather stations have been shown to have a significantly greater wet bias, with some reporting a probability as low as 70% as a certainty.
In 2002, computer scientist Eric Floehr started analyzing historical weather prediction data on a website called ForecastWatch. He found that the commercial forecasts were biased and the National Weather Service forecasts weren’t. His findings, though known within the meteorology community for some time, was first popularized in Nate Silver’s 2012 book ‘The Signal and the Noise.’ According to Silver, the phenomenon is due to skewed incentives: if the correct low probability of precipitation is given, viewers may interpret the forecast as if there were no probability of rain, and then be upset if it does rain. Forecasters are compensating for the fact that people have greater loss aversion than they think they do (and are especially prone to miscalculate their cost-loss ratio when it is low). Silver quotes Dr. Rose of The Weather Channel as saying, ‘If the forecast was objective, if it has zero bias in precipitation, we are in trouble.’
The Daily Omnivore 