Wet bias

from Wikipedia, the free encyclopedia

Wet bias ([ wĕt baiəs ], English, roughly “wet bias ”) is the phenomenon that some meteorologists state in their weather forecasts (mostly deliberately) the probabilities of precipitation (especially for rain) higher than they believe the probability (and state it is higher than the probability resulting from empirical data), with the aim of compensating for the reverse bias of many users when interpreting the information. The US television station The Weather Channel has empirically shown, and has also admitted, a wet bias when the probabilities of precipitation are low(for example, a 5 percent probability is predicted as a 20 percent), but not high probabilities of precipitation (so a 60 percent probability is also reported as a 60 percent probability). Some local TV stations have been shown to have a significantly stronger wet bias , often reporting a 100 percent chance of precipitation when it only rains 70% of the time.

discovery

In 2002, Eric Floehr, then a postgraduate student at Ohio State University , began collecting historical data from weather forecasts for the United States from the non-commercial state National Weather Service , The Weather Channel and AccuWeather , and published the data on ForecastWatch.com. Floehr found that the predictions of the two commercial providers have a bias: They regularly predict a higher probability of precipitation than the frequency of precipitation actually is. The National Weather Service's predictions were without bias, whereas the Weather Channel's predictions were biased on low probabilities of precipitation: when the Weather Channel predicted a 20% probability of precipitation, historically it rained only 5% of the time, while one reported 70% probability of precipitation already corresponds to the actual frequency of precipitation. Blogger Dan Allan noted that the Weather Channel also has a bias on the high end: a probability of 90% or higher is rounded up to 100%. On the other hand, local TV stations tend to consistently exaggerate the probability of precipitation (unless they forecast a probability of 0%, in which case it still rained 10% of the time). The wet bias , although informally known within the weather community for some time, community Weather was first done correctly known outside of Nate Silver in 2012 the released and book The Signal and the Noise (German title: The calculation of the future ).

The term " wet bias " is used because it is a systematic error (English " systematic error " or " bias ") in the direction that the weather is wetter than it actually is.

Reasons for a wet bias

According to Silver, the Weather Channel has admitted that it reports low precipitation probabilities exaggeratedly high. This is justified by incentives in this direction: If the correct (low) value for the probability of precipitation were given, viewers could interpret the forecast in such a way that there is no probability of rain and then get upset when it does rain. In other words, the tendency for those watching their loss aversion (English loss aversion ) underestimated and thus their cost-to-loss ratio (English cost-loss ratio ) calculate wrong when the rainfall probability is low, compensates the Weather Channel by intentionally giving too high probabilities. Silver quotes Dr. Rose of the Weather Channel : "If the forecast was objective, if it had zero bias in precipitation, we'd be in trouble."

Individual evidence

  1. ↑ a b c d Nate Silver : The Weatherman Is Not a Moron . New York Times . September 7, 2012. Retrieved May 24, 2014.
  2. ↑ Why everyone hates the weatherman . September 27, 2012. Retrieved May 24, 2014.
  3. ↑ a b Dan Allan: Wet Bias . Retrieved May 24, 2014.
  4. ↑ a b c d e f Nate Silver: The Signal and the Noise: Why So Many Predictions Fail , ISBN 978-1594204111 . , Pages 131-136
  5. ↑ ForecastWatch: Accuracy Defined . Retrieved May 24, 2014.
  6. ↑ Eric Bickel, Seong Dae Kim: Verification of The Weather Channel Probability of Precipitation Forecasts . In: Monthly Weather Review . 136, No. 12, December 2008, pp. 4867-4881. doi : 10.1175 / 2008MWR2547.1 .
  7. ^ A b Icon Forecast Bias and Pleasant Surprises . ForecastAdvisor. September 19, 2012. Retrieved May 24, 2014.