A Weather Machine
A Weather Machine
As computers and models become more sophisticated, weather forecasting is growing increasingly independent of humans. Even the text of this analysis could eventually be produced by a computer. What then is the contribution of humans in the process of weather forecasting? If the contribution is merely communicating the forecast, then meteorology as a physical science is only important in providing the tools to improve models. People are not usually interested in the physical consistency of their forecasts, rather, they are interested in the forecasts themselves and their accuracy. Once a model is capable of producing a forecast, the communication of the forecast to the public proceeds independently from the physical principles on which the forecast was based.
One can make a few points in favor of humans. A posteriori explanations of a weather event can usually be found, even when all model forecasts failed (yes, in this respect meteorology seems no different from economics). That means we retain at least the ability to explain mechanistically a weather event, even in the unfavorable case that no model gives the right answer. A stronger point in favor of humans is the fact that some humans consistently rank better than models in weather forecasting contests. For several years, these people have used models to help themselves achieve better accuracy, while obviously adding something else. Disentangling that ‘something else’ is what allows meteorologists to improve models and produce better forecasts. But only to make the whole process even more human-independent.
Extended Forecast
Tonight: Partly cloudy to mostly clear. Much colder. Low 17°F (-8°C).
Tomorrow: Mostly sunny. Cold. High 26°F (-3°C).
Sunday: Partly cloudy to cloudy. Much warmer. Chance of rain. Low 18°F (-8 C) High 41°F (5°C).