Probabilistic Forecasts
When reviewing the other forecast types in this tutorial, you’ll notice that all of them are deterministic (i.e., non-probabilistc) values. They have a quantitative value that can be verified with observations. Probabilistic forecasts on the other hand, do not provide a specific value relating directly to the variable (precipitation, wind speed, etc.). Instead, probabilistic forecasts provide, you guessed it, a probability of a certain event occurring. The most common probabilistic forecasts are for precipitation. Given the spatial and temporal irregularities any precipitation type could display during accumulation, probabilities are a better option for numerical models to provide for the general public. It allows the general public to decide for themselves if a 60% forecasted chance of rain is enough to warrant bringing an umbrella to the outdoor event, or if a 40% forecasted chance of snow is too much to consider going for a long hike. Consider the alternative where deterministic forecasts were used for precipitation: would it be more beneficial to hear a forecast of no precipitation for any probability less than 20% and precipitation forecasted for anything greater than 20% (i.e. a categorical forecast)? Or maybe a scenario where the largest amount of precipitation is presented as the precipitation a given area will experience with no other information (i.e. a continuous forecast)? While they have other issues which will be explored in this section, probabilistic forecasts play an important part in weather verification statistics. Moreover, they require special measures for verification.
