Models Make Mistakes

When meteorologists make forecasts, they rely on many tools. One of these key tools are weather models. There are various weather models to choose from, which each have a different range of dates.  These models use not only the current conditions, but also climatology to predict just what will be predicted to happen up to a few weeks out. Since this is based on historical movement of storms and what previous results, the end product may end up being rather off when looking at an event beyond five days away from an event.

An example of this just recently happened for people located in the Midwest and some portions of the Southern Great Plains.  On the 06 Z run on December 5th, 2019, the GFS, one type of model, predicted that there would be 28 inches of snow that fell in parts of west Tennessee and western Kentucky.  This type of snow is not heard of in these parts of the country, so it sent people into a frenzy.

The very next day, the 06 Z run said that around the same time of this event, there would not be any snow for west Tennessee and western Kentucky.  If anything were to fall, then it would only be a trace amount at the most.

So here’s the question: why do two model runs differ so much over the course of 24 hours?  This can factor greatly because of the data that is put into a model can vary, and if someone puts in bad data, then a bad model run will occur.  Even the best meteorologists make mistakes, so if someone else does not catch that mistake, then the end product will not only be wrong this far out, but for the near future as well.

Another factor that leads to differing results would be related to the historical tracks of storms and climatology. If a storm takes a certain path around the same time of the year, then any model that attempts to forecast beyond five days will rely on what has happened in the past.  Much of the knowledge that deals with weather came from noticing that these patterns brought this type of weather over and over again. So if it happened once, then there’s always a chance of it happening once more. Models tend to rely on the calculated pattern that the occurrence will happen again each time they forecast this far out because the data only predicts the weather a few days out, not weeks in advance.
​
With this example, both of these models were predicting an event that was over a week and a half away, so the forecast from this far out cannot be fairly relied on.  However, many people see a forecast like the first model and worry about what will happen. In the next part of this series, there will be some tips that forecasters use to establish a hypothesis of what could happen in the future.

For more information on other tools used to make forecasts, make sure to check out www.globalweatherclimatecenter.com/weather-observations!
 
©2019 Weather Forecaster Shannon Sullivan
​
(Source: College DuPage GFS Model Runs)