Have you ever wondered how meteorologists come up with their forecast? If you're an avid weather watcher, you probably already know that we use computer model data to help us discern the forecast details.This link takes you to one of the sites I use every day in forecasting!

It usually takes me about 1 hour to look through all the model data and develop my forecast. The most important thing that I'm looking for can be summarized by calling it "the 2 Cs".They are consensus, and consistency. If you have both consensus and consistency your model data, then you can have a lot of confidence in your forecast. But if you have neither, the forecaster has a lot of thinking to do! In this article, I'll tell you what each of these terms means to me, and how you can spot them, too.

Consensus is good agreement between different models. In other words, when 2 different forecast models are predicting the same thing will happen at the same time, we have consensus. The picture below is of the WRF forecast for 6 hours of precipitation, valid at 12Z Tuesday (8am):

is good agreement between different models. In other words, when 2 different forecast models are predicting the same thing will happen at the same time, we have consensus. The picture below is of the WRF forecast for 6 hours of precipitation, valid at 12Z Tuesday (8am):

College of DuPage Meteorology

And here's the 6-hour precipitation forecast from the GFS model, valid at the same time:

?

College of DuPage Meteorology

As you can see, the projections are similar, with the heaviest rainfall along the Appalachians in West Virginia, but they are somewhat different from each other. Forecasts differ between different models because the equations that govern the atmosphere are incredibly complicated. In fact, we literally don't have computers that are fast enough to process all the atmospheric equations in a reasonable timeframe for operational use. So, each forecast model has to omit some data or parameters. Despite this, you'll often see some agreement between the models, especially when the weather is fairly stagnant or calm.

If the models are in good agreement with each other on key features, like the movement of a trough or the placement of a precipitation bullseye, it helps build our confidence in the forecast. But there's something else we're looking for...

We also need consistency. When a particular model is giving us the same forecast solution, run after run, it's a huge confidence-booster for your forecast. Compare the WRF precip forecast again from the 12Z model run:

College of DuPage Meteorology

to the WRF's 18Z model run from yesterday:

College of DuPage Meteorology

You can see that the more recent model run (the 12Z from today) shows the heaviest rain further east than the model run from yesterday, which placed the bullseye of precipitation in eastern Ohio!

Consistency is often harder to achieve than consensus between different models. That's because with each new model initialization (for most models, this happens every 6 hours), the computersare ingesting new initial conditions based on satellite data, sounding data from balloons (every 12 hours), and surface observations. In theory, theforecast should be consistent, but our model data's grid points are spaced apart too widely to account for all the features that can disrupt our weather. These features can include mountains, lakes and other water features, pockets of severe weather, farmland or forest land, and even cities.

Winter storms are often the hardest for our model data to resolve, which explains why your TV forecasters are so reluctant to give you a firm snowfall forecast! Ultimately, until we can get more data points into our model initializations, and until we build computers that are capable of handling the massive amount of math required to resolve the conditions, our forecasts will continue to have discrepancy and error. In the meantime, we forecasters are just doing the best we can with today's forecasting technology!