Friday, January 31, 2014

Don't Believe Snowfall Forecasts A Week Out


This pattern gets very tricky with a slightly southerly flow (WSW to ENW) this weekend and especially over the next week. The probability for more "panhandle snows" is growing.


In this age of the internet where every possible computer model and projection is available 24/7 on hundreds of websites and apps, anyone can post snapshots of these computer outputs especially when a rumor of heavy snow gains steam (next week). The problem with this is twofold. First, most people believe that these neat looking maps mean more accurate forecasts especially farther out.  Secondly, most people treat these snapshots as Gospel and believe that these computer model depictions are actual forecasts. Nothing could be further from the truth. The phrases "could happen" with "mostly likely" to happen become interchangeable. Both phrases couldn't be more different. Wxbrad has an excellent post on this here.

"Could happen" scenarios bubble up from the seeds nestled in a common computer model misconception. Many believe that they provide forecasts. Not true. They are only guides. Nothing more. They come in two forms or outputs. One is called a deterministic solution, the other is an ensemble. A deterministic is ONLY ONE possible outcome with a set of initial conditions. An ensemble is a blend or average of a bunch of deterministic solutions. It smooths out the extremes giving us an average of the multiple outputs. The problem lies within the INITIAL CONDITIONS. Not all weather conditions are known for each geographic point in an area (in this case, northern Ohio). So there are data gaps before the model even runs. So unless you know the weather conditions for each and every point in northern Ohio (say every 50 or 100 feet) the computer output will have gaps. The overall solution will be estimated and smoothed out where these gaps exist.  Imagine the gaps between Cleveland and Westlake or Akron and Medina. Actual solutions for these gaps are not known so the output is approximated.

We learn to use both the ensemble and deterministic in our assessment of each weather pattern and forecast for the area. Computer models are only a guide, an interpretation of the weather scenario, nothing more. We learn the pros, cons and biases inherent in each model and apply them to the current weather situation. Meteorologists deal in "mostly likely" scenarios after taking all of the available data and computer models into account.  WE create forecasts that are "mostly likely" to occur. Look how the computer output (GFS) for snowfall accumulation for next week has changed in both depth and location over the last 3 days.


Be careful of Facebook posts that show solutions more than 5 days out. This is only one possible solution with one set of initial conditions showing a "could happen" scenario. Don't get me wrong. Its fun to speculate. But "Could happen" scenarios are not a healthy way of looking at the weather (just look at the devolution of Facebook comment threads). "Mostly likely" scenarios are safer and work better in the long run.

Trust the meteorologist not the computer.

No comments: