Why forecasts don’t always agree
Published 7:00 am Tuesday, February 20, 2018
By Skip Rigney
Have you ever checked the weather app on your smartphone only to hear a different forecast on the radio or television a few minutes later? Maybe you have more than one weather app on your phone and have checked them both only to discover that one is calling for rain and the other sunny skies.
Can’t those weather people get their act together?
This past weekend my wife and I attended an outdoor wedding of a family member in Houston. Last Tuesday my sister-in-law texted how glad she was that the forecast high for Saturday was 71. My wife texted her sister back saying that I had told her only minutes earlier that the forecast was for a much cooler day with the high reaching only 61.
My sister-in-law wanted to know why I thought it would be so cool, because her weather apps were saying low 70s. Here’s the explanation that I shared with her.
When you use a weather “app” and type in a location, the forecast you get is heavily dependent on the output from a mathematical model of the physics of the atmosphere, the calculations of which have been performed on a computer.
The companies that field the weather apps use output from several models. Those models differ in how they simulate the physics of the atmosphere, how finely or coarsely they do their spatial calculations, and which and how observations are fed into the model. Some models are run by the companies and some by government organizations such as the National Weather Service or the European Center for Medium-Range Weather Forecasting.
One powerful statistical approach is when a single physics model is run tens of times with each run starting with slightly different initial conditions. These are known as ensembles. Meteorologists know that even with large quantities of excellent observations to input into the model, those observations have errors, and there are areas with no observations. The averages of ensembles sometimes produce a better forecast than the single model run that used the best observations available.
Some companies take output from several physics models and ensembles and statistically average them to come up with the forecast you see on your phone’s app. The forecast displayed on your weather app is usually computer-generated output that has not been reviewed by a meteorologist for your specific location.
The forecasts issued by National Weather Service (NWS) Forecast Offices around the country are different because they add the expertise of a human to the process for the specific locations in their area of responsibility.
Twice each day an NWS forecaster in Slidell looks at the output from several models and model ensembles. Based on his or her meteorological knowledge, current weather observations, radar, satellite, and experience with how the models have been performing recently for the local area and in similar situations, the forecaster puts together his or her best estimate of what the weather will be over the next seven days tailored for the various regions across southeast Louisiana and southern Mississippi.
Whose forecast is best? In most situations, they’re very similar. However, in tricky situations such as the one in Houston this past weekend when a weak, slow-moving front was near the area, the models, and thus the apps, often give different answers, especially more than a few days in advance.
In those cases, and also when severe weather is a possibility, I like to check the forecast put together by the local expert at the NWS. You can get those at www.weather.gov .
The models and apps all agree about this week. It’s going to be warm.