A Suggestion to Young Meteorologists
Below is a screen capture of the 6-day ECMWF forecast temperature for the Chiefs - Patriots game played yesterday evening at Arrowhead Stadium. The forecast was for -14°F. [The wind chill was forecasted to be -24°!]
The actual temperature at 6pm was 19°... an error of 31°F. It was a terrible forecast. Had I thought there was even a slight chance that temperatures would have been that cold, you would have read about it here on the MSE Consulting Blog. You didn't because I was highly confident that forecast would be way off.
Why do I bring this up?
Even thought I am retired from weather forecasting, I care deeply about the future of the profession. Based on what I saw in the media, there were many meteorologists who bought into these excessively cold forecasts for the game and were broadcasting them about this time last week.
Advice to young meteorologists: In the wise words of SUNYA Professor Tom Stewart, an expert in how forecasts are made, "more models increase meteorologists' confidence but they do not improve accuracy." Focus on one or two mesoscale models for short-term forecasts and one model for longer term forecasts. Learn the biases of these models.
I focus on the ECMWF model. Because I knew it over-forecasts temperature extremes, I knew not to believe the extreme temperature forecasts because the other conditions needed to bring Arctic temperatures to Kansas City (solid snow cover across central Canada and the central northern Great Plains, detached Polar Vortex far enough south, etc.) did not exist.
It is impossible to keep track of the biases of numerous models. You'll add a lot more value if you focus on really learning how to use 2-3 of them. Remember: the case for keeping human forecasters in the loop is so they can add value to the output of the models. If all you do is regurgitate the model output, you may, in the long run, be putting yourself out of a job.
Why do I bring this up?
Even thought I am retired from weather forecasting, I care deeply about the future of the profession. Based on what I saw in the media, there were many meteorologists who bought into these excessively cold forecasts for the game and were broadcasting them about this time last week.
I focus on the ECMWF model. Because I knew it over-forecasts temperature extremes, I knew not to believe the extreme temperature forecasts because the other conditions needed to bring Arctic temperatures to Kansas City (solid snow cover across central Canada and the central northern Great Plains, detached Polar Vortex far enough south, etc.) did not exist.
It is impossible to keep track of the biases of numerous models. You'll add a lot more value if you focus on really learning how to use 2-3 of them. Remember: the case for keeping human forecasters in the loop is so they can add value to the output of the models. If all you do is regurgitate the model output, you may, in the long run, be putting yourself out of a job.
Comments
Post a Comment