Sunday, April 29, 2018

Jack D's Question Re Bizarre Point Forecasts

Background related to this question was started back in posts on April 18 and 19 - detailed, worded forecasts are available for the millions of NWS grid point forecasts that are essentially model-generated. I forwarded Jack's questions along to a number of people I know who are familiar with NWS forecast procedures. The responses I received covered the question about why some of the grid forecasts end up so bizarre, especially when complicated weather events are likely. However, the responses also covered a number of related forecast issues. The majority of responses I received were from shift-working forecasters, although a couple were from folks now retired. A condensed summary of what I received is presented below.
Why do the grids lead to such bizarre worded forecasts?

Models are getting really good – but programs and algorithms that convert model output to text are not very good. Forecast wording issues reflect problems with the post-processing. Baseline worded forecasts from the various algorithms (in which Regional and higher levels seem to have no interest) are basically considered “good enough.” There is minimal investment in meteorologically sound, post-processing systems. Absurd public text forecasts occur because there no longer is importance given (at different Headquarters levels) to issuing worded forecasts that make sense. Why not average model output over three-hours, or use “wintry mix” instead of things like “rain, then snow, then rain mixed with snow, then freezing rain”? There need to be more people working to make these kind of improvements.

Conflicting responsibilities

A related issue is that forecaster attention is increasingly being shifted/mandated to “IDSS” (Impact-based Decision Support Services) and various social media and away from direct involvement in the daily forecast process.

Forecasters are stuck in a trap re providing best product for public but they are also required to produce “pleasing” maps.

Today it's more about collaboration, chat rooms, webinars, conference calls, Facebook, etc., and direct interaction with users in real time during events. The overall trend in forecast services isn't surprising. Many could see the proverbial writing on the wall many years (even decades) ago.

Human vs Automated Forecasts

Fewer people in our urbanized truly need to know about weather details –the small fraction of people who need detail probably know better sources than NWS text forecasts.

Models have improved steadily while forecasters have not (meteorological cancer as per Snellman and many others).

Most of time human forecasters can’t do better than blended models on days 4 to 7. Waste of time for forecasters to try to improve base variables in the forecast except in frontal/convective situations and western US complex terrain effects.

No problems with blended model forecasts producing the forecast beyond day 3.

Fully automated forecasts are the future in public sector forecasts. In public sector forecasting emphasis ($$) has mostly been in NWP with little investment in humans.

It is sad but true – the negative momentum against human forecasters is unstoppable.

1 comment:

  1. Totally agree Bob. The golden age of forecasting where products were made by man is giving way to the amazing advancements in our lifetime of NWP. It is just a matter of time before most personalized forecast will be available from competing sources, if not already. The sometimes bizarre wording of the grid forecasts diminishes the credibility of the product. Staffing at the WFO level precludes the feasibility of usually just one forecaster assuring quality of the grid generated forecast being delivered for the entire area of their responsibility. You see that all the time in the diverse local climates of the mountainous west.