From the Adjusting Our Expectations Department -- A Weather Model Incident
[climate-change
mathematics
modeling
]
Climate change is making weather prediction a bit more complicated these days, and climate scientists are having a hard time keeping up. Last week, a climate scientist reported that an important weather prediction model was giving absurd results for West Coast weather that was ten days away.
Posting as heads-up to weather modeling world (not as a realistic prediction of future conditions): the (new) GFS has occasionally been spitting out completely absurd surface temperatures for CA's Central Valley. 18z run today shows 128-130F. Anyone know what's up? pic.twitter.com/qmhP3eeGuF
— Daniel Swain (@Weather_West) June 18, 2021
Notable here is a modeling professional describing model output as “not realistic” and “completely absurd”. This means that model was doing something unexpected and that its output was unbelievable.
Granted, a 10-day forecast is less reliable than a 10-hour forecast. Mathematical models are better at accurately predicting events that are near their input data. In the case of temporal weather forecasts, this means the near-future predications are more accurate than far-future predictions. And certainly the modeler here knows that. And yet he describes the model output as “completely absurd” and puts out a call for help in his efforts to identify the source of the problem.
Three days later, he’s made little progress to his goal of finding a problem.
Seems as if the ECMWF is now doing something similar. Either both models share some kind of similar surface scheme bias/error that involves sporadic positive/self-reinforcing surface temp feedback during extreme events (still by far most likely), or...hmm. https://t.co/exwvekuPSW
— Daniel Swain (@Weather_West) June 21, 2021
This tweet is a bit hard to parse. My takeaway is this: two different models are giving the same prediction (now a seven-day farecast). If these two models are built differently, then this decreases the likelihood that the model output is the result of some “error” or “mistake”. It is evidence supporting the idea the original forecast was not a mistake.
Paging anyone in numerical weather modeling world who has insights, esp. folks at @NOAAResearch, @NWSWPC, @NWSEMC, etc. This is turning into a potential public communications problem, especially because there could actually be very high-impact heat event buried in these errors.
— Daniel Swain (@Weather_West) June 21, 2021
Soon, the modelers are shrugging about the stubbornly consistent model output.
Update: models not really backing off on this one. Highest confidence in extreme and possibly historically hot temps this weekend/next week is across Oregon, but could also extend across Washington and northern CA depending on how things evolve. #CAwx #ORwx #WAwx https://t.co/kMLQKhkVQ1
— Daniel Swain (@Weather_West) June 22, 2021
A day later, the weather forecaster are going with the flow.
A historical heat wave is coming for the Pacific Northwestern states next week. High temperatures to soar well into the 100Fs... pic.twitter.com/RzasEt7JVp
— Doc V (@MJVentrice) June 23, 2021
Even the National Weather Service is rolling with it, now.
Extreme heat wave expected this weekend, likely breaking several records. Make sure you're prepared for the heat!
— NWS Portland (@NWSPortland) June 23, 2021
Stay hydrated
Complete outdoor activities in the morning, or wait until the evening
Do NOT leave kids or pets in a parked car#orwx #wawx #pdxtst #heatwave pic.twitter.com/rZ0FJgK3aF
And online news services are writing about the extremem predictions.
A "historic" and potentially deadly heat wave is on tap for the Pacific Northwest this weekend with never-before-seen temperatures possible in cities like Portland and Spokane. https://t.co/5BXuIMsKlH
— Axios (@axios) June 23, 2021
Looks like we’re in for a hot one!
Update: NOAA, too (24 June)
This modeling “issue” has even made it into a NOAA “Area Forecast Discussion” for the Portland, OR, area. They refer to it as “hype”.
There has been a lot of (well-justified) hype lately regarding the
high temperatures shown by many models (and the National Blend of Models) Sunday and Monday, with some showing all-time high temperature records being obliterated by 5 to 10 degrees. While models seem to be well in agreement of +28 deg C to +31 deg C for 850 mb temps during that period. The all-time record for 850 mb temperatures at KSLE per SPC sounding climatology is +28.3 deg C. So, just considering 850 mb temps, we would be in the ballpark of all-time record highs. But there are additional considerations...
So remarkable.
Update: Climate.gov, writes (2 July)
As we move past this remarkable climate event, people are writing about it. Here’s what Tom Di Liberto writes for Climate.gove: “Astounding heat obliterates all-time records across the Pacific Northwest and Western Canada in June 2021”.
It’s a good write up, and it comes close to mentioning the disproportionate impact heat waves have on our neighbors who have been marginalized through policies like redlining and the like. This is an issue we need to be thinking about and making policy that damps the negative impact of climate change on those neighbors.
Di Liberto’s article does link to an article on work being done by NOAA to understand climate impact on these populations. The article “NOAA and communities to map heat inequalities in 11 states” is worth a read.