OceanStWx

Seems the most likely given the strength of the winds, but so far all I've seen is still under investigation. I think they said the Sunset Fire in Hollywood was suspicious.

Congrats Isles of Shoals.

I think Pacific Palisades in particular was a perfect storm of hazards are natural, disasters are man-made. It is an old neighborhood, one without the same building standards for fire as newer CA construction (fire resistant building material, safe zones around structures where no fuel sources can be planted, etc). They had prepared with three 1,000,000 gal water tanks for just that neighborhood, but the fire was so large and intense that they couldn't refill them fast enough. Fire fighting all over the city reduced the water pressure necessary to pump water back up into Pacific Palisades. It is also outside of the typical fire season and CA was lacking the usual resources at their disposal (seasonal fire fighters, and big tanker planes shared with Australia currently in their fire season). Then you have the age old dispute between CA and feds about who is to manage the land area (since the feds own way more of the land than CA does), since nobody wants to actually fund it. Altadena is its own unique thing too because they are usually sheltered from a typical Santa Ana. But this was a different beast altogether. More of a pure downslope windstorm than a gap wind event like most Santa Anas. So that was not an area where people were accustomed to fire. And there is just the fact that 80-100 mph winds are going to make fire do things that you just can't stop. It's not like the Palisades fire was contained, it just ran out of stuff to burn on its way to the ocean.

The interesting thing about these CA fires is it's all driven by boom/bust rainfall patterns. Two years ago they got abundant rain, which drove fine fuel growth, which dried out quickly as they entered another short term drought this year. Another interesting aspect is that conditions were so extreme, that even prescribed burning of these fuels would likely not have mattered. 60-90 mph wind gusts will push a fire through any type of fuel.

Just a six fingered circle jerk waiting to happen.

For the time being, AI is going to be trained off of what has happened in the past. And much of that reanalysis type stuff. In order to really suss out the details AI will need the full resolution of the atmosphere, and so it will suffer the same problems as traditional modeling in that respect. I think we'll see far more advancement in the pattern recognition type machine learning (think Colorado State severe weather stuff).

Pretty sure PWM did the same that summer. They just couldn't get cooler than 70 at night, but the thermometer was definitely reading about a degree too high.

In an era when our "cold" months are like -1.0 below normal, February 2015 was -12.8 at BOS.

Honestly not very often. When it starts getting into bias correction, that's when it becomes a problem. Or if we start setting records with data we think is suspect.

Seems like a great way to have @dendrite hijack some spare parts for his backyard.

BOS does not average a daily snowfall of 12" per season, it's something less than 1 per season. But if you accounted for snowfall splitting midnight, it's probably around 1 per season. Never had BOS had more than 2 daily snowfalls of 12" in a season (at that time it was just 60-61 and 77-78, 17-18 has now joined that list). 2015 had 4!

Good news @dendrite, we just ripped the temp sensor out of DAW altogether. The part is on order, but rather than send out bad data we're opting for no data (which may actually speed up the delivery of the part).

It was truly incredible. Every time Kevin cracked open an IPA, BOS got another 18".

I think Sam Lillo reran that pattern and found it was like a 1:1,000,000 chance occurrence for BOS.

I think the ECMWF is showing one possible way forward. You have a good model and you come up with an AI version that can run quickly, and you do large ensembles quickly, many times per day. You can end up with some really good probabilistic guidance that way.

The plan was to run with an entirely new model (RFS), but I think the teams have realized that the FV3 core does not scale well to the resolution that 60 hour run should have. So I think we're trying to come up with a new core that will lead to more successful results.

I remember the final nail in the coffin was when the NAM gave in and forecast 4"+ QPF for PWM. The old EE rule in full effect.

We are! The ultimate goal is to develop a new coupled weather model with an ensemble suite that is larger than the current GEFS. In addition there will be a more mesoscale model (think hybrid NAM/HRRR) which runs to 60 hours.

There were some LOL NWS snow forecasts based on those Euro runs. Just never a good idea to put 3 ft in your forecast until you see the whites of its eyes.

Is this like the spank bank for weather modeling?

A distant cousin giving you the middle finger of snow.

Let's turn it into a NorLun.

Takes AEMATT to a new level.

ASOS is the final say on QPF, but we always use the observers to check the snow reports, just in case a spider fart caused the sensor to throw in a SNBxx group.

No days off for her and @tunafish