January 2024 Weather in the PNW (Part II)

[TigerWoodsLibido]

47 minutes ago, Front Ranger said:

Is that where the regional line is? I thought it was Roseburg.

I'd say down to at least the entire Willamette Valley, which is a tributary of the Columbia River. Umpqua Basin is debatable since it can be tough to get a surface boundary over the Kalapuya pass to Drain, Elkton, Yoncalla etc. And the Umpqua goes directly into the Pacific Ocean. If it makes it to Roseburg then it's unquestionably regional and to Medford is extremely regional. 

[BLI snowman]

16 minutes ago, SeanNyberg said:

my God, this forum has devolved into the worst of the worst. It's always emotional and reactionary in here, but it seems that the rational and seasoned members have pulled back to make way for the kids.

I'm going to step back for the day too. It's just too much uninformed nonsense.

I am excited for a lot of you guys, because it seems like this is your first time tracking a PNW winter storm. I remember my first time.

Post of the decade. Fred should make this required reading for every new member and pin this at the very top of the forum so we can be reminded. 

[RCola]

2 minutes ago, BlvuSumit said:

This would be a compromise that would make 80% of this forum happy. It's so smoothed out by the two camps of solutions though that it gives a false impression of a regional event unfortunately. 

My money is on a GEM solution that pounds PDX and Seattle is on the fringe.

[iFred]

28 minutes ago, westcoastexpat said:

A word of caution with the snow totals maps, especially for Seattle south, the GFS has a period of sleet/mixed precip and will count it all as accumulating snow even though it would actually do the reverse (i.e. eat away at accumulations)

We've seen rain get counted in those totals before as well. I think your earlier call for a trace at best is still the best call.

[MR.SNOWMIZER]

2 minutes ago, BLI snowman said:

Post of the decade. Fred should make this required reading for every new member and pin this at the very top of the forum so we can be reminded. 

Funny thing is what he is complaining about is the magic of the forum and why it works.

[Ken in Wood Village]

Here  is the next storm 🌬🌨

[Front Ranger]

7 minutes ago, BlvuSumit said:

How much of this blend is GFS?

[iFred]

23 minutes ago, SeanNyberg said:

my God, this forum has devolved into the worst of the worst. It's always emotional and reactionary in here, but it seems that the rational and seasoned members have pulled back to make way for the kids.

I'm going to step back for the day too. It's just too much uninformed nonsense.

I am excited for a lot of you guys, because it seems like this is your first time tracking a PNW winter storm. I remember my first time.

I've sent you the SSH keys and a link to the cloud project. You are now the owner of the forum. Being serious, check the email you registered with. You'll need to accept the cloud project details before tomorrow or the link will expire. Once there, you will need to add a payment method. You can delete all of the snapshots and the two VMs used for tooling. The folder structure is setup to allow you to migrate to a bucket or block storage if you're so daring.

[RCola]

3 minutes ago, Front Ranger said:

How much of this blend is GFS?

Probably 3 extra shots of Druncle.   But seriously, the weightings change depending on the current skills of the models. GFS did score in the eastward progression of the TPV, so maybe it is getting a higher weight.

 

[Perturbed Member]

18z GRAF. Looks like it sobered up. Doesn't matter too much though, I think we have a pretty decent idea of the range of possibilities tonight. 

[LowerGarfield]

39 minutes ago, Snowlova said:

I think a part 3 to this forum topic tomorrow or tonight will be fitting to complete imo what will be one of the most legendary arctic blast forum discussions ever, as a trilogy

Then someone will make an ill-advised low budget part 4 featuring the icon that no one likes.

[Kolk1604]

1 minute ago, Front Ranger said:

How much of this blend is GFS?

I posted this earlier, probably 50 pages ago lol 

31 models, maybe it's even? 

 

 

[Rubus Leucodermis]

2 minutes ago, Front Ranger said:

How much of this blend is GFS?

Probably too much.

[SnowWillarrive]

Just now, Perturbed Member said:

18z GRAF. Looks like it sobered up. Doesn't matter too much though, I think we have a pretty decent idea of the range of possibilities tonight. 

That looks similar to the WRF

[TacomaWx]

10 minutes ago, MR.SNOWMIZER said:

I said this 4 days ago and he attacked me saying i had no hope and knew nothing. I let it go because i didn't care.

He had been annoying me for awhile too…but I didn’t want to cause an issue. Hopefully that’s his last post ever. 

[NorthBend RainEnthusiast]

12 minutes ago, Front Ranger said:

How much of this blend is GFS?

"The NBM ingests and bias corrects a host of weather elements and determines the optimum weights to be assigned to each model when generating a final blended forecast. Simply put, the NBM takes whatever models are available from 31 different model systems (for rainfall forecasts a total of 171 inputs) and combines the information into a single forecast. Intense research into model performance over time has given today’s meteorologists a baseline of known model biases.  Verification of the NBM output has shown that this methodology consistently outperforms individual model runs over time leading to a more accurate and consistent dataset of available guidance."

"On an hourly basis, the NBM starts with the hourly HRRR model , adds each of these other listed models above on a regular schedule, and compares their forecasts to something called URMA, which is an error-corrected measurement of actual observed conditions six hours previously.  It basically sees which model is doing better from actual measurements six hours earlier and weights that towards a better composite forecast. "

More detail here: 

"At the most fundamental level the NBM blending technique is relatively straightforward and can be broken down into three primary steps: (1) Collating model input data, (2) calibrating each model’s input using the most recent URMA observations, and (3) finally blending the bias corrected guidance usually through objective weights (Fig. 5).

Currently, the NBM uses almost 30 different model inputs ranging from the global scale to mesoscale from a variety of Meteorological Centers (Fig. 6) and runs hourly. When model data arrives, the NBM processes that data and it is absorbed in the next hour’s NBM solution. Given the sheer number of input models there are times of the day where several models are processed for any particular hour, while at other hours only two or three short-term models may be processed. For this reason there may be some NBM cycles where NBM guidance beyond a specific time horizon is not updated.

For many of the NBM weather elements, a decaying average bias correction algorithm (Cui et al. 2012) is used to calibrate the input model data prior to the blending process (with one notable exception —NBM QMD precipitation products). This algorithm works by comparing a model’s recent past performance with the verifying URMA observation and adjusts the bias correction factor accordingly. These bias correction factors are updated daily as a function of model, cycle, weather element, and projection. This same methodology is also applied when generating MAEfactors used in objectively weighting each model’s contribution to the NBM’s final blended solution. Many of the continuous weather elements are bias corrected and objectively weighted in this manner.

However, for those non-Gaussian weather elements (e.g., ceiling height and visibility) which are not suited to be bias corrected in a linear fashion, expert weights are used. These expert weights have been predetermined through various retrospective tests and are likewise a function of model, cycle, weather element, and projection. Following the blending process, select weather elements are quality control checked for consistency, such as the temperature always being equal to or greater than the dew point temperature. Care has also been taken to minimize temporal inconsistencies in NBM guidance by temporally interpolating model data into hours for which no model data is available."

[Frontal Snowsquall]

12 minutes ago, BlvuSumit said:

Regional snow event!

[Cloud]

29 minutes ago, SeanNyberg said:

my God, this forum has devolved into the worst of the worst. It's always emotional and reactionary in here, but it seems that the rational and seasoned members have pulled back to make way for the kids.

I'm going to step back for the day too. It's just too much uninformed nonsense.

I am excited for a lot of you guys, because it seems like this is your first time tracking a PNW winter storm. I remember my first time.

Cool story bro! 

[Front Ranger]

7 minutes ago, RCola said:

This would be a compromise that would make 80% of this forum happy. It's so smoothed out by the two camps of solutions though that it gives a false impression of a regional event unfortunately. 

My money is on a GEM solution that pounds PDX and Seattle is on the fringe.

I think some PS folks will be pleasantly surprised by what happens with the Arctic front. 

PDX to OLM definitely in prime position after that.

[GHweatherChris]

31 minutes ago, SeanNyberg said:

my God, this forum has devolved into the worst of the worst. It's always emotional and reactionary in here, but it seems that the rational and seasoned members have pulled back to make way for the kids.

I'm going to step back for the day too. It's just too much uninformed nonsense.

I am excited for a lot of you guys, because it seems like this is your first time tracking a PNW winter storm. I remember my first time.

 

[Doinko]

1 minute ago, Front Ranger said:

I think some PS folks will be pleasantly surprised by what happens with the Arctic front. 

PDX to OLM definitely in prime position after that.

Hoping for a good storm here. I think I'm in a decent spot

[RentonHill]

[Brian_in_Leavenworth]

From what I've read, the NBM isn't simply taking every ensemble run and making a mean out of all of them. Nor does it "favor" one model over another or weigh one model over another.

It sounds like it weighs all of them, but instead of a simple mean, it takes into account all known biases and tendencies of each model and then blends it all in.

And over time it has improved, partly because models have improved, and partly because it better understands model tendencies and biases and better takes them into account 

 

[RCola]

9 minutes ago, NorthBend RainEnthusiast said:

"The NBM ingests and bias corrects a host of weather elements and determines the optimum weights to be assigned to each model when generating a final blended forecast. Simply put, the NBM takes whatever models are available from 31 different model systems (for rainfall forecasts a total of 171 inputs) and combines the information into a single forecast. Intense research into model performance over time has given today’s meteorologists a baseline of known model biases.  Verification of the NBM output has shown that this methodology consistently outperforms individual model runs over time leading to a more accurate and consistent dataset of available guidance."

"On an hourly basis, the NBM starts with the hourly HRRR model , adds each of these other listed models above on a regular schedule, and compares their forecasts to something called URMA, which is an error-corrected measurement of actual observed conditions six hours previously.  It basically sees which model is doing better from actual measurements six hours earlier and weights that towards a better composite forecast. "

Awesome info. Thanks!

If I understand correctly, what I said above is not quite right. In reality, GFS would only be given a higher weight in NBM only if it was outperforming other models in the past 6 hours. 

Honestly, I get the need to check in every 6 hours for the mesoscale models like HRRR, but the global models should have an entirely different skill score based on a time scale of days and not hours. I wonder what the mathematical details really are to the weights. 

[TigerWoodsLibido]

5 minutes ago, Frontal Snowsquall said:

Regional snow event!

Almost but it just misses the south valley so a widespread event but not quite regional.

[MWG]

From NWS of Medford. I'll take 60% Bob!

[iFred]

Just now, Brian_in_Leavenworth said:

From what I've read, the NBM isn't simply taking every ensemble run and making a mean out of all of them. Nor does it "favor" one model over another or weigh one model over another.

It sounds like it weighs all of them, but instead of a simple mean, it takes into account all known biases and tendencies of each model and then blends it all in.

And over time it has improved, partly because models have improved, and partly because it better understands model tendencies and biases and better takes them into account 

 

The bias accounted for are very very very very simplified. The usefulness though is that because the heavy lifting is moved to other models, it had the best (re: safest) "modeled" terrain influences. I remember coming across a paper when the HRRR came out stating that some NWS offices were using the NBM for nowcasting.

[Frontal Snowsquall]

8 minutes ago, NorthBend RainEnthusiast said:

"The NBM ingests and bias corrects a host of weather elements and determines the optimum weights to be assigned to each model when generating a final blended forecast. Simply put, the NBM takes whatever models are available from 31 different model systems (for rainfall forecasts a total of 171 inputs) and combines the information into a single forecast. Intense research into model performance over time has given today’s meteorologists a baseline of known model biases.  Verification of the NBM output has shown that this methodology consistently outperforms individual model runs over time leading to a more accurate and consistent dataset of available guidance."

"On an hourly basis, the NBM starts with the hourly HRRR model , adds each of these other listed models above on a regular schedule, and compares their forecasts to something called URMA, which is an error-corrected measurement of actual observed conditions six hours previously.  It basically sees which model is doing better from actual measurements six hours earlier and weights that towards a better composite forecast. "

Heavily weighted towards the NOAA models. I’m surprised there’s no UKMET in there.

[Front Ranger]

9 minutes ago, NorthBend RainEnthusiast said:

"The NBM ingests and bias corrects a host of weather elements and determines the optimum weights to be assigned to each model when generating a final blended forecast. Simply put, the NBM takes whatever models are available from 31 different model systems (for rainfall forecasts a total of 171 inputs) and combines the information into a single forecast. Intense research into model performance over time has given today’s meteorologists a baseline of known model biases.  Verification of the NBM output has shown that this methodology consistently outperforms individual model runs over time leading to a more accurate and consistent dataset of available guidance."

"On an hourly basis, the NBM starts with the hourly HRRR model , adds each of these other listed models above on a regular schedule, and compares their forecasts to something called URMA, which is an error-corrected measurement of actual observed conditions six hours previously.  It basically sees which model is doing better from actual measurements six hours earlier and weights that towards a better composite forecast. "

Thanks. Secret sauce. 

[SnowWillarrive]

9 minutes ago, TacomaWx said:

He had been annoying me for awhile too…but I didn’t want to cause an issue. Hopefully that’s his last post ever. 

I’m fine with him. It’s fun having different characters here and he is enthusiastic about Puget Sound.

[RCola]

6 minutes ago, Front Ranger said:

I think some PS folks will be pleasantly surprised by what happens with the Arctic front. 

PDX to OLM definitely in prime position after that.

I hope your gut instinct is right. I missed out on the arctic front events here the past few years, so I have little reference to work from.

 

What I do understand are overruning events, where the common theme is a more expansive precip shield. If we get a GEM track, the lower half of the sound could still get flakes Saturday. 

[MikeInEverett]

2 minutes ago, SnowWillarrive said:

I’m fine with him. It’s fun having different characters here and he is enthusiastic about Puget Sound.

I mostly agree, but that last post was just so out of left field.

[RCola]

13 minutes ago, NorthBend RainEnthusiast said:

"The NBM ingests and bias corrects a host of weather elements and determines the optimum weights to be assigned to each model when generating a final blended forecast. Simply put, the NBM takes whatever models are available from 31 different model systems (for rainfall forecasts a total of 171 inputs) and combines the information into a single forecast. Intense research into model performance over time has given today’s meteorologists a baseline of known model biases.  Verification of the NBM output has shown that this methodology consistently outperforms individual model runs over time leading to a more accurate and consistent dataset of available guidance."

"On an hourly basis, the NBM starts with the hourly HRRR model , adds each of these other listed models above on a regular schedule, and compares their forecasts to something called URMA, which is an error-corrected measurement of actual observed conditions six hours previously.  It basically sees which model is doing better from actual measurements six hours earlier and weights that towards a better composite forecast. "

More detail here: 

"At the most fundamental level the NBM blending technique is relatively straightforward and can be broken down into three primary steps: (1) Collating model input data, (2) calibrating each model’s input using the most recent URMA observations, and (3) finally blending the bias corrected guidance usually through objective weights (Fig. 5).

Currently, the NBM uses almost 30 different model inputs ranging from the global scale to mesoscale from a variety of Meteorological Centers (Fig. 6) and runs hourly. When model data arrives, the NBM processes that data and it is absorbed in the next hour’s NBM solution. Given the sheer number of input models there are times of the day where several models are processed for any particular hour, while at other hours only two or three short-term models may be processed. For this reason there may be some NBM cycles where NBM guidance beyond a specific time horizon is not updated.

For many of the NBM weather elements, a decaying average bias correction algorithm (Cui et al. 2012) is used to calibrate the input model data prior to the blending process (with one notable exception —NBM QMD precipitation products). This algorithm works by comparing a model’s recent past performance with the verifying URMA observation and adjusts the bias correction factor accordingly. These bias correction factors are updated daily as a function of model, cycle, weather element, and projection. This same methodology is also applied when generating MAEfactors used in objectively weighting each model’s contribution to the NBM’s final blended solution. Many of the continuous weather elements are bias corrected and objectively weighted in this manner.

However, for those non-Gaussian weather elements (e.g., ceiling height and visibility) which are not suited to be bias corrected in a linear fashion, expert weights are used. These expert weights have been predetermined through various retrospective tests and are likewise a function of model, cycle, weather element, and projection. Following the blending process, select weather elements are quality control checked for consistency, such as the temperature always being equal to or greater than the dew point temperature. Care has also been taken to minimize temporal inconsistencies in NBM guidance by temporally interpolating model data into hours for which no model data is available."

Amazing info. Even more helpful. Definitely checking out that paper later. It does sound like the global models do get treated differently to some extent.

[Port Angeles Foothiller]

3 minutes ago, SnowWillarrive said:

I’m fine with him. It’s fun having different characters here and he is enthusiastic about Puget Sound.

Yeah, but he publicly sh*t his pants with that last post.  

[theRunner]

I'm pretty uninformed but essentially GFS favors SEA region for snow and ECMWF favors PDX? If ECMWF were to be correct would Seattle essentially see no precip? Praying for the NBM lol 

[RCola]

5 minutes ago, Frontal Snowsquall said:

Heavily weighted towards the NOAA models. I’m surprised there’s no UKMET in there.

No ICON either, which eliminates the farthest south solution we had. Also no AI models yet. I expect that to change in 5 years.

[Parrot]

6 minutes ago, RCola said:

I hope your gut instinct is right. I missed out on the arctic front events here the past few years, so I have little reference to work from.

 

What I do understand are overruning events, where the common theme is a more expansive precip shield. If we get a GEM track, the lower half of the sound could still get flakes Saturday. 

I'm no expert but as far as I was aware, arctic front events range from very marginal to localized at best. And the models are only calling for a trace with the front. Getting snow going into an event is not necessarily an event caused by the arctic front. But please corect me if I'm a bit off base...

[Cloud]

What are the D and E For the euro? 

[MR.SNOWMIZER]

Just now, theRunner said:

I'm pretty uninformed but essentially GFS favors SEA region for snow and ECMWF favors PDX? If ECMWF were to be correct would Seattle essentially see no precip? Praying for the NBM lol 

Even with a solution like the euro Seattle would see some snow. The storms that carry a southern track tend to spread precip further north than what the model shows. 

[Port Angeles Foothiller]

2 minutes ago, theRunner said:

I'm pretty uninformed but essentially GFS favors SEA region for snow and ECMWF favors PDX? If ECMWF were to be correct would Seattle essentially see no precip? Praying for the NBM lol 

Basically. Seattle will see some snow from the arctic front in both solutions.