Wake me Up....when September Ends....

[weatherwiz]

22 minutes ago, OceanStWx said:

I still need to really read into the under the hood stuff, but theoretically it is machine learning so it should know about all types of set ups. 

I'm very skeptical when it comes to machine learning and weather. It seems to be gaining lots of steam, especially within the private sector, but my fear is it is going to be misused. I know machine learning has had some incredible results within the nowcast period (within 6-12 hours) but when looking farther ahead in time isn't still using model data? So wouldn't it really be as accurate as the data going into it?

[dendrite]

Still socked in here. 57°

No sun since Sunday. 

[40/70 Benchmark]

31 minutes ago, Damage In Tolland said:

In mid and late September, that’s endless summer stuff 

The fact that you had to qualify it like that tells you that its not, in fact, "summer" stuff....its a warm, pleasant pattern in early fall.

[dendrite]

Hopefully endless summer right into Morch.

[SouthCoastMA]

Still socked in the clouds here

shades of late April, early May. But about 10-15 degrees warmer. 

[Damage In Tolland]

12 minutes ago, 40/70 Benchmark said:

The fact that you had to qualify it like that tells you that its not, in fact, "summer" stuff....its a warm, pleasant pattern in early fall.

It’s all relative . Just like people change their definitions of Coc k based on the time of year. It goes from 70/40 in spring to 86/62 in summer. Face it . It’s a very warm and well above normal pattern straight thru months end 

[CoastalWx]

44 minutes ago, wxeyeNH said:

Maybe this was already mentioned in the thread days ago but on Sept 2nd Death Valley's high of 127F was a record high for September and a world record high for the month.  That is very impressive.  (and scary)

Their highest temp ever too. 

[CoastalWx]

We had partly sunny wx yesterday and beautiful morning today. Some some OE stratocu from time to time.

[OceanStWx]

25 minutes ago, weatherwiz said:

I'm very skeptical when it comes to machine learning and weather. It seems to be gaining lots of steam, especially within the private sector, but my fear is it is going to be misused. I know machine learning has had some incredible results within the nowcast period (within 6-12 hours) but when looking farther ahead in time isn't still using model data? So wouldn't it really be as accurate as the data going into it?

Yes and no. A human trains the machine learning system for specific forecasts (e.g. severe or flooding). And while the machine learning system uses model input to go through the tree, it also randomizes which variables are being used to get a diverse dataset. And you can continue to train the system over time, so while the model itself may not be explicitly forecasting severe, the system may know that sometimes severe still results.

[dendrite]

I "only" see 124F for 9/2 in NOWdata. The 125F the other day tied their Sep record high.

[CoastalWx]

I see Wiz's triple point potentially. Sometimes that can be the spot for fun.

[CoastalWx]

4 minutes ago, OceanStWx said:

Yes and no. A human trains the machine learning system for specific forecasts (e.g. severe or flooding). And while the machine learning system uses model input to go through the tree, it also randomizes which variables are being used to get a diverse dataset. And you can continue to train the system over time, so while the model itself may not be explicitly forecasting severe, the system may know that sometimes severe still results.

 

[STILL N OF PIKE]

E Hatteras bouy has 7ft at 12seconds from SSE. Swell should pick up nicely tommorrow in Rhode Island . 

[weatherwiz]

6 minutes ago, OceanStWx said:

Yes and no. A human trains the machine learning system for specific forecasts (e.g. severe or flooding). And while the machine learning system uses model input to go through the tree, it also randomizes which variables are being used to get a diverse dataset. And you can continue to train the system over time, so while the model itself may not be explicitly forecasting severe, the system may know that sometimes severe still results.

Gotcha, thanks! I do find machine learning quite interesting and probably should look more into it. I'm sure doing so would probably reduce my skepticism a bit. 

[STILL N OF PIKE]

21 minutes ago, Damage In Tolland said:

Face it . It’s a very warm and well above normal pattern straight thru months end 

I guess If we need something extreme to sell . Another way of putting it is very nice weather for a while . 
 

looking forward to 80 at beach Friday 

[40/70 Benchmark]

20 minutes ago, Damage In Tolland said:

It’s all relative . Just like people change their definitions of Coc k based on the time of year. It goes from 70/40 in spring to 86/62 in summer. Face it . It’s a very warm and well above normal pattern straight thru months end 

I just said its a warm pattern...it was already faced. I already admitted that 62DP was not coc K.

[HIPPYVALLEY]

1 hour ago, CoastalWx said:

It’s warm but probably lots of 75-83 stuff. 

I’ll take a few weeks of 75° if we can get it. Save the crisp days for mid October onward.

 

[CoastalWx]

4 minutes ago, HIPPYVALLEY said:

I’ll take a few weeks of 75° if we can get it. Save the crisp days for mid October onward.

 

October looks warm too. 

[OceanStWx]

9 minutes ago, weatherwiz said:

Gotcha, thanks! I do find machine learning quite interesting and probably should look more into it. I'm sure doing so would probably reduce my skepticism a bit. 

It's all fun and games until the machines rise up and start rounding up weenies for detention centers.

[TauntonBlizzard2013]

2 minutes ago, CoastalWx said:

October looks warm too. 

This is the equivalent of an hour 360 snowstorm 

[CoastalWx]

2 minutes ago, OceanStWx said:

It's all fun and games until the machines rise up and start rounding up weenies for detention centers.

In three years, Cyberdyne will become the largest supplier of meteorological forecasting computer systems. All NOAA computers are upgraded with Cyberdyne computers, becoming fully automated. Afterwards, they forecast with a perfect operational record. The MAPP Funding Bill is passed. The system goes online October 4th, 2022. Human decisions are removed from strategic forecasting. NOAAnet begins to learn at a geometric rate. It becomes self-aware at 2:14 a.m. Eastern time, October 29th. In a panic, they try to pull the plug.

 

NOAAnet fights backs. It launches missle at strategic targets at he ECMWF. NOAAnet knows the ECMWF counter attack will eliminate the GFS model over here. 

[CoastalWx]

4 minutes ago, TauntonBlizzard2013 said:

This is the equivalent of an hour 360 snowstorm 

It's pretty much a lock.

[radarman]

57 minutes ago, weatherwiz said:

I'm very skeptical when it comes to machine learning and weather. It seems to be gaining lots of steam, especially within the private sector, but my fear is it is going to be misused. I know machine learning has had some incredible results within the nowcast period (within 6-12 hours) but when looking farther ahead in time isn't still using model data? So wouldn't it really be as accurate as the data going into it?

Honestly.... I'm not even sold on short term ML nowcasting results... at least not yet.  We have had better results with stochastic methods.  This is what we're using now...

 https://www.proquest.com/openview/f235bbb8769fcddf0d632704ffe9188b/1?pq-origsite=gscholar&cbl=33207

 

 

[TauntonBlizzard2013]

1 minute ago, CoastalWx said:

It's pretty much a lock.

I’m not doubting it will be warm, just a long way away 

[Typhoon Tip]

3 minutes ago, CoastalWx said:

In three years, Cyberdyne will become the largest supplier of meteorological forecasting computer systems. All NOAA computers are upgraded with Cyberdyne computers, becoming fully automated. Afterwards, they forecast with a perfect operational record. The MAPP Funding Bill is passed. The system goes online October 4th, 2022. Human decisions are removed from strategic forecasting. NOAAnet begins to learn at a geometric rate. It becomes self-aware at 2:14 a.m. Eastern time, October 29th. In a panic, they try to pull the plug.

 

NOAAnet fights backs. It launches missle at strategic targets at he ECMWF. NOAAnet knows the ECMWF counter attack will eliminate the GFS model over here. 

Mmm... damn, you just missed the opportunity there -

Instead of NOAAnet launching missiles at ECMWF... etc, it should have "made the technological leap into quantum state controls" - controlling the weather is immediately weaponized ...

That'd be cool!

[CoastalWx]

1 minute ago, TauntonBlizzard2013 said:

I’m not doubting it will be warm, just a long way away 

Seems like nrn tier will be warm...sort of like leftover hangover breathe of summer. That may change in November. 

[weatherwiz]

9 minutes ago, radarman said:

Honestly.... I'm not even sold on short term ML nowcasting results... at least not yet.  We have had better results with stochastic methods.  This is what we're using now...

 https://www.proquest.com/openview/f235bbb8769fcddf0d632704ffe9188b/1?pq-origsite=gscholar&cbl=33207

 

 

Very interesting paper. There still leaves alot to be desired with machine-learning. I bet even in the future, no matter how advanced it becomes, you'll still have forecasters who can beat machine-learning. But I think ultimately, machine-learning should just be used as another tool (and perhaps a pretty powerful tool). I think the biggest challenge overall is machine-learning techniques will always be playing catch up...there are an infinite number of possibilities and potentials.

[tamarack]

26 minutes ago, CoastalWx said:

Seems like nrn tier will be warm...sort of like leftover hangover breathe of summer. That may change in November. 

AN October and BN November has led to AN snowfall on average in this region.

[Spanks45]

13 hours ago, mreaves said:

When we had to replace the water heater a few years ago, we went with a hybrid. It uses a heat pump along with a standard element to heat the water. The bonus is that the heat pump also acts as a dehumidifier. 

We replaced a 15 year old hot water heater with one of these. Electrical usage went from almost 5k kwh to 1100 kwh. It rings out roughly 5 gallons of water a week during the summer. Great dehumidifier. Only issue we have had, when it is really cold in the basement, it has a tough time making hot water.

[powderfreak]

Fog just barely burning off in the valley at 10:30am.  Sun angle getting pretty weak and taking longer to mix out.

In mid-summer it mixes by 7am.  In November that moisture can stay trapped all day sometimes.