Model output discussion - 20th Feb onwards

[IPredictASnowStorm]

reload after reload

[Paul]

Ignoring the specifics for now, as they're sure to change. Just picking out Thursday as a random day, and looking at how consistent the models have been so far in their forecasts for it, the last 8 runs have been fairly solid. 

Particularly from the GFS

Ensembles similar, particularly last 4 runs. 

ECMWF not as consistent, but last 3 runs singing from a close enough hymn sheet to the GFS. 

Growing confidence of the overall picture, but it's also easy to see how it wouldn't take the biggest change to take the coldest air away from the bulk of the country, so nothing set in stone.

[fergieweather]

12 minutes ago, Tim Bland said:

Ground temp could be around freezing along with -5c upper air temp and low DP. Hence why it shows lying snow in areas below 300m 

But it's by definition NOT 'lying snow'. It's the rainfall equivalent accumulation using a PPN total to snow algorithm. That isn't the same as products showing *settled* snow and herein lies one of the major issues with GFS snow charts. Only bespoke HRES products (such as UKV) can adequately synthesise that through detailed modelling of, eg, the forecasted depth temp/land class/topography. Yes, there will inevitably be times when with no marginality to mitigate settling, the GFS amounts will indeed replicate what we have underfoot. But we can't be prescriptive on any solution being the correct one at this range. 

[Paul]

12 minutes ago, fergieweather said:

But it's by definition NOT 'lying snow'. It's the rainfall equivalent accumulation using a PPN total to snow algorithm. That isn't the same as products showing *settled* snow and herein lies one of the major issues with GFS snow charts. Only bespoke HRES products (such as UKV) can adequately synthesise that through detailed modelling of, eg, the forecasted depth temp/land class/topography. Yes, there will inevitably be times when with no marginality to mitigate settling, the GFS amounts will indeed replicate what we have underfoot. But we can't be prescriptive on any solution being the correct one at this range. 

That's not strictly correct, it's not a PPN total to snow algorithm, it's from a data field directly in the GFS - water equivalent snow depth. So whilst higher res models (and not just Met office ones I may add!) are clearly more capable of picking out topographical features/land classes etc, that is a model forecast of snow depth rather than an attempt to show that just using precip totals and precip type.

[Bristawl Si]

6 minutes ago, fergieweather said:

But it's by definition NOT 'lying snow'. It's the rainfall equivalent accumulation using a PPN total to snow algorithm. That isn't the same as products showing *settled* snow and herein lies one of the major issues with GFS snow charts. Only bespoke HRES products (such as UKV) can adequately synthesise that through detailed modelling of, eg, the forecasted depth temp/land class/topography. Yes, there will inevitably be times when with no marginality to mitigate settling, the GFS amounts will indeed replicate what we have underfoot. But we can't be prescriptive on any solution being the correct one at this range. 

Thanks Ian, for clarifying. 

[fergieweather]

14 minutes ago, Paul said:

That's not strictly correct, it's not a PPN total to snow algorithm, it's from a data field directly in the GFS - water equivalent snow depth. So whilst, higher res models (and not just Met office ones I may add!) are clearly more capable of picking out topographical features/land classes etc, that is a model forecast of snow depth rather than an attempt to show that using precip totals.

Paul, the point I'm making here is that the granularity of GFS snow charts doesn't take account of the critical elements for *settling*. Its showing what would fall into a gauge (as snow... but could, for all we know, melt minutes later). To adequately resolve risk of snow actually lying on the deck, it needs modelling (not just UKV, as you note) that synthesises depth temps etc.

PS I think the sole issue where there is clear agreement is how the coming week quite evidently has a risk of snow for parts of the south....!

Edited February 20, 2016 by fergieweather

[Paul]

7 minutes ago, fergieweather said:

Paul, the point I'm making here is that the granularity of GFS snow charts doesn't take account of the critical elements for *settling*. Its showing what would fall into a gauge (as snow... but could, for all we know, melt minutes later). To adequately resolve risk of snow actually lying on the deck, it needs modelling (not just UKV, as you note) that synthesises depth temps etc. 

It's actually not quite doing that though as it does attempt to do just that - it'll often show categorical snow but won't show snow depth for that precip. As I understand it, the model also calculates snow melt in each time period to create the final figure,  based on my understanding the snow depth figure is essentially snow fallen minus snow melted. 

All that said though I do totally agree that even at 0.25 degrees it's very broad brushed and higher res models are the way better at modelling it. It's one of the reasons we've always marked those snow depth charts as experimental. 

[Frosty.]

Isn't it great that we are discussing potential accumulations of snow! The models look really interesting with a colder spell arriving next week from the north..let's get the cold air in place and see where this goes...I think somewhere in the UK is going to get a significant dumping next week.

[fergieweather]

2 minutes ago, Paul said:

It's actually not quite doing that though as it does attempt to do just that - it'll often show categorical snow but won't show snow depth for that precip. As I understand it, the model also calculates snow melt in each time period to create the final figure,  based on my understanding the snow depth figure is essentially snow fallen minus snow melted. 

All that said though I do totally agree that even at 0.25 degrees it's very broad brushed and higher res models are the way better at modelling it. It's one of the reasons we've always marked those snow depth charts as experimental. 

...but is it measuring snow melt based just on modelled 2m temperature? Dunno. Appreciate the parallel suite might be 'cleverer'. 

[Allseasons-Si]

23 minutes ago, Karl. said:

Isn't it great that we are discussing potential accumulations of snow! The models look really interesting with a colder spell arriving next week from the north..let's get the cold air in place and see where this goes...I think somewhere in the UK is going to get a significant dumping next week.

It is great Karl and it's about  bloody time that we see some snow now,it has been an eternity for some,especially the south.but it is still 5 days away and things will change but there is some concistancy to drop this low se in the favour that we want to see it,the charts have been churning out scenarios simular to the ones back in 2010/2013 but we need to get them to t0

on another note,there is another warming at the end of the gfs run,snow in april anyone?

 

Edited February 21, 2016 by Allseasons-si

[fergieweather]

PS Paul, as an aside: with UKV already sold to (at least 3) commercial forecast companies in UK and - I'm told - MOGREPS and GloSea5 soon to follow (this year?), is Netweather likely to become a subscriber for any of that output? I'd imagine any or all three suites would prove popular fare?

[Nouska]

25 minutes ago, Paul said:

It's actually not quite doing that though as it does attempt to do just that - it'll often show categorical snow but won't show snow depth for that precip. As I understand it, the model also calculates snow melt in each time period to create the final figure,  based on my understanding the snow depth figure is essentially snow fallen minus snow melted. 

All that said though I do totally agree that even at 0.25 degrees it's very broad brushed and higher res models are the way better at modelling it. It's one of the reasons we've always marked those snow depth charts as experimental. 

There are five snow parameters for the GFS on the Instant weather maps site - surface snow depth and surface snow depth change are two of them - these are usually quite different to the snow ones usually linked on the thread. Is this the model calculating the actual accumulation and then melt/compaction of previously fallen snow?

http://www.instantweathermaps.com/GFS-php/showmap-eurosfc.php?run=2016022018&time=3&var=SNOD&hour=192

NB if linking any images, they cannot be hotlinked - need to use something like Imgur.

An example 

[March]

47 minutes ago, Karl. said:

Isn't it great that we are discussing potential accumulations of snow! The models look really interesting with a colder spell arriving next week from the north..let's get the cold air in place and see where this goes...I think somewhere in the UK is going to get a significant dumping next week.

Snow has been just 5 days away since the first week of Jan! 

[Abyss]

GFS and UKMO look pretty similar at T144

GFS

UKMO

 

[Lukesluckybunch]

Yes both quite similar,ukmo slightly better I would say with cold more deep further south and east into the continent! But both are good!

[IDO]

The hi-res models for snow on Tuesday are underwhelming, Arpege has it in the Channel, wRF just one glancing blow and GFS all rain:

It looks like the further north the low is the more likely it will be rain.

As for the front crossing on Friday the GFS has stalled this just off Ireland and it no longer crosses the UK:

  Low Sunday: 

The low on Sunday in the south looks marginal at the moment. So for anyone north of the M4 a cold week but dry though showers possible from the northerly flow. ECM looks to stall that front at the end of the week as well as a wedge of HP repels it (what the majority of GEFS showed yesterday).

Again corrections still possible but a move towards a wedge of HP at T120 rather than the first diving low. All the GEFS agree on the second diving low that builds the disrupting trough, though scope for its path to change remains,  somewhere within the mean would be a good guess:

   London GEFS: 

A cool week ahead for the south but no cold blast by anyone's play book.

These temporary amplified Atlantic ridges do not look like pattern changes from the last 2-3 runs, just the repeated attempt by the Azores to ridge North, thwarted by the PV lobe to the NW. So a couple of topplers in the next 10 days which was what GFS was charting 5 days ago. The difference is that it is a colder upper flow so we keep in a more seasonal temp spectrum. 

 

Edited February 21, 2016 by IDO

[Interitus]

A quick clarification of the GFS snow forecast capabilities from the NCEP documentation -

Quote

Snow Cover

Snow cover is obtained from an analysis by NESDIS (the IMS system) and the Air Force, updated daily. When the snow cover analysis is not available, the predicted snow in the data assimilation is used. Precipitation falls as snow if the temperature at sigma=.85 is below 0 C. Snow mass is determined prognostically from a budget equation that accounts for accumulation and melting. Snow melt contributes to soil moisture, and sublimation of snow to surface evaporation. Snow cover affects the surface albedo and heat transfer/capacity of the soil, but not of sea ice. See also Sea Ice, Surface Characteristics, Surface Fluxes, and Land Surface Processes.

http://www.emc.ncep.noaa.gov/GFS/doc.php#snocov

The NESDIS snowcover data can be seen here for example - http://satepsanone.nesdis.noaa.gov/northern_hemisphere_multisensor.html

Other data assimilation is from station reports for example - not sure how different the various model initial conditions will be after the data assimilation stage.

As this section hints, the handling of snow is far from naive affecting surface characteristics (albedo) and processes - not all directly ice/snow related but come back into consideration when <100% snow cover -

Quote

Land Surface Processes

(The land surface was updated in the 2005 T382 GFS upgrade. View the specifics here or below.)
The Noah LSM includes the following specific upgrades, relative to the older OSU LSM: 
1) Increase from two soil layers (10, 190 cm thick) to four (10, 30, 60, 100 cm). 
2) Add frozen soil physics. Specifically, simulate the amount of liquid (unfrozen) soil moisture, including super-cooled liquid, as a new state variable. The difference of the traditional total soil moisture state (sum of liquid and frozen) and the liquid soil moisture state yields the amount of frozen soil moisture. Include the impacts of soil freezing and thawing on soil heat sources/sinks, vertical movement of soil moisture (including impact on root uptake of soil moisture), soil thermal conductivity and heat capacity, and surface infiltration of precipitation. 
3) Add the depth of the snowpack as a new prognostic state variable. Together with the traditional prognostic state of snowpack water equivalent (SWE), one can obtain the snowpack density as the ratio of SWE to the physical depth. 
4) Change the function for calculating snow cover fraction as a function of SWE. Require higher values of SWE to achieve given values of snow cover fraction, such as 100 percent snow cover. 
5) For snow cover fraction values of less than 100 percent, increase the contribution of the non-snow covered surface to the surface sensible, latent, and ground heat fluxes. In particular, drop the condition that all the evaporative demand at the surface be satisfied from the snowpack whenever any nonzero snowpack is present. This change and the previous change (#4) substantially eliminates the systematic bias of early snowpack depletion in the previous GFS. 
6) Allow spatially varying root depth (2 meters for forest vegetation classes, 1 meter for all other vegetation classes), rather than fixed 2 meters for all vegetation classes. 
7) Drop the lower bound of 0.50 for the fraction of green vegetation cover. This allows unconstrained seasonality of the green vegetation cover, still specified from daily interpolation of the 5-year, global monthly climatology of the AVHRR-based, 0.144-degree green vegetation fraction (GVF) fields produced by NESDIS. 
8] Modify the functional dependence of direct surface evaporation from the top soil surface. This modification parameterizes the thin, dry topsoil "crust" (less than 1 cm) that forms under drying conditions and amplifies the decrease of direct surface evaporation for increasingly drier soils. 
9) Enable all four rather than just two of the temporally varying resistance terms in the widely used canopy resistance approach, known as the "Jarvis" formulation, which is presented in detail in section 3.1.2 of Chen et al. (1996). Specifically, enable both the water vapor deficit term and the air temperature stress term, which respond to the near-surface air humidity and air temperature, respectively. Retain the soil moisture deficit and the solar insolation stress terms. 
10) Change the dependence of soil thermal conductivity on soil moisture to a less non-linear function that yields less extreme values for dry and moist soils. This significantly reduced the systematic bias of having too little (too much) ground heat flux in the presence of very dry (very moist) soils. 
11) Improve the ground heat flux formulations under conditions of non-deep snowpack and non-sparse vegetation, yielding less ground heat flux under non-sparse vegetation and more ground heat flux under shallow/patchy snowpack. 
12) Modify the formulation for the infiltration of precipitation into the soil column, by parameterizing the sub-grid variability of precipitation rate and soil moisture content. This change produces more surface runoff under non-saturated soil conditions and for lower precipitation rates.
13) Allow the surface emissivity to be less than unity over snowpack. Use the snow cover fraction to weight the surface emissivity between the value of 0.95 for snow cover fraction of 100 percent and 1.0 for zero snow cover.

http://www.emc.ncep.noaa.gov/GFS/doc.php#surflux

How accurate this is with regards to grid resolution and topography of our small islands is questionable, but it's a far from simplistic alternative to specialist products.

Edited February 21, 2016 by Interitus

[Man With Beard]

Great post Interitus. Thanks.

Onto the models, confidence is surely rising this morning in a genuine cold spell, and I must take issue with suggestions it won't be cold because we get 3 days of virtual nationwide 850s of sub -7C (as low as -9/-10 in places), and then aside a moment or two on the south coast, values never recover above -6 for almost a week - bearing in mind that once to cold gets in, it takes a mild push to get rid of it at ground level. Looking at the GFS 0.25, between this Wednesday and next Tuesday, surface maximums are between 3 and 5C but this would just be for an hour or two a day in sustained sunshine - generally, temps will be close to freezing throughout, and lower on days with cloud cover.

All in all, this is what I call a proper cold spell coming up, and what's interesting is the amount of mobility in the pattern - I think snow simply has to come at some point for most. Still a chance of low pressure being too far west allowing a SEly, but this possibility seems to be receding this morning.

[fergieweather]

6 minutes ago, Man With Beard said:

All in all, this is what I call a proper cold spell coming up, and what's interesting is the amount of mobility in the pattern - I think snow simply has to come at some point for most. Still a chance of low pressure being too far west allowing a SEly, but this possibility seems to be receding this morning.

In contrast to GFS, the 00z EC manages to keep any meaningful precip away from the south of the UK all through this week (and thus offers no snow at all), but it's clearly a fickle situation with low confidence in many aspects. What joy....

[Polar Maritime]

Yes agree there MWB a cold spell for sure with -7/-8 850's at least over the UK more or less throughout the run from the GFS, With surface cold building from N/NEly flows giving ice days with a bitter wind chill at times. Thermals at the ready!

Edited February 21, 2016 by Polar Maritime

[gottolovethisweather]

30 minutes ago, fergieweather said:

In contrast to GFS, the 00z EC manages to keep any meaningful precip away from the south of the UK all through this week (and thus offers no snow at all), but it's clearly a fickle situation with low confidence in many aspects. What joy....

 

 

No low confidence in Temperature predictions though I'd suggest. It will be cold nationwide, whether it can be deemed very cold is questionable but I've seen projected maximums next week of only 5c down South which is nippy. Of course, these are town/city Temperatures which suggests a cold enough spell for nationwide wintry precipitation if only it were apparent. This prospect in itself feels me with joy and is a good enough reason to be watching this week's forecasting prospects very closely.

 

Edited February 21, 2016 by gottolovethisweather

[Paul]

8 hours ago, fergieweather said:

PS Paul, as an aside: with UKV already sold to (at least 3) commercial forecast companies in UK and - I'm told - MOGREPS and GloSea5 soon to follow (this year?), is Netweather likely to become a subscriber for any of that output? I'd imagine any or all three suites would prove popular fare?

Funny enough I was wondering whether they were available as yet not so long ago. They do sound interesting, although potentially budget busting too! I'll be keeping my ear out for their availability for sure.

[karlos1983]

7 minutes ago, Paul said:

Funny enough I was wondering whether they were available as yet not so long ago. They do sound interesting, although potentially budget busting too! I'll be keeping my ear out for their availability for sure.

I'd pay a premium to see those charts  bet there wouldn't be many other sites that could compete with that.

 

[tinybill]

http://www.meteociel.fr/modeles/gfse_cartes.php?mode=2&ech=6

 

 

morning  all  the  gfs is looking interesting this  morning   dont put the winter stuff away just yet

Edited February 21, 2016 by tinybill

[winterof79]

Liable to change but I would not fancy a trip to the south of Spain thanks