Which Model for rainfall - A laymans guide ?

Discodog · 26-11-2011 11:37am #1

I monitor rainfall forecast daily as I spend a lot of time outdoors. Recently I have noticed big variations between different forecast sites - especially yr.no & XC. I know that XC uses the latest GFS.

I use the Weatheronline expert charts but it seems that the only rainfall prediction is via the GFS model. I also use the EHMI site where their rainfall prediction is based on HIRLAM.

So who do you think is the best at predicting where & when rain will fall ? Is there one model that is accepted as being more accurate ?

Su Campu · 26-11-2011 11:52am

Weatheronline also has the NAE, which is the hi-res son of the UKMO, as well as all the others. Compare this with the GFS and HiRLAM and you should get a good overview of what should happen.

It should be noted, however, that no model is absolutely perfect when it comes to precip forecasts, but the LAMs will give the best output.

maquiladora · 26-11-2011 11:55am

I like the NAE (available on weatheronline). It has a higher resolution that the global models like the GFS.

Discodog · 26-11-2011 12:08pm

Thanks. I can see what you mean when I compare the GFS & NAE. Is a HIRLAM rainfall prediction available on Weatheronline - is it called something else ? I find that the EHMI HIRLAM rainfall animation to be pretty accurate - it's a pity that the map is a bit small.

For me it's the timing of the rain that is more important than amounts. I am forever looking at the radar & trying to extrapolate as to when or whether a lump of rain will hit my location.

maquiladora · 26-11-2011 12:11pm

Discodog wrote: »

Thanks. I can see what you mean when I compare the GFS & NAE. Is a HIRLAM rainfall prediction available on Weatheronline - is it called something else ? I find that the EHMI HIRLAM rainfall animation to be pretty accurate - it's a pity that the map is a bit small.

For me it's the timing of the rain that is more important than amounts. I am forever looking at the radar & trying to extrapolate as to when or whether a lump of rain will hit my location.

HIRLAM :

http://www.meteociel.fr/modeles/hirlam_cartes.php?ech=6&code=0&mode=2&carte=0

Discodog · 26-11-2011 12:18pm

Thanks but I find that the EHMI map is much clearer & easy to use. Other web sites would do well to copy this:

http://www.emhi.ee/index.php?ide=19,394,416,418

EDIT: HIRLAM may not be so good ! The EHMI shows Ireland as dry & we have torrential rain here !

I have also just realised that NAE only covers 48 hours ahead.

According to yr.no. I might as well give up on next week whereas XC is far more optimistic.

Su Campu · 26-11-2011 12:43pm

Discodog wrote: »

Thanks. I can see what you mean when I compare the GFS & NAE. Is a HIRLAM rainfall prediction available on Weatheronline - is it called something else ? I find that the EHMI HIRLAM rainfall animation to be pretty accurate - it's a pity that the map is a bit small.

The NAE's right there on the same page as the GFS!!

http://www.weatheronline.co.uk/cgi-bin/expertcharts?LANG=en&MENU=0000000000&CONT=ukuk&MODELL=nae&MODELLTYP=1&BASE=-&VAR=prty&HH=3&ARCHIV=0&ZOOM=0&PERIOD=&WMO=

Discodog · 26-11-2011 12:46pm

Su Campu wrote: »

The NAE's right there on the same page as the GFS!!

http://www.weatheronline.co.uk/cgi-bin/expertcharts?LANG=en&MENU=0000000000&CONT=ukuk&MODELL=nae&MODELLTYP=1&BASE=-&VAR=prty&HH=3&ARCHIV=0&ZOOM=0&PERIOD=&WMO=

I know - I referred to looking at it in a previous post

It was HIRLAM that I was trying to find..

Su Campu · 26-11-2011 12:47pm

Discodog wrote: »

I know - I referred to looking at it in a previous post

It was HIRLAM that I was trying to find..

Oh duh.....I'm still half asleep!!! :rolleyes:

Su Campu · 26-11-2011 12:51pm

By the way, yr.no is the HiRLAM for Europe, and ECMWF for the rest of the world.

http://www.yr.no/place/Norway/Sogn_og_Fjordane/V%C3%A5gs%C3%B8y/Kr%C3%A5kenes/advanced_map.html

Discodog · 26-11-2011 12:53pm

Su Campu wrote: »

Oh duh.....I'm still half asleep!!! :rolleyes:

I was walking the dogs along the Galway coast - that wakes you up

I guess that part of my frustration is trying to extrapolate the radar. Last night I was sure that a shower would miss me so I walked the dogs .....& came home drenched :eek:

Last Monday a huge lump of rain moved up towards Galway & I was about to give up & go home - but at the very last minute it changed direction !

Redsunset · 26-11-2011 12:53pm

I don't get what your saying Disco,

NAE shows exactly what was forecast.
NOON

MET RADAR

Discodog · 26-11-2011 12:57pm

Su Campu wrote: »

By the way, yr.no is the HiRLAM for Europe, and ECMWF for the rest of the world.

http://www.yr.no/place/Norway/Sogn_og_Fjordane/V%C3%A5gs%C3%B8y/Kr%C3%A5kenes/advanced_map.html

I used to rely on yr.no. but their longer range forecasts seem way off recently. There have been a series of forecasts where they, at the week end, predict a dire week ahead but then it's not so bad.

It may sound daft but XC seem to be the optimists & yr.no. the pessimists.

Discodog · 26-11-2011 1:04pm

redsunset wrote: »

I don't get what your saying Disco,
NAE shows exactly what was forecast.

Yes but I have only discovered NAE today

But, as I have already pointed out, NAE only covers 48 hours. On Sunday evening I will be trying to plan the week. I will be trying to guess which days will be best for which task. So I have to use sites that give longer range predictions.

In any given week I might have say three days where I need some dry weather. So it's case of trying to balance the forecasts & the timing of rain can be critical.

Redsunset · 26-11-2011 1:07pm

Oh ok.

As far as Im aware no higher resolution model goes past 48 hours.

Maybe Met Eireann have something but the time gets too far ahead after 48 hours to be called hi res in my book

Discodog · 26-11-2011 1:15pm

I appreciate that Sun eve is too early to be sure about rainfall on the following Thursday/Friday. I was just wondering if one model might tend to be more accurate. I guess that I need to monitor them over a period of time & see.

I can't recall a thread where anyone has said that they use a particular site because it is the most reliable. Of course there is MTC forecast but that is not detailed enough to really help.

Su Campu · 26-11-2011 2:33pm

Just note that all the different sites are all posting the same model data, just with different graphics, and there is no human input. So no one site will be better than another. The only exception is the Hirlam, which does have slightly different output depending on the country it's generated in. The Spanish AEMET hirlam on meteociel goes out to 60 hours.

Redsunset · 26-11-2011 2:41pm

The Spanish AEMET hirlam on meteociel goes out to 60 hours.

Oh yes i forgot bout that one.:)

Discodog · 26-11-2011 2:54pm

Su Campu wrote: »

Just note that all the different sites are all posting the same model data, just with different graphics, and there is no human input. So no one site will be better than another. The only exception is the Hirlam, which does have slightly different output depending on the country it's generated in. The Spanish AEMET hirlam on meteociel goes out to 60 hours.

But there is a huge difference in the graphics & predicted rainfall amounts so there must be some individual interpretation.

That 60 hr is too depressing - Monday looks dire !

Su Campu · 26-11-2011 3:06pm

Discodog wrote: »

But there is a huge difference in the graphics & predicted rainfall amounts so there must be some individual interpretation.

There's no individual interpretation. The various models each have their own set of characteristics, resolution, physics, etc. Some are hydrostatic, some are sigma, some are spectral, etc., so hence the variationn in output. Part of a forecaster's job is to estimate which ones are the more likely outcomes for a particular setup, and indeed some may even disregard the models completely if they're totally off the wall. But this only happens when generating human forecasts, not the charts we see on meteociel, etc., which are just a set of numbers automatically plotted in graphical format on a map.

The ECMWF is usually taken as the most reliable overall, and the Hirlam takes its initial data from the ECMWF. The NAE takes its data from the UKMO, the WRF takes its from the GFS (see the UCD link posted in the Useful Sites thread). At the moment, the setup is highly uncertain when looking for fine details for the end of the week, as the strength of the strong Atlantic jet at the moment is open to small changes, and there are few upper data there where they're needed. These variations extend to the surface charts and magnify, leading to the variations in rainfall predictions a few days down the road. If we had a different setup the models may be a lot more in agreement, but hey, thet's the best we can do for the time being.

Discodog · 26-11-2011 3:12pm

Su Campu wrote: »

T

The ECMWF is usually taken as the most reliable overall, and the Hirlam takes its initial data from the ECMWF.

Weatheronline does not appear to have a precipitation section for ECMWF. Do they produce one or does one have to use HIRLAM for rain ?

Su Campu · 26-11-2011 3:18pm

Discodog wrote: »

Weatheronline does not appear to have a precipitation section for ECMWF. Do they produce one or does one have to use HIRLAM for rain ?

I know, that's annoying. The ECMWF release very few freely-available data to the public, but MET Eireann do have ECMWF precipitation in their 5 Day Charts section.

Discodog · 26-11-2011 3:24pm

Su Campu wrote: »

I know, that's annoying. The ECMWF release very few freely-available data to the public, but MET Eireann do have ECMWF precipitation in their 5 Day Charts section.

I know that I criticise ME but I have to say that the 5 day is poor. I thought that it would be invaluable but it has turned out to be really inaccurate & I have given up using it. By the way if you have influence in ME please ask them to highlight the outline of Ireland in the 3 hour - the graphics mask the coastline.

For example the ME 5 day shows Galway as dry at 12.00 Monday. The Spanish HIRLAM shows Galway as deluged at the same time. So if I am planning work ME indicates that I might get a least a few hours whereas HIRLAM says no chance.

Su Campu · 26-11-2011 3:26pm

If it's inaccurate then it's not their fault, they're just publishing the model data.

Redsunset · 26-11-2011 3:28pm

The Iceland site is ECM
http://www.vedur.is/vedur/spar/atlantshaf/#teg=urkoma

Discodog · 26-11-2011 3:33pm

redsunset wrote: »

The Iceland site is ECM
http://www.vedur.is/vedur/spar/atlantshaf/#teg=urkoma

Thanks. My Sunday evening planning is going to take even longer as I whiz around websites.

Su Campu · 26-11-2011 3:34pm

It's the exact same data, just a different site.

Redsunset · 26-11-2011 3:39pm

Su Campu wrote: »

It's the exact same data, just a different site.

Of course it is,but from what I gather Disco is trying to find a site more appealing to the their eye.

Su Campu · 26-11-2011 3:43pm

If you zoom out the ME charts to Atlantic View then it's very clear, and the geographical area is more relevant to Ireland. Sometimes I think there's a refusal by some to acknowlege when ME do something right.

Redsunset · 26-11-2011 3:51pm

Not by me anyway,I think Met Eireann do a fantastic job and yes i do often use the Atlantic charts on their site to get a feel for what lies ahead.

No ME bashing here.

Deep Easterly · 26-11-2011 4:06pm

For what it's worth, my first port of call would be to look at the EMHI & DMI models. As Su noted above, each (European) country has its own slightly different version of the HIRLAM model which tends to be very specific for that particular country.

Regarding the EMHI, I tend to find this version of the HIRLAM model is not so great handling frontal precipitation events such as is the case today! whereas the DMI version handles the same far better I think.

Having said that, the EMHI model I think is far superior in handling convective, mesoscale type features both in terms of intensity and location.

I have noticed though that the lower res (30km grid) ECMWF model can sometimes be quicker than even the HIRLAM model on picking up on sudden changes at a synoptic level that might cause an area of precipitation to take a more different path than what its higher res counterpart might have shown (an example of this happened earlier this week in fact)

As has already been said, the NAE is another quality hi res model and does quite well with convective features especially I think. The GFS I think is improving a lot over the last couple of years re prep forecasts and is always worth having a look. In fact, I have noticed that this model has sometimes handled rain events better than all the others over the last year.

Su Campu · 26-11-2011 4:28pm

redsunset wrote: »

Not by me anyway,I think Met Eireann do a fantastic job and yes i do often use the Atlantic charts on their site to get a feel for what lies ahead.

No ME bashing here.

Sorry Red, I didn't mean you, though it may have come across like that!

Discodog · 27-11-2011 9:47pm

When do Weatheronline update - they are still showing the 12 noon chart. Shouldn't the 1800 be up by now ?

OK now this is getting really confusing. Both Yr.no. & DMI use HIRLAM yet yr.no show Galway as dry tomorrow & DMI show it as wet !

Harps · 27-11-2011 9:56pm

Assume you mean the GFS charts? The 18z isn't released until about 9.30pm, takes about an hour for the whole thing to come out

Su Campu · 27-11-2011 11:58pm

Discodog wrote: »

When do Weatheronline update - they are still showing the 12 noon chart. Shouldn't the 1800 be up by now ?

OK now this is getting really confusing. Both Yr.no. & DMI use HIRLAM yet yr.no show Galway as dry tomorrow & DMI show it as wet !

The yr model seems to be an earlier run than the DMI one, but taking their output for midday tomorrow, they're pretty similar.

Yr.no

DMI

Discodog · 28-11-2011 1:10am

This, issued at 7pm, gives a different picture. A little rain midday but basically dry till 1800. XC (gfs) gives a similar forecast whilst NAE has it as wet. I am going to wake up & see !

Discodog · 28-11-2011 9:57am

On the GFS & NAE charts they state that the numbers refer to how much rain will fall in a 6 hour period. When does the period run from in relation to the forecast time ?

Su Campu · 28-11-2011 11:02am

If the chart is for 12UTC, then it means the total rainfall that will have fallen in the 6 hours up to then (i.e. from 0600 to 1200).

Deep Easterly · 28-11-2011 12:39pm

Discodog wrote: »

This, issued at 7pm, gives a different picture. A little rain midday but basically dry till 1800. XC (gfs) gives a similar forecast whilst NAE has it as wet. I am going to wake up & see !

Usually when yr.no update around 7pm or 7am it is the 'rest of the world' data that has been updated (ie, that latest ECMWF run). Those hourly graphs are what the HIRLAM model they use are forecasting and I think are only updated twice a day.

Bottom line is though that you can look too much into something. Rainfall models will never ever be totally spot on so it is fruitless trying to find one particular model that does. They only serve as a guide and an estimation of where rain, and how heavy it may be, at any one time. This is why we need human input and interpretation of these such as we see on weather forecasts etc. People may criticize weather forecasts for being to general but that is the best they can ever be because it is not a precise science!

If it is any help, the DMI 6hr maps give a pretty good idea of where rainfall may occur during each 6 hour period with a 24hr period:

http://www.dmi.dk/dmi/index/danmark/vejrkort/eurprec1.htm

PS, Since you do work outdoors, it is no harm actually looking at the sky also, always gives a good idea if it about to rain or not. Better than any hi res model!!

Discodog · 28-11-2011 5:46pm

A big thank you to everyone for your detailed answers - I should have posted this thread years ago. Hopefully it will guide others. I do study the Sky & cloud patterns etc but especially on the West Coast things move so fast. Also a lot of systems seem to slide up almost parallel to the coast so it can be touch & go whether Galway is wet or dry. Without the radar I would have no idea whether the rain is a 20 min shower or a 2 hour burst !.

Su Campu · 28-11-2011 7:52pm

Another way of predicting rainfall extent is by using the 700 hPa relative humidity charts. For frontal rain anyway, the rain usually occurs within areas where the 700 hPa RH is 90% or greater.

Case in point - this evening:

Discodog · 30-11-2011 8:00pm

I emailed yr.no & asked about the discrepancy between their map & the pictogram. I got a refreshingly honest reply:

Hello.

Yes, this is not good!
The maps are a composite of different model runs, and there is a delay in these data. So while the meteograms are updated frequently, the precipitation in these maps are at certain times from an older model.

In the near future we'll change all our maps at yr.no, and we'll try to sort this out.

Best regards
Havard Larsen
met.no

Which Model for rainfall - A laymans guide ?

Comments