Tuesday Sloppy Tuesday! And Finally A Look At Model Resolution

Today

As promised another sloppy day, hint, probably more of the same Wednesday.

Here is short term high-resolution liquid totals from 6 am this morning until midnight everything should be done by then.

Screenshot 2020-03-10 at 12.39.13 PM

Screenshot 2020-03-06 at 9.31.20 AM

And snow

Screenshot 2020-03-10 at 12.40.34 PM

Screenshot 2020-02-22 at 1.24.12 PM

As I said, probably more of the same tomorrow, I will talk more about that, well, probably tomorrow!

 

The Long-Awaited Simple Yet Complex Discussion on Model Resolution

Over the last couple of years, I have talked about model resolution and model accuracy as if it is the sole problem with our forecasting problems in SW Colorado. The reality is it is just not that simple. Land-based data collection linked into the NWS will be a problem for us forever.

That being said there have been some very recent developments in the last few weeks that will aid the supercomputers behind the US GFS model and the European model. probably not coincidentally NOAA and The ECMWF announced within days of each other, large investments to greatly increase the supercomputing power of the GFS and the Euro model.

NOAA especially is being immodest with their claims. They are promising the most accurate Weather model in the world within three years. They are going to triple their supercomputing power to a maximum capacity of 40 petaflops. Although that should fire up the geeks out there, I want to try to put this in perspective for the rest of us.

One petaflop is equal to one thousand million million (ie one quadrillion) floating data point operations per second! Will that get the job done around here? It certainly will help, but as I have said, bad data in, bad data out. That being said, land-based data collection is being less heavily relied upon as we flood the atmosphere with satellites. That still doesn’t change the fact that we cannot get a radar. BTW, I don’t know what exactly is happening with that, it is the topic that no one talks about anymore, and I am as in the dark as you are. It might be a great topic for someone at the Herald to follow up on.

Talk about going off on a tangent… Let’s talk about the simple and complicated issue of model resolution. I want to keep things pretty basic, I want you to be able to look at one the weather maps and say, Wow! look at all that snow, oh wait, that is the low-resolution Canadian Model, what does the higher-resolution Euro say? Maybe I should look at the very high-resolution 3km NAM to see what it says.

The three models I use the most as you know are the Euro, GFS, and Canadian. At least up until 12-24 hours before the precipitation is supposed to start. After that, I usually rely on a blend of the Nam 3km and the Euro.

Simply put resolution indicates how large of an area the model samples all parameters to determine the forecast. The LOWER the number the HIGHER the resolution.

Here are the resolutions (in kilometers) for the common models I use.

Canadian 15km  (9.3 miles)

US GFS 13km  (8.1 miles)

Euro 9km (5.6 miles)

US High-Resolution NAM-nest 3km (1.9 miles)

All of the models use different shapes of the grids that cover the size/length of the resolution they are equipped with. Since they vary, for illustrative purposes only, let’s use a circle setting the radius equal to the resolution distance.

So let’s take a look at the terrain differences, using downtown Durango as a focal point, feeding in the resolutions among the models. Hopefully, this will start to make immediate sense to you.

Here is the “grid” (again for illustrative purposes only) starting with the Canadian’s 15 km resolution model. You will see the little red (crude) circle below left, which is at 10,100 feet in elevation. The model is trying to balance elevations between roughly 6,200 feet and 10,100 feet. The yellow circle is imperfect because Google Earth Pro tracks over terrain.

Durango15km

Next up the GFS at 13 km resolution, that little red circle top right is 9,600 feet.

Durango13km

Here is the Euro 9km, the red circle top right is at 8,100 feet.

durango9km

Finally, the NAM 3km, that circle left side in blue is 7,600 feet.

Durango3km

It’s Tuesday, things are looking more cautiously optimistic for, just maybe, some significant precipitation starting Friday. So now that you have seen differences in resolution let’s apply that using the three different models (the 3km model only stretches to 60 hours so there is no data yet).

The following maps I am just using in conjunction with what we just learned about resolution, don’t take these as a solid forecast yet.

I have another helpful hint to “find” Durango on these maps below using this as a guide.

La Plata County has a unique shape at its western border, I have circled it in red, you should be able to always identify it on the weather maps. If you draw a line through the county right in the middle it will always run right through downtown Durango. Here it is on Google maps. You can see the airport SE of downtown, which is where the models list the numerical output.

LaPlata

Canadian precipitation beginning Friday morning until the last drop.

Screenshot 2020-03-10 at 11.54.03 AM

Screenshot 2020-03-06 at 9.31.20 AM

The GFS

Screenshot 2020-03-10 at 12.19.40 PM

Screenshot 2020-03-06 at 9.31.20 AM

Euro

Screenshot 2020-03-10 at 12.24.58 PM

Here is the obvious question, I will ask it for you. Why not just use the highest resolution model won’t that be the most accurate? The answer is not that simple. You have to keep in mind that these models do not just predict rain and snow, they predict storm track, they predict upper-level trough position, cold fronts, jet orientation. and many many other parameters. So you have to give enough room in the resolution to pick up those features. That is why I never, ever use a high-resolution (3km) model longer than 24 hours out.

The European model has always scored highest in skills scores, but they don’t judge it based on precipitation, they judge it on trough position at 18,200 feet otherwise known as 500 millibars.

That is why I use all three models from 7-5 days out, then I try to weed them out based on consistency, At some point within 24 hours of arrival, I start to focus on the 3km model, blended (usually) with the Euro. Hopefully, this has provided you with a better understanding of model resolution. Some of those maps look pretty encouraging but it is still too early to get too excited yet! Next Update Wednesday

If you have not yet read about the contest yet, make sure you check it out by clicking here: End Of The Season Announcement-The Contest

Click here and donate for a chance to win!

Click here to contact

Leave a Reply