How Did The Models Perform?

[Zenarcher]

Thats brilliant Weather Master.

 

Thumbs up.

 

Thanks also forgot to mention in the post on how it works. Each timeframe 24, 48, 72 hours etc wasn't tested just once or twice but every day through December this allowed me to gather a good accurate and consistent record of how they were performing.

 

First I have to say good work to Weather Master. That was a very fun post.

 

But it is true that model verifications are made on a much larger scale, and not point wise, because that makes no real sense whatsoever. But I am sure you know that.

 

Yes its a different way to do it I think its a good way of seeing how well the models can predict the weather for certain locations. Perhaps next time I could run the experiment over the course of 2 months for more data to analyse.

[Recretos]

That is a good idea. Tho I think 1 month is quite enough, since each month has its own climatological features, and you will have less work to do. If its not too much of a problem, you could take 5 or 10 random points, and just make the average result of all the locations. Would certainly give more merit to the end result.

 

And again, great work.

[Zenarcher]

That is a good idea. Tho I think 1 month is quite enough, since each month has its own climatological features, and you will have less work to do. If its not too much of a problem, you could take 5 or 10 random points, and just make the average result of all the locations. Would certainly give more merit to the end result.

 

And again, great work.

 

Thanks for this idea next time I'll choose a few different locations and test them over the course of a month then make up the graphs and post them here. I'll try and get more weather models involved as well.

[Vorticity0123]

First, what I want to say is (falling in repetition), nice work! Really interesting to have such model comparison material available.

 

However, I was wondering, what kind of data do you use for comparison? The reason for the question is that each individual parameter is influenced by different meteorological scales and synoptics. To give a few examples:

 

The temperature at 850 hPa is mainly influenced by large-scale weather systems, and not by small-scale boundary-layer processes (i.e. it is not directly influenced by the surface).

 

The surface temperature, on the other hand, is influenced by synoptic processes as well as by boundary layer dynamics (presence/absence of St-cloudiness, radiation, to name but a few).

 

If you would like to expand your experiment a bit, you could try to split some variables based on the processes that influence it (if possible to acquire them). In this way, it might be possible to judge models on for example boundary-layer processes and large-scale synoptics, and high-pressure weather and low pressure weather. Although it could be pretty difficult to assess what variable belongs where, it could deliver great insights in the various domains of model verification .

Edited January 25, 2015 by Vorticity0123