Pretty pictures and saying goodbye

Pretty pictures and saying goodbye

So it actually happened. I’m sitting here, on a rainy day wearing a sweater and thinking summer is over. The summer of HPC is over at least, if the sun ever comes out again is questionable as well (thank you to London weather).
But what have we been up to in the last few weeks of the project?

Performance of MPAS atmosphere for a regional cut. Notable is the superlinear speedup and efficiency above 60% even after the run time plateaus.

We ran our final experimental set up of MPAS. This time instead of running full global simulations we only used a regional cut, this is a fairly common method in atmospheric modelling and allows for fast running of very detailed set ups. We chose a cut above the western Mediterranean, with varying cell sizes between 60 and 3km, running between 1 and 8 hours. The results of this showed were very promising for MPAS. we saw both good strong (see the figure to the right) and weak scaling behaviour, which allowed us to simulate 8 hours of weather in under 20 minutes! We go into more detail in our final presentation that you can watch here.

Animation of changing wind speeds for a 8 hour simulation starting at 6pm on January 1st 2010. Note: negative wind speed meaning southward direction

This run was also our chance to test out the new visualisation script we developed in our project. MPAS used to rely on NCL, a plotting language created by NCAR who also developed MPAS, for any visualisation. They recently announced however a stop of development on NCL to move all plotting towards Python. Since a model is of limited use without the possibility of visualising the outputs in a good way that humans understand, we decided to write visualisation scripts that can be used for MPAS. This worked very well and allowed us to create beautiful animations of the regional experiment. My favourite one is shown here, this is the wind speeds of an 8 hour run. The main feature is a persistently strong wind in the Gulf of Lion. We believe this is the Mistral, a persistent wind pattern bringing cold strong winds to the South of France. Moreover, when looking at where the Alps should be, the wind almost seems blocked by them. As an Earth Scientist, I really enjoy staring at this, finding patterns and trying to explain what they show. It also shows very well just how powerful these kind of weather models are.

Overall, I really enjoyed my time working on this project. I have never imagined myself working on a supercomputer, especially not on one of the most powerful machines in Europe. Wanting to go into research in my later career, this helped me discover a whole new side of research and I’m very thankful for getting this opportunity. Having had only very little knowledge about computing, it was at times quite challenging but I learned so much about programming. Also, whenever I truly didn’t know how to solve any issues, our amazing mentors Evgenij Belikov and Mario Antonioletti helped guide us through it all.
Even though we sadly weren’t able to work on this project in Edinburgh, but had to work remotely it was a lot of fun and a great show of how important cooperation across Europe is for research.
So this is it from me for the Summer of HPC, thank you to all of you who have read through my articles and came with me on this journey!

Tagged with: , , , , , ,

Leave a Reply

Your email address will not be published. Required fields are marked *


This site uses Akismet to reduce spam. Learn how your comment data is processed.