2024 - The First Annual Data Assimilation Grand Prix

2024 - The First Annual Data Assimilation Grand Prix#

Below you can see a recap from last year’s First Annual Data Assimilation Grand Prix.

(And yes, the name of the contributor in first place was indeed a Rick Roll)

The winners#

Our winners in the category ensemble mean RMSE were:

πŸ₯‡ 1st place: Simon van Leent / Rick Roll (particle filter)

πŸ₯ˆ 2nd place: Hakan Zor / Zorro (ensemble Kalman filter)

πŸ₯‰ 3rd place: Bagas Mahadika A. P. (particle filter)

Our π˜„π—Άπ—»π—»π—²π—Ώπ˜€ in the category π—²π—»π˜€π—²π—Ίπ—―π—Ήπ—² π—₯π— π—¦π—˜ are:

πŸ₯‡ 1st place: Simon van Leent / Rick Roll (particle filter)

πŸ₯ˆ 2nd place: Bagas Mahadika A. P. (particle filter)

πŸ₯‰ 3rd place: Daniel van der Hoorn / Lightning McQueen (particle filter)

Honorable mentions go to Anne Grooten (particle filter), whose filter maintained a solid second place until it tragically diverged in the last 25 timesteps of the (unseen) validation period. Truly an emotional rollercoaster.

The model#

So, what was the model actually doing? The true dynamics follow several steps:

  1. Initialization: Place a random circle in the domain.

  2. Move a random arc distance along the circle’s circumference.

  3. Create a new circle which shares a tangent with the old circle. Switch the path over to the new circle. Return to step 2.

A demonstration of this principle is shown below:


../_images/2024_Grand_Prix_model.gif

Fig. 80 The true model followed a circle, and randomly switched to a new circle that shares a tangent.#


Contents