Variational data assimilation with deep prior – Arthur Filoche

Our next seminar will be held on Wensday the 5th of October 2022, at 14h30 ECT, -not on the 30th of September as previously advertised-, at the Pierre et Marie Curie campus of Sorbonne Université, in seminar room 105 of LiP6, located on the first floor of the corridor 25/26 (easier access through tower 26).

The seminar can also be followed remotely through zoom here: 
https://cnrs.zoom.us/j/99096155743?pwd=ZmlNb3BYWGFRWE5lSnlBdjNRRDRGZz09

Password : n5jBHd 

You can ask questions during and after the talk, in the slack channel.


Arthur Filoche’s talk is entitled:

« Variational data assimilation with deep prior »

Data Assimilation remains the operational choice when it comes to
forecast and estimate Earth’s dynamical systems, and proposes a large panel of methods to optimally combine a dynamical model and observations
allowing to predict, filter, or smooth system state trajectory.

The classical variational assimilation cost function is derived from
modelling errors prior with uncorrelated in times Gaussian distribution.
The optimization then relies on errors covariance matrices as
hyperparameters.
But such statistics can be hard to estimate particularly for background
and model errors. In this work, we propose to replace the Gaussian prior
with a deep convolutional prior circumventing the use of background
error covariances.

To do so, we reshape the optimization so that the initial condition to
be estimated is generated by a deep architecture. The neural network is
optimized on a single observational window, no learning is involved as
in a classical variational inversion.
The bias induced by the chosen architecture regularizes the proposed
solution with the convolution operators imposing locality.
 From a computational perspective, control parameters have simply been
organized differently and are optimized using only the observational
loss function corresponding to a maximum-likelihood estimation.

We propose several experiments highlighting the regularizing effect of
deep convolutional prior. First, we show that such prior can replace
background regularization in a strong-constraints 4DVar using a shallow
water model. We extend the idea in a 3DVar set-up using spatio-temporal
convolutional architecture to interpolate sea surface satellite tracks
and obtain results on par with optimal interpolation with fine-tuned
background matrix. Finally, we give perspective toward applying the same
method in weak-constrained 4DVar removing the need for model-errors
covariances but still enforcing correlation in space and time of model
errors.


Biographic notice:
Arthur Filoche is a Ph.D. student at the LiP6 of Sorbonne Université in France, under the supervision of Dominique Béréziat, Julien Brajard, and Anastase Charantonis. His research interests lie in combining deep learning and data assimilation

Eddy Detecting Neural Networks: harnessing visible satellite imagery and altimetry for operational oceanography – Evangelos Moschos

Our next seminar will be held on Wednesday the 11th of May 2022, at 14h00 ECT, at the Pierre et Marie Curie campus of Sorbonne Université, in seminar room 105 of LiP6, located on the first floor of the corridor 25/26 (easier access through tower 26).


The seminar can also be followed remotely through zoom here: https://zoom.us/j/98859268451 

You can ask questions during and after the talk in the slack channel: https://tinyurl.com/AI4CLIMATESLACK


Evangelos Moschos’ talk is entitled:

« Eddy Detecting Neural Networks: harnessing visible satellite imagery and altimetry for operational oceanography »

Mesoscale Eddies are oceanic vortices with a typical radius of the order of 20–80 km. They live for days, months or even years, trapping and transporting heat, salt, pollutants, and various biogeochemical components from their regions of formation to remote areas. Eddies have a primordial role in the oceanic circulation modifying the surface currents as well as the mixed layer depth of the ocean. Thus, eddy detection and tracking is an emergent thematic of operational oceanography with advances in the last 10 years.

Eddies can be tracked on altimetry-derrived Sea Surface Height (SSH) and geostrophic velocity currents through standard detection methods. However the strong spatio-temporal interpolation of the altimetric observations raises uncertainty of detection. Satellite imagery, on the visible and infrared spectrum, contains signatures of eddies, which despite their fine-scale resolution are too complex to be processed by geometric based methods.

Machine Learning/Computer Vision methods have proved very prominent in exploiting complex remote sensing information, such as the eddy signatures on satellite imagery. We build a Convolutional Neural Network which can accurately detect the position, shape and form of mesoscale eddies in satellite Sea Surface Temperature (SST) images. Our CNN only misses 3% of coherent structures, with more than 20km radius and a clear signature on the Sea Surface Temperature, compared with a 34% miss rate of standard eddy detection methods on altimetric maps. Additionally, while standard altimetric detection has a 10% false positive rate (“ghost eddies”) the neural network detects less than 1% of ghosts.

By combining detections on data stemming from different sensors (here SSH & SST) we do also provide a set of reference nowcast (real-time) detections with almost zero uncertainty. We focus here on an application of validation of operational numerical oceanographic models in the Mediterranean Sea through the comparison of their outputs with the reference detections, allowing to quantify their nowcast error and pick-and-choose between different numerical on a certain region.

Biographic notice:
Evangelos Moschos is a PhD student at the Laboratoire de Météorologie Dynamique, Ecole Polytechnique, France. After an engineering diploma in the Athens Polytechnic, he developed a keen interest in bridging computer vision methods with earth observation and operational oceanography. Co-founder of Amphitrite, a start-up harnessing AI to provide maritime stakeholders with real-time, reliable oceanic data.