Assimilation of drifter data in the East Mediterranean Sea

Participants : Julien Brajard, Milad Fakhri [CNRS, Lebanon] , Daniel Hayes [Oceanography Centre, Cyprus] , Leila Issa [Lebanese American University, Lebanon] , Laurent Mortier [LOCEAN] , Pierre-Marie Poulain [Oceanography Institute of Trieste, Italy] .

Surface velocity fields of the ocean in the Eastern Levantine Mediterranean are estimated by blending altimetry and surface drifters data. The method is based on a variational assimilation approach for which the velocity is corrected by matching real drifters positions with those predicted by a simple advection model, while taking into account the wind effect. The velocity correction is done in a time-continuous fashion by assimilating at once a whole trajectory of drifters using a sliding time window. A divergence-free regularization term is added to the cost function mnimized during the assimilation process in order to constrain the velocity field. First results show that with few drifters, the method improves the estimation of the surface velocity: an eddy between the Lebanese coast and Cyprus is better assessed and the values of velocities along the Lebanese coast are more accurate.

Traffic simulation

Participants : Vivien Mallet, Vincent Aguiléra [CEREMA] , Ruiwei Chen [CEREA] .

The ANR project ESTIMAIR aims at propagating uncertainties in the complete simulation chain of air quality at urban scale. A key step in the chain lies in traffic assignment and the computation of the corresponding emissions. We take part to the simulation of traffic in the streets of Clermont-Ferrand metropolitan area, with the dynamic traffic assignment model LADTA. The simulations are evaluated against observations from loop counters and also against the simulations of the reference static model VISUM.

From the traffic assignment, the emissions are computed for nitrogen dioxide and particulate matter, using COPERT IV formulae. Preliminary work shows large uncertainties in the emissions due to the fleet composition.

Observation of noise pollution

Participants : Vivien Mallet, Raphaël Ventura, Valérie Issarny [MiMove] , Pierre-Guillaume Raverdy [Ambientic] , Fadwa Rebhi [MiMove] .

Exposure to noise pollution is highly variable in space. As a consequence, it is very difficult to determine individual exposure using only numerical simulations of noise levels. Together with the MiMove Inria project-team, we take part to the SoundCity project that aims at collecting noise observations from smartphones and better evaluating the individual exposure. We assist MiMove in the development of an Android application that automatically senses noise along the day and collects the data (when the user agrees) for the improvement of simulated noise maps. Clime especially contributes to the calibration of the application. Comparisons between the measurements of smartphones and a sound meter allow us to estimate the bias of the main smartphones available on the market.

The SoundCity application was launched in July 2015 with Bernard Jomier, deputy mayor responsible for health, disability, and relations with Paris public hospital system, during a press conference organized by Paris City. The application received a positive coverage in the media, so that the application gained about 2500 users. About one million observations are collected every four days and ongoing work tries to process these data to correct Paris noise maps.

Evaluation of fire models

Participants : Jérémy Lefort, Vivien Mallet, Jean-Baptiste Filippi [CNRS] .

In the field of forest fires risk management, important challenges exist in terms of people and goods preservation. Answering to strong needs from different actors (firefighters, foresters), researchers focus their efforts to develop operational decision support system tools that may forecast wildfire behavior. This requires the evaluation of model performance.

We carry out the evaluation of several fire propagation models based on over 500 real fires. We use the data as they would be available in operational conditions, so as to avoid any tuning that would be incompatible with real-time forecasting. The study shows significant performance difference between the models, despite the poor data quality.