Data Science Seminar Series: Florence Troude and Alexandre Thomasson

on the October 18, 2018

15:45 to 17:00
The Data Science Seminar Series bring high profile researchers to Université Grenoble Alpes to show the increasing role of data science in modern research. It is open to all researchers of Université Grenoble Alpes. For the 1st seminar of this year, we will welcome Florence Troude and Alexandre Thomasson from Atmo Auvergne-Rhône-Alpes.
Florence Troude and Alexandre Thomasson from Atmo Auvergne-Rhône-Alpes will give a talk on “Daily air pollution forecast”. The event will take place on Thursday 18 October at IM2AG building. The event will be followed by a coffee break.


Atmo Auvergne-Rhône-Alpes is the air quality observatory over the Auvergne-Rhône-Alpes region accredited by the Ministry of Ecology, Sustainable Development, and Energy.
Daily, pollutant maps are forecasted for the next two days over the region to inform the local authorities and citizens.
Meteorological fields are first calculated at the regional scale with the WRF model. These hourly weather fields are used to compute air pollutant dispersion concentrations. They are calculated with the chemistry transport model CHIMERE  which is a deterministic mesoscale eulerian model. A nesting of domains is applied allowing to downscale from large scale to the innermost domain with a 3 by 3 km horizontal resolution.
The second step is a statistical one:  NO2, PM10, O3 concentrations at measurement locations (~40) are forecasted using a machine learning technique. The random forest  method is applied with a 4 years dataset training period (2013 – 2016). Temperature, wind and past concentrations are part of the predictors. These forecasts are then used to adjust the first predicted maps: an ordinary kriging  method is applied as well as a downscaling process to obtain 1x1 km maps. The latter are used for estimating whether alert pollutant thresholds are exceeded and informing the local authorities and citizens.
The third step of the platform consists in performing urban scale pollutant dispersion using the SIRANE  model. These runs take into account the explicit description of road/railway/airport emissions, the main industrial emitters and the regional scale concentrations firstly computed.
Finally, regional and urban scale results are combined to obtain concentration forecasts described with a high resolution (10 by 10 meters) over larger urban agglomerations.

Related document
Atmo AURA air quality network

Related articles

Published on October 25, 2018

Practical informations


Bâtiment IM2AG, building F, Amphi 18, campus universitaire de Grenoble