MSCA-COFUND-CLEAR-Doc-PhD Position#CD22-51: Air quality evolution at the local scale, in the absence of a calibration reference point

Recent years following the COVID19 pandemic. Air quality has become an important concern in major cities around the world. Currently, the measurement of air quality is governed by European regulations. It is often assessed on a regional scale using sets of fixed measuring stations coupled with geospatial observation systems.

The current measurement approach has the advantage of being able to cover an important measurement area. However, such a method does not allow to take into account measurement anomalies detected locally at the scale of a human body.

This need for accurate, real-time information at the individual level has accelerated the development of portable measuring devices.

Currently, autonomous and geolocatable pollution sensors with integrated SIM card seem to stand out (PM 2.5; PM10 for example). These sensors have the advantage of being easily accessible with a human-sized range. One of the major problems encountered in air quality monitoring concerns the analysis and correction of pollution measurements when moving away from the reference sensor. This problem is become more important when measurements are made in alleys or between buildings or obstacles high enough to interfere with the signals emitted by the reference sensors. Other operational problems have also been identified: lack of reliable geospatial data to serve as a reference at the local scale, increasing need for accuracy of continuous flow measurements and the difficulty of acquiring data using continuous flow emission sensors in constrained environments, especially urban.

For several years, the GERS GIE laboratory has developed a competence in the study of the particle flight mechanism in a thin boundary layer under environmental stress. This multiphysics approach (soil mechanics, fluid mechanics) was first used to study the degradation of unpaved road under traffic loading (Sediki et al., 2016), to deal with visual nuisance problems due to dust emissions (Le Vern et al. 2022), and to optimize the use of water for dust suppression on construction sites (Sédiki et al. 2022) (Rayssac et al. 2022). This skill was then transposed to the study of aerosol behavior (Victor, 2021) (Morin, 2022), at the laboratory and in situ scales before being extended to the study of air quality.

By extension, with a multidisciplinary approach, the perception of exposure risks, the impact of emissions on air quality at local and regional scales and the correlations between lung diseases and dust-borne pathogens are proving to be a real fundamental need and have been integrated in the approach.

The GERS-GIE laboratory has equipment for control of air quality at laboratory and in situ scale. At laboratory scale, a subsonic wind tunnel equipped with a stereo velocimetry system using high frequency stereo particle images velocimetry (HF PIV stereo) and a LIF ratiometric option is available for wake flows modeling.

To solve the scientific and technological challenges mentioned above, the thesis aims at improving the reliability of air quality metrology at the local (metric) scale by a IA (deep learning) combined with a probabilistic approach.

Air quality measurement data collected from a previous in situ measurement campaign in the Pays de la Loire region will be used to feed the learning, modeling and validation while taking into account the sets of climatic parameters and the immediate environment parameter of the measurement area: weather, building, ...

To optimize the resolution in time and space, the geolocalization parameters of the sensors will be used to allow a continuous comparison between: the measurement values at the regional scale, the one measured by the sensor and the one predicted between two consecutive time increments.

The adopted correction methodology will be modeled by a probabilistic approach and then validated with wind tunnel tests to highlight the mechanism(s) governing the diffusion of particles in the wake of the sensor. This approach aims to define the suitable correction coefficient of measurement. An additional in situ measurement campaigns will be carried out by applying the correction coefficient to assess the level of uncertainty observed.