Post doctoral position at ONERA to develop Bayesian inverse algorithms for the retrieval of aerosol lidar products
The global warming concerns imply the study of atmospheric aerosols, including soot aggregates emitted
by aircraft engines. Particulate matter emitted by aircraft remains in the upper troposphere and lower
stratosphere and affects the global radiative budget: ice condensates around soot aggregates nuclei to form
contrails. Besides, soot is a factor of health damage by entering the lung cells. Real-time monitoring and
characterization of soot aggregates required accurate determination of soot particle size and morphology.
International aviation agencies (FAA, EPA, and EASA), industrials and scientists are working together on new
particulate matter certifications for aircraft engines, which are more accurate and reliable regarding
One goal of PROMETE research project is to address the need of active remote-sensing of fine particulate
matter emitted from aircraft engine (e.g. soot and other particulate emissions). LiDAR (LIght Detection And
Ranging) is an active remote sensing technique for measuring the backscattered light from particles or
molecules in the atmosphere. Inversion of the lidar signal is a well-known ill-posed problem. A wide range of
techniques have been proposed to estimate the lidar products (i.e. backscattering and extinction profiles) [1,
2]. Those techniques usually require prior knowledge of the atmosphere constituents to ensure a stable
One way to tackle the challenging inversion problem of lidar signals is to use non-linear optimal estimation
methods to retrieve the radiative and microphysical properties of aerosols . This Bayesian framework
provides a formal way of handling the ill-posedness of the retrieval problem and its associated uncertainties.
For several decades, it has been successfully applied to the analysis of passive remote sensing detectors ,
but it has also recently been adapted to Raman lidar observations, with promising results . In order to
assess the relevance of the estimation, it is essential to account for all sources of uncertainty: model
parameter, measurement errors… in the retrieval of the aerosols properties and optimal estimation has the
advantage of providing a full uncertainty budget on a profile-by-profile basis.
The objective of this postdoc is to develop an inversion algorithm for lidar signals with high spatial and
temporal resolution, especially in the frame of short-range lidar measurements. For example, Bayesian
inversion (such as optimal estimation) seems to be well suited for our problem with some adaptations. As a
matter of fact, the distribution of noise is not necessarily Gaussian for low signal-to-noise ratio observations,
and optimal estimation cannot be directly applied in this case. Moreover, in order to account for large amount
of temporal data, it can be interesting to take advantage of machine learning methods instead of performing
repeated calls to a complex forward model. Some ideas in this direction have been proposed very recently [5-
8], such as using quantile regression neural networks to estimate the a posteriori distribution of remote
sensing retrievals .
The candidate will be in charge of the following activities:
1) Providing a state-of-the-art about lidar inverse and optimal estimation algorithms.
2) Developing retrieval algorithms for both synthetic lidar signal and measured lidar signals from
metrological validation campaigns
This includes making use of and improving a forward lidar model, mainly to account for multiple scattering,
modeling the experimental error, and implementing and testing the inversion retrieval algorithms.
The candidate will also have to analyze the sensitivity of the retrievals to the a priori distribution of aerosols
profiles and to find the more appropriate ones.
3) Dissemination of the results and publications.
The post doc will be involved at all steps of the PROMETE project. He/she will investigate how to develop
new inverse methods dedicated to short-range lidar signal with high spatial and temporal resolution, but will
also be associated to the development of the forward lidar model, in collaboration with DRDC (Defence
Research and Development Canada). In order to validate the proposed inverse algorithm, several synthetic
test cases based on our lidar simulator will be considered. The performance of the new algorithm will also be
tested and assessed on real data from metrological validation campaigns.
• New inversion algorithms for lidar signals with high spatial and temporal resolution
• Validation of the lidar signal processing and inversion workflow
• Dissemination of the results and publications.
 Klett, J. D., Appl. Optics, 20, 211–220, 1981.
 Fernald, F. G., Appl. Optics, 23, 652–653,1984.
 Povey, A. C, Atmos. Meas. Tech., 7, 757–776, 2014.
 Rodgers, C. D., Inverse Methods for Atmospheric Sounding: Theory and Practice, 2011.
 Efremenko, D. S. et al I. J. Of Remote Sensing, 38, 1-27, 2017
 Qin W., Remote Sensing, 10, 1022, 2018
 Pfreundschuh S., Atmos. Meas. Tech., 11, 4627-4643, 2018
 Adler, J., Oktem O., arXiv:1811.05910v1, 2018
Start: 2nd semester 2019 for 12 months
Host Laboratory at ONERA:
Department : DOTA
Location (ONERA center) : Toulouse or Palaiseau
Contact : Romain Ceolato (DOTA/IODI) / Sidonie Lefebvre (DOTA/MPSO)
Phone : +33 5 62 25 26 17 / +33 1 80 38 63 76
Email : Romain.Ceolato@onera.fr / Sidonie.Lefebvre@onera.fr
(c) GdR 720 ISIS - CNRS - 2011-2019.