Description of the thesis topic

Background
Ice crystals present in clouds play a central role in the climate system, particularly through their influence on precipitation formation and the radiative balance of the atmosphere. Accurate characterization of their physical properties (size, shape, mass, phase, concentration, etc.) is essential for improving their representation in numerical models—whether for weather forecasting, climate simulations, or aircraft icing modelling. Furthermore, better knowledge of these properties helps refining the interpretation of data obtained from both active and passive remote sensing.

PhD Topic
This PhD project aims to quantitatively analyse the microphysical and morphological properties of ice crystals observed in atmospheric clouds. For several decades, the scientific community has successfully used hydrometeor imagers OAP (optical array probes)—capable of producing thousands (even up to ten thousand) images per second using the principle of shadowgraphy. The OAP imagers are particularly used for airborne measurement campaigns.
These OAP imagers, which vary in resolution and measurable size range, allow extracting valuable information from 2D images: dimensions, mass, density, phase, and more recently, morphological classification using convolutional neural networks (recently developed in the lab).
However, the properties inferred from 2D images are affected by the arbitrary projection of a 3D crystal onto a plane, thus depending on its orientation. This reduction in information can introduce significant biases, which have been little studied to date.
The main objective of this thesis is to quantify the impact of ice crystal orientation on the retrieval of their microphysical and morphological properties from 2D images. This methodological work is a first attempt within the scientific community.

Approach and Methodology
A realistic 3D simulation model of hydrometeor shapes—representative of those observed in clouds—has recently been developed in the laboratory. This model can generate crystals with various morphologies (monomers and aggregates), resulting from the main growth modes: vapor deposition, riming, or aggregation.
Further improvements to the crystal generation with the model should be based on real image datasets obtained from:
• the MASC (Multi-Angle Snowflake Camera), which provides three simultaneous views of the same crystal;
• the two stereo systems of the 2DS imager (2D Stereo Probe), which provides two orthogonal views of the same crystal.
These data will allow comparison between properties extracted from 2D images and the "truth" of known simulated 3D structures. The impact of orientation will be analysed across various retrieved parameters:
• Dimensional spectra (number, mass, effective density)
• 2D geometric properties (area, perimeter, aspect ratio, roughness)
• Morphological classification

Perspectives
The knowledge acquired during this PhD will be applied to real datasets from various contexts:
• Arctic clouds
• Deep convection
• Industrial studies on hydrometeors (aeronautics, icing)
This work aims to provide robust tools and references for correcting or more precisely interpreting data derived from 2D imaging, especially under the contrasting environmental conditions in which atmospheric clouds form.

Work Context

The PhD student will work at LaMP in Clermont-Ferrand (France).
LaMP (approximately 35 scientists and engineers, 10-15 PhD students is a laboratory recognized both in France and internationally for its expertise in observing and modeling cloud microphysics.
This 3-year PhD project requires collaboration with scientists from the MNP working group (Cloud and Precipitation Microphysics).

Additional Information

We are looking for a motivated PhD student to investigate above scientific questions. The candidate should hold a Master of Science or equivalent in a field relevant to the proposed research field (e.g. Environmental Sciences, Applied Mathematics, or Statistics). Good working skills on either image analysis, advanced data processing (e.g. neural network, clustering) or statistics applied to experimental data will be an asset for this position. We expect the candidate to be proficient in a programming language (Python, Matlab, etc…). Proficiency in English is a prerequisite.