Uncertainty Quantification in SuperCam IRS Data: A Bayesian MCMC Approach

• Simplex optimization: rewrote and adapted a simplex method to handle physical constraints and bounds (abundances, grain sizes), added a penalty term to the cost function to enforce the constraint ∑ abundances = 1, and implemented a probabilistic reset to mitigate local minima.
• Laboratory spectra testing: validated the procedure on known mixtures (magnesite / nontronite / olivine), which revealed and allowed correction of biases (notably a simple multiplicative gain factor ≈0.9 applied to the real spectrum, significantly improving RMS fits). In some laboratory cases, RMS dropped from ~1.11% to 0.47% after correction.
• Application to Martian targets: applied the improved method to Martian sites (e.g., Hastaa — Sol106 and Chiniak — Sol565), testing different mineral lists and analyzing the robustness of MCMC-derived distributions (4000 iterations, 40 walkers in the runs presented).

‍

• Technical / Scientific: radiative transfer modeling (Shkuratov), spectral fitting methodologies, Martian spectral interpretation.
• Computational: Python programming (scipy, implementation of optimization algorithms and MCMC), handling of numerical constraints and cost function optimization.
• Methodological & Soft Skills: critical reading of scientific literature, autonomous project management (balancing theory and practice), communicating results to supervisors and the research team.

‍

‍

‍

‍

‍