Titan.
Open, GPU-accelerated Cassini VIMS pipeline converting 15 TB of raw cubes into publication-grade maps of Titan’s stratosphere—data Dragonfly and the broader community can use for flight-planning, climate modelling, and discoveries.
Titan · Planetary Science
Open Cassini VIMS analysis pipeline
Since Nov 4 2020
Spectral Intelligence Pipeline
Ingest — raw Cassini VIMS cubes losslessly decompressed · Denoise & Map — co-added frames, cylindrical reprojection, striping noise removed via limb-mask stacking · Boundary Detection — shift-subtract ±6° latitude then fit 6-order polynomial; derivative yields NSA latitude column-wise · ML — scikit-learn models refine tilt vs. wavelength, SVM flags outliers · Acceleration — CuPy CUDA kernels on RTX 3090, ~14× speed-up.
Thesis
Most planetary pipelines still rely on 1990s IDL. By rebuilding in modern Python (pyVIMS + scikit-image) and offloading heavy math to a single CUDA workstation, one undergrad compressed months of manual reduction to hours.
Scientific Impact
The limb analysis confirmed and presaged the lower stratospheric dichotomy observed in [Vashist et al. 2023], showing how upper-atmosphere structure evolves seasonally. This provides a more complete vertical picture of Titan's global haze dynamics, linking high-altitude phenomena to seasonal changes deeper in the atmosphere. The work offers a new method for tracking global atmospheric layers independently.
Limb Comparisons
Analyzed limb brightness profiles from 24 targeted Cassini/VIMS flybys (2004–2017) to probe the vertical structure of Titan's haze. We processed low-phase angle cubes, extracted N/S transects 30° from the equator, and fit them to a quadratic limb darkening law to quantify brightening/darkening coefficients across wavelengths and time.
DPS ’23 Abstract
Presented at the AAS Division for Planetary Sciences 2023.
Core Technologies
Python 3.10 · pyVIMS · NumPy / CuPy · scikit-image · scikit-learn · Matplotlib · Jupyter