Martian Winds

Python Matlab CIV LaTeX Image Analysis Image Processing Data Visualisation Astropy Matplotlib NumPy Pandas NetCDF4 scikit-learn

Understanding wind patterns on Mars is essential for future human exploration and for advancing our knowledge of planetary climate dynamics. Despite their importance, Martian winds remain largely unknown. This project analyzes six image sequences from the Emirates Mars Mission’s EXI 320 nm ultraviolet observations to derive detailed wind field maps.
Using a cloud-tracking method called Correlation Image Velocimetry (CIV), wind fields were calculated with time intervals of 10–30 minutes. Various image pre-processing techniques were tested to optimize CIV performance, with contrast limited adaptive histogram equalization (CLAHE) producing the most reliable results.

View on GitHub

Download Thesis

raw image mapped to longitudes and latitudes

overview of the selected areas

Comparison of Image Processing Techniques

comparison of different image processing techniques

This project involved multiple stages: literature review, data exploration, data preparation, image processing, Correlation Image Velocimetry (CIV), result analysis, and plotting. The raw data came in FITS files, which were explored, prepared, and processed in Python using libraries such as Astropy, Matplotlib, NumPy, Pandas, NetCDF4, and scikit-learn. For the CIV, a cross-correlation–based technique was executed in UVMAT, a MATLAB-based interface.

screenshot of UVMAT interface

resulting velocity fields

CIV performance analysis

The CIV workflow consisted of several steps:

CIV1: Initial velocity fields were calculated using specified interrogation window and shift box parameters; the grid size defined the number of vectors displayed.
FIX1: Error correction and filtering of vectors above a pre-defined correlation coefficient.
PATCH1: Smoothness filtering through interpolation and determination of the maximum vector size.
CIV2, FIX2, PATCH2: The process was repeated using the vectors obtained from CIV1, allowing for refined parameter adjustments and accounting for rotation and deformation.

Velocity fields were then exported and analyzed in Python, where final results were visualized and plotted to provide detailed insights into Martian wind patterns. The project also included careful documentation of methodology, results, and discussion, and was compiled as a thesis in LaTeX format.

bandpass filter indicating atmospheric waves

visualisation of zonal means

The analysis of Martian wind fields revealed several key insights. While image processing techniques improved the visibility of clouds, they did not significantly outperform the original images. However, using Contrast Limited Adaptive Histogram Equalization (CLAHE) allowed for the detection of additional velocity vectors, demonstrating the potential of enhancing sub-regions within the data.
The choice of CIV parameters and input image material had a strong influence on the final results, showing that parameter tuning plays a critical role in deriving accurate velocity fields.
The study also identified several limitations. These include uncertainties in vector direction due to the 3–4 km pixel resolution, variations in solar illumination, and potential spacecraft position instabilities.

Finally, the work suggests promising future directions. Alternative pre-processing methods, particularly those focusing on sub-region enhancement, could further improve CIV performance. Additionally, the appearance of atmospheric wave patterns in high-pass filtered images points to opportunities for new insights into Martian atmospheric dynamics.