TESTING NORMALIZATION SCHEMES OF UNCALIBRATABLE NIR SPECTRAL IMAGES OF MARS USING PRINCIPAL COMPONENT ANALYSIS

Sarah Jayne Kessler

Abstract


Our research is part of a larger program to measure the water content in Martian clouds over diurnal, seasonal, and interannual timescales. This requires recovering a surface spectral model independent of the atmospheric spectral response. This is done using Principle Component Analysis (PCA), which has been shown to be fairly uniform across all timescales. Much of the data used come from ground-based near infrared (NIR) imaging; however, some of the data were uncalibratable due to the absence of comparison standard-star measurements. To determine if the data were still useful, we tested three normalization schemes: data mean, data median, and spot-spectra. We then preformed PCA on the normalized data to look for trends in the eigenvectors. The primary question being: are the PCA results still uniform even in uncalibrated data? We present here the results of our consistency analysis. Most of it was done through simple observations of the graphs of the PCA eigenvectors of the different normalizations. We find the mean and median normalizations show too much variability across all time scales and so are considered less superior in comparison to the spot-spectra normalized data, which showed much greater uniformity over time. We will supplement our qualitative analysis using a quantitative measure of uniformity based on the average chi-squared value between eigenvectors and their median—the greater the average, the more non-uniform they are. Our research shows that data previously determined uncalibratable due the absence of a comparison star may still be useful in further research. This will allow us to extend the study of Martian clouds to days of less-than-ideal observing conditions.


Keywords


Astronomy; PCA; Mars

Full Text: PDF

Refbacks

  • There are currently no refbacks.