Using PCA With the 2 μm Band Excluded Over Normalization Schemes to Analyze Uncalibratable NIR Spectral Images of Mars

Phillip Chandler


In our research analyzing the water content of Martian clouds, much of the ground-based spectral imaging data collected was deemed uncalibratable due to the absence of standard star measurements to compare to our images of Mars. One of our goals in this research is to use principle component analysis (PCA) to produce a surface spectral model independent of the atmospheric spectral response. When the PCA was used and analyzed over the whole 1.5–4.1 μm spectrum, we noted that the absorption feature from 1.9–2.2 μm was showing up in all eigenvectors, implying it was major component of the surface spectral signature. However, we know that the atmospheric gas is really the dominant reason for this feature so the question was raised: if the 1.9–2.2 μm band were removed, would the resulting PCA eigenvectors show more consistent trends, leading to a better surface model? However, this spectral region is also where water ices are active, so the secondary question is: if the 1.9–2.2 µm band were removed, would it adversely affect our ability to model Martian clouds? In order to address the first question, we performed PCA tests across three different types of normalization excluding the 1.9–2.2 μm band. The three normalization schemes are: disc mean, disc median, and spot-sectra. We will present the results of these three tests: a comparison of the consistency of the eigenvectors across diurnal and seasonal time scale and to an overall median set of eigenvectors.


Mars; Atmosphere; Water

Full Text: PDF


  • There are currently no refbacks.