IRMACS: The Interdisciplinary Colloquium: "The application of remote sensing as a tool for locating mass graves"

Abstract

Remote sensing, in the form of in-situ, airborne and satellite imagery has been relatively untested as an investigative tool in the location of mass graves. We focus on examining the spectral signature of an experimental grave from in-situ and airborne hyperspectral data in a tropical moist forest environment. With over one hundred bands in the visible- shortwave infrared regions of the electromagnetic spectrum we can examine specific features that may be characteristic of the spectral response of the grave versus disturbed soil from the in-situ data or land covers such as pasture, forest, exposed soil, etc. from the imagery. We examined the spectral response of the grave and a control site over a period of sixteen months from in-situ data with a combination of feature selection followed by pattern classification. The airborne imagery, collected one month following experimental set-up was examined in n-dimensional space to isolate the spectral response of the grave. We found a clear separability between the spectra of the grave and the disturbed soil throughout the sixteen month period. We also found similar distinct characteristics in the spectral response of the grave versus other targets in the landscape from the imagery (one month time period). Vegetation regeneration was severely inhibited on the grave in comparison to disturbed soil. At the sixteen month period when the regeneration was sufficient to collect leaf samples, we examined in detail leaf-level spectra from grass growing on and off the grave with the feature selection and pattern classification; the spectra were separable with minimal error.