Classification

We also conducted an unsupervised classification to try to determine if ewaste has a unique spectral signature that could be extracted from the visible image. Using the same image from December 2012, we did a 10 class classification using the IsoData tool. As one can tell from the resulting map below (which excludes the red color, water, and the burgundy color, dirt), this did not yield an identifiable “ewaste” class.

In the image below, which includes the red color in the classification, it appears that red may correspond to E-waste.  

However, doing a classification of the same image from farther away demonstrates that this is most likely not the case. From the visible image below, taken in December of 2012, we did an unsupervised classification.

If the red color in the unsupervised classification below were to indicate E-waste or E-waste contaminated water, it would have probably flowed out from the river in a polluted bloom. Rather, the red color is probably just water since it is the overall color of the river, if you look closely, and it is the color of the deeper ocean water.

However, the unsupervised classification of the image without red also yields few results. There is more pink in the areas around the polluted river, but pink is also a highly ubiquitous color in this image so it is difficult to identify what exactly it corresponds to. There is also a possibility that blue and green could indicate e-waste pollution since they are in and around the dump site, the river, and where the river feeds into the ocean, but they are also elsewhere in the image as well. Accordingly, no single class corresponding to e-waste's spectral signature could be determined using this procedure.