Remote Sensing Combined with Field Spectroscopy for the Detection and Monitoring of Heavy Metal Contamination from Informal E-waste Recycling

Electronic waste (e-waste) is one of today's fastest growing waste streams. Made up of discarded electronics, e-waste disposal is complex. However, e-waste also provides economic opportunity through the processing and extraction of precious metals. Sometimes referred to as "urban mining," this recycling operates informally or illegally and is characterized by dangerous practices such as, open-pit burning, acid leaching, and burning of low value wastes. Poorly controlled e-waste recycling releases dangerous contaminants, especially heavy metals, directly to the surface environment where they can infiltrate water resources and spread through precipitation events. Despite growing recognition of the prevalence of unregulated e-waste processing, systematic data on the extent and persistence of the released contamination is still limited. In general, contamination is established through techniques that provide only a snapshot in time and in a limited geographic area. Here we present preliminary results from attempts to combine field, laboratory, and remote sensing studies toward a systematic remote sensing methodology for e-waste contamination detection and monitoring. The ongoing work utilizes a tragic "natural experiment," in which over 500 e-waste burn sites were active over more than a decade in a variety of agricultural, residential, and natural contexts. We have collected over 100 soil samples for which we have both XRF and ICP-AES measurements showing soil Pb concentrations as high as 14000 ppm. We have also collected 480 in-situ reflectance spectra with corresponding soil samples over 4 field transects of areas with long-term burn activity. The most heavily contaminated samples come from within the burn sites and are made up of ash. Field spectra of these samples reflect their dark color with low overall reflectance and shallow spectral features. These spectra are challenging to use for image classification due to their similarity with other low-reflectance parts of the image (e.g., shadows). We have begun to distinguish shadows from the dark burn site centers by automatically detecting and masking shadows. This will allow us to utilize images taken at different times and our in-situ field spectral results to develop a method for monitoring contaminant spread from these complex point sources.