In Appalachia, it is not uncommon to see the effects that mountaintop removal mining have on the surrounding ecosystems. As someone living in the great Mountain State of West Virginia, one of the projects that caught my interest when I started my internship at SkyTruth was their work mapping the extent of surface mines in central Appalachia. There are many risks that come with this type of mining. Lung cancer, kidney disease, and birth defects are more likely in areas with high exposure to the toxins produced by mountaintop mining. There are currently no government funded long-term health studies in progress. The last one, started during the Obama administration, was shut down in August 2017.
SkyTruth has been using the Normalized Difference Vegetation Index (NDVI) to detect surface mines. Basically, this algorithm measures vegetation intensity over an area. So, areas with low NDVI scores – or low vegetation levels – show up as mines. Seems simple enough. But the challenge to using NDVI to detect surface mines is that roads, parking lots, lakes, and buildings also have very low levels of vegetation, so they can confuse the algorithm into believing that they’re actually mines. The two NDVI images below – one showing Charleston, West Virginia, and another showing a surface mine in central Appalachia – show how difficult it can be for the algorithm to tell things apart.
Compare that image with an image of surface mines in central Appalachia:
Both of these images show areas with low NDVI scores (or low vegetation intensity). Low NDVI scores are indicated by the darker shades in each of the images. The areas that have a high NDVI score are colored in lighter/whiter shades. Using these images as an example, it’s easy to see how the algorithm that we’re using can get confused. Since things can look alike, we needed to figure out a way to help the algorithm determine which things were roads and buildings and which things were actually mines.
SkyTruth’s surface mine mapping work relies on the use of a data mask to separate out the region’s urban areas, water features, and roads. The mask is used to block out areas from the analysis which have a similar spectral signature to mines (basically roads, buildings, water, parking lots, etc.). Since the surface mine mapping project is updated annually, the mask needs to be updated annually, too. First, I downloaded the area files, which are provided by US Census Bureau, and then I buffered the roads and water features by 60 meters. The buffer is to ensure that these areas are not picked up by the algorithm. The image below shows the mask that will cover the features that have a low vegetation index that could potentially be incorrectly identified as a surface mine.
This next image shows how the mask covers an area like Charleston, West Virginia and blocks the algorithm from detecting it as a surface mine. The other features in the image that have low vegetation intensity – like potential mines – are still visible as darkly areas in the image.
SkyTruth, Appalachian Voices, and scholars at Duke University recently published the first-ever annual footprints of mountaintop mining in central Appalachia between 1985 and 2015. You can learn more about it from SkyTruth’s lead author, Christian Thomas, or you can read the whole paper in PLOS ONE. The updated map will give SkyTruth, Appalachian Voices, and scholars at Duke University, the most current information about the footprint of surface mining in Appalachia, and it will allow them to update their annual footprints with 2016 and 2017. I hope that this map will help inform the public about where surface mines in Appalachia are located, and that it will show people just how much of the Appalachian region is affected.