Fig 1.
Project methodology flow chart.
Fig 2.
Summary list of predictor variables.
Fig 3.
U.S. Energy Information Administration (EIA) coal supply regions within the Appalachian Landscape Conservation Cooperative boundary used in this project.
Counties contained within EIA coal supply regions (Northern, Central and Southern Appalachian, Eastern Interior/Illinois) were determined from the EIA Annual Energy Outlook [2]. The area modeled was further limited to the intersection of the coal supply regions with generalized coal field boundaries for the United States, obtained from the U.S. Geological Survey [25]. This figure shows the intersection of the coal supply regions with actual coal field boundaries. The boundary of the Appalachian Landscape Conservation Cooperative is shown as a thin blue line, obtained from the U.S. Fish and Wildlife Service.
Fig 4.
Surface mining probability from Random Forests model results.
Model extent was limited to the known extent of coal in the region (coal field extents obtained from U.S. Geological Survey) [25]. Random Forests model result ranges from 0 (lowest modeled probability of future surface mining activity) to 100 (highest probability), shown here in a blue to red color ramp.
Fig 5.
Importance of predictor variables measured as out-of-bag mean decrease in accuracy.
Based on Random Forests model results.
Table 1.
Supply regions and probability (prob) results.
Fig 6.
Future surface mining spatial footprint in order to meet coal production estimates for low coal production scenario through 2035 (based on EIA GHG25+low gas price scenario) [2]. Areas with predicted new surface mining through the year 2035 with modeled probability < 0.9 (based on Random Forests results) are shown in orange, areas with modeled probability > 0.9 are shown in dark red. Total area (km2) required to support new surface coal production were determined from an area to production ratio of 1.15 million short tons of coal production per km2 disturbed based on Lutz et al. [19].
Fig 7.
High coal production scenario.
Future mining footprint for coal production through 2035 (based on EIA low coal production cost scenario or high coal production). Future surface mining spatial footprint in order to meet coal production estimates for high coal production scenario through 2035 [2]. Areas with predicted new surface mining through the year 2035 with modeled probability < 0.9 (based on Random Forests results) are shown in orange, areas with modeled probability > 0.9 are shown in dark red.
Table 2.
Production scenarios by region.
Fig 8.
Dark purple areas indicate results of Random Forests modeling as areas with high likelihood of future surface mining (probability 0.90 and above) with mapped coal fields (green).
Fig 9.
Overburden of coal seam comparison to random forests model results.
In general, coal seams with higher amounts of overburden indicate increased costs for mining and recovery of coal resources. Coal seam overburden data were obtained from U.S. Geological Survey [40] for selected seams including the Pittsburgh coal seam (within the Appalachian basin) and the Baker-Danville coal seam (within the Illinois basin), shown here. Modeled areas of high probably of future surface coal mining (Random Forests probability > = 0.90) are shown for comparison, and are indicated by dark purple.
Fig 10.
Remaining coal reserves for West Virginia counties comparison to Random Forests model results.
Remaining reserves (surface and underground) for WV counties were obtained on a county basis from the WV Coal Association [45]. Modeled areas of high probably of future surface coal mining (Random Forests probability > = 0.90) are shown for comparison, and are indicated by dark purple.
Fig 11.
Newly permitted surface mines in West Virginia comparison to Random Forests model results.
Newly permitted surface mines were obtained from the WV Department of Environmental Protection, and include only surface mine permits with status indicating the mine has been permitted, but active mining has not yet begun as of 2013. Modeled areas of high probably of future surface coal mining (Random Forests probability > = 0.90) are shown for comparison, and are indicated by dark purple.
Table 3.
Comparison of results to new permits.