Changes in the extent of surface mining and reclamation in the Central Appalachians detected using a 1976–2006 Landsat time series
Introduction
Quantifying the temporal and spatial patterns of land cover/land use change (LCLUC), as well as its consequences for ecological, hydroclimatological, and socioeconomic systems on Earth, is a central focus of land change science (Turner et al., 2003). Among the variety of different land conversions (e.g., deforestation, urbanization, etc.) that has received scientific attention, the removal of forest cover has been shown to dramatically affect hydrological processes such as evapotranspiration, canopy interception, and runoff at scales ranging from small plots to large river basins (Eshleman, 2004). In the case of urbanization, the loss of vegetative cover in combination with changes in soil infiltration capacity due to disturbance, has been shown to significantly enhance the flood generation potential of watersheds with substantial consequences for human well-being (Beighley and Moglen, 2002, Hollis, 1975, Rose and Peters, 2001, Sauer et al., 1983).
Modern techniques of surface mining using heavy equipment can produce dramatic alterations in land cover, both ecologically and hydrologically (Simmons et al., 2008). In forested regions such as the Appalachian Mountains, surface mining for bituminous coal since World War II has led to a widespread transformation of the mountainous landscapes. Reclamation of surface mines, as mandated by the Surface Mine Control and Reclamation Act of 1977 (SMCRA), has not resulted in restoration of pre-mining hydrologic characteristics (Negley & Eshleman, 2006) or ecological functions (Simmons et al., 2008). SMCRA requires mine operators to reclaim strip mines to the approximate original contours of the landscape, and to restore appropriate permanent vegetation types. In general, reclamation has thus involved replacement and grading of the overburden (topsoil and other near-surface materials) using large earthmovers, followed by seeding with grasses and other herbaceous vegetation. The end result is the transformation of native forests and their associated soils into predominantly herbaceous-covered minelands with reduced soil infiltration capacity due principally to surface compaction (Bell et al., 1994, Bussler et al., 1984, Chong and Cowsert, 1997, McSweeney and Jansen, 1984, Negley and Eshleman, 2006).
Although comparative hydrological research in the Appalachians shows relatively minor changes in annual water balances (e.g., annual evapotranspiration and annual runoff) due to deforestation and surface mining (Simmons et al., 2008), several studies show higher peak and total storm runoff from mined compared to forested Appalachian watersheds (Bonta et al., 1997, Negley and Eshleman, 2006). These results raise the specter that surface mining and reclamation in the Appalachian Mountain region may be increasing the risk of flooding hazards, thus underlining a need for more careful, objective, and quantitative estimation of the trajectory and spatial dimension of mined land conversions in this region. Knowledge of the extent of mining and reclamation within watersheds is critical to managing or mitigating the potential impacts of surface mining on downstream settlements. However, spatial data necessary to characterize the timing and extent of mining and reclamation in the region are either unavailable or unreliable. For example, the most reliable information – GIS layers of state issued mine permits – have imprecise boundaries and omit many areas identifiable as mines on aerial photographs or satellite imagery. Moreover, these mine permits identify only the date permits were issued and not the actual onset of mining activity, which may occur many years (a decade or more) after the issuance of the permit.
Remote sensing has been used widely to characterize land cover changes relevant to hydrologic functioning. In particular, remote sensing is especially useful for detecting the conversion (e.g. deforestation) of natural vegetation cover to land cover types having higher rainfall runoff rates (DeFries and Eshleman, 2004, de Smedt et al., 2004, Foley et al., 2007, Foody et al., 2004, Genxu et al., 2006, Griffith, 2002, Miller et al., 2002). Likewise, maps of urbanization and surface imperviousness derived from remotely sensed data provide critical inputs for understanding the nature of hydrologic changes and changes in flood dynamics in watersheds experiencing LCLUC (Carlson, 2004, Finch et al., 1989, Shuster et al., 2005, Weng, 2001).
Comparatively fewer studies have comprehensively examined the use of remote sensing to map surface mine extent through time, and as such, there are few studies that incorporate maps derived from remotely sensed data into analyses of the broad-scale effects of mining on watershed hydrology. Rathore and Wright (1993) reviewed some of the early papers and conference proceedings that used Landsat MSS and TM data to map mining and reclamation, and found that the methods for mapping active mines were relatively straightforward and accurate, especially in areas where similar barren cover types are not present. Two papers (Anderson and Schubert, 1976, Anderson et al., 1977) delineated and inventoried active strip mines using MSS data in western Maryland, but found high variability in spectral signatures of mines. Band ratios were found to be most effective for discriminating mined areas. Other studies have used time series Landsat TM and similar image sources to map changes interpreted as having resulted from coal mining (Prakash & Gupta, 1998) and there is an extensive literature on mapping industrial open pit mines and associated tailings (e.g., Hagner and Rigina, 1998, Latifovic et al., 2005), but most studies are generally qualitative and do not report map accuracy.
Despite success at mapping active mines using satellite imagery, the mapping of mine reclamation has been considerably more troublesome (Rathore & Wright, 1993). The most successful studies for mapping reclaimed mines have used aerial photography, in which context and pattern have been used to infer process. However, Irons and Kennard (1986) demonstrated some capacity for Landsat TM imagery from Pennsylvania to distinguish bare mine spoil, grass on mine spoil and trees on mine spoil from agriculture, water and forest on a single image date, although overall map accuracy rates were low (62.3%). Also in coal mining areas of Pennsylvania, Guebert and Gardner (1989) used SPOT data to map classes associated with differing infiltration rates on mines and reclaimed mines (70% accuracy), but did not attempt to distinguish mined/reclaimed areas from other cover types. Parks, Petersen, and Baumer (1987) found that MSS was useful for monitoring active mines, and that TM and simulated SPOT imagery could be used to characterize the spectral characteristics of reclaimed mines, though not necessarily the spatial pattern.
Several broad-scale studies have attempted to quantify overall rates of land cover change during the Landsat era (Loveland et al., 2002 and related papers). Loveland et al. (2003, Figure 6-3) demonstrated through this approach that conversion to mining was the single largest land cover change in the Central Appalachian region between 1973 and 2000. Other major types of land cover change during that time include conversion to clear-cut/regeneration and to grass/shrub (possibly indicative of mine reclamation), indicating that forest clearance accounted for more than two-thirds of the landscape conversion in the region during the last three decades. However, the methods and accuracy for mapping mined lands as presented in Loveland et al. (2003) are not presented, but are presumed to be based on the 1992 National Land Cover Classification (NLCD) map following Loveland et al. (2002) and Griffith et al. (2003). The NLCD maps (Vogelmann et al., 2001) for U.S. Environmental Protection Agency Ecoregion 3 (Central Appalachians) have a commission error of 65% and omission error of 48% for mine-relevant classes, pointing to the limitations of using NLCD for characterizing land cover processes associated with mining.
Although studies have documented the use of remote sensing to map surface mining with varying levels of success, maps of mine reclamation are rare, in large part because reclaimed mines are spectrally indistinguishable from grasslands or pastures. From the hydrologic perspective, however, reclaimed mines do not function like natural grasslands or pastures, and retain many of the hydrologic properties of impervious mines. The biological, chemical and physical process required for soil development to pre-mining conditions occurs very slowly over century time scales. As such, the mapping of reclaimed mines is at least as important as mapping active mines for studies of watershed hydrology. Active mines are generally transient features of landscapes, and are typically minor in extent at any given time (Negley, 2002). In contrast, reclaimed mineland is generally persistent, with the area gradually increasing as new mines are opened and subsequently reclaimed. Although reclaimed mines may not be distinguishable from pastures or grasslands based on spectral characteristics alone, a logical approach using temporal sequences of land cover change could be used to properly label reclaimed mines and other grass-dominated cover types. This suggests the use of logical decision trees to identify land use classes based on transition trajectories of land cover maps through time. Although such an approach has not been used for mapping surface mine reclamation, the approach has been used to label classes based on their cover transition in studies of LCLUC in areas experiencing other types of land conversion (e.g. Wolter et al., 2006).
The objective of our study is to quantify patterns of LCLUC (e.g. conversion of forest to mined and reclaimed mine lands) in eight study watersheds within the Central Appalachian Region of the Eastern U.S. during a 30-year time period (1976–2006). This work provides the basis to understand changes in hydrologic response through time in these river basins as a consequence of continued mining and mine reclamation. To distinguish active mines and reclaimed minelands from spectrally similar classes, we developed simple land cover classifications on four image dates (1976, 1987, 1999 and 2006) and employed a logical decision tree based on class transitions and ancillary data such as GIS layers of mine permits (from the 1950s to the present), wetlands, and urban land cover.
Section snippets
Study area
The Appalachian Mountain region forms the headwaters for several major U.S. rivers, including the Potomac, Susquehanna, and Ohio Rivers, and provides water resources to tens of millions of residents throughout the eastern and central U.S. The Central Appalachian Ecoregion (Fig. 1) in western Pennsylvania, western Maryland, eastern West Virginia, southwestern Virginia, eastern Kentucky, and north-central Tennessee is also rich in other natural resources: (1) wood from productive temperate
Land cover change
Across our eight study watersheds, the area of active mines was highest in 1976 (1.8% of the land area) and relatively stable in 1987 and 1999 before declining to 0.5% in 2006 (Table 1, Fig. 3). In the most heavily mined watershed, Georges Creek in Maryland, active mines peaked at 5.4% around 1976, declining to around 2% or slightly less in later years (Table 1). Reclaimed mines increased in area following 1976, balancing the decline in active mines and forest area. The amount of reclaimed land
Discussion
The ability to map the extent of active mines and reclaimed mines is essential to understanding the long-term ramifications of mining on ecosystem services (Simmons et al., 2008), especially those associated with watershed hydrology (Guebert & Gardner, 1989). In related research within our study area, Negley and Eshleman (2006) found infiltration rates nearly two orders of magnitude higher in unmined forests compared to reclaimed grasslands (30 cm h− 1 vs. 0.3 cm h− 1); infiltration on active
Conclusion
Surface mining for coal and subsequent mine reclamation has been extensive in watersheds throughout the Appalachian Mountains. Our watersheds are likely representative of the larger region, but should be generalized with caution because some areas, especially in southern West Virginia, are in fact experiencing much more widespread surface mining. Application of a Landsat-based, temporal filtering approach to assessing LCLUC in these and other heavily mined areas worldwide could greatly
Acknowledgements
This research was funded by NASA Energy and Water Cycle Grant NNG06GC83G to KNE and PAT. Thanks to Brian McCormick for assistance with data acquisition and processing. Author contributions: PAT and KNE designed the study. DPH, CCK, KMDB and PAT developed the specific methods reported in this paper. DPH and CCK assembled the data sets and wrote code to implement the methods. DPH, PAT and BEM analyzed the output results, and PAT wrote the manuscript with contributions from all of the other
References (51)
- et al.
Infiltration in reclaimed mined land ameliorated with deep tillage treatments
Soil & Tillage Research
(1997) - et al.
Predicting locations sensitive to flash flooding in an arid environment
Journal of Hydrology
(2004) - et al.
Detection of forest decline in Monchegorsk area
Remote Sensing of Environment
(1998) - et al.
Assessing land cover change resulting from large surface mining development
International Journal of Applied Earth Observation and Geoinformation
(2005) - et al.
Land use land cover change in the U.S. great lakes basin 1992 to 2001
Journal of Great Lakes Research
(2006) - et al.
ERTS-1 data applied to strip mining
Photogrammetric Engineering and Remote Sensing
(1976) - et al.
Landsat imagery for surface-mine inventory
Photogrammetric Engineering and Remote Sensing
(1977) - et al.
Trend assessment in rainfall-runoff behavior in urbanizing watersheds
Journal of Hydrologic Engineering
(2002) - et al.
Morphological-characteristics of reconstructed prime farmland soils in Western Pennsylvania
Journal of Environmental Quality
(1994) - et al.
Impact of coal-surface mining on three Ohio watersheds: Surface water hydrology
Journal of the American Water Resources Association
(1997)