Differentiating mine-reclaimed grasslands from spectrally similar land cover using terrain variables and object-based machine learning classification

Abstract

(2015) Differentiating mine-reclaimed grasslands from spectrally similar land cover using terrain variables and object-based machine learning classification, International Journal of Remote Sensing, 36:17, Incorporating ancillary, non-spectral data may improve the separability of land use/ land cover classes. This study investigates the use of multi-temporal digital terrain data combined with aerial National Agriculture Imagery Program imagery for differentiating mine-reclaimed grasslands from non-mining grasslands across a broad region (6085 km 2 ). The terrain data were derived from historical digital hypsography and a recent light detection and ranging data set. A geographic objectbased image analysis (GEOBIA) approach, combined with two machine learning algorithms, Random Forests and Support Vector Machines, was used because these methods facilitate the use of ancillary data in classification. The results suggest that mine-reclaimed grasslands can be mapped accurately, with user's and producer's accuracies above 80%, due to a distinctive topographic signature in comparison with other spectrally similar grasslands within this landscape. The use of multitemporal digital elevation model data and pre-mining terrain data only generally provided statistically significant increased classification accuracy in comparison with post-mining terrain data. Elevation change data were of value, and terrain shape variables generally improved the classification. GEOBIA and machine learning algorithms were useful in exploiting these non-spectral data, as data gridded at variable cell sizes can be summarized at the scale of image objects, allowing complex interactions between predictor variables to be characterized.

A Review of Fine-Scale Land Use and Land Cover Classification in Open-Pit Mining Areas by Remote Sensing Techniques

“…References opencast stope, mineral processing land, dumping site 3 [35] reclaimed herbaceous vegetation, reclaimed woody vegetation, barren (including haul roads, active quarries, land disturbed by mining) 3 [22][23][24] mine-reclaimed grassland (including reclaimed land within mine sites and valley fills dominated by herbaceous vegetation) 1 [25] open stope, stripping area, waste-dump area, mine industrial area 4 [26] active, disturbed vegetation and pasture, rehabilitation, remnant, spoil/waste, and water management 6 [27] coal deposit, over burden dump, mine dump 3 [28] tree cover, dense grass, sparse grass, bare ground 4 [29] coal mining excavation cities, coal dump areas 2 [30] active mine, reclaimed mine (including grass, woodland, forest) 2 [16] opencast mining (coal), overburden dump 2 [12] bare tailings, exposed lime, water/wet tailings 3 [31] tailings zone, dry vegetation zone, transition zone, vegetated islands 4 [32] mine, dump, coal stockpile 3 [33,34] …”

“…ZiYuan-3 stereo satellite imagery [41] RapidEye [24] RapidEye, LiDAR [25,26] Landsat 5 TM, Landsat 8 OLI [28,30] Aerial imagery from Bing Maps, Google Maps imagery, NAIP [27,29] SPOT-5 imagery, DEM from aerial photographic stereo pairs [31] Landsat TM, Landsat ETM+, Landsat ETM [32] Landsat MSS, Landsat TM, Landsat ETM+ [18] Landsat MSS, Landsat TM, IRS LISS-II [14] Airborne high spatial and spectral resolution Compact Airborne Spectrographic Imager, hyperspectral data acquired with the Probe-1 airborne imager [33] Hyperspectral data acquired with the TRWIS III [34] IKONOS, QuickBird [35,36] HR images often have four bands; thus, identifying land cover at a fine scale using only spectral curves becomes difficult. In previous research employing HR images, the widely used spectral band and several spectral indexes, such as vegetation index [30,35], were considered as the input features of the classification process.…”

“…One research team introduced Gaussian low-pass filtering [24], geometric features, a digital surface model [54], textural features [26], and terrain features (slope, roughness, and elevation) [27] with RapidEye and LiDAR to enhance the classification accuracy of land cover in mining areas. The grey-level co-occurrence matrix (GLCM) [54] is the most popular for describing spatial properties, particularly in VHR imagery [55].…”

“…For details, see Table 1. Maxwell and Warner [27] designed two land cover classes to differentiate mine-reclaimed grasslands (Table 1). Guan et al [28] formed seven classes for land use classification and change detection in a coal mining area: open stope, stripping area, waste-dump area, mine industrial area, farmland, urban area, and the original landscape (grassland) ( Table 1).…”

“…For LCCMA, the following remote sensing data have been utilized: (1) Landsat images [14,18,28,30,32]; (2) high-resolution images such as ZiYuan-3 [27,41], RapidEye [24][25][26], SPOT-5 [31], IKONOS [35,36], and QuickBird [35,36]; (3) airborne aerial images [24,29,33,34]. For details, see Table 2.…”