Detecting Decadal Land Cover Changes in Mining
Regions based on Satellite Remotely Sensed Imagery:
A Case Study of the Stone Mining Area in
Luoyuan County, SE China
Zhaoming Zhang, Guojin He, Mengmeng Wang, Zhihua Wang, Tengfei Long, and Yan Peng
Abstract
Mining regions often undergo abrupt and extensive land cover
changes that pose serious environmental and social impacts.
In this study, decadal land cover changes in stone mining
areas of Luoyuan County, southeastern China from 2001 to
2010 were examined based on multi-source satellite remote
sensing imageries including
ALOS
,
SPOT2
, and Landsat-7.
Object-oriented classifications combined with decision tree
and retrospective approaches are employed to extract land
cover and change information for the ten-year period. The
study results show that the stone mining area nearly quadru-
pled over the ten-year period. It is found that the digging area
accounts for only 14.3 percent of the stone mining region.
However, mine dumps and tailings occupy the majority of the
region, a remarkable characteristic distinct from other mining
regions. Reclaimed land in the mined region is very limited.
An evident increase in the extent of urbanized land cover is
also observed in the study area for the last decade.
Introduction
The mining industry has grown quickly over the last decade
in China. Mining activities cause numerous eco-environ-
mental problems that lead to ecological degradation and
environmental pollution in mining areas, limiting regional
sustainable development (Townsend
et al.
, 2009). In addition
to being the only available data source in many areas, satel-
lite remote sensing technology has other advantages, such
as acquiring data with sufficient area coverage and temporal
frequency. These features render remote sensing technolo-
gies suitable for studying and monitoring primary impacts
of surface mining at low costs. Remote sensing technologies
have been recognized as promising tools of mining activity
monitoring by several researchers (e.g., Prakash and Gupta,
1998, Olthof and King, 2000, Latifovic
et al
., 2005). However,
in regional-scale studies, mining regions are often merged
with other land cover classes, such as “cleared land,” “built-
up land” and “other” land classes (Sonter
et al
., 2014). This
approach may have been applied because medium or coarse
spatial resolution remote sensing images were used, whereas
mining operations often occur at small spatial scales relative
to other land cover changes (such as deforestation and urban
sprawl). Thus, distinguishing mining activities as a separate
land cover class can be time-consuming and potentially im-
precise (Sonter
et al
., 2013). Such problems can be addressed
through the use of high-spatial resolution satellite technolo-
gies. When using high spatial resolution satellite imagery,
land cover data extraction for mining regions is more accurate
and efficient. In this study, pan-sharpened
ALOS
(Advanced
Land Observing Satellite) high-resolution satellite imagery are
used to extract land cover data for the stone mining region of
Luoyuan county, China.
High spatial resolution remote sensing images allow us
to obtain more detailed data and to improve mining region
mapping techniques. Remote sensing classification meth-
ods can be generally classified into two groups: pixel-based
approaches and object-oriented approaches. Conventional
per-pixel methods use only spectral information of individual
pixels, and information extracted using this approach does
not provide sufficient detail for high spatial resolution mining
region mapping. Object-oriented classification approaches
need to be adopted in order for satisfactory results to be ob-
tained (Lucieer
et al
., 2005, Bhaskaran
et al
., 2010, Kassouk
et
al
., 2014). Unlike pixel-based classifications, object-oriented
methods involve segmenting images into object segments and
then classifying these segments using spectral, spatial, tex-
tural, relational and contextual methods. Classification And
Regression Trees (
CART
) were first presented by Breiman
et al
.
in 1984. Hansen
et al
. (1996) introduced a
CART
decision tree
classifier for the examination of remotely sensed data.
CART
decision tree techniques effectively address remote sensing
land cover classification problems by virtue of their flexibility,
intuitive simplicity, and computational efficiency (Friedl
et
al
., 1997). A
CART
decision tree classifier was utilized in this
study. Additionally, a retrospective approach was employed
to conduct land cover change analyses in mining area from
2001 to 2010.
This study applies an object-oriented classification method
and decision tree classifier and retrospective approaches
to extract land cover and change information for a mining
region. A work flow is proposed, and high accuracy land
cover and change data are extracted for the mined region. The
article concludes with a summary of change monitoring pro-
cesses that have occurred in stone mining areas of Luoyuan
County.
Zhaoming Zhang, Guojin He, Zhihua Wang, Tengfei Long, and
Yan Peng are with the Institute of Remote Sensing and Digital
Earth, Chinese Academy of Sciences, Beijing 100094, China
(
).
Mengmeng Wang is with the Institute of Remote Sensing and
Digital Earth, Chinese Academy of Sciences, Beijing 100094;
and the University of Chinese Academy of Sciences, Beijing
100049, China.
Photogrammetric Engineering & Remote Sensing
Vol. 81, No. 9, September 2015, pp. 745–751.
0099-1112/15/745–751
© 2015 American Society for Photogrammetry
and Remote Sensing
doi: 10.14358/PERS.81.9.745
PHOTOGRAMMETRIC ENGINEERING & REMOTE SENSING
September 2015
745
