How Did Deciduous Rubber Plantations
Expand Spatially in China’s Xishuangbanna Dai
Autonomous Prefecture During 1991–2016?
Chiwei Xiao, Peng Li, and Zhiming Feng
Abstract
Rubber plantations have experienced continuous expan-
sion in Xishuangbanna, in southwestern China, since the
1950s. However, the question of how the establishment of
adventive rubber trees has spatially
of the Asian tropics in recent decad
stood. Here, a robust phenology-bas
the change rate of the Landsat-deriv
ratio—was modified based on tri-wi
tion, defoliation, and foliation) and then used to discriminate
deciduous rubber plantations from other land cover types by
applying a threshold of 1.0 and combining Landsat-based
forest masks every five years during 1991–2016. Deciduous
rubber plantations increased more than 6.6 times, approxi-
mately 3074 km
2
in 2016, or at an annual rate of about 8%
(104 km
2
/year) in Xishuangbanna over that period, show-
ing two typical expansion trends toward both higher (over
1000 m) and lower (below 600 m) elevations and increas-
ingly spread to borderlands with Laos and Myanmar.
Introduction
In recent decades, a substantial expansion of deciduous
rubber plantations has occurred in mainland Southeast Asia
(
MSEA
) and southern China where they historically have been
considered unsuitable (Ziegler
et al.
2009). Meanwhile, China
has since 2001 become the world’s largest importer and con-
sumer of natural rubber (Liu
et al.
2013). It has been reported
that the rapid development of Chinese rubber production has
occurred predominantly in Xishuangbanna Dai Autonomous
Prefecture (Xishuangbanna) within Yunnan Province since
the 1980s (Chen
et al.
2016), gaining a great deal of atten-
tion nationally and internationally (Li
et al.
2012; Qiu 2009).
Large-scale expansion of deciduous rubber plantations has
been responsible for significant eco-environmental effects
on biodiversity conservation (Li
et al.
2007), carbon seques-
tration (Li
et al.
2008), and water conservation (Qiu 2009).
Therefore, the detection and mapping of updated and histori-
cal information on deciduous rubber plantations is of great
significance to the scientific community and serves as a pre-
requisite data basis for research into rubber-related impacts.
To date, a number of studies have used remotely sensed
data to map deciduous rubber plantations in Xishuangbanna
at some temporal points with various kinds of satellite data.
However, these monitoring studies either lack longitudinal
analysis (Fan
et al.
2015; Kou
et al.
2015; Zhai
et al.
2018) or
use coarse spatial resolution data, such as Moderate Resolu-
tion Imaging Spectroradiometer (MODIS, 250 m; Li and Fox
2012; Senf
et al.
2013). Also, other time-series historical
products of deciduous rubber plantations neglect the updated
information (Liu
et al.
2013; Chen
et al.
2016; Kou
et al.
2018)
and cannot meet the needs of policy-makers in a timely man-
ner. Therefore, timely and updated geospatial data sets on
ns at finer spatial resolution (e.g.,
ently, a phenology-based single-
m dense Landsat Time Series (
LTS
)
tic Mapper (
TM
), Enhanced
TM
Plus
nd Imager (
OLI
)—was developed to
track the stand-age information of deciduous rubber planta-
tions in Xishuangbanna (Beckschäfer 2017). Therefore, the us-
age of Landsat intra- and interannual image properties related
to phenology for identifying deciduous rubber plantations has
become an important research trend (Dong
et al.
2013; Chen
et al.
2016; Xiao
et al.
2019), especially after the introduction
of free access to
LTS
imagery (Woodcock
et al.
2008). Com-
pared with traditional classifiers such as decision trees (Sun
et al.
2017), object-oriented (Zomer
et al.
2014), and super-
vised maximum likelihood based approaches (Li
et al.
2008),
phenological algorithms have already demonstrated great
potential for monitoring deciduous rubber plantations with
LTS
-based vegetation indices—including the Normalized Dif-
ference Vegetation Index (
NDVI
), Enhanced Vegetation Index,
and Land Surface Water Index—and the Normalized Burn
Ratio (
NBR
), as well as the nonvisible spectral bands (e.g.,
near-infrared [
NIR
] and shortwave infrared [
SWIR
]; Dong
et al.
2013; Fan
et al.
2015; Kou
et al.
2015). In addition, the half-
year-long dry spells in the tropics greatly facilitate Landsat
sensors’ acquiring cloud-free or low-cloud-cover (
CC
; 30% and
less) observations (Li
et al.
2018; Xiao
et al.
2018). Therefore,
LTS
data (i.e.,
TM
,
ETM+
, and
OLI
) can be a robust data source
for mapping deciduous rubber plantations within critical time
windows during the dry season.
We have developed a Landsat-derived bi-temporal method
with the
NBR
to map mature rubber plantations in Xishuang-
banna (Li
et al.
2015).
NBR
is more sensitive to phenological
changes in vegetation cover and canopy moisture of decidu-
ous rubber plantations. The bi-temporal algorithm is based
on the unique phenological features of deciduous rubber
plantations, including defoliation and the formation of new
leaves over two months during the peak of the dry season.
This approach has greatly improved the mapping accuracy of
deciduous rubber plantations with less data demand via the
incorporation of intensive phenological information—that is,
from single-date (Dong
et al.
2013; Kou
et al.
2015) to bi-tem-
poral (Li
et al.
2015). However, an unsettled question of how
the establishment of adventive rubber trees obtained from a
Chiwei Xiao, Peng Li, and Zhiming Feng are with the Institute
of Geographic Sciences and Natural Resources Research,
Chinese Academy of Sciences, Beijing, China; and the
College of Resources and Environment, University of Chinese
Academy of Sciences, Beijing, China (
.
Photogrammetric Engineering & Remote Sensing
Vol. 85, No. 9, September 2019, pp. 687–697.
0099-1112/19/687–697
© 2019 American Society for Photogrammetry
and Remote Sensing
doi: 10.14358/PERS.85.9.687
PHOTOGRAMMETRIC ENGINEERING & REMOTE SENSING
September 2019
687
