ArticlePDF Available

Patterns of tree-cover loss along the Indonesia–Malaysia border on Borneo Patterns of tree-cover loss along the Indonesia–Malaysia border on Borneo

Authors:

Abstract

Borneo Island is experiencing rapid tree-cover loss. This loss has been quantified for the Indonesian part of the island at Landsat spatial resolution, but no recent study exists that extends across the border into Malaysia. This research focused on quantifying patterns of tree-cover loss in the Indonesia–Malaysia border zone on Borneo. The methods used for quantifying 2000–2010 tree-cover loss within 20 km on either side of the border are an internally consistent mapping algorithm used on Landsat imagery and a local indicator of spatial autocorrelation to quantify the concentration of loss. Within the 20 km zone on either side of the border, tree-cover loss rates in lowlands were high in the two countries (19.8% and 14.4%, in Indonesia and Malaysia, respectively), but rates in the Malaysian uplands were an order of magnitude higher than in the Indonesian uplands (2.95% and 0.25%, respectively). Clusters of tree-cover loss in the Malaysian uplands were considerably larger than in the Indonesian uplands.
Full Terms & Conditions of access and use can be found at
http://www.tandfonline.com/action/journalInformation?journalCode=tres20
Download by: [UNSW Library] Date: 12 May 2016, At: 01:02
International Journal of Remote Sensing
ISSN: 0143-1161 (Print) 1366-5901 (Online) Journal homepage: http://www.tandfonline.com/loi/tres20
Patterns of tree-cover loss along the
Indonesia–Malaysia border on Borneo
Mark Broich , Matthew Hansen , Peter Potapov & Michael Wimberly
To cite this article: Mark Broich , Matthew Hansen , Peter Potapov & Michael Wimberly (2013)
Patterns of tree-cover loss along the Indonesia–Malaysia border on Borneo, International
Journal of Remote Sensing, 34:16, 5748-5760, DOI: 10.1080/01431161.2013.796099
To link to this article: http://dx.doi.org/10.1080/01431161.2013.796099
Published online: 17 May 2013.
Submit your article to this journal
Article views: 312
View related articles
Citing articles: 2 View citing articles
International Journal of Remote Sensing, 2013
Vol. 34, No. 16, 5748–5760, http://dx.doi.org/10.1080/01431161.2013.796099
Patterns of tree-cover loss along the Indonesia–Malaysia
border on Borneo
Mark Broicha*, Matthew Hansenb, Peter Potapovb, and Michael Wimberlyc
aClimate Change Cluster, University of Technology, Sydney, Ultimo, NSW 2007, Australia;
bDepartment of Geographical Sciences, University of Maryland, College Park, MD 20742, USA;
cGeographic Information Science Center of Excellence, South Dakota State University, Brookings,
SD 57007, USA
(Received 15 May 2011; accepted 4 April 2013)
Borneo Island is experiencing rapid tree-cover loss. This loss has been quantified for the
Indonesian part of the island at Landsat spatial resolution, but no recent study exists that
extends across the border into Malaysia. This research focused on quantifying patterns
of tree-cover loss in the Indonesia–Malaysia border zone on Borneo. The methods used
for quantifying 2000–2010 tree-cover loss within 20 km on either side of the border
are an internally consistent mapping algorithm used on Landsat imagery and a local
indicator of spatial autocorrelation to quantify the concentration of loss. Within the
20 km zone on either side of the border, tree-cover loss rates in lowlands were high
in the two countries (19.8% and 14.4%, in Indonesia and Malaysia, respectively), but
rates in the Malaysian uplands were an order of magnitude higher than in the Indonesian
uplands (2.95% and 0.25%, respectively). Clusters of tree-cover loss in the Malaysian
uplands were considerably larger than in the Indonesian uplands.
1. Introduction
Monitoring of humid tropical forest extent and change has progressed rapidly, with some
countries, such as Brazil, having implemented systematic annual updates of humid tropi-
cal forest disturbance (INPE 2010). Forest monitoring has been conducted in the past for
areas in South East Asia (GOI/World Bank 2000; Achard et al. 2002; Page et al. 2002; van
der Werf et al. 2004; Mayaux et al. 2005; Langner, Miettinen, and Siegert 2007; Hansen
et al. 2009; Dorais and Cardille 2011; FAO 1996, 2005, 2010). In these studies, tree-cover
change has been either mapped using coarse spatial resolution data (Page et al. 2002; van
der Werf et al. 2004; Langner, Miettinen, and Siegert 2007; Dorais and Cardille 2011),
estimated using sample-based approaches (FAO 1996, 2005; Achard et al. 2002; Hansen,
Stehman, et al. 2008; Mayaux et al. 2005; Hansen et al. 2009) or mapped for specific coun-
tries or in-country regions using higher spatial resolution data (GOI/World Bank 2000;
Uryu et al. 2008; Gaveau, Epting, et al. 2009; Gaveau, Wich, 2009). Sample-based esti-
mates can provide accurate and precise tree-cover change rates but do not result in spatially
explicit information. Change maps based on coarse spatial resolution imagery are valu-
able for detecting hotspots of change (Broich et al. 2009; Hansen et al. 2009; Dorais and
Cardille 2011) but are known to be inaccurate for area estimation (Morton et al. 2005),
*Corresponding author. Email: mark.broich@gmail.com
© 2013 Taylor & Francis
Downloaded by [UNSW Library] at 01:02 12 May 2016
International Journal of Remote Sensing 5749
where higher spatial resolution is required. The persistent cloud cover over much of insular
Southeast Asia is the main reason that has long precluded monitoring with higher spatial
resolution optical imagery. Robust monitoring approaches providing Landsat spatial reso-
lution maps of tree-cover loss have only recently been demonstrated for Indonesia (Broich,
Hansen, Potapov, et al. 2011; Potapov et al. 2012). However, up-to-date change maps with
a spatial resolution of Landsat multispectral bands or maps better depicting both sides of
the border between Indonesia and Malaysia on Borneo do not exist.
1.1 Approach
This article examines the application of a Landsat-based tree-cover extent and loss mapping
algorithm to the island of Borneo, and quantifies the variation of tree-cover loss rates and
patterns along the border between Indonesia and Malaysia. We provide an up-to-date depic-
tion of the tree loss from 2000–2010 and identify statistically significant clearing clusters
quantifying the spatial distribution of tree-cover loss on both sides of the Indonesia–
Malaysia border on Borneo. We investigate whether the rate of tree-cover loss over the
past ten years in the upland and lowlands near the border in Malaysia were higher than on
the Indonesian side, and if upland and lowland tree-cover loss patterns near the border in
Malaysia were more spatially concentrated than on the Indonesian side.
Similar to the border area land-cover change quantifications undertaken by other
researchers (Kuemmerle et al. 2006, 2007; Soto-Berelov and Madsen 2011; Southworth
et al. 2011), we aim to identify cross-border differences in tree-cover loss rates and pat-
terns. We aim to identify differences that are likely due to dissimilarities in societal,
economic, and political conditions as well as to remoteness from infrastructure and markets
in the respective countries rather than ecological factors, as the areas along the border are
ecologically similar (Kuemmerle et al. 2006, 2007).
We first map tree-cover loss at Landsat spatial resolution using an internally consis-
tent, automated mapping approach. From the resulting map, we derive tree-cover loss rates.
The derived tree-cover loss map also serves as the input for a local indicator of spatial
autocorrelation (LISA) cluster detection algorithm. The LISA serves to identify statisti-
cally significant clusters of tree-cover loss, which allows the quantitative analysis of spatial
patterns of tree-cover loss.
2. Methods
2.1. Study area and definitions
Borneo’s island area is shared between three nations: Indonesia, Malaysia, and Brunei.
The 1600 km-long border between Indonesia and Malaysia runs mostly through forested
uplands. The forests of Borneo are important for climate change mitigation, biodiver-
sity conservation, watershed protection, the livelihood of local communities, and for the
national economies of Malaysia and Indonesia (FWI/GFW 2002; Sodhi et al. 2004; Myers
et al. 2000). The biodiversity value of the uplands of Borneo is internationally considered
as significant. While the lowlands have only limited areas under legal protection, many and
some large protected areas are located in remote uplands close to the border, predominantly
on the Indonesian side (IUCN and UNEP 2009).
Borneo is a hotspot of tree-cover loss (Achard et al. 2002; Mayaux et al. 2005; Hansen,
Stehman, et al. 2008, 2009). In 2000, the remaining Intact Forest Landscapes (Potapov et al.
2008) near the border were smaller in the Malaysian part than in the Indonesian part of the
island. This can be interpreted as more ‘past’ tree-cover loss up until 2000 (Figure 1).
Downloaded by [UNSW Library] at 01:02 12 May 2016
5750 M. Broich et al.
Figure 1. Map of tree-cover loss 2000–2010 (red) on a 2000 Landsat composite (bands 5/4/7as
R/G/B) showing the area close to the border (yellow) on Borneo. Indonesia and Malaysia are
located south and north of the border, respectively. Year 2000 Intact Forest Landscapes (Potapov
et al. 2008) are shown as black outlined crosshatch. Centre of figure is 1 50 N, 114 20 E.
We chose a 20 km buffer on either side of the Indonesia–Malaysia border on Borneo
as our study area to evaluate differences in tree-cover loss patterns near the border.
We subdivided the study area into upland and lowland that we assumed to be ecologi-
cally homogeneous. We defined lowland as <300 m elevation following the definition of
Global Forest Watch (FWI/GFW 2002) and areas of 300 m elevation as upland (Figure 2).
We delineated the zones based on digital elevation data obtained by the Shuttle Topography
Radar Mission (SRTM; Rabus et al. 2003; USGS 2006).
We used a tree cover and tree-cover loss definition that conforms with our previous
work (Broich, Hansen, Potapov, et al. 2011; Broich, Hansen, Stolle, et al. 2011) defining
tree cover as >25% canopy cover of trees 5 m in height. We measured tree-cover loss
without regard to land use, therefore including forest, plantation, and oil palm estate change
dynamics. Based on the definition of tree cover, we used a tree-cover loss definition that
describes the conversion of tree cover to non-tree cover and represents a stand-replacement
disturbance at a 60 m spatial resolution. By the term ‘tree-cover loss’, we stress that we
did not differentiate between a temporary loss of tree cover and a permanent conversion to
a land use without tree cover. Therefore, ‘tree-cover loss’ represents gross rather than net
area and rate. Intensive logging that does lead to a tree cover/nontree cover conversion at
the Landsat pixel scale qualifies as tree-cover loss. ‘Selective logging’, which is commonly
defined as tree harvesting that does not result in conversion of tree cover to non-tree cover,
is not included in our definition of tree-cover loss.
2.2. Mapping method
To map tree-cover loss at 60 m × 60 m spatial resolution for the 2000–2010 time inter-
val, we ingested 2300 Landsat Enhanced Thematic Mapper Plus (ETM+) images and
10 years of Moderate Resolution Imaging Spectroradiometer (MODIS) imagery following
Downloaded by [UNSW Library] at 01:02 12 May 2016
5751 International Journal of Remote Sensing
Figure 2. Map showing the border (yellow) and the 20 km buffer around the border with upland
(blue) and lowland areas (white). Centre of figure is 1 50 N, 114 20 E.
the methods described in Hansen, Roy, et al. (2008) and Broich, Hansen, Potapov, et al.
(2011). We used a supervised decision-tree algorithm (Breiman 1996; Prasad, Iverson,
and Liaw 2006) to classify per pixel time series into either tree-cover loss or no change
areas. In this method, we selected all ETM+ imagery in the US Geological Survey/Earth
Resources Observation and Science Center (USGS/EROS) archive from 2000 to 2010 with
less than 50% cloud cover in any quarter of the image according to image metadata.
We resampled the Landsat data to 60 m spatial resolution to compensate for the effect
of residual misalignment between images. Each Landsat image was normalized using the
MODIS-mapped dark tree cover as a reference (Broich, Hansen, Potapov, et al. 2011;
Hansen, Roy, et al. 2008). This radiometric normalization was necessary to use generic,
decision tree-based per pixel classification algorithms. Such generic algorithms were used
to flag cloud contamination and estimate the probability of the uncontaminated pixel being
tree cover. Input variables of the generic classifier consisted of the Landsat bands, band
ratios, and topographic indices derived from the SRTM digital elevation data (Rabus et al.
2003; USGS 2006). To derive the normalization reference target, we used MODIS 16-day
composites of the MODIS land bands (Vermote, El Saleous, and Justice 2002) and the
10,780–11,280-nm thermal band. We classified the MODIS data into a map of persistently
dark tree cover at a spatial resolution of 250 m × 250 m. The map was derived from a
regression tree algorithm following the method of the standard vegetation continuous field
product of the MODIS land science team (Hansen et al. 2003). For the purpose of normal-
ization, the MODIS-mapped, persistently dark tree cover over the 2000–2010 interval was
assumed to be spectrally invariant. Due to limited acquisitions per year and persistent cloud
cover over the study area, cloud-free ETM + images are highly unlikely. To allow complete
cover mapping of tree-cover loss, cloud-free observations acquired over ten years were
assembled into per-pixel time series of tree-cover probability. These per-pixel time series
were then characterized into tree-cover loss and no change using a decision tree algorithm
(Broich, Hansen, Potapov, et al. 2011). The resulting map shows clearings that were created
and main logging roads that were constructed between 2000 and 2010.
Downloaded by [UNSW Library] at 01:02 12 May 2016
5752 M. Broich et al.
2.3. Cluster detection
Using the mapped tree-cover loss as input, we identified clusters of tree-cover loss by means
of LISA, the local Moran’s I statistic (Anselin 1995). The local Moran’s I statistic allows
the identification of areas with significant spatial clusters relative to the mean value i n the
area under investigation.
The local Moran’s I value for each cell in the study is computed as (Equation 1):
Ii = (xi xbar) wij(xj xbar) , (1)
where xi represents the value at the current cell, xj represents the value at a neighbouring
cell, xbar represents the average value of all cells in the study area and wij represents a row
standardized weight matrix. Resulting local Moran’s I values can be interpreted as follows:
values greater than 0 indicate positive local autocorrelation and values less than 0 indicate
negative local autocorrelation. Positive local autocorrelation can result from cells with high
values in the neighbourhood of cells with high values (high-high) or cells with low values in
the neighbourhood of cells with low values (low-low). Statistical significance was assessed
using a randomization test in which the spatial location of cells are randomly per muted n
times to generate a reference distribution for each cell under the null hypothesis of no local
autocorrelation.
The local Moran’s I index has been used in several studies associated with tropical tree
cover and tree-cover loss. For example, Southworth, Munroe, and Nagendra (2004) used
the index with the normalized difference vegetation index (NDVI) to examine the patterns
of landscape fragmentation for a forested study area in western Honduras. The index was
also used by Chaves et al. (2008) to identify clustering of disease risk factors that are
associated with forest clearance. Sunderlin et al. (2008) used a bi-variate local Moran’s I
index in various tropical countries to identify regions with different relationships between
rural poverty and forest.
We calculated the local Moran’s I of tree-cover loss. Because of computational con-
straints, we computed the index and associated significance levels for percentage of
tree-cover loss within 1 km cells. Because uplands and lowlands are known to have dif-
ferent tree-cover loss dynamics (Hansen et al. 2009; Mayaux et al. 2005), we computed the
Moran’s I statistic separately for uplands and lowlands. We defined the local spatial neigh-
bourhood of each cell as the eight adjacent cells that share an edge or vertex (queen’s rule).
Data were logarithmically transformed and tests of statistical significance were carried out
using a permutation test where n = 999. We selected from the local Moran’s I statistics
all 1 km cells that had a local Moran’s I value >1, exhibit ‘high-high’, and had a signifi-
cance level of <0.01. We conducted a neighbourhood analysis to indentify the statistically
significant 1 km cells that belonged to the same cluster. We then computed and compared
the cluster size distribution for uplands and lowlands of both the Indonesian and Malaysian
sides of the border.
3. Results
The analysis compared rates of loss over the period 2000–2010 and examined the size
distribution of statistically significant clusters of tree-cover loss in the uplands and lowlands
of Indonesia and Malaysia within a 20 km buffer on either side of the border. The uplands
of the study area account for 63% of the total area.
Downloaded by [UNSW Library] at 01:02 12 May 2016
5753 International Journal of Remote Sensing
Table 1 shows the areas within the study area belonging to uplands and lowlands in
Indonesia and Malaysia and further subdivides these areas into protected area, and area
within year 2000 Intact Forest Landscapes (IFL).
Almost 42% of the study area was identified as IFL in 2000. A larger proportion of the
uplands (61%) compared to lowlands (almost 5%) still belonged to IFL in 2000. Of the
Malaysian part of the study area, 27% qualified as ILF in 2000 compared to 55% on the
Indonesian side of the border. Almost 28% of the study area is legally protected. A larger
part of the uplands is protected (38%) compared to the lowlands (8.5%). While almost half
of the Indonesian part of the study area is protected, only 5% of the Malaysian part protects
forests from clearing.
3.1. Tree-cover loss rates and patterns
Upland areas within 20 km of the border accounted for 16% of the overall tree-cover loss
in the border area of both countries from 2000 till 2010. The tree-cover loss rate in the
uplands within 20 km of the border was 1.6%, with the uplands of Malaysia experiencing a
higher rate of tree-cover loss (2.95%) than the uplands of Indonesia (0.25%) within 20 km
of the border. The overall tree-cover loss rate in lowlands within 20 km of the border was
17.3%. The loss rate within 20 km of the border in Indonesia was higher than the loss rate
in Malaysia (19.8% and 14.4%, respectively).
We analysed the spatial concentration of tree-cover loss patterns in the uplands and
lowlands of Indonesia and Malaysia within 20 km of the border in terms of cluster size
distribution. Tree-cover loss patterns in the Malaysian uplands were more concentrated
than in the Indonesian uplands as indicated by the cluster size distribution (Figure 3(a)).
The largest Indonesian upland cluster covered 44 km2, while the largest cluster in the
Malaysian uplands encompassed 858 km2. Tree-cover loss patterns in the Malaysian
lowlands appeared more concentrated than in Indonesian lowlands as indicated by the
cluster size distribution (Figure 3(b)). However, the largest cluster in Indonesia’s lowlands
(583 km2) was considerably larger than in Malaysia’s lowlands (152 km2).
4. Discussion
In this research we compared tree-cover loss rates and patterns in the lowlands and uplands
along the border between Malaysia and Indonesia using Landsat data from 2000 to 2010.
Tree-cover loss rates in lowlands were similar between the two countries compared to
uplands where differences were large. While uplands experienced lower tree-cover loss
rates than lowlands, the upland area of Malaysia experienced loss rates an order of magni-
tude higher than the Indonesian uplands close to the border. Additionally, the cluster size
distribution analysis showed that loss is more concentrated in Malaysia’s uplands compared
to Indonesia’s.
We carried out the local Moran’s I cluster detection separately for the upland and the
lowland as clusters are identified relative to the mean value in the area under investigation.
The known differences between upland and lowland in terms of land use also encouraged
this separate analysis. The uplands are characterized by small isolated loss patches on both
sides of the border, larger loss areas mostly in Malaysia, and extensive areas of no change
(Figure 4). Conversely, the lowlands are characterized by continuous areas with no change,
many small loss patches in certain areas, and patches of large-scale loss. These lowland
loss patterns are probably caused by small- and medium-scale agricultural activities, and
by the development of large oil palm and commercial timber plantations. Examples of loss
Downloaded by [UNSW Library] at 01:02 12 May 2016
Table 1. Areas (km2) within the study area inside the 20 km buffer around the border belonging to uplands and lowlands in Indonesia and Malaysia
with percentages of area within protected area and year 2000 IFL.
Percentage of protected area within Percentage of year 2000 IFL within
Entire area within buffer buffer buffer
Malaysia Indonesia Both countries Malaysia Indonesia Both countries Malaysia Indonesia Both countries
Lowland area 9174 9827 19,001 3.7 12.9 8.5 6.6 2.7 4.6
Upland area 17,946 18,490 36,436 6.1 69.0 38.0 38.2 83.5 61.2
Total area 27,120 28,317 55,437 5.3 49.5 27.9 27.5 55.4 41.8
5754 M. Broich et al.
Downloaded by [UNSW Library] at 01:02 12 May 2016
5755 International Journal of Remote Sensing
Figure 3. Cluster size distribution for Indonesia (squares) and Malaysia (triangles) in (a) uplands
and (b) lowlands within 20 km either side of the border. The distribution is described as quartiles
(q25 is the first quartile, med is the median, and q75 is the third quartile).
Figure 4. Tree-cover loss clusters (black outlined crosshatch). The border is shown in yellow and
the 20 km buffer is s hown in blue. Indonesia and Malaysia are located south and north of the border,
respectively. Centre of figure is 2 13 N, 114 48 E.
patterns likely due to small- and medium-scale agricultural activities that are not flagged by
the clustering method can be seen in Figure 5 as smaller red patches. The use of the local
Moran’s I cluster detection aided the visual interpretation of the per-pixel tree-cover loss
mapping results and allowed the quantitative analysis of the tree-cover loss patterns.
Timber and palm oil production are specifically important for the economies of east-
ern Malaysia and Indonesian Borneo. These broad-scale factors have been associated with
high tree-cover loss rates (FAO 2001; FWI/GFW 2002; Rautner and Hardiono 2005; ITTO
2006). In Indonesia, the rapid expansion in the processed wood products sector has created
a critical imbalance between legal wood supplies and consumption that has caused illegal
Downloaded by [UNSW Library] at 01:02 12 May 2016
5756 M. Broich et al.
Figure 5. Example for tree-cover loss clusters in lowlands and uplands (black outlined hatch and
blue outlined areas, respectively). The border is shown in yellow and the 20 km buffer with upland
and lowland areas are shown in blue and white, respectively. Centre of figure is 1 2 N, 111 29 E.
clearing of natural forest (FWI/GFW 2002; Uryu et al. 2008). Likewise, eastern Malaysia
has a large wood-processing capacity but has insufficient local timber supply to cover the
resulting demand (Rautner and Hardiono 2005; EIA 2001). Besides being major players in
the tropical timber and timber products market, Indonesia and Malaysia are the world’s two
largest producers of palm oil (Sheil et al. 2009). The expansion of oil palm, which is grown
mostly at low elevations, has been related to large-scale deforestation (Sheil et al. 2009).
Interestingly, the largest overall cluster in the study area (833 km2) was not located in
the lowlands but in the uplands of Malaysia, in an area bordering a relatively disturbance-
free Indonesian upland landscape (Figure 4). While the lowlands of the two countries near
the border are somewhat alike in terms of tree-cover loss rates and patterns, the uplands of
Malaysia stand out with their higher rate and more concentrated loss patter n compared to
that on the Indonesian side of the border.
These findings suggest that Malaysian upland forests close to the border have been more
intensively used than their Indonesian counterpart, which is in line with and augments the
IFL maps of the year 2000. Tree-cover loss typically takes place only in accessible locations
in the lowland and have, with diminishing lowland resources and increasing timber prices,
expanded to the sloped terrain of the uplands (Ross 2001; Jomo, Khoo, and Chang 2004).
Our findings suggest that logging in Malaysia has now expanded to reach the remote border
area, and this result provides an update on tree-cover loss relative to the 2000 IFL map
(Figure 1). The lack of large tree loss clusters close to the border in Indonesia may be due
to the more remote and less accessible location of the Indonesian uplands compared to
those on the Malaysian side.
Inaccessible upland forest is known to be less likely to be cleared (Gaveau, Wich, et al.
2009; Rautner and Hardiono 2005). Factors that limit ‘accessibility’ include steep terrain
slope and high elevation, long travel distance to the edge of the forest, to existing roads and
logging roads as well as legal forest protection status and high level of law enforcement
Downloaded by [UNSW Library] at 01:02 12 May 2016
5757 International Journal of Remote Sensing
(Kaimowitz and Angelsen 1998; Mas 2005; Andam et al. 2008; Gaveau, Epting, et al. 2009;
Gaveau, Wich, 2009). Many protected areas are located in the remote uplands. The propor-
tion of Malaysia’s uplands on Borneo within the study area that is under legal protection is
smaller compared to Indonesia’s (IUCN and UNEP 2009). Therefore, the combination of
factors that may make the Indonesian uplands near the border less ‘accessible’ compared to
their Malaysian counterpart could be the greater distance to past tree-cover loss, the pres-
ence of fewer logging roads, and the protection status. We assume that these broad-scale
factors, which have been related to tree-cover change in the two countries, also influence
the study area along the Indonesia–Malaysia border on Borneo. Local factors for the study
area are not specified or quantified in the literature.
Our results describe only the area under investigation, specifically a 20 km buffer on
either side of the border and may not be representative of all of Borneo, let alone the entire
nations of Indonesia and Malaysia. Besides the difference in accessibility, the quantified
difference in rates and patterns of gross tree-cover loss may be due to Indonesia having
more remaining forested land area in the nation as a whole.
Numerous studies in different parts of the globe have highlighted the importance of
analysing the boundary between countries, where ecological conditions are similar and
differences between broadscale socio-economic factors, accessibility factors, and policies
translate directly into differences in land-cover change (Kuemmerle et al. 2006, 2007; Soto-
Berelov and Madsen 2011; Southworth et al. 2011). According to Geist and Lambin (2002)
and Linderman et al (2005), local scale factors are certainly important drivers of land cover
change, yet the influence of broad-scale political and socio-economic factors may dominate
change dynamics (Lambin et al. 2001).
For the study area, research investigating forest change is missing. Our study con-
tributes to closing this knowledge gap by quantifying tree-cover loss rates and patterns
in the border area based on internally consistent algorithms and by quantifying the pat-
tern of tree-cover loss in the lowlands compared with the uplands. To further narrow the
knowledge gap in the uplands of Borneo, more research into both the local and broader
scale drivers of tree-cover loss is needed. We suggest that future research should collect
spatial data on accessibility and link these with patterns of tree-cover loss. Investigating,
for example, aspects of trans-border illegal logging remains a challenge for future research
and requires identifying the agents responsible for the detected tree-cover loss. While the
rates and patterns of gross lowland tree-cover loss i n the two countries are similar, these
lowland results need to be viewed with caution because they do not distinguish between
permanent and temporary loss associated with different land use. In the uplands, the vast
majority of tree-cover loss is likely due to clearing of natural forest rather than the harvest-
ing of oil palm or pine plantations that occur mostly in lowlands. The disaggregation of the
change signal by forest type and land use requires further research.
5. Conclusion
This article investigates tree-cover loss patterns along the Indonesia–Malaysia border on
Borneo. Specific advances over previous work are the quantification of tree-cover loss with
an internally consistent mapping method at an unprecedented spatial resolution of 60 m,
followed by the identification of statistically significant clusters of mapped tree-cover loss
over the period 2000–2010. These complementary approaches were used to compare and
contrast tree-cover loss on the Indonesian and Malaysian sides of the border on Borneo.
The use of the local Moran’s I index to identify clusters of mapped tree-cover loss assisted
the visual interpretation and qualitative analysis of the spatial patterns.
Downloaded by [UNSW Library] at 01:02 12 May 2016
5758 M. Broich et al.
Tree-cover loss rates in Malaysia’s uplands close to the border were found to be an order
of magnitude higher than in neighbouring Indonesia. Tree-cover loss in Malaysia’s uplands
close to the boarder also occurred in larger clusters than in Indonesia’s uplands. Our finding
that Malaysia is experiencing high and intensive tree-cover loss along the upland border
with Indonesia may be explained by the imbalance between the large wood-processing
capacity and insufficient timber supply, combined with the smaller area of remaining forest
area of Malaysian relative to Indonesian Borneo. For Malaysia, this may indicate a possible
end to high wood production due to overexploitation. In Indonesia, overexploitation has
been documented in lowlands and uplands but may not have reached the remote uplands of
Indonesian Borneo just yet. Future research to quantify both the type of forest utilization
and the local and broader scale drivers of loss is regarded as highly relevant.
Acknowledgements
Funding from the National Aeronautics and Space Administration supported this research under grant
NNG06GD95G managed under the NASA Land Cover Land Use Change programme (manager: Dr
Garik Gutman).
References
Achard, F., E. D. Hugh, H.-J. S. Stibig, P. Mayaux, J. Gallego, T. Richards, and J. P. Malingreau. 2002.
“Determination of Deforestation Rates of the World’s Humid Tropical Forests.” Science 297:
999–1002.
Andam, K. S., P. J. Ferraro, A. Pfaff, G. A. Sanchez-Azofeifa, and J. A. Robalino. 2008. “Measuring
the Effectiveness of Protected Area Networks in Reducing Deforestation.” Proceedings of the
National Academy of Sciences of the United States of America 105: 16089–16094.
Anselin, L. 1995. “Local Indicators of Spatial Association LISA.” Geographical Analysis 27:
93–115.
Breiman, L. 1996. “Bagging Predictors.” Machine Learning 24: 123–140.
Broich, M., M. C. Hansen, P. V. Potapov, and S. V. Stehman. 2011. “Time-series Analysis of Multi-
resolution Optical Imagery for Quantifying Forest Cover Loss in Sumatra and Kalimantan,
Indonesia.” Journal of Applied Earth Observation and Geoinformation 13: 277–291.
Broich, M., M. C. Hansen, F. Stolle, P. V. Potapov, B. Arunarwati, and B. Adusei. 2011. “Remotely
Sensed Forest Cover Loss Shows High Spatial and Temporal Variation Across Sumatera and
Kalimantan, Indonesia 2000–2008.” Environmental Research Letters 6. Accessed March 15,
2013. http://iopscience.iop.org/1748-9326/6/1/014010/fulltext/
Broich, M., S. V. Stehman, M. C. Hansen, P. V. Potapov, and Y. E. Shimabukuro. 2009. “A
Comparison of Sampling Designs for Estimating Deforestation from Landsat Imagery: A Case
Study of the Brazilian Legal Amazon.” Remote Sensing of Environment 113: 2448–2454.
Chaves, L. F., J. M. Cohen, M. Pascual, and L. M. Wilson. 2008. “Social Exclusion Modifies Climate
and Deforestation Impacts on a Vector-Borne Disease.” Plos Neglected Tropical Diseases 2.
Accessed March 15, 2013. http://www.plosntds.org/article/info%3Adoi%2F10.1371%2Fjournal.
pntd.0000176
Dorais, A., and J. Cardille. 2011. “Strategies for Incorporating High-Resolution Google Earth
Databases to Guide and Validate Classifications: Understanding Deforestation in Borneo.”
Remote Sensing 3: 1157–1176.
EIA (Environmental Investigation Agency). 2001. Timber Trafficking: Illegal Logging in Indonesia,
South East Asia and International Consumption of Illegally Sourced Timber. London:
Environmental Information Agency.
FAO (Food and Agriculture Organization). 1996. Forest Resources Assessment 1990: Survey of
Tropical Forest Cover and Study of Change Processes. Forestry Paper No. 130. Rome: Food
and Agriculture Organization of the United Nations.
FAO (Food and Agriculture Organization). 2001. Global Forest Resources Assessment 2000 Main
Report. Forestry Paper No. 140. Rome: Food and Agriculture Organization of the United Nations.
Downloaded by [UNSW Library] at 01:02 12 May 2016
5759 International Journal of Remote Sensing
FAO (Food and Agriculture Organization). 2005. State of the World’s Forests. Rome: Food and
Agriculture Organization of the United Nations.
FAO (Food and Agriculture Organization). 2010. Global Forest Resources Assessment. Rome: Food
and Agriculture Organization of the United Nations).
FWI/GFW (Forest Watch Indonesia/Global Forest Watch). 2002. The State of the Forest. Bogor:
Forest Watch Indonesia/Global Forest Watch.
Gaveau, D. L. A., J. Epting, O. Lyne, M. Linkie, K. Kumara, M. Kanninen, and N. Leader-Williams.
2009. “Evaluating Whether Protected Areas Reduce Tropical Deforestation in Sumatra.” Journal
of Biogeography 36: 2165–2175.
Gaveau, D. L. A., S. Wich, J. Epting, D. Juhn, M. Kanninen, and N. Leader-Williams. 2009.
“The Future of Forests and Orangutans (Pongo Abelii) in Sumatra: Predicting Impacts of Oil
Palm Plantations, Road Construction, and Mechanisms for Reducing Carbon Emissions from
Deforestation.” Environmental Research Letters 4. Accessed March 15, 2013. http://iopscience.
iop.org/1748-9326/4/3/034013/fulltext/
Geist, H. J., and E. F. Lambin. 2002. “Proximate Causes and Underlying Driving Forces of Tropical
Deforestation.” Bioscience 52: 143–150.
GOI (Government of Indonesia)/World Bank. 2000. Deforestation in Indonesia: A Review of the
Situation in 1999. Jakarta: Government of Indonesia/World B ank.
Hansen, M. C., R. S. DeFries, J. R. G. Townshend, M. Carroll, C. Dimiceli, and R. A. Sohlberg.
2003. “Global Percent Tree Cover at a Spatial Resolution of 500 Meters: First Results of
the MODIS Vegetation Continuous Fields Algorithm.” Earth Interactions 7. Accessed March
15, 2013. http://journals.ametsoc.org/doi/abs/10.1175/1087-3562%282003%29007%3C0001%
3AGPTCAA%3E2.0.CO%3B2
Hansen, M. C., D. P. Roy, E. Lindquist, B. Adusei, C. O. Justiceand, and A. Altstatt. 2008. “A Method
for Integrating MODIS and Landsat Data for Systematic Monitoring of Forest Cover and Change
in the Congo Basin.” Remote Sensing of Environment 112: 2495–2513.
Hansen, M. C., S. V. Stehman, P. V. Potapov, B. Ar unarwati, F. Stolle, and K. Pittman. 2009.
“Quantifying Changes in the Rates of Forest Clearing in Indonesia from 1990 to 2005 Using
Remotely Sensed Data Sets.” Environmental Research Letters 4. Accessed March 15, 2013. http:/
/iopscience.iop.org/1748-9326/4/3/034001/fulltext/
Hansen, M. C., S. V. Stehman, P. V. Potapov, T. R. Loveland, G. R. G. Townshend, R. S. Defries,
and K. W. Pittman. 2008. “Humid Tropical Forest Clearing from 2000 to 2005 Quantified by
Using Multitemporal and Multiresolution Remotely Sensed Data.” Proceedings of the National
Academy of Sciences 105: 9439–9444.
ITTO (International Tropical Timber Organization). 2006. Status of Tropical Forest Management
2005. ITTO Technical Series No 24. Yokohama: International Tropical Timber Organization.
IUCN (International Union for Conservation of Nature), and UNEP (United Nations Environment
Programme). 2009. The World Database on Protected Areas (WDPA). Cambridge: UNEP-
WCMC.
Jomo, K. S., K. J. Khoo, and Y. T. Chang. 2004. Deforesting Malaysia: The Political Economy and
Social Ecology of Agricultural Expansion and Commercial Logging. London: Zed Books.
Kaimowitz, D., and A. Angelsen. 1998. Economic Models of Tropical Deforestation: A Review.
Bogor: Centre for International Forestry Research.
Kuemmerle, T., P. Hostert, V. C. Radeloff, K. Perzanowski, and I. Kruhlov. 2007. “Post-Socialist
Forest Distubance in the Carpathian Border Region of Poland, Slovakia, and Ukraine.” Ecological
Applications 17: 1279–1295.
Kuemmerle, T., V. C. Radeloff, K. Perzanowski, and P. Hostert. 2006. “Cross-border Comparison of
Land Cover and Landscape Pattern in Eastern Europe Using a Hybrid Classification Technique.”
Remote Sensing of Environment 103: 449–464.
Lambin, E. F., B. L. Turner, H. J. Geist, S. B. Agbola, A. Angelsen, J. W. Bruce, and O. T. Coomes.
2001. “The Causes of Land-use and Land-cover Change: Moving Beyond the Myths.” Global
Environmental Change 11: 261–269.
Langner, A., J. Miettinen, and F. Siegert. 2007. “Land Cover Change 2002–2005 in Borneo and the
Role of Fire Derived from MODIS Imagery.” Global Change Biology 13: 2329–2340.
Linderman, M. A., L. An, S. Bearer, G. He, Z. Ouyang, and J. Liu. 2005. “Modeling the Spatio-
temporal Dynamics and Interactions of Households, Landscapes, and Giant Panda Habitat.”
Ecological Modelling 183: 47–65.
Downloaded by [UNSW Library] at 01:02 12 May 2016
5760 M. Broich et al.
Mas, J. F. 2005. “Assessing Protected Area Effectiveness Using Surrounding (buffer) Areas
Environmentally Similar to the Target Area.” Environmental Monitoring and Assessment 105:
69–80.
Mayaux, P., P. Holmgren, F. Achard, H. Eva, H. J. Stibig, and A. Branthomme. 2005. “Tropical Forest
Cover Change in the 1990s and Options for Future Monitoring.” Philosophical Transactions of
the Royal Society B: Biological Sciences 360: 373–384.
Morton, D. C., R. S. DeFries, Y. E. Shimabukuro, L. O. Anderson, D. P. Espirito-Santo, M. C. Hansen,
and M. Carroll. 2005. “Rapid Assessment of Annual Deforestaion in the Brazilian Amazon Using
MODIS Data.” Earth Interactions 9. Accessed March 15, 2013. http://journals.ametsoc.org/doi/
abs/10.1175/EI139.1
Myers N., R. A. Mittermeier, C. G. Mittermeier, G. A. B. DaFonseca, and J. Kent. 2000. “Biodiversity
Hotspots for Conservation Priorities.” Nature 403: 853–858.
INPE (National Intitute For Space Research). 2010. Brazilian National Intitute For Space Research.
Accessed February 2013. http://www.inpe.br/ingles/
Page, S. E., F. Siegert, J. O. Rieley, H. D. V. Boehm, A. Jaya, and S. Limin. 2002. “The Amount of
Carbon Released from Peat and Forest Fires in Indonesia During 1997.” Nature 420: 61–65.
Potapov, P. V., S. Turubanova, M. C. Hansen, B. Adusei, M. Broich, A. Altstatt, L. Mane, and C. O.
Justice. 2012. “Quantifying Forest Cover Loss in Democratic Republic of the Congo, 2000–2010,
with Landsat ETM+ Data.” Remote Sensing of Environment 122: 106–116.
Potapov, P. V., A. Yaroshenko, S. Turubanova, M. Dubinin, L. Laestadius, C. Thies, and D. Aksenov.
2008. “Mapping the World’s Intact Forest Landscapes by Remote Sensing.” Ecology and Society
13. Accessed March 15, 2013. http://www.ecologyandsociety.org/vol13/iss2/art51/
Prasad, A. M., L. R. Iverson, and A. Liaw. 2006. “Newer Classification and Regression Tree
Techniques: Bagging and Random Forests for Ecological Prediction.” Ecosystems 9: 181–199.
Rabus, B., M. Eineder, A. Roth, and R. Bamler. 2003. “The Shuttle Radar Topography Mission
a New Class of Digital Elevation Models Acquired by Spaceborne Radar.” ISPRS Journal of
Photogrammetry and Remote Sensing 57: 241–262.
Rautner, M., and M. Hardiono. 2005. Treasure Island at Risk. Status of Forest, Wildlife and Related
Threats on the Island of Borneo. Technical Report. Frankfurt am Main: WWF, Germnay.
Ross, M. L. 2001. Timber Booms and Institutional Breakdown in Southeast Asia. Cambridge:
University Press.
Sheil, D., A. Casson, E. Meijaard, M. van Nordwijk, J. Gaskell, J. Sunderland-Groves, K. Wertz, and
M. Kanninen. 2009. The Impacts and Opportunities of Oil Palm in Southeast Asia: What Do We
Know and What Do We Need to Know? Occasional Paper No. 51. Bogor: Centre for International
Forestry Research (CIFOR).
Sodhi, N. S., L. P. Koh, B. W. Brook, and P. K. L. Ng. 2004. “Southeast Asian Biodiversity: An
Impending Disaster.” Trends in Ecology & Evolution 19: 654–660.
Soto-Berelov, M., and K. D. Madsen. 2011. “Continuity and Distinction in Land Cover Across a
Rural Stretch of the U.S.-Mexico Border.” Human Ecology 39: 509–526.
Southworth, J., M. Marsik, Y. Qiu, S. Perz, G. Cumming, F. Stevens, K. Rocha, A. Duchelle, and
G. Bar nes. 2011. “Roads as Drivers of Change: Trajectories Across the Tri-National Frontier in
MAP, the Southwestern Amazon.” Remote Sensing 3: 1047–1066.
Southworth, J., D. Munroe, and H. Nagendra. 2004. “Land Cover Change and Landscape
Fragmentation comparing the Utility of Continuous and Discrete Analyses for a Western
Honduras Region.” Agriculture, Ecosystems & Environment 101: 185–205.
Sunderlin, W. D., S. Dewi, A. Puntodewo, D. Muller, A. Angelsen, and M. Epprecht,2008. “Why
Forests Are Important for Global Poverty Alleviation: A Spatial Explanation.” Ecology and
Society 13. Accessed March 15, 2013. http://www.ecologyandsociety.org/vol13/iss2/art24/
Uryu, Y., C. Mott, N. Foead, K. Yulianto, A. Budiman, F. Setiabudi, and F. Takakai. 2008.
Deforestation, Forest Degradation, Biodiversity Loss and CO2 Emission in Riau, Indonesia.
Jakarta: WWF Indonesia Technical Report.
USGS (US Geological Survey). 2006. Shuttle Radar Topography Mission, 3 Arc Second, Filled
Finished-A, 2.0. Global Land Cover Facility, University of Maryland, College Park, Maryland,
February 2000.
Van der Werf, G. R., J. T. Randerson, J. G. Collatz, L. Giglio, P. S. Kasibhatla, A. F. Arellano, S. C.
Olsen, and E. S. Kasischke. 2004. “Continental-scale Partitioning of Fire Emissions During the
1997 to 2001 El Nino/La Nina Period.” Science 303: 73–76.
Vermote, E. F., N. Z. El Saleous, and C. O. Justice. 2002. “Atmospheric Correction of MODIS Data
in the Visible to Middle Infrared: First Results.” Remote Sensing of Environment 83: 97–111.
Downloaded by [UNSW Library] at 01:02 12 May 2016
... Despite being known for extremely high species richness, Bornean tropical forests are rapidly disappearing (Broich et al., 2013;Gaveau et al., 2014). Forests in montane regions that are remote by distance or steep terrain, are among the last remaining habitats for the unique flora and fauna on the island (Scriven et al., 2015). ...
Article
Anthropogenic pressure in tropical montane forests is rapidly increasing, becoming a major threat to these complex ecosystems. Studies have shown that the wide variety of human activities in tropical uplands results in different ecological responses of secondary forests, but basic information on the disturbance impacts and underlying recovery processes is lacking. Here, we compared structural characteristics and tree community composition of old growth forest and secondary forests in a montane region of Sabah, Malaysia, which experienced five different anthropogenic disturbances. We also investigated the use of metrics from spectral trajectories of a Landsat time series (LTS) change detection algorithm (LandTrendr) to identify characteristics of disturbance events and their linkage to the recovery of tree community composition, with field validation. Five LTS metrics—time since the greatest disturbance (TSD), magnitude of disturbance (MD), distance to undisturbed forests (d_UND), recovery indicator (RI), and years to recovery (Y2R) were derived and were related to field-based tree community composition. Our analysis revealed a gradient of recovery patterns in community composition and structural attributes among forest disturbance types, suggesting the importance of community composition as an indicator of forest recovery. Among derived LTS metrics, TSD, MD, d_UND, and Y2R 100% were significantly related with the similarity in community composition. Our results suggest that spectral trajectories from LTS can serve as a useful predictor of community composition change in recovering stands. This approach provides an efficient means for developing systematic conservation strategies for high-elevation regions in the tropics, where human-modified landscapes are expanding.
... Even the mangrove forests are increasingly threatened by deforestation activities [23]. The anthropogenic disturbances have extended to the montane rainforests near the international borders between Malaysia and Indonesia [24]. These activities have adversely affected biodiversity [11] and resulted in aboveground biomass or carbon losses in montane rainforests [25]. ...
Article
Full-text available
Tropical forests play an important role in carbon storage, accumulating large amounts of carbon in their aboveground and belowground components. However, anthropogenic land-use activities have increasingly threatened tropical forests, resulting in accelerated global greenhouse gas emissions. This research aimed to estimate the carbon stocks in soil, organic layer, and litterfall in tropical montane forests under three different land uses (intact forest, logged-over forest, and plantation forest) at Long Mio, Sabah, Malaysia. Field data were collected in a total of 25 plots from which soil was randomly sampled at three depths. Litterfalls were collected monthly from November 2018 to October 2019. The results showed that the soil in the study area is Gleyic Acrisol, having pH values ranging between 4.21 and 5.71, and high soil organic matter contents. The results also showed that the total soil carbon stock, organic layer, and litterfall is higher in the intact forest (101.62 Mg C ha−1), followed by the logged-over forest (95.61 Mg C ha−1) and the plantation forest (93.30 Mg C ha−1). This study highlights the importance of conserving intact forests as a strategy to sequester carbon and climate change mitigation.
... The characteristics cited above result in high production of biomass per hectare (Matsumara, 2011) and high input of nutrients via litter where the trees are grown (Hedge et al., 2013). The species' fast growth occurs when submitted to unfavorable conditions, such as prolonged dry seasons and low-fertility soils, with pH around 4.0 (Souza et al.,2010;Broich et al., 2013). A. mangium is recommended to planting in agroforestry systems and also for honey production (Oliveira, 2017). ...
Article
Full-text available
This study aimed to assess the infl uence of the seasonality on the bark tannins content of Acacia mangium trees grown in the Northeastern Brazilian Region and the eff ect of soil preparation on the results. Two experimental plots of 1.0 ha each were submitted to diff erent soil preparation methods, with two diff erent intensities. The experimental design consisted of four treatments, two types of soil preparation and, two diff erent bark collection seasons (end of the rainy and dry seasons). The bark of the trees was collected in each treatment and the contents of condensed tannins were determined. For each experimental treatment, 15 trees were harvested and debarked. Bark material was submitted to extraction with hot water, obtaining the total solids content (TSC), Stiasny index (I), and the condensed tannins content (CTC). There was no infl uence of the soil preparation method on the TSC, I, and CTC. However, there was a signifi cant diff erence in these parameters for tree bark collected in the rainy season, with higher values. The less intensive soil preparation method is recommended due to its lower cost, and bark should be collected at the end of the rainy season for the best yield of condensed tannins.
... (Sarawak) in which rates of loss are 10-times greater than across the border [5]. Much of the forest loss and degradation could affect cover, structure, and carbon stocks or biomass of the remaining forests. ...
Article
Full-text available
Monitoring anthropogenic disturbances on aboveground biomass (AGB) of tropical montane forests is crucial, but challenging, due to a lack of historical AGB information. We examined the use of spaceborne (Shuttle Radar Topographic Mission Digital Elevation Model (SRTM) digital surface model (DSM)) and airborne (Light Detection and Ranging (LiDAR)) digital elevation data to estimate tropical montane forest AGB changes in northern Borneo between 2000 and 2012. LiDAR canopy height model (CHM) mean values were used to calibrate SRTM CHM in different pixel resolutions (1, 5, 10, and 30 m). Regression analyses between field AGB of 2012 and LiDAR CHM means at different resolutions identified the LiDAR CHM mean at 1 m resolution as the best model (modeling efficiency = 0.798; relative root mean square error = 25.81%). Using the multitemporal AGB maps, the overall mean AGB decrease was estimated at 390.50 Mg/ha, but AGB removal up to 673.30 Mg/ha was estimated in the managed forests due to timber extraction. Over the 12 years, the AGB accumulated at a rate of 10.44 Mg/ha/yr, which was attributed to natural regeneration. The annual rate in the village area was 8.31 Mg/ha/yr, which was almost 20% lower than in the managed forests (10.21 Mg/ha/yr). This study identified forestry land use, especially commercial logging, as the main driver for the AGB changes in the montane forest. As SRTM DSM data are freely available, this approach can be used to estimate baseline historical AGB information for monitoring forest AGB changes in other tropical regions.
... However, we posit that for other tree cover loss-dominated areas, carbon loss should be present. For tropical America, Africa, and Asia, Baccini et al. estimate no carbon loss for 68%, 76% and 72% of Landsat-derived tree cover loss-dominant cells; we identify these as errors of omission (Figs. 1 to 3) (12,13). ...
Article
Baccini et al . (Reports, 13 October 2017, p. 230) report MODIS-derived pantropical forest carbon change, with spatial patterns of carbon loss that do not correspond to higher-resolution Landsat-derived tree cover loss. The assumption that map results are unbiased and free of commission and omission errors is not supported. The application of passive moderate-resolution optical data to monitor forest carbon change overstates our current capabilities.
... In recent years, Earth-observation (EO) has provided a more accurate and better picture of the global rate and geographical distribution of deforestation (Skole and Tucker, 1993;Defries et al., 2002;Miettinen et al., 2011), highlighting Southeast Asia, and in particular Indonesia, as of major concern . Within Indonesia, the loss of moist tropical forests on the islands of Borneo and Sumatra, primarily due to the expansion of industrial palm oil plantations, has been well documented (Broich et al., 2011(Broich et al., , 2013Margono et al., 2012;Shevade et al., 2017), but far less attention has been directed towards Java. Indeed, the forests of Java have not received bespoke study and are frequently omitted from published statistics, in part due to the relative sparsity of forest cover remaining since Dutch colonial rule in the eighteenth and nineteenth centuries (Smiet, 1990). ...
... In Borneo, most lowland primary forest has been lost as a result of deforestation and forest degradation over the past 40 years (Langner et al. 2012, Gaveau et al. 2014. The remaining uplands rainforests are severely threatened by increasing anthropogenic activities, particularly in the uplands of the Malaysian Bor-neon near Indonesia, where rates of loss are ten-times greater than across the border (Broich et al. 2013). ...
... In Borneo, most lowland primary forest has been lost as a result of deforestation and forest degradation over the past 40 years (Langner et al. 2012, Gaveau et al. 2014. The remaining uplands rainforests are severely threatened by increasing anthropogenic activities, particularly in the uplands of the Malaysian Bor-neon near Indonesia, where rates of loss are ten-times greater than across the border (Broich et al. 2013). ...
Article
Full-text available
Accurately quantifying the above-ground carbon stock of tropical rainforest trees is the core component of “Reduction of Emissions from Deforestation and Forest Degradation-plus” (REDD+) projects and is important for evaluating the effects of anthropogenic global change. We used high-resolution optical imagery (IKONOS-2) to identify individual tree crowns in intact and degraded rainforests in the mountains of Northern Borneo, comparing our results with 50 ground-based plots dispersed in intact and degraded forests, within which all stems > 10 cm in diameter were measured and identified to species or genus. We used the dimensions of tree crowns detected in the imagery to estimate above-ground biomasses (AGBs) of individual trees and plots. To this purpose, preprocessed IKONOS imagery was segmented using a watershed algorithm; stem diameter values were then estimated from the cross-sectional crown areas of these trees using regression relationships obtained from ground-based measurements. Finally, we calculated the biomass of each tree (AGBT, in kg), and the AGB of plots by summation (AGBP, in Mg ha⁻¹). Remotely sensed estimates of mean AGBT were similar to ground-based estimates in intact and degraded forests, even though small trees could not be detected from space-borne sensors. The intact and degraded forests not only had different AGB but were also dissimilar in biodiversity. A tree-centric approach to carbon mapping based on high-resolution optical imagery, could be a cheap alternative to airborne laser-scanning.
... As for Sumatra and Kalimantan, forest loss rates assessed by Landsat imagery were found to be 2.86 Mha (2.86% of the land area) from 2000 to 2005, with the highest concentration of deforestation in Riau and Kalimantan Tengah provinces (Broich et al., 2011a,b). Lowland forest loss in Borneo Island was intense on both sides of the Indonesia-Malaysia border; however, upland forests in Indonesia were almost undisturbed, compared to logged upland forests in Malaysia (Broich et al., 2013). Sumatra and Kalimantan uplands in particular experienced increases in annual clearing rates from the 1990s to the 2000s, possibly in response to a reduced lowland forest resource base (Hansen et al., 2009). ...
Article
Estimating aboveground biomass changes in tropical mountains is difficult due to the small-scale anthropogenic land use activities such as selective logging. This study examined how multi-temporal airborne LiDAR data could estimate AGB changes in Borneo's tropical montane forest. Using airborne LiDAR data acquired in 2012 and 2017, we compared direct and indirect approaches to estimating the AGB changes. The direct method predicts the AGB change directly based on differences in LiDAR variables between the two time points whereas, the indirect method first constructs a model for predicting the AGB for each time point and then estimates the changes. The direct approach produced a model with an adjusted R² of 0.321 and a relatively high RMSE (6.37 Mg/ha/year; relative RMSE: 134.36%). On the other hand, annual AGB changes derived from the indirect approach had a low RMSE value (1.413 Mg/ha/year; relative RMSE: 29.80%) and were strongly correlated with the field AGB changes (R² = 0.988). We estimated the AGB changes using the indirect approach to be −7.49 Mg/ha/year for AGB loss and 8.91 Mg/ha/year for AGB gain. We identified land use conversion as the primary driver of AGB changes in the montane forest since the rate of AGB decrease in state-land was higher than in the managed forest. The LiDAR-based approach provides high-resolution estimates of AGB changes by enlarging field plots to more extensive area coverage, facilitating the adoption of incentive-based carbon conservation mechanisms.
Article
Full-text available
The first results of the Moderate Resolution Imaging Spectroradiometer (MODIS) vegetation continuous field algorithm's global percent tree cover are presented. Percent tree cover per 500-m MODIS pixel is estimated using a supervised regression tree algorithm. Data derived from the MODIS visible bands contribute the most to discriminating tree cover. The results show that MODIS data yield greater spatial detail in the characterization of tree cover compared to past efforts using AVHRR data. This finer-scale depiction should allow for using successive tree cover maps in change detection studies at the global scale. Initial validation efforts show a reasonable relationship between the MODIS-estimated tree cover and tree cover from validation sites.
Article
Common understanding of the causes of land-use and land-cover change is dominated by simplifications which, in turn, underlie many environment-development policies. This article tracks some of the major myths on driving forces of land-cover change and proposes alternative pathways of change that are better supported by case study evidence. Cases reviewed support the conclusion that neither population nor poverty alone constitute the sole and major underlying causes of land-cover change worldwide. Rather, peoples’ responses to economic opportunities, as mediated by institutional factors, drive land-cover changes. Opportunities and
Article
Bagging predictors is a method for generating multiple versions of a predictor and using these to get an aggregated predictor. The aggregation averages over the versions when predicting a numerical outcome and does a plurality vote when predicting a class. The multiple versions are formed by making bootstrap replicates of the learning set and using these as new learning sets. Tests on real and simulated data sets using classification and regression trees and subset selection in linear regression show that bagging can give substantial gains in accuracy. The vital element is the instability of the prediction method. If perturbing the learning set can cause significant changes in the predictor constructed, then bagging can improve accuracy.
Article
Southeast Asia has the highest relative rate of deforestation of any major tropical region, and could lose three quarters of its original forests by 2100 and up to 42% of its biodiversity. Here, we report on the current state of its biota and highlight the primary drivers of the threat of extinction now faced by much of the unique and rich fauna and flora of the region. Furthermore, the known impacts on the biodiversity of Southeast Asia are likely to be just the tip of the iceberg, owing to the paucity of research data. The looming Southeast Asian biodiversity disaster demands immediate and definitive actions, yet such measures continue to be constrained by socioeconomic factors, including poverty and lack of infrastructure. Any realistic solution will need to involve a multidisciplinary strategy, including political, socioeconomic and scientific input, in which all major stakeholders (government, non-government, national and international organizations) must participate.
Article
Forest cover and forest cover loss for the last decade, 2000–2010, have been quantified for the Democratic Republic of the Congo (DRC) using Landsat time-series data set. This was made possible via an exhaustive mining of the Landsat Enhanced Thematic Mapper Plus (ETM +) archive. A total of 8881 images were processed to create multi-temporal image metrics resulting in 99.6% of the DRC land area covered by cloud-free Landsat observations. To facilitate image compositing, a top-of-atmosphere (TOA) reflectance calibration and image normalization using Moderate Resolution Imaging Spectroradiometer (MODIS) top of canopy (TOC) reflectance data sets were performed. Mapping and change detection was implemented using a classification tree algorithm. The national year 2000 forest cover was estimated to be 159,529.2 thou-sand hectares, with gross forest cover loss for the last decade totaling 2.3% of forest area. Forest cover loss area increased by 13.8% between the 2000–2005 and 2005–2010 intervals, with the greatest increase occur-ring within primary humid tropical forests. Forest loss intensity was distributed unevenly and associated with areas of high population density and mining activity. While forest cover loss is comparatively low in protected areas and priority conservation landscapes compared to forests outside of such areas, gross forest cover loss for all nature protection areas increased by 64% over the 2000 to 2005 and 2005 to 2010 intervals.