You are on page 1of 24

Author’s Accepted Manuscript

Assessment of ALOS-2 PALSAR-2 L-band and


Sentinel-1 C-band SAR backscatter for
discriminating between large-scale oil palm
plantations and smallholdings on tropical peatlands

Aslinda Oon, Khanh Duc Ngo, Rozilah Azhar,


Adham Ashton-Butt, Alex Mark Lechner, Badrul
Azhar www.elsevier.com/locate/rsase

PII: S2352-9385(17)30196-9
DOI: https://doi.org/10.1016/j.rsase.2018.11.002
Reference: RSASE188
To appear in: Remote Sensing Applications: Society and Environment
Received date: 18 September 2017
Revised date: 6 July 2018
Accepted date: 4 November 2018
Cite this article as: Aslinda Oon, Khanh Duc Ngo, Rozilah Azhar, Adham
Ashton-Butt, Alex Mark Lechner and Badrul Azhar, Assessment of ALOS-2
PALSAR-2 L-band and Sentinel-1 C-band SAR backscatter for discriminating
between large-scale oil palm plantations and smallholdings on tropical peatlands,
Remote Sensing Applications: Society and Environment,
https://doi.org/10.1016/j.rsase.2018.11.002
This is a PDF file of an unedited manuscript that has been accepted for
publication. As a service to our customers we are providing this early version of
the manuscript. The manuscript will undergo copyediting, typesetting, and
review of the resulting galley proof before it is published in its final citable form.
Please note that during the production process errors may be discovered which
could affect the content, and all legal disclaimers that apply to the journal pertain.
Assessment of ALOS-2 PALSAR-2 L-band and Sentinel-1 C-band SAR
backscatter for discriminating between large-scale oil palm plantations and
smallholdings on tropical peatlands
Aslinda Oon1,2, Khanh Duc Ngo3, Rozilah Azhar2, Adham Ashton-Butt4, Alex Mark Lechner3,
Badrul Azhar5,6
1
Institute of Climate Change, National University of Malaysia, 43000 Bangi, Selangor
Malaysia
2
Department of Civil Engineering, Faculty of Engineering, Universiti Putra Malaysia, 43400
Serdang, Selangor, Malaysia.
3
School of Environmental and Geographical Sciences, University of Nottingham Malaysia
Campus, 43500 Semenyih, Selangor, Malaysia
4
Institute for Life Sciences, University of Southampton, Southampton SO17 1BJ, United
Kingdom
5
Department of Forest Management, Faculty of Forestry, Universiti Putra Malaysia, 43400
Serdang, Selangor, Malaysia.
6
Biodiversity Unit, Institute of Bioscience, Universiti Putra Malaysia, 43400 Serdang,
Selangor, Malaysia.
*Corresponding author. b_azhar@upm.edu.my

Abstract
Oil palm agriculture is rapidly expanding across the tropics, particularly on peatlands to
meet increasing global demand for palm oil based products. Oil palm production systems
can be divided into two broad categories of management system: large-scale monoculture
plantations and smallholdings. Both categories are separated by large differences in
environmental and social impacts. These oil palm production systems are often
characterized by different agricultural practices and vegetation characteristics and therefore
land cover. To date, there are no examples of radar remote sensing studies in oil palm
production landscapes assessing differences between large-scale plantations and
smallholdings. Here, we investigate whether these management systems have distinct radar
signatures that can be identified through backscattering intensity using ALOS (Synthetic
1
Aperture Radar) – PALSAR (Phased Array L-band Synthetic Aperture Radar) L-band and
Sentinel-1 C-band Synthetic Aperture Radar (SAR). SAR has been shown to be superior to
other remote sensing sensors in the tropics for monitoring oil palm expansion due to its all-
weather capabilities. In this study we measured backscattered intensity of 196 plots planted
with oil palm that were established on peatland in Peninsular Malaysia. Our results
indicated that backscattered intensity was significantly influenced by the management
systems. We found that canopy and soil moisture was greater in smallholdings compared to
large-scale plantations. With the exception of HV polarization method, season had
significant effect on backscattered intensity. Irrespective of management systems, canopy
and soil moisture was greater in wet months compared to dry months. Our findings suggest
that ALOS-2-PALSAR-2 L-band and Sentinel-1 C-band have great potential to discriminate oil
palm production landscapes managed under different management systems. Further
investigation is needed to determine whether the current findings are consistent for oil
palm in mineral soils.

Keywords: Management systems; canopy moisture; polarization; radar; soil moisture;


vegetation structure

Introduction
Oil palm agriculture is rapidly expanding throughout the tropics to meet rising global
demand for palm oil based products such as cooking oil, cosmetics and biofuel (Bicalho et
al., 2016; Permpool et al., 2016). The expansion of oil palm production landscapes in
producing countries have modified the original landscapes from native forests, such as peat
swamp forest to large-scale monoculture plantations (Wicke et al., 2011; Vijay et al., 2016).
The area of land under oil palm cultivation is forecasted to increase in the coming years
(Pirker et al., 2016; Furomo & Aide, 2017) and will have profound environmental and socio-
economic effects in regions undergoing oil palm expansion (Azhar et al., 2013; Nurulita et
al., 2015; Silva et al., 2016; Tulloch et al., 2016; Adila et al., 2017).
Oil palm production landscapes are not homogenous at the landscape level. Production
landscapes can be classified into two broad management categories: large-scale plantations

2
and smallholdings (Azhar et al., 2015, 2017) (Fig. 1). Both management systems are common
in oil palm producing countries including those in Southeast Asia, Western Africa, and South
America (Bissonnette & De Koninck, 2017; Castellanos‐Navarrete & Jansen, 2018; Ordway et
al., 2017). Large-scale plantations are commonly characterized by monoculture farming
system, uniform crop stand age, planted area of more than 50 ha and are operated by large
plantation companies (e.g. Sime Darby, FGV, and IOI) using advanced machinery and high
levels of agricultural inputs (e.g. pesticides and fertilizers) (Azhar et al., 2015). In contrast,
smallholdings typically feature monoculture or polyculture (main crop is planted alongside
with subsistence crops) management systems, uniform or variable crop stand age,
plantations of less than 50 ha and management by individual farmers with often limited use
of modern farming technology (Azhar et al., 2015; RSPO 2017). In addition, ground cover is
more heterogenous in smallholdings than that large-scale plantations (Gutiérrez-Vélez &
DeFries, 2013). More than 3 million smallholders earn a living from palm oil globally and
they account for about 40% of total global palm oil production, making smallholders
significant contributors to the oil palm industry (RSPO, 2017) and need to be more widely
considered in creating a more sustainable oil palm industry.
There is an urgent need to monitor the expansion of oil palm agriculture, particularly in
the tropical biodiversity-hotspot regions where oil palm agriculture commonly takes place in
habitats such as tropical peatlands in Malaysia and Indonesia (Wicke et al., 2011).
Stakeholders (e.g. government agencies) require reliable, up-to-date and accurate remote
sensing data to facilitate decision-making with respect to land use and land cover change.
High to moderate spatial resolution radar remote sensing is particularly useful as the tropics
tend to have high cloud cover preventing the use of passive multispectral remote sensing
(i.e. Landsat), while radar supports all-day and all-weather Earth observation regardless of
cloud cover or time of day (Kuplich et al., 2000; Cheng et al., 2016).
Synthetic Aperture Radar (SAR) sensors are valuable tools for studying vegetation using
volumetric scattering (Rahman & Sumantyo, 2010; Santos et al., 2004; Simard et al., 2000).
To date, studies on land use/land cover mapping using radar data have been able to
discriminate between forest stands and oil palm plantations (Morel et al. 2011; Gutiérrez-
Vélez & DeFries, 2013; Dong et al., 2014) and between different crops such as rice paddy,

3
rubber and oil palm (Rosenqvist, 1996; Miettinen & Liew, 2011; Yusoff et al., 2017).
However, little is known about the utility of radar for discriminating between different oil
palm management systems. An important step in developing classification methods for this
purpose is characterising the backscatter radar signatures associated with these land cover
types (Li et al., 2015).
Large-scale plantations and smallholdings may have distinct radar signatures that reflect
crop productivity, aboveground biomass and phenology. Backscatter data holds information
that can be used to measure vegetation structure as it interacts with leaves, branches and
trunks (Horning et al., 2010; Pohl & Loong, 2016). Oil palms with their characteristic
structures such as palm fronds have radar signals which differ compared to other vegetation
such as rubber trees (Rosenqvist, 1996). For example, SAR images at L-band are more
sensitive to changes in the structure of the canopies, particularly the fronds, than other
bands (Teng et al., 2015). Radar mapping of oil palm production landscapes would enable
stakeholders to identify the land use/land cover of regions of interest and improve the
monitoring of crop expansion. Nevertheless, all previous SAR studies in oil palm production
landscapes have not considered the difference between large-scale plantations and
smallholdings (Rosenqvist, 1996; Morel et al., 2011, 2012; Li et al., 2015; Yusoff et al., 2017).
For example, in Southeast Asia most existing wide-scale land cover mapping which includes
oil palm (e.g. Miettinen et al., 2012, 2015; Bryan et al., 2013) have not differentiated
between the two types of oil palm management schemes, even though their social-
environmental footprints are vastly different and differentiating between the two are
critical for managing their impacts.
This research aims to investigate the utility of radar for characterizing oil palm production
landscape on tropical peatlands, operated under different management systems. We used a
case study in Peninsula Malaysia based on ALOS-2 PALSAR-2 L-band from three consecutive
years (2014, 2015, and 2016) and all available Sentinel-1 imagery from five consecutive
years (2014, 2015, 2016, 2017 and 2018). The objectives of this study were: 1) to determine
the backscattered intensity of radar images representing large-scale plantations and
smallholdings; 2) to assess whether this holds for different sensors and; 3) whether there
are differences associated with wet and dry seasons. Our study methods followed Miettinen

4
and Liew (2011) who specifically focused only on differences in backscattered intensities as
precursor to land cover mapping. We conclude by discussing our results in terms of
suitability of SAR for other locations, land cover mapping and potential areas for future
research.

Materials and methods


Study area
The study sites for this research were located at Tanjung Karang, Sekincan and Kuala
Selangor in the state of Selangor, on the west coast of Peninsular Malaysia (Fig. 2). These
areas are characterized by flat and low-lying peatland adjacent to the coast (between
3°25'4.79"N, 101°11'27.15"E and 3°16'41.11"N, 101°27'35.14"E). Originally, the sites were
peat swamp forest before intensive conversion into agricultural lands in the 1950s.

Remote sensing data pre-processing and processing


ALOS-2 PALSAR-2 L-band imagery
Six scenes of ALOS-2 PALSAR-2 L-band (50 m) imagery with fine beam dual (FBD)
polarization (horizontal transmitting and horizontal receiving (HH) and horizontal
transmitting and vertical receiving (HV)) were acquired in the ascending path from 2014 to
2016 for this study. PALSAR-2 imagery was tested in this study as it uses the L-band
frequency which has a long wavelength (15–30 cm) that penetrates the canopy before being
scattered back to the radar receiver and thus will be affected by the vertical structure of
vegetation (Horning et al., 2010). The radar data was obtained in November 2014, January
2015 and September 2016. PALSAR-2 images were pre-processed to obtain backscatter
intensity using Sentinel Application Platform (SNAP) version 5.0 provided by European Space
Agency (ESA) (Fig. 3 & 4). Radiometric calibration was applied. Speckle filtering was
performed using multi-temporal speckle, Lee filter with 5 × 5 kernel size. The HH and HV
normalized radar backscattered coefficients (σ0 in decibels (dB)) were derived by converting
from digital number (DN) values with the following equation (Rosenqvist et al., 2007): σ0
(dB) = 10 × log10DN2 + CF, where CF represents the absolute calibration factor of −83 dB.

5
Backscatter intensity at HH and HV polarization were stored in decibel (db) unit to permit
comparison. All images were geometric correction referred to ancillary ground data.

Sentinel-1 C-band imagery


All available Sentiel-1 imagery was acquired for the study area using the Google Earth
Engine (GEE) platform (https://earthengine.google.com/); pre-processed and mosaicked and
as backscatter coefficient (σ°) in decibels (dB). The Sentinel-1 C-band SAR IW
(Interferometric Wide Swath) Level 1 GRDH (ground-range detected, high resolution) was
utilised in the IW mode operated in the dual polarization mode (VV + VH). Sentinel-1
imagery on GEE are pre-processed using the SENTINEL-1 Toolbox (S1TBX)
(https://sentinel.esa.int/web/sentinel/toolboxes/sentinel-1) which includes thermal noise
removal, radiometric calibration and conversion to backscatter coefficient. The GEE pre-
processed Sentinel-1 imagery was then processed using a Lee filter as described for the
ALOS-2 imagery. Sentinel-1 C-band has a shorter wavelength than Palsar 2 at 5.6 cm,
meaning it is less likely to penetrate the canopy (Lillesand et al., 2008). All available imagery
from October 2014 to May 2018 were processed (Table 1). As the footprint of Sentinel 1
varies over time in some cases imagery for the complete study area were unavailable.

Sample plots
We systematically selected 98 plots of mature oil palm stands (age > 6 years) in large-scale
plantations and smallholdings (Fig. 4). To validate the management system of each plot,
‘ground truthing’ was conducted by inspecting the study areas in March 2017. For each plot,
we extracted the backscattered intensity of HH, HV, VH and VV polarization data using 16 x
16 pixels equivalent to 1 ha or 100 m x 100 m (Nguyen et al. 2016). The two sites were
chosen as they qualitatively and visually from the ground assessment appeared very
different (Fig. 4) in terms of their physical characteristics and provide good representations
of the types of land cover which exist on smallholdings or large-scale plantations. The
qualitative assessment was based on our expert knowledge derived from visiting and
conducting field-based research on numerous sites across Peninsula Malaysia.

6
Statistical analyses
Following similar methods used in a study by Miettinen and Liew (2011) we examined the
relationship between backscattered intensities in agricultural landscapes with oil palm
management systems and seasonality. We first qualitatively assessed differences in
backscatter responses using descriptive statistics and then by comparing the distribution of
values. For the Sentinel-1 imagery in which a time-series of data was available we plotted
mean backscatter values over time for both oil palm systems to assess potential temporal
variability that might be associated with seasonal differences. In addition, for each site we
divided the sites into even and odd plots to allow for a simple visualisation of the variability
in site backscatter characteristics. Next, we performed generalized linear mixed models
(GLMMs) which is an extension of generalized linear models (e.g. logistic regression) that
include both fixed and random effects (hence mixed models) (Schall, 1991; Bolker et al.,
2009) in Genstat 12 (VSN International, Hemel Hempstead, UK). We generated four
separated models based on polarization method. Models used a Normal distribution with an
identity‐link function. The backscattered intensity was the response variable. We fitted oil
palm management system (i.e. large-scale plantation and smallholding) and season (i.e. wet
and dry) as fixed effects. Plot location was included as a random factor (Piepho et al., 2003;
Bolker et al., 2009). The final models were selected by sequentially adding explanatory
variables to the initial model.

Results
Backscattering intensity measured using Palsar HH polarization
Our results revealed that the backscattered intensity was not significantly influenced by oil
palm management system (Wald = 0.05; p = 0.827). However, season played significant role
in determining the backscattered intensity (coefficient for wet month = 1.6929; Wald =
20.75; p < 0.001), particularly during wet months (Table 2; Fig. 6).

Backscattering intensity measured using Palsar HV polarization


We found that oil palm management system had significant effect on the backscattered
intensity (coefficient for smallholdings = 0.8712; Wald = 5.18; p = 0.023). Smallholdings had

7
greater backscattered intensity than large-scale plantations (Table 2; Fig. 5). The
backscattered intensity was not significantly determined by season (Wald = 0.47; p = 0.493).

Backscattering intensity measured using Sentinel 1 VH polarization


Our analysis indicated that the backscattered intensity was significantly determined by oil
palm management system (coefficient for smallholdings = 0.7436; Wald = 562.18; p < 0.001)
and season (coefficient for wet month = 0.2042; Wald = 43.97; p < 0.001). Smallholdings had
higher backscattered intensity compared to large-scale plantations (Table 2; Fig. 5). The
backscattered intensity was greater in wet months than dry months (Table 2; Fig. 6).

Backscattered intensity measured using Sentinel 1 VV polarization


Similar to VH polarization, we found that oil palm management system significantly
influenced the backscattered intensity (coefficient for smallholdings = 1.0029; Wald =
771.12; p < 0.001). Season (coefficient for wet month = 0.5329; Wald = 224.33; p < 0.001)
had significant effect on backscattered intensity. Large-scale plantations had lower
backscattered intensities compared to smallholdings (Table 2; Fig. 5). Consistent with VH
polarization, the backscattered intensity was greater in wet months than dry months (Table
2; Fig. 6).

Discussion
Management system influences backscattered intensity in oil palm production landscapes
Radar remote sensing is capable of discriminating between different crops (Rosenqvist,
1996; Miettinen & Liew, 2011). Using Palsar HV, and Sentinel VH and VV polarizations our
main findings suggest that there is a significant difference in backscattered intensity of
mature oil palm stands between different oil palm management systems with small-scale oil
palms having higher average backscatter intensity values. The differences in backscattered
intensity between large-scale plantations and smallholdings may be because both
smallholdings often consist of multiple tree species besides oil palm that may be able to
maintain canopy and soil moisture. The smallholdings were apparently well-drained,

8
provisioned by systematic flood-controlled waterways whereas the large-scale plantations
had limited waterways.
Longer wavelength L-band (λ = 23.6 cm) can penetrate deeper than the shorter
wavelength C-band (λ = 5.6 cm), although penetration depth is reduced by the moisture of
the target as an inverse function of water content (Lillesand et al., 2008). Backscatter values
from Sentinel with its shorter wave lengths are therefore likely to be more representative of
the characteristics of the canopy as opposed to the C-band. This difference may account for
why using ALOS-2-PALSAR-2 L-band based on HH polarization a significant difference was
not detected.
Both canopy and sensor characteristics influence radar backscatter from vegetation
(Lillesand et al., 2008; Nikolaevan et al., 2015). The canopy parameters that affect the
backscatter are the canopy structure (size, orientation, and distribution of scattering
surfaces within the canopy such as trunks, branches and foliage), the surface roughness and
slope, the dielectric constant or moisture content and the orientation of the canopy
(Nikolaevan et al., 2015). Different plant species are characterized by different vegetation
structure characteristics. Existing studies on radar remote sensing have successfully
differentiated between forested areas and oil palm planted areas (Morel et al., 2011, 2012).
Furthermore, differences can be observed in newly planted versus mature oil palm
plantations at around 4 dB in the L-band JERS-1 data (Rosenqvist, 1996), however, this
difference has not been tested statistically. A related study by Yusoff et al. (2017) found
that oil palm did not demonstrate any useful phenological characteristics for discriminating
between abandoned and non-abandoned lands by using ALOS - 1 and 2 (Advanced Land
Observing Satellite-1 and 2) and PALSAR data. Similarly, a study by Tan et al. (2013) found
that differences in leaf area index of mature oil palm trees stabilizes at around 10 years of
age and discriminating the variation in age between mature oil palm trees is complicated.
It is worth noting that cross polarization Sentinel VH is sensitive to vegetation parameters
(Gao et al., 2013; Hosseini et al., 2015). Although various thresholds of Palsar HV
backscatters have been shown to be sensitive to forest structure in the tropical region
(Thapa et al., 2014), our findings have shown the similar outcomes for discriminating oil
palm management systems. In addition, further research is needed to determine whether

9
the current findings are similar to oil palm planted on mineral soil and not specific to
peatland soils that accumulate carbon, have high water tables and low decomposition rates
(Carlson et al., 2015; Dommain et al., 2015).
Finally, as there are a range of factors which limit the capability of monitoring oil palms
landscapes and their characteristics (i.e. agricultural methods, age, abandonment) with
radar alone, the application of multiple sensor remote sensing including SAR and optical
using data fusion can provide the relevant, timely and accurate information required for
monitoring oil palm expansion (Pohl, 2014). Cheng et al. (2017), for example, did not
recommend the use of PALSAR images alone to distinguish oil palm from other land-
cover/use types. In addition, other studies have demonstrated the utility of multiple
satellite sensors for oil palm mapping (Gutiérrez-Vélez & DeFries, 2013).

Backscattered intensity differed significantly between dry and wet months irrespective of
management systems
Radar backscatter is affected by the amount of moisture in vegetation and soil (Lucas et al.,
2010) as well as precipitation (Morel et al., 2011). Our results were likely to be influenced by
wet or dry season. Backscattered intensities measured during 2014 were likely highest as
the radar data was gathered primarily in the primary rainy season (i.e. October-December).
In contrast, backscattered intensities measured during 2018 were probably lowest as the
radar data was gathered in the dry season (i.e. January-February and May-September).
Regional temperature and precipitation can also influence the accumulation of carbon in
peatlands by controlling both plant productivity and organic matter decomposition (Yu et
al., 2009; Charman et al., 2013). Increasing the moisture content limits the penetration of
the radar signal through a vegetation canopy or into soil (Lucas et al., 2010). Our results
align with similar studies which have shown that backscatter information can be useful for
investigating canopy and soil moisture changes (Nikolaevan et al., 2015).

Challenges for mapping and monitoring oil palm with remote sensing
Our study indicates that both smallholders and large-scale plantations have distinct
backscattered intensity (Fig. 5-6) suggesting that land cover mapping with ALOS-2-PALSAR-2

10
L-band based on HV polarization method and Sentinel-1 SAR based on VH and VV
polarization methods is likely to be feasible. These two land uses have fundamentally
different impacts on the environment, and management and planning need to treat them
separately. For example, many large-scale plantations are signatories to Round table on
Sustainable Palm Oil (RSPO) and companies must comply with a set of environmental and
social criteria for certification (Azhar et al., 2017). Thus, mapping approaches which fail to
make this distinction between the two types of oil palm plantations will provide suboptimal
information for natural resource planning. A key challenge for remote sensing in general is
to address semantic and biophysical differences in land cover classification in order to
reconcile land use (i.e. how humans use the land) with land cover (biophysical properties
the land) (Comber et al., 2005; Lechner et al., 2012).

Conclusions
Our study is the first to examine the canopy and soil moisture based on backscattered
intensity of mature oil palm stands in large-scale plantations and smallholdings established
on peatland. Oil palm expansion under different management systems was characterized by
different backscattered intensity. Hence, the application of radar remote sensing imagery
for monitoring oil palm expansion by agricultural system - either large-scale plantations or
smallholdings - on peatland is feasible. However, further research is required for oil palm
planted in areas with other soil types such as mineral soil.

Acknowledgements
We thank Abdul Rashid Mohamed Shariff for assisting us to access the radar image. We are
also grateful to Biswajeet Pradhan for providing technical supports to conduct the research
as well as constructive comments to improve our manuscript.

References
Adila, N., Sasidhran, S., Kamarudin, N., Puan, C. L., Azhar, B., & Lindenmayer, D. B. (2017).
Effects of peat swamp logging and agricultural expansion on species richness of native
mammals in Peninsular Malaysia. Basic and Applied Ecology, In press.

11
Azhar, B., Lindenmayer, D. B., Wood, J., Fischer, J., Manning, A., Mcelhinny, C., & Zakaria, M.
(2013). The influence of agricultural system, stand structural complexity and landscape
context on foraging birds in oil palm landscapes. Ibis, 155(2), 297-312.
Azhar, B., Saadun, N., Puan, C. L., Kamarudin, N., Aziz, N., Nurhidayu, S., & Fischer, J. (2015).
Promoting landscape heterogeneity to improve the biodiversity benefits of certified
palm oil production: evidence from Peninsular Malaysia. Global Ecology and
Conservation, 3, 553-561.
Azhar, B., Saadun, N., Prideaux, M., & Lindenmayer, D. B. (2017). The global palm oil sector
must change to save biodiversity and improve food security in the tropics. Journal of
environmental management, 203, 457-466.
Bolker, B. M., Brooks, M. E., Clark, C. J., Geange, S. W., Poulsen, J. R., Stevens, M. H. H., &
White, J. S. S. (2009). Generalized linear mixed models: a practical guide for ecology and
evolution. Trends in Ecology & Evolution, 24(3), 127-135.
Bicalho, T., Bessou, C., & Pacca, S. A. (2016). Land use change within EU sustainability
criteria for biofuels: The case of oil palm expansion in the Brazilian Amazon. Renewable
Energy, 89, 588-597.
Bissonnette, J. F., & De Koninck, R. (2017). The return of the plantation? Historical and
contemporary trends in the relation between plantations and smallholdings in Southeast
Asia. The Journal of Peasant Studies, 44(4), 918-938.
Bryan, J. E., Shearman, P. L., Asner, G. P., Knapp, D. E., Aoro, G., & Lokes, B. (2013). Extreme
differences in forest degradation in Borneo: comparing practices in Sarawak, Sabah, and
Brunei. PloS one, 8(7), e69679.
Carlson, K. M., Goodman, L. K., & May-Tobin, C. C. (2015). Modeling relationships between
water table depth and peat soil carbon loss in Southeast Asian
plantations. Environmental Research Letters, 10(7), 074006.
Castellanos‐Navarrete, A., & Jansen, K. (2018). Is oil palm expansion a challenge to
agroecology? Smallholders practising industrial farming in Mexico. Journal of Agrarian
Change, 18(1), 132-155.
Charman, D. J., et al. (2013), Climate-related changes in peatland carbon accumulation
during the last millennium, Biogeosciences, 10, 929–944.

12
Cheng, Y., Yu, L., Cracknell, A. P., & Gong, P. (2016). Oil palm mapping using Landsat and
PALSAR: a case study in Malaysia. International Journal of Remote Sensing, 37(22), 5431-
5442.
Cheng, Y., Yu, L., Zhao, Y., Xu, Y., Hackman, K., Cracknell, A. P., & Gong, P. (2017). Towards a
global oil palm sample database: design and implications. International Journal of
Remote Sensing, 38(14), 4022-4032.
Comber, A., Fisher, P., & Wadsworth, R. (2005). What is land cover?. Environment and
Planning B: Planning and Design, 32(2), 199-209.
Dommain, R., Cobb, A. R., Joosten, H., Glaser, P. H., Chua, A. F., Gandois, L., ... & Harvey, C.
F. (2015). Forest dynamics and tip‐up pools drive pulses of high carbon accumulation
rates in a tropical peat dome in Borneo (Southeast Asia). Journal of Geophysical
Research: Biogeosciences, 120(4), 617-640.
Dong, J., Xiao, X., Sheldon, S., Biradar, C., Zhang, G., Duong, N. D., ... & Moore III, B. (2014). A
50-m forest cover map in Southeast Asia from ALOS/PALSAR and its application on forest
fragmentation assessment. PloS one, 9(1), e85801.
Furumo, P. R., & Aide, T. M. (2017). Characterizing commercial oil palm expansion in Latin
America: land use change and trade. Environmental Research Letters, 12(2), 024008.
Gao, S., Niu, Z., Huang, N., & Hou, X. (2013). Estimating the Leaf Area Index, height and
biomass of maize using HJ-1 and RADARSAT-2. International Journal of Applied Earth
Observation and Geoinformation, 24, 1-8.
Gutiérrez-Vélez, V. H., & DeFries, R. (2013). Annual multi-resolution detection of land cover
conversion to oil palm in the Peruvian Amazon. Remote Sensing of Environment, 129,
154-167.
Horning, N., Robinson, J. A., & Sterling, E. J. (2010). Remote sensing for ecology and
conservation: a handbook of techniques. Oxford University Press.
Hosseini, M., McNairn, H., Merzouki, A., & Pacheco, A. (2015). Estimation of Leaf Area Index
(LAI) in corn and soybeans using multi-polarization C-and L-band radar data. Remote
Sensing of Environment, 170, 77-89.

13
Kuplich, T. M., Salvatori, V., & Curran, P. J. (2000). JERS-1/SAR backscatter and its
relationship with biomass of regenerating forests. International Journal of Remote
Sensing, 21(12), 2513-2518.
Lechner, A.M., Langford, W.T., Bekessy, S.A., Jones, S.D., 2012. Are landscape ecologists
addressing uncertainty in their remote sensing data? Landscape Ecology. 27, 1249–1261.
Li, L., Dong, J., Njeudeng Tenku, S., & Xiao, X. (2015). Mapping oil palm plantations in
Cameroon using PALSAR 50-m orthorectified mosaic images. Remote Sensing, 7(2),
1206-1224.
Lillesand, T.M., Kiefer, R.W., Chapman, J., 2008. Remote Sensing and Image Interpretation,
fourth. ed. John Wiley & Sons Inc., New York.
Lucas, R., Armston, J., Fairfax, R., Fensham, R., Accad, A., Carreiras, J., ... & Metcalfe, D.
(2010). An evaluation of the ALOS PALSAR L-band backscatter—Above ground biomass
relationship Queensland, Australia: Impacts of surface moisture condition and
vegetation structure. IEEE Journal of Selected Topics in Applied Earth Observations and
Remote Sensing, 3(4), 576-593.
Miettinen, J., Shi, C., Tan, W. J., & Liew, S. C. (2012). 2010 land cover map of insular
Southeast Asia in 250-m spatial resolution. Remote Sensing Letters, 3(1), 11-20.
Miettinen, J., & Liew, S. C. (2011). Separability of insular Southeast Asian woody plantation
species in the 50 m resolution ALOS PALSAR mosaic product. Remote Sensing
Letters, 2(4), 299-307.
Miettinen, J., Liew, S. C., & Kwoh, L. K. (2015, October). Usability of Sentinel-1 Dual
Polarization C-Band Data for Plantation Detection in Insular Southeast Asia. In 36th Asian
Conference on Remote Sensing (ACRS2015), Manila, October (pp. 19-23).
Morel, A. C., Saatchi, S. S., Malhi, Y., Berry, N. J., Banin, L., Burslem, D., ... & Ong, R. C.
(2011). Estimating aboveground biomass in forest and oil palm plantation in Sabah,
Malaysian Borneo using ALOS PALSAR data. Forest Ecology and Management, 262(9),
1786-1798.
Morel, A. C., Fisher, J. B., & Malhi, Y. (2012). Evaluating the potential to monitor
aboveground biomass in forest and oil palm in Sabah, Malaysia, for 2000–2008 with

14
Landsat ETM+ and ALOS-PALSAR. International Journal of Remote Sensing, 33(11), 3614-
3639.
Nguyen, L. V., Tateishi, R., Nguyen, H. T., Sharma, R. C., To, T. T., & Le, S. M. (2016).
Estimation of tropical forest structural characteristics using ALOS-2 SAR data. Adv.
Remote Sens, 5, 131-144.
Nikolaeva, E., Sykioti, O., Elias, P., & Kontoes, C. (2015, October). Multi-temporal intensity
and coherence analysis of SAR images for land cover change detection on the Island of
Crete. In SPIE Remote Sensing (pp. 964208-964208). International Society for Optics and
Photonics.
Nurulita, Y., Adetutu, E. M., Kadali, K. K., Zul, D., Mansur, A. A., & Ball, A. S. (2015). The
assessment of the impact of oil palm and rubber plantations on the biotic and abiotic
properties of tropical peat swamp soil in Indonesia. International Journal of Agricultural
Sustainability, 13(2), 150-166.
Ordway, E. M., Naylor, R. L., Nkongho, R. N., & Lambin, E. F. (2017). Oil palm expansion in
Cameroon: Insights into sustainability opportunities and challenges in Africa. Global
Environmental Change, 47, 190-200.
Permpool, N., Bonnet, S., & Gheewala, S. H. (2016). Greenhouse gas emissions from land use
change due to oil palm expansion in Thailand for biodiesel production. Journal of
Cleaner Production, 134, 532-538.
Piepho, H. P., Büchse, A., & Emrich, K. (2003). A hitchhiker's guide to mixed models for
randomized experiments. Journal of Agronomy and Crop Science, 189, 310-322.
Pirker, J., Mosnier, A., Kraxner, F., Havlík, P., & Obersteiner, M. (2016). What are the limits
to oil palm expansion?. Global Environmental Change, 40, 73-81.
Pohl, C. (2014). Mapping palm oil expansion using SAR to study the impact on the CO2 cycle.
In IOP Conference Series: Earth and Environmental Science (Vol. 20, No. 1, p. 012012).
IOP Publishing.
Pohl, C., & Loong, C. K. (2016, April). In-situ data collection for oil palm tree height
determination using synthetic aperture radar. In IOP Conference Series: Earth and
Environmental Science (Vol. 34, No. 1, p. 012027). IOP Publishing.

15
Pohl, C., Kanniah, K. D., & Loong, C. K. (2016, July). Monitoring oil palm plantations in
Malaysia. In Geoscience and Remote Sensing Symposium (IGARSS), 2016 IEEE
International (pp. 2556-2559). IEEE.
RSPO 2017. Accessed from https://www.rspo.org/smallholders/rspo-smallholders-definition
Rosenqvist, Å. (1996). Evaluation of JERS-1, ERS-1 and Almaz SAR backscatter for rubber and
oil palm stands in West Malaysia. Remote Sensing, 17(16), 3219-3231.
Rosenqvist A, Shimada M, Ito N, Watanabe M (2007) ALOS PALSAR: A Pathfinder mission for
global-scale monitoring of the environment. Ieee Transactions on Geoscience and
Remote Sensing 45: 3307–3316.
Schall, R. (1991). Estimation in generalized linear models with random effects. Biometrika,
78(4), 719-727.
Silva, S. J., Heald, C. L., Geddes, J. A., Austin, K. G., Kasibhatla, P. S., & Marlier, M. E. (2016).
Impacts of current and projected oil palm plantation expansion on air quality over
Southeast Asia. Atmospheric Chemistry and Physics, 16(16), 10621-10635.
Tan, K. P., Kanniah, K. D., & Cracknell, A. P. (2013). Use of UK-DMC 2 and ALOS PALSAR for
studying the age of oil palm trees in southern peninsular Malaysia. International Journal
of Remote Sensing, 34(20), 7424-7446.
Teng, K. C., Koay, J. Y., Tey, S. H., Lim, K. S., Ewe, H. T., & Chuah, H. T. (2015). A Dense
Medium Microwave Backscattering Model for the Remote Sensing of Oil Palm. IEEE
Transactions on Geoscience and Remote Sensing, 53(6), 3250-3259.
Tulloch, V. J., Brown, C. J., Possingham, H. P., Jupiter, S. D., Maina, J. M., & Klein, C. (2016).
Improving conservation outcomes for coral reefs affected by future oil palm
development in Papua New Guinea. Biological Conservation, 203, 43-54.
Vijay, V., Pimm, S. L., Jenkins, C. N., & Smith, S. J. (2016). The impacts of oil palm on recent
deforestation and biodiversity loss. PLoS One, 11(7), e0159668.
Wicke, B., Sikkema, R., Dornburg, V., & Faaij, A. (2011). Exploring land use changes and the
role of palm oil production in Indonesia and Malaysia. Land use policy, 28(1), 193-206.
Yu, Z., D. W. Beilman, and M. C. Jones (2009), Sensitivity of northern peatland carbon
dynamics to Holocene climate change, in Carbon Cycling in Northern

16
Peatlands, Geophys. Monogr. Ser., vol. 184, edited by A. Baird et al., pp. 55–69,
AGU, Washington, D. C.
Yusoff, N. M., Muharam, F. M., Takeuchi, W., Darmawan, S., & Abd Razak, M. H. (2017).
Phenology and classification of abandoned agricultural land based on ALOS-1 and 2
PALSAR multi-temporal measurements. International Journal of Digital Earth, 10(2), 155-
174.

Figure 1. Large-scale plantation (left) creates homogenous landscape, characterized by oil


palm monocultures and similar stand age, whereas smallholding (right) are characterized by
heterogenous landscape, planted with one or more commodity crops (e.g. oil palm and
coconut) and varied/similar stand age.

17
Figure 2. Regions of interest (i.e. Kuala Selangor) in the state of Selangor on the west coast
of Peninsular Malaysia.

18
Figure 3. Images pre-processing and processing flow chart.

Figure 4. Example image of PALSAR displaying the location of study area in Peninsular
Malaysia.

19
20
Figure 5. Histograms of (a) Sentinel HH, (b) Sentinel HV, (c) Palsar VH and (d) Palsar VV
backscattered intensity under different oil palm management systems.

21
Figure 6. Histograms of (a) Sentinel HH, (b) Sentinel HV, (c) Palsar VH and (d) Palsar VV
backscattered intensity under different seasons.

Table 1. Sentinel-1 data available for the study area from google earth engine and number
of scenes per year.
Year Smallholding Large-scale plantation
2014 4 3
2015 10 10
2016 13 14
2017 25 26
2018 8 10
22
Table 2. Predicted means of backscattered intensity across different oil palm management
systems and image years.

Explanatory variables Backscattered intensity (dB)


HH HV VH VV
polarization polarization polarization polarization

Management Large-scale -9.445 -17.35 -15.24 -15.23


system plantation
Smallholding -9.368 -16.48 -14.49 -14.22

Season Dry -10.253 -17.05 -14.97 -14.99


Wet -8.560 -16.78 14.76 -14.46

23

You might also like