International Journal of Applied Earth Observation and Geoinformation
Integration of carbon conservation into sustainable forest management using high resolution satellite imagery: A case study in Sabah, Malaysian Borneo
Highlights
▸ Landsat-based development of biomass model for tropical forest over hilly terrain. ▸ Comparison of 2 forest concessions in Sabah with different management types. ▸ One concession was sustainably managed and the other conventionally logged. ▸ Influence of forest management on above-ground biomass was analyzed. ▸ Replacing conventional with reduced-impact logging preserves 10.5 t C/ha per year.
Introduction
Pressure on tropical forests has increased tremendously in the past two decades (Fuller, 2006). Insular Southeast Asia experienced a deforestation rate of 1.0% per year between 2000 and 2010 (Miettinen et al., 2011), and a detailed analysis of Borneo revealed an annual rate of 1.7% between 2002 and 2005 (Langner et al., 2007). According to Kitayama (2008) most of the forests on Borneo are permanent forest estates, so called “production forests”, which have been commercially logged more than twice, mainly using a conventional logging (CL) technique.
Even though CL results in the removal of only a certain number of economically valuable trees per hectare, the process can destroy more than 50% of all trees (Sist et al., 2003). The number of trees extracted under CL is only limited by girth and harvesting cycle (Sist et al., 1999). Additionally, a lot of collateral damage is caused to the surrounding vegetation by the falling of large trees and the use of heavy machinery (Cannon et al., 1998).
The prevailing application of CL in Borneo has led to vast areas of degraded forests. This has severe impact on biodiversity as the lowland forests of Borneo are considered to have the highest plant species richness worldwide (Kier et al., 2005). In addition, forest degradation causes considerable carbon emissions into the atmosphere (Achard et al., 2004, Asner et al., 2005, Asner et al., 2010), which accounts together with deforestation for at least 12% of global anthropogenic CO2 emissions, thus being the second largest anthropogenic emission source (van der Werf et al., 2009).
In contrast to CL, reduced-impact logging (RIL) mitigates the physical impacts on the ground, to the remaining standing trees, and ecosystem as a whole by using a combination of pre-harvest census, controlled felling, lowered allowable cut, and regulated machinery use (Sist et al., 2003). In combination with longer cutting cycles as applied under next-generation sustainable forest management (SFM) RIL also helps to preserve carbon (Pinard and Putz, 1996, Putz et al., 2008).
Regardless of the obvious ecological advantages, SFM is currently practiced in less than 5% of all natural permanent forest estates in the tropics (ITTO, 2006). A study by Putz et al. (2000) discusses the most common motives of commercial loggers for not applying RIL, with reasons ranging from elevated costs to lack of governmental incentives. However, the application of SFM can be financially rewarded through forest certification, thus leading to an improved market access and increased unit log-price (Lagan et al., 2007).
Reducing Emissions from Deforestation and Forest Degradation (REDD) is an approach to mitigate CO2 emissions due to land cover conversions in the tropics (UNFCCC, 2007). When including the concept of additionality, which creates incentives for additional actions in forest conservation, SFM practices are promoted and will in effect lead to higher permanent carbon stocks (Kitayama, 2008). As the implications of these incentives on the enhancement of carbon stocks have to be monitored, it is necessary to obtain information about the degradation status as well as the above-ground biomass (AGB) of a forest.
Remote sensing offers a practical way to assess information on the condition of forests in large and inaccessible areas (Saatchi et al., 2007). Unfortunately, measuring forest degradation and deriving AGB estimates is technically more demanding than monitoring deforestation processes (DeFries et al., 2007). Traditionally, biomass estimation has been based on either radar or optical remote sensing technology.
The advantage of radar in contrast to passive optical data is the ability to acquire data irrespective of haze and the persistently cloudy weather conditions in the humid tropics (Asner, 2001). However, the signal of all available radar sensors tends to saturate at a lower value than the actual AGB volumes of tropical rain forests, which are among the most carbon-dense and structurally complex ecosystems in the world, and additionally there are also increased errors in mountainous terrain (Gibbs et al., 2007). Therefore, estimation of AGB in tropical forests is very challenging. Morel et al. (2011) describes the limitations of L-band data to estimate forest degradation in several forest reserves in Sabah. Engelhart et al. (2011) demonstrated that a multi-temporal L- and X-band model can be successfully used to estimate even high AGB values in the peatlands of Central Kalimantan. However, this study also pointed out the crucial role of the extremely flat topography of that area, thus avoiding problems due to shadow and layover effects.
To overcome the problem of sensor saturation, light detection and ranging (LiDAR) sensors have to be considered, which have the potential to obtain AGB measurements even in the tropics. Although large-scale applications are not feasible due to the narrow swath of view and the high costs of data acquisition, the use of LiDAR instead of field inventory for calibration purposes of satellite data is nevertheless a promising approach (Asner et al., 2010). Satellite-based LiDAR data has also been successfully used to derive a pan-tropical carbon map (Saatchi et al., 2011).
In contrast to radar data, passive optical remote sensing technologies cannot penetrate a closed forest canopy and therefore need to rely on secondary vegetation features for biomass estimation, making optical remote sensing-based methods more restricted to empirical approaches. However, artifacts due to a heterogeneous topography are less problematic as with radar data. A lot of research in the humid tropics has been undertaken using Landsat data (Foody et al., 2001, Foody et al., 2003, Lu, 2005, Steininger, 2000, Tangki and Chappell, 2008). With a spatial resolution of 30 m, each Landsat Thematic Mapper (TM) or Enhanced Thematic Mapper (ETM+) pixel value integrates reflectance from several tree crowns. The density of these trees and their physical characteristics due to species composition and age contribute to the spectral properties of each pixel. The more mature a tropical rain forest, the more the canopy becomes uneven (Weishampel et al., 2001), resulting in higher levels of self-shadowing by emergent trees, which is an important parameter to distinguish younger from mature forests (Adams et al., 1995). Subsequently, the AGB of a forest can be estimated based on information on the stage of succession and the level of degradation (Houghton, 2005).
The objective of our research was to analyze the impact of different forest management practices on the AGB using optical remote sensing techniques supported with field data. The effects of CL and SFM with RIL on the AGB were analyzed in two production forests in Borneo dominated by mixed dipterocarp rain forest. In contrast to other studies, our biomass model was applied at three different times, totally encompassing a period of 16 years, thereby enabling evaluation of long term effects of different forest management practices. Annual AGB change rates were derived and analyzed to determine to what extent next-generation SFM can conserve carbon in comparison to the business as usual situation of CL.
Section snippets
Study site
The study area consists of two adjacent production forests, Deramakot and Tangkulap, situated in Sabah, Malaysian Borneo. The climate is characterized by frequent rainfall and high temperatures throughout the year and both forest reserves are dominated by lowland mixed dipterocarp forest with varying degrees of degradation. While Tangkulap is less hilly, most of Deramakot is characterized by a heterogeneous topography.
The forest management units (FMU) of Deramakot and Tangkulap were licensed
Forest degradation-based AGB model
Based on the result of the CCFScorr index of the year 2000, analyzing an area of 81 778 ha of lowland dipterocarp forest, a regression analysis with the AGB values of the PSP was done, in which two second-order polynomial AGB models were fitted. The coefficients of both models are shown in Table 1.
Even though model 1 is more complex, it is not significantly better than model 2 (ANOVA, p < 0.05). And as the AIC is also slightly higher in model 1, model 2 was further used in this study (Fig. 1).
Validation of AGB model
The
Discussion
Climate-change mitigation is one of several important ecosystem services provided by tropical rain forests. The international framework of REDD+ is expected to include the concept of additionality, rewarding incentives for supplemental activities toward an enhancement of forest carbon stock, such as SFM. In order to implement REDD+ in a post-2012 climate regime and to survey the impact of such additional activities, a transparent monitoring and verification system is crucially needed (
Acknowledgements
We would like to thank all staff members of the Forest Research Centre, Sabah Forestry Department, Sabah, Malaysia, as well as of the Deramakot Forest Office. Special thanks go to P. Lagan as well as to A. Rawinder for their crucial support in the field. This work was partly supported by the Global Environment Research Funds (F-071 and D-1006) of the Ministry of the Environment, Japan, to Kanehiro Kitayama and by the JSPS Postdoctoral fellowship program for foreign researchers (P 09097) to
References (50)
- et al.
Classification of multispectral images based on fractions of endmembers – application to land-cover change in the Brazilian Amazon
Remote Sensing of Environment
(1995) - et al.
Summary of current radiometric calibration coefficients for Landsat MSS, TM, ETM+, and EO-1 ALI sensors
Remote Sensing of Environment
(2009) An improved dark-object subtraction technique for atmospheric scattering correction of multispectral data
Remote Sensing of Environment
(1988)- et al.
Earth observations for estimating greenhouse gas emissions from deforestation in developing countries
Environmental Science & Policy
(2007) - et al.
Predictive relations of tropical forest biomass from Landsat TM data and their transferability between regions
Remote Sensing of Environment
(2003) - et al.
Topographic normalization of Landsat TM images of forest based on subpixel Sun–Canopy–Sensor geometry
Remote Sensing of Environment
(1998) - et al.
Interpretation and topographic compensation of conifer canopy self-shadowing
Remote Sensing of Environment
(2008) - et al.
Estimating aboveground biomass in forest and oil palm plantation in Sabah, Malaysian Borneo using ALOS PALSAR data
Forest Ecology and Management
(2011) - et al.
Reduced-impact logging in Indonesian Borneo: some results confirming the need for new silvicultural prescriptions
Forest Ecology and Management
(2003) - et al.
Biomass variation across selectively logged forest within a 225-km2 region of Borneo and its prediction by Landsat TM
Forest Ecology and Management
(2008)
Improved estimates of net carbon emissions from land cover change in the tropics for the 1990
Global Biogeochemical Cycles
Cloud cover in Landsat observations of the Brazilian Amazon
International Journal of Remote Sensing
Selective logging in the Brazilian Amazon
Science
High-resolution forest carbon stocks and emissions in the Amazon
PNAS
Biomass of tropical forests of south and southeast Asia
Canadian Journal of Forestry Research
Tree species diversity in commercially logged Bornean Rainforest
Science
Above-ground biomass and productivity in a rain forest of eastern South America
Journal of Tropical Ecology
Net primary production in tropical forests: an evaluation and synthesis of existing field data
Ecological Applications
Policy Brief REDD Policy Scenarios and Carbon Markets
Aboveground biomass retrieval in tropical forests – the potential of combined X- and L-band SAR data use
Remote Sensing of Environment
Mapping the biomass of Bornean tropical rain forest from remotely sensed data
Global Ecology & Biogeography
Tropical forest monitoring and remote sensing: a new era of transparency in forest governance?
Singapore Journal of Tropical Geography
Monitoring and estimating tropical forest carbon stocks: making REDD a reality
Environmental Research Letters
Timber yield from second entry logging in the lowland mixed dipterocarp forest of Deramakot, Sabah
The Malaysian Forester
Cited by (37)
Estimating aboveground biomass changes in a human-modified tropical montane forest of Borneo using multi-temporal airborne LiDAR data
2022, Remote Sensing Applications: Society and EnvironmentForest canopy structure and reflectance in humid tropical Borneo: A physically-based interpretation using spectral invariants
2017, Remote Sensing of EnvironmentCitation Excerpt :Optical EO data has been used in South East Asian rainforests primarily to map deforestation (Hansen et al., 2013; Margono et al., 2014) and to discriminate basic forest classes i.e., primary (intact) and disturbed (logged) forests (Wijedasa et al., 2012) or, with visual interpretation, forest and oil palm (Carlson et al., 2012; Gaveau et al., 2016). Beyond classification-based analysis, optical EO data has been used, to a smaller extent, in the estimation of forest above ground biomass (Basuki et al., 2012; Langner et al., 2012; Pfeifer et al., 2016; Propastin, 2013), fractional vegetation cover and leaf area index (Pfeifer et al., 2016; Propastin and Erasmi, 2010). The latter type of analysis estimating forest structure as a continuous variable is particularly important to uncover the spatial heterogeneity of modern forest landscape in insular South East Asia, the result of widespread forest degradation (i.e. the second “D” in REDD +).
The mixing ratio of tree functional groups as a new index for biodiversity monitoring in Bornean production forests
2017, Forest Ecology and ManagementCitation Excerpt :All pixels above 650 m asl were excluded to constrain our analysis to the lowland mixed dipterocarp forests. After masking all non-forest pixels, reflectance values were categorized into 5 strata using a level slice method (Fujiki et al., 2016; Imai et al., 2014; Langner et al., 2012). Ten 20 m-radius circular plots (1257 m2 in area) were established in each stratum in each FMU (i.e., a total of 50 plots in each FMU).
Remote sensing of forest degradation in Southeast Asia-Aiming for a regional view through 5-30 m satellite data
2014, Global Ecology and Conservation