Aboveground biomass retrieval in tropical forests — The potential of combined X- and L-band SAR data use

https://doi.org/10.1016/j.rse.2011.01.008Get rights and content

Abstract

In the context of reducing emissions from deforestation and forest degradation (REDD) and the international effort to reduce anthropogenic greenhouse gas emissions, a reliable assessment of aboveground forest biomass is a major requirement. Especially in tropical forests which store huge amounts of carbon, a precise quantification of aboveground biomass is of high relevance for REDD activities. This study investigates the potential of X- and L-band SAR data to estimate aboveground biomass (AGB) in intact and degraded tropical forests in Central Kalimantan, Borneo, Indonesia. Based on forest inventory data, aboveground biomass was first estimated using LiDAR data. These results were then used to calibrate SAR backscatter images and to upscale the biomass estimates across large areas and ecosystems. This upscaling approach not only provided aboveground biomass estimates over the whole biomass range from woody regrowth to mature pristine forest but also revealed a spatial variation due to varying growth condition within specific forest types. Single and combined frequencies, as well as mono- and multi-temporal TerraSAR-X and ALOS PALSAR biomass estimation models were analyzed for the development of accurate biomass estimations. Regarding the single frequency analysis overall ALOS PALSAR backscatter is more sensitive to AGB than TerraSAR-X, especially in the higher biomass range (> 100 t/ha). However, ALOS PALSAR results were less accurate in low biomass ranges due to a higher variance. The multi-temporal L- and X-band combined model achieved the best result and was therefore tested for its temporal and spatial transferability. The achieved accuracy for this model using nearly 400 independent validation points was r² = 0.53 with an RMSE of 79 t/ha. The model is valid up to 307 t/ha with an accuracy requirement of 50 t/ha and up to 614 t/ha with an accuracy requirement of 100 t/ha in flat terrain. The results demonstrate that direct biomass measurements based on the synergistic use of L- and X-band SAR can provide large-scale AGB estimations for tropical forests. In the context of REDD monitoring the results can be used for the assessment of the spatial distribution of the biomass, also indicating trends in high biomass ranges and the characterization of the spatial patterns in different forest types.

Research Highlights

► We analyzed X- and L-band SAR data to estimate tropical forest aboveground biomass. ► Field biomass reference data were upscaled using LiDAR measurements. ► The multi-temporal L- and X-band combined model achieved the best result. ►The model is valid up to 307 t/ha. ►Spatial distribution of AGB is indicated over the whole biomass range up to 614 t/ha.

Introduction

Considering global climate change, carbon, as an element of the greenhouse gas carbon dioxide (CO2), plays a major role in trapping thermal radiation from sunlight and reducing the Earth's space release of energy (Read et al., 2001). The rise of atmospheric CO2 from about 280 ppm (in 1880) to 388 ppm (in 2009) was mainly caused by extensive burning of fossil fuels. Deforestation and land cover change caused 20% of global anthropogenic CO2 emissions in the 1990s and 12% in 2008 (Le Quere et al., 2009). The total amount of global emissions from deforestation, forest degradation, and peatland fires is determined to be about 15% of global anthropogenic CO2 emissions from 1997 to 2006 (van der Werf et al., 2009).

Tropical forests cover approximately 15% of the Earth's land surface (FAO, 2009, Page et al., 2009) and contain up to 40% of the terrestrial carbon (Page et al., 2009). The main carbon pools are typically the living aboveground biomass (AGB) and the dead mass of litter, woody debris and soil organic matter (Gibbs et al., 2007). In Southeast Asia's peat swamp forests, both AGB and belowground thick peat deposits make a significant contribution to the carbon reservoir (Page et al., 2009). Indonesia alone, where 48.8% of the land's surface is covered by forest, comprises a total carbon stock of 61 Gt to 63 Gt whereby aboveground forests store 6 Gt (FAO, 2009) and belowground peatlands comprise between 55 Gt (Jaenicke et al., 2008) and 57 Gt (Page et al., 2011).

Southeast Asia features the highest rate of deforestation worldwide at 1.3% per year (Achard et al., 2002, FAO, 2000, Langner and Siegert, 2009) and is therefore a prime target for reducing emissions from deforestation and forest degradation (REDD) initiatives. REDD is a key element emphasized at the United Nations Climate Change Conference in Copenhagen in December 2009 for the reduction of CO2 emissions. The most important requirements for a successful implementation of REDD are accurate biomass and carbon estimations and the subsequent monitoring of the forest carbon pool.

AGB or carbon stocks in tropical forests can be monitored in various ways (FAO, 2000, Gibbs et al., 2007, Goetz et al., 2009). The most accurate way of AGB retrieval is destructive sampling through harvesting, drying and weighting the living biomass, hereby assuming a carbon content of dry biomass from 50% (Goetz et al., 2009, Malhi et al., 2004). While this method is very precise, it is cost- and time-consuming and in tropical forests often impractical. Forest inventories are usually accomplished to collect in-situ biomass data. Allometric equations of ground-based measurements, such as diameter at breast height (DBH) and total tree height are used to extrapolate plot values to forest stands (Chave et al., 2005, FAO, 1997, IPCC, 2006). Forest inventories are also expensive and time-consuming, especially in remote and inaccessible tropical forests, but extrapolation of the results is reasonably accurate. The simplest but least precise method to derive AGB maps is the biome approach whereby forest type specific biomass values are linked to land cover map classes (Goetz et al., 2009). Land cover maps can be generated by remote sensing technology, and biomass values are available in the published literature (for example: Lasco, 2002, Waldes and Page, 2001).

Observations and measurements by satellites have nowadays become the primary source for estimating AGB in tropical forests (Lu, 2006). Since no remote sensing instrument can directly measure either biomass or carbon content, additional in-situ data is required for establishing a relationship between the remote sensing signal and the biomass (Rosenqvist et al., 2003). Various optical remote sensors have bands in the infrared, which allow the discrimination between different forest types and other land covers. However, frequent cloud coverage in the inner tropics often hampers the acquisition of high-quality data. Another major disadvantage is the low saturation level of the spectral bands and the derived spectral indices regarding the biomass estimation (Gibbs et al., 2007).

LiDAR (light detection and ranging) instruments mounted on airplanes send out active laser pulses and measure various echoes of the signal. This determines the three-dimensional vertical structure of vegetation in great detail (Goetz et al., 2009). LiDAR has been proven to allow accurate estimates of tree height, canopy closure and AGB (Duncanson et al., 2010, Kronseder et al., in preparation, Zhao et al., 2009). The disadvantage is that the acquisition of airborne LiDAR data requires sophisticated technical equipment, is very expensive especially in remote areas and has, therefore, not often been used in tropical forests (Gibbs et al., 2007).

Spaceborne synthetic aperture radar (SAR) sensors such as the L-band ALOS PALSAR, the C-band ERS/SAR, RADARSAT/SAR or ENVISAT/ASAR and the X-band TerraSAR-X instrument are also active systems, transmitting microwave energy at wavelengths from 3.0 (X-band) to 23.6 cm (L-band). The major advantage of all SAR systems is their weather- and daylight-independency. Their penetration depth and therefore their ability to measure biomass mainly depend on the wavelength: the longer the wavelength, the deeper the penetration (Le Toan et al., 2001, Henderson and Lewis, 1998). The ability to measure biomass is additionally affected by other sensor properties such as the polarization and the incidence angle, but also by the land cover and terrain properties, i.e. roughness and dielectric constant (Lu, 2006).

Various studies have analyzed the retrieval of AGB using radar data in tropical regions (Gama et al., 2010, Kuplich et al., 2005, Mitchard et al., 2009, Pandey et al., 2010) and Santos et al., 2006. Longer wavelengths have proven to be more useful because of an increasing backscatter range with changing biomass (Castro et al., 2003, Dobson et al., 1992, Lu, 2006, Luckman et al., 1997). These biomass estimations are valid up to a certain threshold where saturation occurs, i.e. where the slope of the biomass/backscatter coefficient curve approaches zero (Lucas et al., 2007, Mitchard et al., 2009).

The X- and C-band backscatter saturate at low biomass levels (30–50 t/ha). Pandey et al. (2010) found high r² values for a C-band backscatter/biomass relationship up to 250 t/ha. For L-band backscatter published saturation levels range from 40 t/ha (Luckman et al., 1997; Imhoff, 1995) to 150 t/ha (Kuplich et al., 2005, Lucas et al., 2007, Imhoff, 1995; Mitchard et al., 2009). Austin et al. (2003) stated that the L-band saturation level is possibly up to 600 t/ha.

Three different methods of calculating the saturation threshold are found in published literature. Luckman et al. (1997) established the saturation of an asymptotic backscatter/biomass relationship at 90% and 95% of the asymptotic intensity values. Rauste (2005) determined the saturation at regression analyses in a series of steps leaving a part of the stem volume range out. Watanabe et al. (2006) chose an arbitrary slope threshold of the backscatter/biomass curve of 0.01 dB ha/t.

In general, the saturation level not only depends on the SAR frequency but also on the amount of reference data. In the above mentioned SAR studies the correlation between AGB and SAR backscatter signal was often assessed using only a very limited set of in-situ data. Kuplich et al., 2005, Watanabe et al., 2006, Santos et al., 2006, Lucas et al., 2007 made AGB predictions in the tropics on the basis of SAR imagery in combination with approximately 51 in-situ reference data, whereas Pandey et al., 2010, Mitchard et al., 2009 analyzed up to 191 field inventory data. The latter found a limitation of biomass/backscatter correlation at 400 t/ha. Only Solberg et al. (2010) used upscaled biomass reference data from airborne LiDAR measurements, calibrated to in-situ data, resulting in estimations valid up to 250 t/ha with interferometric X-band data for pine and spruce forests.

In this study we investigated the capability of two different SAR sensors, TerraSAR-X (X-band) and ALOS PALSAR (L-band) in single and combined use as well as the use of mono- and multi-temporal data to retrieve AGB in tropical forest ecosystems in Kalimantan, Indonesia. The main focus is hereby on the impact of the different frequencies in regard to the biomass estimation capabilities for intact (high biomass) and degraded (low biomass) tropical forests using a novel approach. The upscaling approach included field inventory data and LiDAR based AGB estimations.

Section snippets

Study area

The principal study area of 280,062 ha is located in the Indonesian province Central Kalimantan on the island of Borneo and encompasses parts of the Mega Rice Project (MRP) (Block C) near Palangka Raya and the upper catchment of the river Sebangau (Fig. 1). It includes two types of peat swamp forests, regrown peat swamp forest on fire scars and heavily degraded forests after recurrent fire episodes (shrubs), as well as riparian forest and seasonally flooded wetlands. The Sebangau catchment was

Regression modeling

First, the relationship between AGB and backscattering coefficients was investigated for both X- and L-band data separately. By carefully examining the different curve progression of the correlation between biomass and radar signals, the optimal equation for mono- and multi-temporal single frequency regression models was found to be:y=aexpσ0bwhere σ0 = backscattering coefficient and a, b = coefficients depending on sensor and date.

Eq. (6) was fitted after a logarithmic transformation and for a

Discussion

The regression modeling of AGB based on TerraSAR-X and ALOS PALSAR imagery, alone and in combination, show that multi-temporal regression models are more accurate than mono-temporal models featuring higher coefficients of determination and lower root mean square errors. Mono-temporal regression models can be affected by extreme climatic conditions and the spatial and temporal transferability is therefore limited. Multi-temporal models compensate extreme conditions, e.g. heavy rainfall, and are

Acknowledgements

The authors would like to thank the Japan Aerospace Agency (JAXA) for supplying ALOS PALSAR data within the PI project No. 211 as well as Infoterra GmbH and German Space Agency (DLR) for providing TerraSAR-X data and their cooperation within the For-X Project (50 E705, PI: LAN0236). We gratefully acknowledge FFI (Fauna and Flora International) for making field data on biomass available and the National Aeronautics and Space Administration (NASA) for providing GPCP precipitation data free of

References (53)

  • F. Baup et al.

    Radar signatures of Sahelian surfaces in Mali using ENVISAT-ASAR data

    IEEE Transactions on Geoscience and Remote Sensing

    (2007)
  • K. Castro et al.

    Monitoring secondary tropical forests using space-borne data: Implications for Central America

    International Journal of Remote Sensing

    (2003)
  • J. Chave et al.

    Tree allometry and improved estimation of carbon stocks and balance in tropical forests

    Oecologia

    (2005)
  • W. Chen et al.

    Biomass measurements and relationships with Landsat-7/ETM+ and JERS-1/SAR data over Canada's western sub-artic and low artic

    International Journal of Remote Sensing

    (2009)
  • M.C. Dobson et al.

    Seasonal change in radar backscatter from mixed conifer and hardwood forests in northern Michigan

  • M.C. Dobson et al.

    Dependence of radar backscatter on coniferous forest biomass

    IEEE Transactions on Geoscience and Remote Sensing

    (1992)
  • FAO

    Estimating biomass and biomass change of tropical forests: a primer

  • FAO

    Global forest resources assessment 2000

  • FAO

    State of the world's forests 2009

    (2009)
  • T. Fritz

    TerraSAR-X Ground Segment Basic Product Specification Document

    (2009)
  • F.F. Gama et al.

    Eucalyptus biomass and volume estimation using interferometric and polarimetric SAR data

    Remote Sensing

    (2010)
  • H.K. Gibbs et al.

    Monitoring and estimating tropical forest carbon stocks: Making REDD a reality

    Environmental Research Letters

    (2007)
  • S.J. Goetz et al.

    Mapping and monitoring carbon stocks with satellite observations: A comparison of methods

    Carbon Balance and Management

    (2009)
  • F.M. Henderson et al.

    Principles and applications of imaging radar — Manual of remote sensing

    (1998)
  • M.L. Imhoff

    Radar backscatter and biomass saturation — Ramifications for global biomass inventory

    IEEE Transactions on Geoscience and Remote Sensing

    (1995)
  • IPCC

    IPCC guidelines for national greenhouse gas inventories

  • Cited by (223)

    • Forest total and component biomass retrieval via GA-SVR algorithm and quad-polarimetric SAR data

      2023, International Journal of Applied Earth Observation and Geoinformation
    View all citing articles on Scopus
    View full text