Elsevier

Advances in Space Research

Volume 69, Issue 4, 15 February 2022, Pages 1752-1767
Advances in Space Research

Synergistic evaluation of Sentinel 1 and 2 for biomass estimation in a tropical forest of India

https://doi.org/10.1016/j.asr.2021.03.035Get rights and content

Abstract

Spatially explicit measurement of Above Ground Biomass (AGB) is crucial for the quantification of forest carbon stock and fluxes. To achieve this, an integration of Optical and Synthetic Aperture Radar (SAR) satellite datasets could provide an accurate estimation of forest biomass. This will also help in removing the uncertainties associated with the single sensor-based estimation approaches. Therefore, the present study attempts to integrate Sentinel-2 optical data with Sentinel-1 SAR dataset to estimate AGB in the Shoolpaneshwar Wildlife Sanctuary (SWS), Gujarat, India. In this study, two non-parametric machine learning algorithms viz Support Vector Machines (SVMs) with different kernel functions—linear, sigmoidal, radial and polynomial and Random Forest (RF) were employed for the prediction of AGB using different combinations of VV, VH, Normalized Difference Vegetation Index (NDVI) and Incidence Angle (IA). Ground based AGB was estimated through allometric equation at 35 sampling sites with the help of tree height and Diameter at Breast’s Height (DBH). Standalone collinearity analysis among different parameters resulted in poor correlation of AGB with VH (r = 0.05) and IA (r = 0.015), whereas a significantly good correlation with NDVI (r = 0.80) and VV (r = 0.74) were observed. Inclusion of NDVI with VV and VH together also resulted in a better correlation (r = 0.85) than other combinations. The SVM with linear kernel utilizing parametric the combinations of VV + VH + NDVI and VV + VH + NDVI + IA were found to be best performing on the basis of evaluation metrics. The outcome of this study highlighted the significance of machine learning techniques and synergistic use of different remote sensing data for an improved AGB quantification in tropical forests.

Introduction

Tropical forests are recognized as worldwide biodiversity epicentres and apparent modulators governing climate change (Lewis, 2006). These forests undertake a significant task for global carbon storage and sequestration, and hence render imperative solutions for climate change mitigation (Gebeyehu et al., 2019, Houghton, 2005, Ota et al., 2014). Tropical forest embraces about 40% of the global Above Ground Biomass (AGB) (Williams et al., 2009) and this AGB accounts for approximately between 70% and 90% of total forests biomass (Cairns et al., 1997). Intergovernmental Panel on Climate Change (IPCC) has categorized AGB as one of the critical terrestrial ecosystem carbon pools (Eggleston et al., 2006). Carbon stored in AGB, especially within tropical forests is the largest among others, accounting for about 30% of the total terrestrial ecosystem carbon pool (Kumar and Mutanga, 2017). Initiatives such as Reducing Emissions from Deforestation and Forest Degradation (REDD) and REDD+ also demands more accurate quantification of AGB (Gibbs et al., 2007, Koch, 2010, Williams et al., 2009).

Direct conventional means of AGB measurements entails destructive field approach and considered as the most effective, reliable and accurate method (Goetz et al., 2009). This approach is found to be money and labour intensive and seldomly being used nowadays since it requires the cutting of trees. Being the most accurate AGB estimation technique, this method is economically not feasible and it cannot be extended to large geographical areas (Segura and Kanninen, 2005, Seidel et al., 2011, Wang et al., 2011). In many forest studies, this method is applied for building allometric equations for tree specific biomass estimation. Other field based indirect AGB estimation technique includes individual tree specific allometric models that uses tree biophysical parameters like Diameter at Breast Height (DBH) and height as input parameters (Chave et al., 2005, Malhi et al., 2020a). This approach is simple, cost effective and most importantly non-destructive. Limitation of this approach is that it brings an error propagation chain and requires appropriate selection of allometric models to higher accuracy for AGB estimates (Gonzalez de Tanago et al., 2018, Molto et al., 2013).

Quantification of AGB is also feasible by Remote Sensing (RS) and Geographic Information System (GIS) based models employing environmental parameters. Remote sensing works as a great tool for large scale, reliable and accurate spatiotemporal AGB measurements and monitoring, since it holds unique features in terms of data acquisition, huge spatial coverage, repetitivity and digital encoding (Chen and Qi, 2013, Lu et al., 2012, Ranjan et al., 2016). But, this method is not potent enough to provide as accurate estimates as the one observed from destructive method because satellite observations comes with several errors and dependencies (Anand, 2020, Chen and Qi, 2013, Lu et al., 2016). Different types of remote sensing data were explored for AGB estimation in the recent past. Satellite based optical remote sensing data such as Landsat, systeme probatoired’ observation de la terre (SPOT), Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER), China–Brazil Earth Resources Satellite (CBERS), Moderate Resolution Imaging Spectroradiometer (MODIS), Advanced Very-High-Resolution Radiometer (AVHRR), Quickbird, IKONOS, Worldview 2, Ziyuan 3 (ZY-3) have shown their potential in estimation of AGB (Fuchs et al., 2009, Günlü et al., 2014, Kumar and Mutanga, 2017, Lu, 2005, Lu, 2006, Lu et al., 2012, Luther et al., 2006, Song, 2013, Sun et al., 2015, Thenkabail et al., 2004). Whereas, Synthetic Aperture Radar (SAR) datasets such as Advanced Land Observing Satellite (ALOS) Phased Array type L-band Synthetic Aperture Radar (PALSAR) L-Band, Radarsat-2C-Band and Radar Imaging Satellite-1 (RISAT-1) C-Band were also used in number of studies for AGB estimates (Braun et al., 2018, Carreiras et al., 2012, Kumar et al., 2019, Mitchard et al., 2011, Padalia and Yadav, 2017, Rahman et al., 2013, Stelmaszczuk-Górska et al., 2018). Literature also shows that airborne and space-borne Light Detection and Ranging (LiDAR) scanning was also explored highly for measuring different forest parameter (Hajj et al., 2017, Hu et al., 2016, Lefsky et al., 2002, Nie et al., 2017, Sun et al., 2019, Wan-Mohd-Jaafar et al., 2017). Biomass estimation studies also exist on utility of advanced hyperspectral remote sensing for determining spatial variability in AGB (Anand et al., 2020, Brovkina et al., 2017, Laurin et al., 2014, Malhi et al., 2020a, Pandey et al., 2019, Srivastava et al., 2020).

Literature highlights the use of various parametric and non-parametric methods in quantification of AGB using remotely sensed data. Parametric methods like simple or multiple linear regression models (Brosofske et al., 2014, Zhang and Ni-meister, 2014) and non-parametric methods such as decision or regression-tree models are used by several authors (Dube and Mutanga, 2015, Güneralp et al., 2014, Kattenborn et al., 2015, Latifi et al., 2010, López-Serrano et al., 2020). Current focus of the research community is mostly on non-parametric Machine Learning (ML) methods like K-nearest Neighbour (KNN), Support Vector Machine (SVM), Back Propagation Neural Networks (BPNN) and Random Forest (RF). These methods surpass traditional regression methods since these can better identify complex non-linear relationships between ground based biomass and remote sensing parameters, which are infeasible using standard regression models (Chen et al., 2015, López-Serrano et al., 2020, Tian et al., 2014).

The current work aims at retrieving the AGB in tropical forests of Shoolpaneshwar Wildlife Sanctuary (SWS) using an ensemble of optical and microwave datasets that will help in determining more accurate AGB, due to its ability of capturing foliar by optical and woody part by SAR sensor (Dobson et al., 1992). This approach would also help in reducing the hindrance in AGB estimation, which occurs in optical data since it has few demerits related to cloud penetration and early saturation of spectral bands (Gibbs et al., 2007). Microwave data is well known for its cloud penetration power and deeper penetration through vegetation canopies. Several studies emphasized the importance of fusion of remote sensing data in solving the problem of uncertainty incorporated due to use of single sensor. (Anand et al., 2018, Sarker and Rahman, 2011). Additionally, the data coupled with field measurements will improve the precision of AGB estimation (Rosenqvist et al., 2003). Thus, the focus of this study lies in predicting the AGB of tropical forests through the use of optical Sentinel-2 and SAR Sentinel-1 data coupled with field AGB estimates. An attempt is being made to predict AGB by establishing the relationship between in-situ based ground AGB and various optical and microwave parameters namely NDVI (from optical) and VV, VH & IA (from microwave) through different Machine Learning (ML) algorithms.

Section snippets

Study area

The study was conducted within the tropical forests of Shoolpaneshwar Wildlife Sanctuary (SWS) located in Narmada District of Gujarat state, India (Fig. 1). In the year 1989, the sanctuary was established in the geographical area spread of 607.71 km2. This sanctuary is situated between 21°03′ to 21°59′ North latitude and 73°05′ to 74°10′ East longitude at an altitude of 800 m to 900 m above Mean Sea Level (MSL) (Fig. 1). It covers highly dense and diverse forest characterized by species like

In situ biomass measurements

Field AGB values estimated for 35 plots studied in the SWS varied between 7.14 t/ha and 282.79 t/ha (Fig. 3). Average AGB was 118.04 t/ha for the sanctuary and standard deviation calculated was 81.6 t/ha. The plot under investigation comprised of heterogenous species. The species found were Tectona grandis L. f., Morinda pubescens J.E. Smith, Wrightia tomentosa (Roxb.) Roem. & Schult, Terminalia crenulata (Heyne) Roth, Alangium salvifolium (L.f.) Wangerin, Butea monosperma (Lam.) Taub, Cassia

Conclusions

The current vulnerabilities in climate change have drawn the attention of researchers in exploring the direct association of tropical biomass enrichments and atmospheric carbon dioxide to estimate and predict the AGB over regional as well as global scale. The climate change mitigation mechanisms proposed by IPCC, such as REDD and REDD + also requires frequent estimates of AGB indispensably. Furthermore, tropical forest biomass estimation is quite challenging due to problems in situ sampling in

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Acknowledgement

First author gives her sincere thanks to Department of Science and Technology and Science and Engineering Research Board for awarding and funding her with National Post-Doctoral Fellowship (PDF/2017/002620). Authors are very grateful to Ministry of Environment, Forest and Climate Change and Gujarat Forest Department for allowing them to conduct ground measurements in the study area. Thanks also go to forest officials for their generous help in the ground measurements.

References (93)

  • A. Kumar et al.

    Tree diversity assessment and above ground forests biomass estimation using SAR remote sensing: A case study of higher altitude vegetation of North-East Himalayas, India

    Phys. Chem. Earth, Parts A/B/C

    (2019)
  • S. Kumar et al.

    Polarimetric SAR Interferometry based modeling for tree height and aboveground biomass retrieval in a tropical deciduous forest

    Adv. Space Res.

    (2017)
  • Y. Li et al.

    Forest aboveground biomass estimation using Landsat 8 and Sentinel-1A data with machine learning algorithms

    Sci. Rep.

    (2020)
  • J. Luther et al.

    Biomass mapping using forest type and structure derived from Landsat TM imagery

    Int. J. Appl. Earth Obs. Geoinf.

    (2006)
  • E.T. Mitchard et al.

    Measuring biomass changes due to woody encroachment and deforestation/degradation in a forest–savanna boundary region of central Africa using multi-temporal L-band radar backscatter

    Remote Sens. Environ.

    (2011)
  • S. Nie et al.

    Above-ground biomass estimation using airborne discrete-return and full-waveform LiDAR data in a coniferous forest

    Ecol. Ind.

    (2017)
  • H. Padalia et al.

    Evaluation of RISAT-1 SAR data for tropical forestry applications

    Adv. Space Res.

    (2017)
  • Å. Rosenqvist et al.

    A review of remote sensing technology in support of the Kyoto Protocol

    Environ. Sci. Policy

    (2003)
  • V.P. Singh et al.

    An Econometric Time Series Forecasting Framework for Web Services Recommendation

    Proc. Comput. Sci.

    (2020)
  • V.P. Singh et al.

    Neural Net Time Series Forecasting Framework for Time-Aware Web Services Recommendation

    Proc. Comput. Sci.

    (2020)
  • X. Zhu et al.

    Improving forest aboveground biomass estimation using seasonal Landsat NDVI time-series

    ISPRS J. Photogramm. Remote Sens.

    (2015)
  • I. Ali et al.

    Review of machine learning approaches for biomass and soil moisture retrievals from remote sensing data

    Remote Sensing

    (2015)
  • A. Anand

    Sentinel SAR Data and In-Situ-Based High-Resolution Above-Ground Carbon Stocks Estimation Within the Open Forests of Ramgarh District

  • A. Anand et al.

    Use of Hyperion for Mangrove Forest Carbon Stock Assessment in Bhitarkanika Forest Reserve: A Contribution Towards Blue Carbon Initiative

    Remote Sensing

    (2020)
  • A. Anand et al.

    Estimating the change in Forest Cover Density and Predicting NDVI for West Singhbhum using Linear Regression

    Intern. J. Environ. Rehabilitation Conservation

    (2018)
  • P.S. Bharadwaj et al.

    Polarimetric scattering model for estimation of above ground biomass of multilayer vegetation using ALOS-PALSAR quad-pol data

    Phys. Chem. Earth, Parts A/B/C

    (2015)
  • C.M. Bishop

    Pattern recognition and machine learning

    (2006)
  • A. Braun et al.

    Above-ground biomass estimates based on active and passive microwave sensor imagery in low-biomass savanna ecosystems

    J. Appl. Remote Sens.

    (2018)
  • L. Breiman et al.

    Classification and regression trees

    (1984)
  • K.D. Brosofske et al.

    A review of methods for mapping and prediction of inventory attributes for operational forest management

    Forest Sci.

    (2014)
  • M.A. Cairns et al.

    Root biomass allocation in the world's upland forests

    Oecologia

    (1997)
  • J.R. Chave et al.

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

    Oecologia

    (2005)
  • Q. Chen et al.

    Lidar remote sensing of vegetation biomass

    Remote Sens. Na. Resources

    (2013)
  • M.C. Dobson et al.

    Dependence of radar backscatter on coniferous forest biomass

    IEEE Trans. Geosci. Remote Sens.

    (1992)
  • Eggleston, H., Buendia, L., Miwa, K., Ngara, T. and Tanabe, K., 2006. IPCC Guidelines for National Greenhouse Gas...
  • K. Folkesson et al.

    Model-based compensation of topographic effects for improved stem-volume retrieval from CARABAS-II VHF-band SAR images

    IEEE Trans. Geosci. Remote Sens.

    (2009)
  • FSI, 1996. Volume equations for forests of India, Nepal and Bhutan. Forest Survey of India, Ministry of Environment and...
  • G. Gebeyehu et al.

    Carbon stocks and factors affecting their storage in dry Afromontane forests of Awi Zone, northwestern Ethiopia

    J. Ecol. Environ.

    (2019)
  • H.K. Gibbs et al.

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

    Environ. Res. Lett.

    (2007)
  • S.J. Goetz et al.

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

    Carbon Balance Manage.

    (2009)
  • J. Gonzalez de Tanago et al.

    Estimation of above-ground biomass of large tropical trees with terrestrial LiDAR

    Methods Ecol. Evol.

    (2018)
  • A. Günlü et al.

    Estimating aboveground biomass using Landsat TM imagery: A case study of Anatolian Crimean pine forests in Turkey

    Ann. Forest Res.

    (2014)
  • M.E. Hajj et al.

    Interest of integrating spaceborne LiDAR data to improve the estimation of biomass in high biomass forested areas

    Remote Sensing

    (2017)
  • T. Hastie et al.

    The elements of statistical learning: data mining, inference, and prediction, springer series in statistics

    (2009)
  • R. Houghton

    Aboveground forest biomass and the global carbon balance

    Glob. Change Biol.

    (2005)
  • T. Hu et al.

    Mapping global forest aboveground biomass with spaceborne LiDAR, optical imagery, and forest inventory data

    Remote Sensing

    (2016)
  • Cited by (24)

    • Crop type discrimination using Geo-Stat Endmember extraction and machine learning algorithms

      2024, Advances in Space Research
      Citation Excerpt :

      The data from AVIRIS NG was pre-processed from the atmospheric correction point of view for generating L0 and L1 products (radiance to reflectance) (Bhattacharya et al., 2019, Srivastava et al., 2020). The acquired data has a spatial resolution of 4 m with 425 bands ranging between 350 and 2500 nm at an interval of 5 nm that are sensitive to the band regions in the VNIR (400–100 nm) and SWIR (970–2500 nm) regions (Singh et al., 2020b, Malhi et al., 2021). ASD FieldSpec®4 full range spectroradiometer instrument was used to capture leaf reflectance data.

    View all citing articles on Scopus
    View full text