Int. J. Environ. Sci. Technol.

DOI 10.1007/s13762-015-0750-0

REVIEW

A review of radar remote sensing for biomass estimation

S. Sinha • C. Jeganathan • L. K. Sharma •

M. S. Nathawat

Received: 26 July 2013 / Revised: 5 November 2014 / Accepted: 3 January 2015

Ó Islamic Azad University (IAU) 2015

Abstract Forest plays a vital role in regulating climate undulating terrain and in radical conditions. Also, synergic
through carbon sequestration in its biomass. Biomass use of multi-sensor optical and SAR has better potential
reflects the health and environmental conditions of a forest than single sensor. Interferometric (InSAR) and polari-
ecosystem. In context to the climate change mitigation metric (PolSAR) SAR or a combination of the both (Pol-
mechanisms like REDD (reducing emissions from defor- InSAR) serves as effective alternatives. These techniques
estation and forest degradation), an extensive forest mon- could serve as valuable methods for biomass assessment of
itoring campaign is especially important. Remote sensing heterogeneous complex biophysical environments. How-
of forest structure and biomass with synthetic aperture ever, SAR data have its own limitations and complexities.
radar (SAR) bears significant potential for mapping and Identifying, understanding and solving major uncertainties
understanding forest ecological processes. Limitations of in different stages of the biomass estimation procedure are
the conventional forest inventory procedures, like the critical. In this regard, the current study provides a review
extensive cost, labor and time, can be overcome through of radar remote sensing-based studies in forest biomass
integrated geospatial techniques. Optical sensor or SAR estimation.
data are suitable for extracting information about simple
and homogeneous forest stand sites. However, optical Keywords Biomass
Interferometry
Polarimetry

sensors face serious limitations, specifically in tropical SAR
Uncertainty
regions, like the cloud cover that SAR can overcome along
with targeting saturation and penetration aspects. Simul-
taneous use of spectral information and image texture Introduction
parameters improves the biomass assessment over
Forest is defined as an ecosystem dominated by trees and
other woody vegetation with land area [0.5 ha, with
S. Sinha
C. Jeganathan (&)
[10 % canopy cover and not being utilized for agriculture
Department of Remote Sensing, Birla Institute of Technology,
Mesra, Ranchi 835215, India or any non-forest land use (FAO 2001). Essentially
e-mail: jegan_iirs@yahoo.com important to mankind, the forest ecosystems have a control
S. Sinha over climate, streams, soil, oxygen–carbon dioxide bal-
e-mail: sumanrumpa.sinha@gmail.com ance, wood supply, aesthetic diversity, biodiversity and
provide various ecosystem services (Nabuurs et al. 2007).
L. K. Sharma
Forests inherit key information about climate change and
Centre for Land Resource Management, Central University of
Jharkhand, Brambe, Ranchi 835205, India are dynamic in terms of phenology, productivity and
e-mail: laxmikant1000@yahoo.com flammability changes. In tropical forests, increase in res-
piration with warming and drying shows a positive feed-
M. S. Nathawat
back as predicted by coupled climate–carbon models (Field
School of Sciences, Indira Gandhi National Open University
(IGNOU), Maidan Garhi, New Delhi 110068, India et al. 2007), while for boreal and temperate forests, sig-
e-mail: msnathawat@ignou.ac.in nificant range shifts and forest expansion can be potentially

123
 Int. J. Environ. Sci. Technol.

caused due to climate change (Loehle 2000). Coordinated the ecosystems imbibe for almost a third of anthropogenic
temporal continuous forests monitoring can put forward an fossil fuel emissions (Malhi 2002). Forests absorb nearly
evidence of the effect of climate change on natural systems one-twelfth of the total earth’s atmospheric CO2 stock,
and can act as a warning indicator of sudden shifts, for most of which is stored as woody biomass or cycled into
instance, changes in leaf water content before forest fires. the soil and accounts for about 72 % of the earth’s ter-
Deforestation affects the proper functioning of ecosys- restrial carbon storage (Malhi 2002). Deforestation of
tem services and has serious impacts on the meteorology tropical forests destroys carbon sinks and hence posses
and climate change scenario. Anthropogenic activities like threat to future climate stabilization (Stephens et al. 2007).
fossil fuel burning and land use changes including defor- The United Nations Framework Convention on Climate
estation and fires have resulted in teeming release of CO2 Change (UNFCCC) agreed to provide financial incentives
into the atmosphere (Malhi 2002; Lu 2006). Hence, there is to promote emission reduction from deforestation below a
a growing need for forest monitoring and management baseline in developing countries, the concept referred to as
especially for forest stand biomass, forest structure and reduced emissions from deforestation and forest degrada-
biodiversity. Biomass is the living plant and animal tion (REDD) (Gibbs et al. 2007). The concept of REDD
material both aboveground and belowground usually evolved from reduced emissions from deforestation (RED),
expressed as dry weight. Aboveground biomass includes while further concepts of REDD? and REDD??/REALU
all living biomass above the soil including stem, stump, (reducing emissions from all land uses) have developed
branches, bark, seeds and foliage. Forest biomass act as an from REDD revealing the importance and severity of the
indicator of climate change and forest health (Nabuurs concept on a global scale (Gibbs et al. 2007; Plugge et al.
et al. 2007; Kumar et al. 2013). Numerous studies on 2010; Sharma et al. 2013). REDD? involves forest con-
biomass assessment have focused on boreal and temperate servation and sustainable management and forest carbon
forests as mentioned below in Table 2; however, studies on stock enrichment, in addition to the objectives of REDD
tropical forests are limited. This is due to the fact that while REALU deals with the emissions from all the land
tropical forests are complex and dynamic with complex uses and not just restricted to forests. REDD regime
species composition and structure, and environmental includes the following steps: assessment of forest carbon
conditions which is difficult to assess and model. Optical stocks and change over time, quantifying the amount of
remote sensing have been successful in forest biomass CO2 reduction, qualifying for accounting, identifying and
studies but over limited geographical regions. But in the ranking of the relevant causes for human impact on forests
tropical region, where cloud cover problem is predominant, (Sharma et al. 2013), developing a reference baseline for
it could not be really used. In these conditions, radar accounting changes of carbon stocks in forests and lastly
remote sensing provides the best solution as it has several executing a framework for the transfer of benefits at ground
advantages over optical remote sensing as all weather, day level (Plugge et al. 2010; Sharma et al. 2013). Townshend
and night; penetrates clouds, vegetation, dry soil, sand, dry et al. (2012) have shown the utility of global Landsat
snow; sensitive to surface roughness, dielectric properties optical datasets to account for cost-effective monitoring of
and moisture content; sensitive to polarization and fre- the earth’s land cover, forest cover and cover change along
quency; imaging possibility from different types of polar- with specific data inputs from MODIS that can serve as
ized energy (HH, VV, HV and VH); and volumetric important breakthrough in the field of REDD and climate-
analysis. related studies in a global aspect. However, due to high
However, radar remote sensing also has limitations like uncertainty in carbon sink and emissions estimates, the
uncertainties in estimation, saturation, expensive datasets, exact size and cause of the sink remain a matter of unsure.
difficulties in data processing and complex interaction with Uncertainties in the assessment of biomass or carbon
forests. This study attempts to provide a review of historic stock are a major problem to cope up. The exact proportion
developments associated with radar remote sensing espe- of carbon sink from secondary forest regrowth is unknown
cially related to synthetic aperture radar (SAR)-based apart from detailed studies for small areas; hence, the
remote sensing applications for the forested environment estimation suffers a high degree of uncertainty (House
with a focus toward biomass extraction. et al. 2003; Houghton 2005). Therefore, a better under-
standing of global carbon cycle can be made possible
through accurate and reliable methods for assessing forest
Changing carbon scenario and global initiatives biophysical parameters including biomass, which in turn
reveal the carbon sink that would ultimately end up in
Terrestrial ecosystems have a significant role in absorbing sustainable forest and natural resource management. Bio-
and emitting CO2 through vegetation growth and metabo- mass is quantified as a mass of living plant material per unit
lism, and respiration. While functioning as a carbon sink, area and includes above- and belowground living mass and

123
Int. J. Environ. Sci. Technol.

the dead mass of soil litter. Estimation of aboveground logarithmic (Husch et al. 2003). Apparently, remote sens-
biomass (AGB) is rather easy as compared to belowground ing and model-based techniques can be included within the
biomass due to the intricacy involved in field data collec- non-destructive methods. Alternatively, it would be better
tion. Biomass has long-term impacts on carbon cycles, soil to classify the methods under three broad groups of
nutrient allocations, wildlife habitats, etc. To reduce the (a) field-dependent, (b) partial field-dependent and
uncertainty in the measurements, precise estimations at (c) field-independent approaches. Destructive sampling
local to global scales are necessary which explains its roles technique can be included in field-dependent approach,
in environmental sustainability. while the non-destructive, remote sensing and model-based
techniques can be categorized within the partial field-
dependent approach. No method is purely field indepen-
Remote sensing in biomass assessment dent; however, the semiempirical model-based technique
relies mainly on the theoretical basis along with experi-
Maturing observation technologies and intense public mental or observational data, whereas empirical models are
interest in protecting and managing forests provides a developed using experimental or observational data only
necessity and opportunity to explore and better understand (Kumar 2009). On the other hand, remote sensing could be
the global forests. The remote sensing technology (capture field independent, if the kind of biomass variation in a
and analysis of satellite and aerial images) meets the steep particular forest type is known, which can then be extended
logistical challenge of measuring the global forests at local for biomass assessment of similar areas.
to global levels in an accurate, precise, repeatable and The unique characteristics of remote sensing data
economical manner. Remote sensing techniques to estimate obtained with synoptic view, high spatiotemporal resolution
biomass can account for the limitations of sample size, and digital format that allow the handling of fast processing
timeliness, expense and access at a range of scales (Pat- of huge amount of data in addition to the availability of data
enaude et al. 2005). Remote sensing data can effectively for inaccessible forest areas unavailable for field survey.
provide a synoptic view over the large areas and greatly Optical remote sensing data, SAR (microwave) data and
increase efficiency and usefulness of limited conventional LiDAR data are the main three types of remotely sensed
methods (Patenaude et al. 2005). Roy et al. (1994) have data that are used to extract information for biomass and
used merged optical and airborne X-band SAR data for stand parameters. The use of remote sensing technology in
forest stratification and canopy characterization. Lucas the assessment of biomass has proved to be a better alter-
et al. (2008) have hypothesized the synergic use of native to the conventional methods of biomass estimation
hyperspectral and LiDAR data for retrieving forest bio- (Lu 2005, 2006). Biomass estimation using optical remote
mass. The technology offers temporal analysis with syn- sensing data is usually realized by revealing the correlation
optic coverage that accounts for change detection of the between biomass and spectral responses and/or vegetation
global forests and its biophysical parameters over varied indices derived from multispectral images. Lu (2006)
scales and levels. So, it can be used as a tool in AGB classified the techniques used for estimation of AGB into
estimation. Therefore, remote sensing using optical, following three main categories as: (1) field measurement-
microwave, hyperspectral and LiDAR techniques for AGB based methods, (2) remote sensing-based methods and (3)
estimation has increasingly attracted scientific interest. GIS-based methods. Optical remote sensing uses the tech-
Herold et al. (2007) classified the techniques involved in nique of modeling based on biomass–vegetation index
biomass estimation into primarily following four catego- relations to estimate the aboveground biomass, as it uses the
ries: (a) harvest mapping or destructive sampling-based, interactions between the electromagnetic waves with the
(b) non-destructive sampling-based, (c) airborne/space- leaf chemistry or structure to measure the vegetation indices
borne remote sensing-based, (d) model-based (empirical like normalized difference vegetation index (NDVI), leaf
and semiempirical) techniques. The conventional method area index (LAI), etc. (Kumar et al. 2013). On the basis of
includes the harvest or destructive sampling approach that spatial resolution of satellite data, Lu (2006) categorized the
deals with the total removal of vegetation of pre-defined optical sensor data for AGB estimation as fine, medium and
sample unit of the forest area (Husch et al. 2003). The non- coarse spatial resolution.
destructive sampling approach which does not involve tree
harvesting has been the most widely used technique for SAR vis-à-vis optical remote sensing
biomass estimation through in situ measurements. It
includes regression equations with parameters like tree In addition to the benefits provided by remote sensing
height, stem volume and basal area to estimate biomass. already mentioned, SAR offers certain unique capabilities
The most common regression equations used for biomass that have advantages over optical sensors which are as
estimation are linear, quadratic, exponential and follows:

123
 Int. J. Environ. Sci. Technol.

(a) All weather capability (penetration capability through Unlike the optical data, the cost of SAR data is a serious
clouds), i.e., 24-h data constraint in the development of commercial technology
(b) Day and night capability (independent of intensity for AGB estimation. Mission cost of ERS-2 was about 650
and sun illumination angle) million US dollars (USD) and the data prices, varying with
(c) Penetration through vegetation, soil sand and dry the product type, processing level and mode of delivery,
snow to a certain extent ranged from 250 euro for a medium-resolution SAR scene,
(d) Sensitivity to surface roughness, dielectric properties to more than 2,000 euro for a terrain-corrected, geo-coded
and moisture (in liquid or vapor forms) SAR product covering an area of 100 km 9 100 km for
(e) Sensitive to wave polarization and frequency ERS datasets. Radarsat-1 also had a mission cost of 650
(f) Volumetric analysis million dollars, however, without launching expenditures,
(g) Better analysis from inaccessible areas and the data cost is 3,600 Canadian dollars (CAD) per SLC
Forest biomass assessment involves the volumetric analysis scene of 100 km 9 100 km for newly acquired standard
of the vegetation, and this exploits the intrinsic capability products and 1,500 CAD for archive products. Similarly,
of penetration of SAR through the tree canopies which is data charges of Radarsat-2 are 3,600 CAD per SLC scene
completely absent in optical remote sensing. The estima- of single polarized data and 3,800 for dual polarized data
tion is also sensitive to wave polarization and frequency, as for standard products. The rate varies between 3,600 and
higher wavelengths have greater penetration capacity and 8,400 CAD per scene depending on the data acquisition
cross-polarization waves are more sensitive to biomass. mode. ALOS PALSAR archive standard products are
The analysis also varies according to the moisture content priced at 600 euro per scene. The rate of COSMO-SkyMed
of the vegetation. These unique qualities of SAR provide standard products of new acquisition varies between 1,650
better estimation of biomass with reduced uncertainties in and 9,450 euro per scene depending on the acquisition
the assessment, as microwaves saturates at higher levels of mode, while for archive, it ranges between 825 and 4,725
biomass in comparison to optical electromagnetic waves. euro per scene. A scene of RISAT-1 is worth at 12,000 INR
In spite of such high potentials, there are some draw- (Indian Rupees) for Indian nationals. The huge cost
backs of using SAR that needs to be overcome. Compared incurred in SAR data acquisition and initial processing
to optical data, the x- and y-resolutions are not same as have created hurdles in the SAR-based researches, and
range resolution varies with local incident angle. Inter- developing technology of low-cost SAR missions are a
pretation requires good understanding of microwave fre- burning issue that needs attention.
quency interaction with various targets, as data Several parameters are involved in SAR, like surface
interpretation is affected by occurrence speckles and image roughness, dielectric property, azimuth and range resolu-
distortions (overlay, foreshortening and shadows) due to tion, image geometry, distortions, multi-looking, incidence
undulating terrains. Speckle refers to a noise-like charac- angle, polarization, etc. that adds to the complexity of the
teristic produced by coherent systems such as SAR caused data, and hence, data interpretation becomes more difficult
by the constructive and destructive interference of radar and complicated. Therefore, SAR data requires intricate,
return scattered from surfaces or objects on ground which accurate and specific processing so as to extract maximum
makes image interpretation difficult. The speckle texture information from them. Processing of SAR comprises of
depends on SAR wavelength and target spacing. Layover certain complex steps that can be performed in specific
occurs when the radar beam reaches the top of a tall object software. Hence, SAR data acquisition and processing
before it reaches the base. The top of the object is displaced incurs huge cost and occupies huge space. SAR techniques
toward the radar from its true position on the ground and are still in experimental mode, and the techniques need to
lay over the base of the feature. Foreshortening starts with be developed commercially.
the gradual cease of layover at small depression angles.
Under moderate slope conditions, though the radar return SAR for biomass estimation
from foot reaches first followed by the top, the ground
range will be less than the actual ground distance causing Spectral responses recorded in optical images are mainly
compression in the image called foreshortening. Shadows due to the interaction between the solar radiance and forest
occur toward the far range, behind vertical features or stand canopies that serve as a limitation in the ability to
slopes with steep sides. Since the radar beam does not predict forest biomass through optical remote sensing
illuminate the surface, shadowed regions will appear dark technologies. This results in weak correlation between
on an image as no energy is available to be backscattered. biomass and spectral responses (or vegetation indices),
These are some of the general image distortions associated specifically for mature heterogenous forests where the
with the SAR images. spectral responses start to saturate resulting in low

123
Int. J. Environ. Sci. Technol.

sensitivities to branch and trunk biomass, i.e., the bole bands has their own characteristics in relating to forest
biomass. These limitations can be overcome by the use of stand parameters. The X band interacts with the leaves and
SAR remote sensing, which has the additional capability to canopy cover surface, hence suitable for tree canopy sur-
penetrate the cloud cover unlike the optical sensors. The face layer information. The C band penetrates through
unique qualities of the SAR data for forest biomass esti- leaves and are scattered by small branches and underlying
mation make SAR a remarkable technology for forest features. L band has the higher penetrating capacity that
investigations, particularly in the areas with frequent cloud penetrates through the surface layers and is scattered by the
cover. Simultaneously, unlike optical, SAR has specific trunk and the main branches. With the greatest penetration
characteristics like polarization, sensitivity toward mois- capabilities, the P band penetrates into the canopy cover.
ture (dielectric constant), surface roughness, different Most part of P-band backscattering is due to the trunk and
incident angles, higher penetration capabilities, etc. that the trunk–ground interactions. Henceforth, the backscatters
adds to its potentialities. of the L and P band are most related to the biophysical
Carl A. Wiley, a mathematician in Goodyear Aircraft parameters of the trees and are maximally used for forest
Company, Arizona, invented SAR in 1951 (Wiley 1985). biomass-related studies.
SEASAT is the first radar satellite for civilian applications Polarization of the SAR signals are an important
launched in June 27, 1978, by the National Aeronautics and parameter of SAR data that interacts variably due to dif-
Space Administration/Jet Propulsion Laboratory (NASA/ ferent orientations and structures of the features. Polari-
JPL). SAR works on the principle of Doppler effect. zation of the electromagnetic waves refers to the direction
Doppler shift is the frequency shift in electromagnetic of electric field and depends upon the interaction between
waves due to the motion of scatterers toward or away from signals and the reflectors. Microwave sensors emit signals
the observer. Frequency decreases when the source moves in horizontal (H) or vertical (V) polarizations. The four
away from the receiver and vice versa. Doppler radar combinations SAR data polarizations: (1) HH: The emitted
determines the frequency shift through measurement of the and backscattered signals have horizontal polarization. (2)
phase change in electromagnetic waves during a series of HV: The emitted signal has horizontal polarization, and the
pulses. Table 1 gives the list of some selected SAR sensors backscattered signal has vertical polarization. (3) VH: The
till date (Ouchi 2013), along with their spatial resolution. emitted signal has vertical polarization, and the backscat-
There is a unique relation between the spatial and temporal tered signal has horizontal polarization. (4) VV: Both
(revisit) resolution, where revisit time for the wide-swath emitted and reflected signals have vertical polarization
mode can be reduced at the expense of spatial resolution. (Ghasemi et al. 2011).
Cosmo-SkyMed and SAR-Lupe with higher spatial reso- The C, L and P bands are frequently used in most of the
lution have their revisit time of 7 and \10 h, respectively, biomass estimation studies. The longer wavelengths (L and
as compared to the nominal revisit times of 24, 25, 35 and P band) and the HV polarization are most sensitive to AGB
46 days of RADARSAT-2, RISAT-1, ENVISAT ASAR (Luckman et al. 1997; Kurvonen et al. 1999; Sun et al.
and ALOS PALSAR, respectively (Ouchi 2013). 2002). It was found that the co-polarized (HH and VV) data
Table 2 summarizes the numerous works that have at the longer wavelengths, like P band, were sensitive to
revealed the ability of SAR data in estimating forest bio- changing surface conditions (Ghasemi et al. 2011). Cross-
physical parameters, particularly the AGB. The most fre- polarized (HV and VH) backscattering mainly occurs from
quently used methods in biomass estimation may be divided multiple scattering within the tree canopy and is less
into two groups: (1) using backscatter values and (2) inter- affected by the surface condition (Ranson and Sun 1994).
ferometry technique (Ghasemi et al. 2011). It has been pro- Backscattering at longer wavelengths is lower than that
ven that longer wavelengths (L and P band) with HV and VH from C band for low biomass sites, such as grassland, bogs,
polarizations yield better result than short wavelengths (X clear cuttings, areas of forest regeneration and young
and C band) with HH or VV polarizations (Le Toan et al. plantations, and hence, C band is preferred for lower
1992; Dobson et al. 1992). Hyperspectral remote sensing vegetation biomass estimations (Ghasemi et al. 2011). P
also has potential in retrieval of biomass (Treuhaft et al. band gives very low backscattering for surfaces covered
2003). Ghasemi et al. (2013) suggested the wavelet analysis with grass or juvenile plant species since these act as small
to be even more effective in terms of biomass estimation. scattering elements as compared to P-band wavelength of
68 cm to give significant backscattering, while the same
SAR wavelength and polarization surfaces would be rough at C band, resulting to strong
backscattering, where the leaves and small primary bran-
SAR data can be acquired in K, X, C, L and P bands ches are the major scatterers for C band that saturates at
(different wavelengths) with different polarizations having nearly 10 kg/m2 (Ranson and Sun 1994). The limitation of
variety of range and azimuth resolutions. Each of these C band is the inability of much penetration into the canopy

123
 Int. J. Environ. Sci. Technol.

Table 1 Selected SAR sensors for earth observation

Platform/sensor Type Agency/country Band (polarizations) Spatial resolution

AIRSAR Airborne NASA/USA X/C/L (quad) 0.6, 3

ALMAZ-1 Spaceborne USSR S (HH) 8, 15
ALOS PALSAR Spaceborne JAXA/Japan L (quad) 5, 10
C/X-SAR Airborne CCRS/Canada X/C (quad) 0.9, 6
CALABAS Airborne FOA/Sweden HF/VHF 3, 3
Cosmo-SkyMed (4) Spaceborne ASI/Italy X (quad) 1, 1
CP-140 Spotlight SAR Airborne Lockheed Martin/Canada X \1, \1
DBSAR Airborne NASA/USA L 10, 10
EMISAR Airborne DCRS/Denmark C/L (quad) 2, 2
ENVISAT ASAR Spaceborne ESA C (dual) 10, 30
ERS-1/2 Spaceborne ESA C (VV) 5, 25
E-SAR Airborne DLR/Germany X/C/S/L/P (quad) 0.3, 1
F-SAR Airborne DLR/Germany X/C/S/L/P (quad) 0.3, 0.2
Global Hawk Airborne Northrop Grumman/USA X 1.8, 1.8
HJ-1-C Spaceborne China S (VV) 5, 20
I-MASTER Airborne Thales-Astrium/UK Ku \1, \1
Ingara Airborne DSTO/Australia X (quad) 0.15, 0.3
JERS-1 SAR Spaceborne NASDA/Japan L (HH) 6, 18
JERS-1 SAR Spaceborne NASA/USA C/L (quad) 7.5, 13
LiMIT Airborne MIT Lincoln Lab/USA X \1, \1
Lynx Airborne Sandia/USA Ku (quad) 0.1, 0.1
MiniSAR Airborne Sandia/USA Ka/Ku/X 0.1, 0.1
Mini-SAR Airborne TNO/Netherland X 0.05. 0.05
PAMIR Airborne FHR-FGAN/Germany X 0.1, 0.1
PHARUS Airborne TNO-FEL/Netherland C (quad) 1, 3
Pi-SAR Airborne NICT, JAXA/Japan X/L (quad) 0.37, 3
RADARSAT-1 Spaceborne CSA/Canada C (HH) 8, 8
RAMSES Airborne ONERA/France W/Ka/Ku/X/C/S/L/P (quad) 0.12, 0.12
RARDASAT-2 Spaceborne CSA/Canada C (quad) 3, 3
RISAT-1 Spaceborne ISRO/India C (dual) 3, 3
SAR-Lupe (5) Spaceborne Germany X (quad) 0.5, 0.5
SEASAT-SAR Spaceborne NASA/USA L (HH) 6, 25
SIR-A Shuttleborne NASA/USA L (HH) 7, 25
SIR-B Shuttleborne NASA/USA L (HH) 7, 13
SIR-C/X-SAR Shuttleborne DLR/Germany, ASI/Italy X (VV) 6, 10
SRTM Shuttleborne NASA/USA C (dual) 15, 8
SRTM Shuttleborne DLR/Germany X (VV) 8, 19
TanDEM-X Spaceborne DLR/Germany X (quad) 1, 1
TerraSAR-X Spaceborne DLR/Germany X (quad) 1, 1
UAVSAR Airborne NASA/USA L (quad) 1, 1.8
Spatial resolution expressed in meters in the azimuth (single-look) and range directions

and its saturation at around 60–70 tons/ha (Nizalapur et al. about 250 t/ha for stands with simple structure and few
2010). This limitation can be overcome using longer dominant species. The best combination for biomass esti-
wavelength bands which have higher forest canopy pene- mation in deciduous forests is the C and L band with the
tration capability like the L band. The L and P band sat- HH and HV polarization, while the same bands with just
urates usually at 100 t/ha for complex heterogenous HV polarization is the best for coniferous forests (Ranson
tropical forest structures. The saturation level increases to and Sun 1994). L band has the ability to estimate biomass

123
Int. J. Environ. Sci. Technol.

Table 2 SAR techniques and methods for AGB estimation

Study Sensor/datasets Study site Models

Le Toan et al. AirSAR Landes forest, France Regression analysis

(1992)
Beaudoin et al. SAR L band Les Landes Forest, France Adapted theoretical model
(1994)
Ranson and AirSAR Maine, USA Regression analysis
Sun (1994)
Ranson et al. AirSAR Maine, USA Gap-type forest succession model, Canopy
(1997) backscatter models
Luckman et al. ERS-1, JERS-1 and Tapajos, Brazil Forest backscatter model
(1997) SIR-C
Kurvonen JERS-1 and ERS-1 Scandinavian forests, Finland Backscattering model, Inversion algorithm
et al. (1999) SAR
Kuplich et al. JERS-1 Tapajos, Brazil; Southern Cameroon Regression analysis
(2000)
Fransson et al. ERS-1/2, SPOT XS Kattbole, Sweden Regression analysis
(2001)
Santos et al. JERS-1 Amazonia, Brazil Regression analysis
(2002)
Sun et al. SIR-C Siberia Regression analysis
(2002)
Austin et al. JERS-1 SAR New South Wales, Australia Regression analysis
(2003)
Pulliainen ERS1/2 Finland Interferometry, Empirical model
et al. (2003)
Santoro et al. ERS-1/2 Kattbole, Sweden; Tuusula, Finland; Thüringer Wald, Interferometric water cloud model (IWCM)
(2003) Germany; Bois de Boulogne, France
Treuhaft et al. C-band radar Central Oregon Leaf area density (LAD) model
(2003) interferometry,
hyperspectral
Santos et al. Airborne SAR Tápajos River region, Pará state, Brazil Interferometric and polarimetric analysis,
(2004) regression analysis
Rauste (2005) SEASAT, JERS and Selected sites in Sweden, Germany, Finland and Regression analysis
airborne AIRSAR Africa
sensor
Jha et al. ENVISAT ASAR Western Ghats, Karnataka, India Regression analysis
(2006)
Hyde et al. LiDAR, SAR/ Sierra Nevada, California, USA Regression analysis
(2006) InSAR, ETM?,
QB
Lucas et al. AirSAR, LiDAR Queensland, Australia Water cloud model (WCM)
(2006)
Santoro et al. JERS-1 SAR Kattbole, Sweden; Tuusula, Finland; Bolshe- Radiative transfer model
(2006) Murtinsky, Siberia
Kumar (2007) EnviSat ASAR, Dudhwa National Park, India Regression analysis
Landsat ETM
Amini and ALOS AVNIR-2, Northern forests of Iran Neural network
Sumantyo PRISM, JERS-1
(2009) SAR
Neumann PolInSAR data Oberpfaffenhofen, Germany PolInSAR–RVoG modeling, Inversion
(2009) approaches, Model-based polarimetric
decomposition
Kumar (2009) Envisat ASAR Dudhwa National Park, India Interferometric water cloud model (IWCM)
Becek (2009) SRTM Nerang State Forest, Australia; Brunei Darussalam; Canopy gap modeling, Interferometry
Kalimantan, Indonesia; Washington State, USA;
Bavaria, Germany

123
 Int. J. Environ. Sci. Technol.

Table 2 continued
Study Sensor/datasets Study site Models

Nizalapur DLR ESAR Rajpipla, Gujarat, India Regression analysis

et al. (2010)
Yu et al. SRTM, Landsat Maine, USA Biomass algorithms
(2010) ETM, NED
Gama et al. OrbiSAR-1 São Paulo State, Brazil Regression analysis
(2010)
Fatoyinbo and PolInSAR data Mangrove forests, Nigeria Regression analysis
Armstrong
(2010)
Alappat et al. E-SAR Chandrapur Forest Division, Maharashtra, India Regression analysis
(2011)
Le Toan et al. BIOMASS SAR Mawas region, Indonesia; Les Landes, France Regression analysis
(2011)
Hamdan et al. ALOS PALSAR Forest Research Institute Malaysia, Malaysia Regression analysis
(2011)
Wollersheim POLSAR data Petawawa Research Forest, Ontario, Canada Polarimetric analysis
et al. (2011)
Englhart et al. ALOS PALSAR, Central Kalimantan, Borneo, Indonesia Multiple linear regression (MLR), Artificial
(2012) TerraSAR-X neural network (ANN), Support vector
regression (SVR)
Sambatti et al. Airborne X- and Paragominas region, Pará, Brazil Regression analysis
(2012) P-band
interferometry data
Antropov et al. ALOS PALSAR Central Finland Semiempirical forest model, model inversion
(2013)
Carreiras et al. ALOS PALSAR Mozambique, Africa BagSGB model
(2013)
Ghasemi et al. ALOS PALSAR, Temperate deciduous forest Wavelet analysis
(2013) ALOS AVNIR
Hame et al. ALOS PALSAR, Lao PDR, Laos Regression analysis, probability method
(2013) ALOS AVNIR
Peregon and ALOS PALSAR Western Siberia Regression analysis, Water cloud model
Yamagata (WCM)
(2013)

from lower frequencies to 160 Mg/ha, whereas the capa- sensitivities to the environmental conditions (Luckman
bility range of P band is from 100 Mg/ha to 200 Mg/ha et al. 1997; Kurvonen et al. 1999; Sun et al. 2002; Lucas
(Nizalapur et al. 2010). Depending on the forest type, et al. 2006). According to Hoekman and Quinones (1997),
specifically for the boreal forests, L band saturates at about the combination of C and L bands has greater potential than
100–150 t/ha (Shugart et al. 2010), while P band is found using any one of the bands for biomass estimation.
to be sensitive to forest biomass up to a saturation level of
100–300 t/ha (GTOS 2009). The problem of saturation is Biomass estimation methods
dealt by Kasischke et al. (1997) where it is mentioned that
the saturation point is higher for longer wavelengths and The two most widely used approaches used for forest
HV cross-polarization has the maximum sensitivity, while biomass estimations are (1) using backscatter values and
VV co-polarization is least sensitive. Based on the empir- (2) interferometry technique, along with polarimetric
ical relationships between AGB and SAR backscatter, analyses. Detailed list of studies related to biomass esti-
Lucas et al. (2006) established that C, L and P band satu- mation using these techniques is summarized in Table 2.
rate at different levels with even stronger relationships at
higher incidence angles and a larger dynamic range and Biomass estimation using backscatter
consistency of relationships at HV polarizations. The most
worthy band for biomass estimation is the L band as it Regression analysis is the most preferred method for bio-
interacts more with the trunk and branches with minimal mass estimation relating backscatter values to field biomass

123
Int. J. Environ. Sci. Technol.

measurement. This has been tested on various sites of The approach involving interferometry has the potential
coniferous forests of North Florida (USA) and Landes to overcome the saturation problem demonstrated by
(France) with accurate results (Le Toan et al. 1992; Dobson Fransson et al. (2001) as it has a relatively high saturation
et al. 1992). Beaudoin et al. (1994) observed relationship point. The approach has the potential to yield more reliable
between VV and HV backscatter returns with crown bio- results than the traditional single-image approach. In con-
mass, while HH return was linked to both trunk and crown trast, Pulliainen et al. (2003) proposed that the accuracy of
biomass. Using JERS-1 for assessing biomass of regener- this technique was highly dependent on certain factors
ating forests reveals the potential for AGB estimation for which inherit dynamic variation including the site condi-
young forests as well (Kuplich et al. 2000). Longer wave- tions (wind speed, moisture, temperature, etc.). It was
lengths (L and P bands) with cross-polarizations (HV and observed that the estimation accuracy can enhance while
VH) produced better result for biomass-related studies than using multi-temporal SAR images acquired under favor-
short wavelengths (X and C bands) with co-polarizations able conditions. This approach produced relative good
(HH or VV) (Le Toan et al. 1992; Dobson et al. 1992; Lucas results in several studies of boreal forests (Fransson et al.
et al. 2006; Wollersheim et al. 2011; Hamdan et al. 2011). 2001; Pulliainen et al. 2003). Pulse coherent SAR operat-
The significant problem with this method is the satura- ing between 80–120 MHz can be used to measure heavy
tion level of different wavelengths and polarizations rele- forest biomass (Imhoff et al. 2000). Luckman et al. (1997)
vant in several studies. The factors on which the saturation compared the accuracy for biomass estimation using this
levels depend are the wavelengths (different for SAR technique in boreal, temperate and tropical forests, where
bands), polarization (co- and cross-polarizations) and the the best result was obtained in boreal forests using L-band
vegetation stand structure and ground condition charac- images with 1-day interval. Tree heights were estimated
teristics (Ghasemi et al. 2011). Santos et al. (2002) indi- using L band (Mette et al. 2004) and X and P band (Santos
cated the use of the ratio between C and L or P band to et al. 2004) interferometry with improved accuracy in
solve the saturation problem. Likewise, Hoekman and comparison to the backscatter approach. InSAR technique
Quinones (1997) suggested the use of combined C and L can be combined with corresponding hyperspectral optical
bands in coniferous forest. Ranson et al. (1997) applied a remote sensing and LiDAR that can augment the vertical-
synergic approach of forest succession and radar back- structure estimates following the biophysical parameters
scatter models to determine forest biomass and observed including the biomass (Treuhaft et al. 2004). Classification
reasonably good results when biomass was \15 kg/m2. of biomass estimation methods can be grouped in four
Landscape properties like topography, surface water and approaches depending upon the type of data (or sensors)
forest structure are beneficial for estimating forest biomass and the techniques used (Fig. 1).
using SAR showing more accurate results (Austin et al.
2003). Lucas et al. (2010) emphasized on the differences in
surface moisture conditions and vegetation structure for Limitations and uncertainties
developing AGB retrieval algorithms and concluded that
PALSAR (L-band) data acquired during minimal surface Remote sensing systems provide variety of data acquisition
moisture and rainfall showed better estimation of woody modes, such as in spectral, radiometric, spatial and tem-
vegetation AGB over Queensland, Australia. poral resolutions and in polarization and angularity. Rec-
ognizing and understanding the strengths and weaknesses
Biomass estimation using interferometry of different types of sensor data are essential for selecting
suitable sensor data for AGB estimation in a specific study
Interferometry is a technique based on interference of area. In rugged or mountainous regions, topographic fac-
waves. Interference, in physics, is a phenomenon where tors such as slope and aspect can considerably affect veg-
two waves superimpose resulting in a wave of greater or etation reflectance, resulting in spurious relationships
lower amplitude. It refers to the interaction of waves that between AGB and backscattering values. Hence, removal
are correlated with each other, either due to their same of topographic effects is necessary. Approaches have been
source of origin or due to same or nearly same frequency. developed for topographic correction of SAR data (Soja
Interference can either be constructive or destructive et al. 2010). The limitation in spatial and radiometric res-
depending upon the phase difference between the waves. olutions inherent in the remotely sensed data is an impor-
Interferometric synthetic aperture radar (InSAR or IfSAR) tant factor affecting the AGB estimation performance. The
is a radar technique used in geodesy and remote sensing. remote sensors with coarse spatial resolution mainly cap-
This technique simultaneously uses two or more SAR ture canopy information, instead of individual tree infor-
images using differences in the phase of the waves mation. Different sensor data have their own characteristics
returning to the sensor. in reflecting land surfaces, and thus, integration of different

123
 Int. J. Environ. Sci. Technol.

LEVEL 1 LEVEL 2 LITERATURE

Single date Santoro et al. 2006; Yu et al. 2010; Le Toan et al. 2011; Hamdan et al. 2011;
Englhart et al. 2012; etc.
Approach 1: Based
on data type Fransson et al. 2001; Santoro et al. 2003; Kumar 2009; etc.
Multi-temporal/
data pairs

Nizalapur et al. 2010; Hamdan et al. 2011; Wollersheim et al. 2011; Carreiras
Single et al. 2013; etc.

Approach 2: Based
on sensor type Hyde et al. 2006; Saatchi et al. 2007; Amini and Sumantyo 2011; Englhart et
Multiple/
al. 2012; etc.
Synergic

Kurvonen et al. 1999; Rauste 2005; Santoro et al. 2006; Kumar 2007;
Backscatter Neumann 2009; Nizalapur et al. 2010; Englhart et al. 2012; Carreiras et al.
2013; etc.

InSAR Fransson et al. 2001; Becek 2009; etc.

PolSAR Wollersheim et al. 2011; etc.

Approach 3: Based
on SAR processing Pol-InSAR Gama et al. 2010 ; Le Toan et al. 2011; etc.

Backscatter+
Santoro et al. 2003; Kumar 2009; etc.
InSAR

Other: Texture
Amini and Sumantyo 2009, 2011; Ghasemi et al. 2013; etc.
measures,
wavelet
analysis, etc.

Linear- Fransson et al. 2001; Hyde et al. 2006; Kumar 2007; Gama et al. 2010;
regression Rahman and Sumantyo 2012; Sambatti et al. 2012; Peregon and Yamagata
models 2013; Hame et al. 2013 ; etc.
Approach 4: Based
on modeling
Non-linear Kurvonen et al. 1999; Santoro et al. 2006; Kumar 2009; Amini and Sumantyo
2009 ; Wollersheim et al. 2011; Carreiras et al. 2013; Ghasemi et al. 2013; etc.
complex models

Fig. 1 Classification of biomass estimation methods

sources of remotely sensed data may enhance the infor- assume errors to be independent and spatial units to be
mation extraction process. The integration of radar and dependent.
optical sensor data has the potential to improve AGB The five possible sources of uncertainty in remote
estimation because it may reduce the mixed pixels and data sensing analysis are described by Dungan (2002) as: (1)
saturation problems (Ban 2003). variable uncertainty, (2) spatial support uncertainty, (3)
The last decade witnessed some of the major milestone positional uncertainty, (4) model uncertainty and (5)
researches targeting the subject of uncertainty in remote parametric uncertainty.
sensing, GIS, spatial models and geographical sciences The geostatistical term ‘support’ refers to the spatial
(Congalton 1991; Sinha et al. 2012; Liang et al. 2013; Shen resolution. Moreover, these variables also have a temporal
et al. 2015). The three challenges identified by Dungan resolution. SAR remote sensing is no exception regarding
(2002) include (a) the unavailability of appropriate refer- the issue of uncertainty in the analysis in addition to other
ence data due to expensive and time-consuming issues, complexities and ambiguities that are specific to SAR
(b) challenge to represent or visualize spatially varying techniques. Several other limitations of SAR are as fol-
intervals and (c) inadequacy of statistical models that lows: (1) costly data, (2) temporal repeativity, (3) few

123
Int. J. Environ. Sci. Technol.

satellite constellation, (4) no time-composite data as the certain parameters. For example, the canopy height can be
case for optical data, (5) limited area coverage and (6) non- calculated more accurately using laser or LiDAR technol-
availability of global-level coherent datasets of SAR. ogy. Zolkos et al. (2013) proved that airborne LiDAR-
Refinements in handling and processing SAR data can generated AGB models are more accurate than those
improve the analyses which are promising prospects for developed from radar or passive optical data. This is
future researches. attributed to the strong relationship of the LiDAR systems
Finally, it is often difficult to transfer one model to the biomass level even beyond 1,000 t/ha that greatly
developed in a specific study area to other study areas exceeds the normal saturation level of passive optical or
because of the environmental characteristics apart from the radar sensors (Yavasli 2012). Variables that are strongly
limitation of the model itself and the nature of remotely correlated should be used for developing algorithms, and
sensed data. Foody et al. (2003) discussed the problems SAR backscatter coefficients often have shown better
encountered in model transfer. Many factors, such as relationship than optical-derived parameters. However, this
uncertainties in the remotely sensed data (image prepro- selection is a complex process that requires good under-
cessing and different stages of processing), AGB calcula- standing of the interactions among the tested variables and
tion based on the field measurements, the disparity between forest structural and biophysical attributes. Selection of
remote sensing acquisition date and field data collection satellite sensors is hence vital for AGB estimation.
and the size of sample plot compared with the spatial Integration of optical and radar data can reduce the data
resolution of remotely sensed data, could affect the success saturation in optical sensor images. Development of
of model transferability. advanced models using multi-source data is important for
In addition to the sources of uncertainty in remote reducing uncertainties. Methods of data integration, devel-
sensing analysis, there are several possible loopholes for opment of advanced models, better selection and quanti-
uncertainties to slink in biomass assessment through fication of variables are hotspots for future research to
remote sensing. The key prerequisite for developing AGB improve current technology for biomass estimation by use
estimation models is the availability of a high-quality data of radar data. Interferometry and polarimetry are the two
source. Allometric equations for calculating AGB requires radar-based techniques that need to be explored further in
DBH (diameter at breast height) and/or height, which is a AGB studies. Improvements in SAR data resolution are
source of uncertainty (Keller et al. 2001; Ketterings et al. also a matter of research. Using digital beam-forming
2001). Mode of AGB sample collection is another source techniques, wide-swath coverage can be obtained without
of ambiguity (Keller et al. 2001). Satellite, ancillary and degrading the spatial resolution. Hence, a new generation
sample data need to be co-registered accurately for AGB spaceborne SAR systems is planned using this technology,
calculation (Lu 2006). The important sources of uncer- for example, TanDEM-L shall cover a swath width of
tainties in AGB enumeration can be summarized as: 350 km with spatial resolution of 10 m, and the revisit time
will be 8 days (Ouchi 2013). As already mentioned,
1. Mode of AGB sample data collection
interferometry gives information on height and research
2. Selection of proper sample data collection sites
is required for further improvement in this. Combined
3. Atmospheric corrections
polarimetric and interferometric SAR (PolInSAR) has a
4. Registration errors between satellite data and AGB
greater potential for calculating biomass at higher densities
sample data
(Yavasli 2012). With the advent of different SAR missions
5. Incompetence between satellite image pixel size and
in future like follow-up of ALOS and RISAT, NISAR,
sample plot dimensions
MAPSAR, etc., there are further possibilities of improve-
6. Selection of suitable remote sensing-derived variables
ment in this technology for AGB estimation.
to derive relationship with field inventorized AGB
7. Algorithms and equations for developing AGB esti-
mation models Conclusion
The level of uncertainty and the evaluation of the model
performance can be accounted by the coefficient of Estimation of AGB in tropical forests are difficult to carry
determination (R2) and root-mean-squared error (RMSE) out due to high dynamism of these forests as well as
(Deepika et al. 2014). A high R2 or low RMSE value tedious accessibility conditions. Although use of multi-
denotes good-fit between the sample data and the model sensor or multi-resolution data has the potential to improve
developed. Most of the earlier AGB studies suffered from AGB estimation performance, the time and labor involve-
difficulties in collection of field data and resulting incon- ment in image processing will be significantly increased.
sistencies between field measurements and AGB estima- The economic factor will be an important aspect in the use
tion. However, the accuracy can enhance on improving of multi-source remotely sensed data in a large area. In

123
 Int. J. Environ. Sci. Technol.

spite of these limitations, SAR provides an efficient means Congalton RG (1991) A review of assessing the accuracy of
for assessment of AGB and it can overcome important classifications of remotely sensed data. Remote Sens Environ
37:35–46
limitations of optical remote sensing. High wavelength and Deepika B, Avinash K, Jayappa KS (2014) Shoreline change rate
cross-polarization in SAR are more sensitive to biomass. estimation and its forecast: remote sensing, geographical infor-
Texture analysis along with backscatter, interferometry and mation system and statistics-based approach. Int J Environ Sci
polarimetric analysis of SAR improves the estimation of Technol 11(2):395–416
Dobson MC, Ulaby FT, Le Toan T et al (1992) Dependence of radar
biomass. Regression analysis remained the most common, backscatter on coniferous forest biomass. IEEE Trans Geosci
effective and easy-to-use technique for biomass estimation. Remote Sens 30:412–416
Model-based approaches using semiempirical models Dungan JL (2002) Toward a comprehensive view of uncertainty in
including radiative transfer models, WCM and IWCM are remote sensing analysis. In: Foody GM, Atkinson PM (eds)
Uncertainty in Remote Sensing and GIS. Wiley, West Sussex,
used for AGB estimations using SAR. Though LiDAR act pp 25–35
as the most suitable single sensor for biomass estimation, Englhart S, Keuck V, Siegert F (2012) Modeling aboveground
the synergic use of optical and SAR would be the obvious biomass in tropical forests using multi-frequency SAR data—a
choice due to their easy availability, cheaper cost and time comparison of methods. IEEE J Sel Top Appl Earth Obs Remote
Sens 5(1):298–306. doi:10.1109/JSTARS.2011.2176720
of processing. Synergic use of multi-temporal optical, FAO (2001) Global forest resources assessment 2000—main report.
microwave (SAR/InSAR/PolInSAR) and LiDAR can FAO Forestry Paper 140, Food and Agriculture Organization of
potentially be the best combination in remote sensing the United Nations, Rome, pp 363
integrated with the use of model-based approach (semi- Fatoyinbo TE, Armstrong AH (2010) Remote characterization of
biomass measurements: case study of mangrove forests. In:
empirical) for the estimation of biomass. Overall, it can be Momba M, Bux F (eds) biomass. InTech Publishers, Croatia
concluded that AGB can be estimated with reliable accu- Field CB, Buitenhuis ET, Ciais P et al (2007) Contributions to
racy using backscatter intensity values, polarimetric and accelerating atmospheric CO2 growth from economic activity,
interferometric techniques. carbon intensity, and efficiency of natural sinks. Proc Nat Acad
Sci USA (PNAS) 104:18866–18870
Foody GM, Boyd DS, Cutler MEJ (2003) Predictive relations of tropical
Acknowledgments The authors express sincere gratitude to the forest biomass from Landsat TM data and their transferability
editor and reviewers for constructive comments and suggestions to between regions. Remote Sens Environ 85:463–474
improve this paper. The authors wish to acknowledge the support Fransson JES, Smith G, Askne J, Olsson H (2001) Stem volume
from Department of Science and Technology (DST), Government of estimation in boreal forests using ERS-1/2 coherence and SPOT
India, for providing funds under DST/INSPIRE Program (Ref. No. XS optical data. Int J Remote Sens 22(14):2777–2791
DST/INSPIRE FELLOWSHIP/2010/[316]). Gama FF, Santos JR, Mura JC (2010) Eucalyptus biomass and
volume estimation using interferometric and polarimetric SAR
data. Remote Sens 2:939–956
Ghasemi N, Sahebi MR, Mohammadzadeh A (2011) A review on
References biomass estimation methods using synthetic aperture radar data.
Int J Geomat Geosci 1(4):776–788
Alappat VO, Joshi AK, Krishnamurthy YVN (2011) Tropical dry Ghasemi N, Sahebi MR, Mohammadzadeh A (2013) Biomass
deciduous forest stand variable estimation using SAR data. estimation of a temperate deciduous forest using wavelet
J Indian Soc Remote Sens 39(4):583–589 analysis. IEEE Trans Geosci Remote Sens 51(2):765–776
Amini J, Sumantyo JTS (2009) Employing a method on SAR and Gibbs HK, Brown S, Niles JO, Foley JA (2007) Monitoring and
optical images for forest biomass estimation. IEEE Trans Geosci estimating tropical forest carbon stocks: making REDD a reality.
Remote Sens 47(12):4020–4026 Environ Res Lett 2:1–13
Antropov O, Rauste Y, Ahola H, Hame T (2013) Stand-level stem GTOS (Global Terrestrial Observing System) (2009) Biomass—
volume of boreal forests from spaceborne SAR imagery at assessment of the status of the development of the standards for
L-band. IEEE J Sel Top Appl Earth Obs Remote Sens the terrestrial essential climate variables. Rome, p 18. http://
6(1):35–44. doi:10.1109/JSTARS.2013.2241018 www.fao.org/gtos/doc/ECVs/T12/T12.pdf
Austin JM, Mackey BG, van Niel KP (2003) Estimating forest Hamdan O, Aziz HK, Rahman KA (2011) Remotely sensed L-band
biomass using satellite radar: an exploratory study in a SAR data for tropical forest biomass estimation. J Trop For Sci
temperate Australian Eucalyptus forest. For Ecol Manag 23(3):318–327
176:575–583 Hame T, Rauste Y, Antropov O, Ahola HA, Kilpi J (2013) Improved
Ban Y (2003) Synergy of multitemporal ERS-1 SAR and Landsat TM mapping of tropical forests with optical and SAR imagery, Part
data for classification of agricultural crops. Can J Remote Sens II: above ground biomass estimation. IEEE J Sel Top Appl Earth
29(4):518–526 Obs Remote Sens 6(1):92–101
Beaudoin A, Le Toan T, Goze S et al (1994) Retrieval of forest Herold M, Brady M, Wulder M, Kalensky D (2007) Biomass ECV
biomass from SAR data. Int J Remote Sens 15:2777–2796 report. ftp.fao.org/docrep/fao/011/i0197e/i0197e16.pdf
Becek K (2009) Biomass representation in synthetic aperture radar Hoekman DH, Quinones MJ (1997) Land cover type and forest
interferometry data sets. Dissertation, The University of Brunei biomass assessment in the Colombian Amazon. In: Geoscience
Darussalam, Brunei and remote sensing, 1997. IGARSS ‘97. Remote sensing—a
Carreiras JMB, Melo JB, Vasconcelos MJ (2013) Estimating the scientific vision for sustainable development. 1997 IEEE Inter-
above-ground biomass in Miombo savanna woodlands (Mozam- national. IEEE IGARSS 4:1728–1730
bique, East Africa) using L-band synthetic aperture radar data. Houghton RA (2005) Aboveground forest biomass and the global
Remote Sens 5:1524–1548. doi:10.3390/rs5041524 carbon cycle. Global Change Biol 11:945–958

123
Int. J. Environ. Sci. Technol.

House JI, Prentice IC, Ramankutty N, Houghton RA, Heimann M Lucas RM, Lee AC, Bunting PJ (2008) Retrieving forest biomass
(2003) Reconciling apparent inconsistencies in estimates of through integration of CASI and LiDAR data. Int J Remote Sens
terrestrial CO2 sources and sinks. Tellus 55B:345–363 29(5):1553–1577
Husch B, Beers TW, Kershaw JA (2003) Forest mensuration, 4th edn. Lucas RM, Armston J, Fairfax R et al (2010) An evaluation of the ALOS
Wiley, New Jersey PALSAR L-band backscatter—above ground biomass relation-
Hyde P, Dubayah R, Walker W et al (2006) Mapping forest structure ship Queensland, Australia: impacts of surface moisture condition
for wildlife habitat analysis using multi-sensor (LiDAR, SAR/ and vegetation structure. IEEE J Sel Top Appl Earth Obs Remote
InSAR, ETM?, Quickbird) synergy. Remote Sens Environ Sens 3(4):576–593. doi:10.1109/JSTARS.2010.2086436
102:63–73 Luckman A, Baker JR, Kuplich TM, Yanasse CCF, Frery AC (1997)
Imhoff ML, Johnson P, Holford W et al (2000) BioSar (TM): an A study of the relationship between radar backscatter and
inexpensive airborne VHF multiband SAR system for vegetation regenerating forest biomass for space borne SAR instrument.
biomass measurement. IEEE Trans Geosci Remote Sens Remote Sens Environ 60:1–13
38(3):1458–1462 Malhi YP (2002) Forests, carbon and global climate. Phil Trans R Soc
Jha CS, Rangaswamy M, Murthy MSR, Vyjayanthi N (2006) Lond A 360:1567–1591
Estimation of forest biomass using Envisat-ASAR data. Proc Mette T, Papathanassiou K, Hajnsek I (2004) Biomass estimation
SPIE 6410:641002 from polarimetric SAR interferometry over heterogeneous forest
Kasischke ES, Melack JM, Dobson MC (1997) The use of imaging terrain. In: Geoscience and remote sensing symposium
radars for ecological applications—a review. Remote Sens (IGARSS), 2004 IEEE International. Anchorage, AK. IEEE
Environ 59:141–156 IGARSS 1:511–514
Keller M, Palace M, Hurtt G (2001) Biomass estimation in the Nabuurs GJ, Masera O, Andrasko K et al (2007) Forestry. In: Metz B,
Tapajos National Forest, Brazil: examination of sampling and Davidson OR, Bosch PR, Dave R, Meyer LA (eds) Climate
allometric uncertainties. For Ecol Manage 154:371–382 change 2007: mitigation. Contribution of working group III to
Ketterings QM, Coe R, van Noordwijk M, Ambagau K, Palm CA the fourth assessment report of the intergovernmental panel on
(2001) Reducing uncertainty in the use of allometric biomass climate change. Cambridge University Press, Cambridge
equations for predicting aboveground tree biomass in mixed Neumann M (2009) Remote sensing of vegetation using multi-
secondary forests. For Ecol Manage 146:199–209 baseline polarimetric SAR interferometry: theoretical modeling
Kumar NR (2007) Forest cover, stand volume and biomass assess- and physical parameter retrieval. Dissertation, University of
ment in Dudhwa National Park using satellite remote sensing Rennes 1, France
data (optical and EnviSat ASAR). Dissertation, Andhra Univer- Nizalapur V, Jha CS, Madugundu R (2010) Estimation of above
sity, India ground biomass in Indian tropical forested area using multifre-
Kumar S (2009) Retrieval of forest parameters from Envisat ASAR quency DLR-ESAR data. Int J Geomat Geosci 1(2):167–178
data for biomass inventory in Dudhwa National Park, UP, India. Ouchi K (2013) Recent trend and advance of synthetic aperture radar
Dissertation, IIRS, Dehradun, India and ITC, Enschede, with selected topics. Remote Sens 5:716–807. doi:10.3390/
Netherlands rs5020716
Kumar P, Sharma LK, Pandey PC, Sinha S, Nathawat MS (2013) Patenaude GM, Milne R, Dawson TP (2005) Synthesis of remote
Geospatial strategy for tropical forest-wildlife reserve biomass sensing approaches for forest carbon estimation: reporting to the
estimation. IEEE J Sel Top Appl Earth Obs Remote Sens Kyoto Protocol. Environ Sci Policy 8:161–178
6(2):917–923. doi:10.1109/JSTARS.2012.2221123 Peregon A, Yamagata Y (2013) The use of ALOS/PALSAR
Kuplich TM, Salvatori V, Curran PJ (2000) JERS-1/SAR backscatter backscatter to estimate above-ground forest biomass: a case
and its relationship with biomass of regenerating forests. Int J study in Western Siberia. Remote Sens Environ 137:139–146
Remote Sens 21:2513–2518 Plugge D, Baldauf T, Ratsimba HR, Rajoelison G, Köhl M (2010)
Kurvonen L, Pulliainen J, Hallikainen M (1999) Retrieval of biomass Combined biomass inventory in the scope of REDD (reducing
in boreal forests from multitempotal ERS-1 and JERS-1 SAR emissions from deforestation and forest degradation). Madag
images. IEEE Trans Geosci Remote Sens 37(1):198–205 Conserv Dev 5:23–34
Le Toan TB, Beaudoin A, Riom J, Guyon D (1992) Relating forest Pulliainen JT, Engdahl M, Hallikainen M (2003) Feasibility of multi-
biomass to SAR data. IEEE Trans Geosci Remote Sens temporal interferometric SAR data for stand-level estimation of
30(2):403–411 boreal forest stem volume. Remote Sens Environ 85:397–409
Le Toan T, Quegan S, Davidson MWJ et al (2011) The BIOMASS Ranson KJ, Sun G (1994) Mapping biomass of a northern forest using
mission: mapping global forest biomass to better understand the multifrequency SAR data. IEEE Trans Geosci Remote Sens
terrestrial carbon cycle. Remote Sens Environ 115:2850–2860 32:388–396
Liang J, Zeng GM, Shen S et al. (2013) Bayesian approach to quantify Ranson KJ, Sun G, Weishampel JF, Knox RG (1997) Forest biomass
parameter uncertainty and impacts on predictive flow and mass from combined ecosystem and radar backscatter modeling.
transport in heterogeneous aquifer. Int J Environ Sci Technol. Remote Sens Environ 59:118–133
doi:10.1007/s13762-013-0453-3 Rauste Y (2005) Techniques for wide-area mapping of forest biomass
Loehle C (2000) Forest ecotone response to climate change: using radar data. Espoo 2005. VTT Publications, Finland. ISBN
sensitivity to temperature response functional forms. Can J For 951–38–6695–5
Res 30:1632–1645 Roy PS, Diwakar PG, Singh IJ, Bhan SK (1994) Evaluation of
Lu D (2005) Aboveground biomass estimation using Landsat TM microwave remote sensing data for forest stratification and
data in the Brazilian Amazon Basin. Int J Remote Sens canopy characterization. J Indian Soc Remote Sens 22(1):31–44
26:2509–2525 Sambatti JBM, Leduc R, Lübeck D, Moreira JR, Santos JR (2012)
Lu D (2006) The potential and challenge of remote sensing-based Assessing forest biomass and exploration in the Brazilian
biomass estimation. Int J Remote Sens 27(7):1297–1328 Amazon with airborne InSAR: an alternative for REDD. Open
Lucas RM, Cronin N, Lee A et al (2006) Empirical relationships Remote Sens J 5:21–36
between AIRSAR backscatter and LiDAR-derived forest bio- Santoro M, Askne J, Dammert PBG (2003) Tree height estimation
mass, Queensland, Australia. Remote Sens Environ from multi-temporal ERS SAR interferometric phase. Proceed-
100(3):407–425 ing of FRINGE 2003 Workshop, 1–5 Dec 2003, Frascati, Italy

123
 Int. J. Environ. Sci. Technol.

Santoro M, Eriksson L, Askne J, Schmullius C (2006) Assessment of Stephens BB, Gurney KR, Tans PP et al (2007) Weak northern and
stand-wise stem volume retrieval in boreal forest from JERS-1 strong tropical Land carbon uptake from vertical profiles of
L-band SAR backscatter. Int J Remote Sens 27(16):3425–3454 atmospheric CO2. Science 316:1732–1735
Santos JR, Pardi Lacruz MS, Araujo LS, Keil M (2002) Savanna and Sun G, Ranson KJ, Kharuk VI (2002) Radiometric slope correction
tropical rainforest biomass estimation and spatialization using for forest biomass estimation from SAR data in the western
JERS-1 data. Int J Remote Sens 23:1217–1229 Sayani Mountains, Siberia. Remote Sens Environ 79:279–287
Santos JR, Neeff T, Dutra LV et al (2004) Tropical forest biomass Townshend JR, Masek JG, Huang C et al (2012) Global character-
mapping from dual frequency SAR interferometry (X and ization and monitoring of forest cover using Landsat data:
P-Bands). In: Twentieth international society for photogrametry opportunities and challenges. Int J Digit Earth 5(5):373–397
and remote sensing (ISPRS) congress. GeoImagery bridging Treuhaft RN, Asner GP, Law BE (2003) Structure-based forest
continents, Istanbul, v.XXXV, pp 1133–1136 biomass from fusion of radar and hyperspectral observations.
Sharma LK, Nathawat MS, Sinha S (2013) Top-down and bottom-up Geophys Res Lett 30(9):1472. doi:10.1029/2002GL016857
inventory approach for above ground forest biomass and carbon Treuhaft RL, Law BE, Asner GP (2004) Forest attributes from radar
monitoring in REDD framework using multi-resolution satellite interferometric structure & its fusion with optical remote
data. Environ Monit Assess 185:8621–8637. doi:10.1007/ sensing. Biosci 54:561–571
s10661-013-3199-y Wiley CA (1985) Synthetic aperture radars: a paradigm for technol-
Shen Z, Xie H, Chen L, Qiu J, Zhong Y (2015) Uncertainty analysis ogy evolution. IEEE Trans Aerosp Electron Syst AES
for nonpoint source pollution modeling: implications for 21(3):440–443
watershed models. Int J Environ Sci Technol 12:739–746 Wollersheim M, Collins MJ, Leckie D (2011) Estimating boreal forest
Shugart HH, Saatchi S, Hall FG (2010) Importance of structure and its species type with airborne polarimetric synthetic aperture radar.
measurement in quantifying function of forest ecosystems. Int J Remote Sens 32(9):2481–2505
J Geophys Res 115(G2):G00E13. doi:10.1029/2009JG000993 Yavasli DD (2012) Recent approaches in above ground biomass
Sinha S, Sharma LK, Nathawat MS (2012) Tigers losing grounds: estimation methods. Aegean Geographical Journal 21(1):39–51
impact of anthropogenic occupancy on tiger habitat suitability Yu Y, Saatchi S, Heath LS et al (2010) Regional distribution of forest
using integrated geospatial-fuzzy techniques. The Ecoscan height and biomass from multisensor data fusion. J Geophys
1:259–263 115:G00E12. doi:10.1029/2009JG000995
Soja M, Sandberg G, Ulander L (2010) Topographic correction for Zolkos SG, Goetz SJ, Dubayah R (2013) A meta-analysis of terrestrial
biomass retrieval from P-band SAR data in boreal forests. In: aboveground biomass estimation using lidar remote sensing.
Geoscience and remote sensing symposium (IGARSS), 2010 Remote Sens Environ 128:289–298
IEEE International. Honolulu, HI, pp 4776–4779

123