Forest Biophysical Parameter Retrieval Using PolSAR Technique

8th International Conference on Microwaves, Antenna, Propagation & Remote Sensing ICMARS-2012, Jodhpur, INDIA, Dec. 11 – 15, 2012

Forest Biophysical Parameter Retrieval Using PolSAR Technique R. P. Amrutkar #, S. Kumar *, S.P.S. Kushuwaha* and G. D. Bhatt* #

Centre for Space Science and Technology Education in Asia and the Pacific Dehradun- 248001, India 1

[email protected] 3 [email protected] 3 [email protected]

*

Indian Institute of Remote Sensing, ISRO Dehradun- 248001, India 2

[email protected]

Abstract— Synthetic Aperture Radar (SAR) polarimetry is an uprising and nascent area of research in radar remote sensing and it has proved its potential in the area of forest biophysical parameter retrieval. The key intention of the present study was to develop SAR polarimetry based semi-empirical model for Stem volume and Aboveground biomass estimation of forest area and check its reliability in implementation to the study area for the relationships between these forest biophysical parameters and Polarimetric SAR (PolSAR) decomposition components. Five scenes of Quadpol SLC ALOS PALSAR data which has been acquired over Southern Gujarat forest area covering portions of Surat, Valsad, The Dangs and Dadra & Nagar Haveli districts in March, April and May 2010 are used. A field calculated circumference at breast height and height of trees in 0.1 hectare plots designed using stratified random sample method has been collected from Forestry and Ecology Department, IIRS. This field data was utilized for plot-wise stem volume density and aboveground biomass estimation. Coherency matrix based decomposition was carried out, yielding three components and resultant generated restoring polarimetric information. Coherency matrix was utilized for polarization orientation angle shift compensation which demonstrated acknowledgeable overestimation of volume scattering mechanism and underestimation of double bounce scattering mechanism unaltering the odd bounce scattering information after decomposition. Coherency matrix was utilized for Radar Vegetation Index and backscattered HH, HV and VV polarimetric channels generation. HV backscattered polarimetric channel and volume scattering component were found to be best correlated with forest parameters. Radar Vegetation Index was correlated with the field retrieved stem volume density and aboveground biomass resulting coefficient of determination i.e., R2 = 0.573 and R2 = 0.617 respectively, which resulted in stating moderate potential for study area into consideration. Semiempirical Water Cloud Model was implemented for execution of parameter retrieval using the field estimated stem volume and aboveground biomass which proved its potential by illustrating higher coefficient of determination. Keywords- Coherency Matrix, Polarization Orientation Angle, Radar Vegetation Index, Water Cloud Model

I. INTRODUCTION Forest parameter retrieval is a very tedious and time consuming job using traditional methods especially for dense forests and forests which are not easily accessible for human beings. In spite of the availability of radar data under any weather conditions, the retrieval of biophysical parameters has been frequently carried out using optical data, mainly because the interaction between the radar signal and the vegetation is more complex than with the optical signal. The SAR image understanding and processing is a challenge as well. It is more difficult to establish the biophysical relationships between the SAR information and the targets with which it interacts due to its mathematical calculation complexity. Thus, more work is required in the retrieval of biophysical parameters using radar satellite data which have been limited by the type of data. Hence, SAR polarimetry is being implemented for forest parameter estimation. Previous studies have analysed the potential of single and dual polarized SAR data in stem volume density (SVD) and aboveground biomass estimation (AGB) [27], [28]. Present study intends to study the potential of SAR polarimetry in SVD and AGB retrieval. SAR has operational advantages over optical sensors for rapid disaster assessment because of its day/night acquisition capability, the ability to “see through” smoke, clouds and dust, and the sidelooking viewing geometry, which is an advantage as whenever data collection directly above the site would prove dangerous [16]. Integration of radar and optical sensor data also demonstrated the potential to improve AGB estimation as it reduced the mixed pixels and data saturation problems [23]. For accomplishment of the study purpose, SAR fully polarimetric data in Single Look Complex (SLC) format has to be decomposed for extraction of the polarimetric information in all scattering mechanisms like Braggs surface scattering, double bounce scattering and Fresnel’s volume scattering. The slope variability of terrain, the structure and the shape of different scatterers under radar beam encounter the orientation angle shift which gives rise to over estimation of volume scattering and hence forest types. Huynen (1970) is the pioneer of orientation angle. The orientation angle is

128

8th International Conference on Microwaves, Antenna, Propagation & Remote Sensing ICMARS-2012, Jodhpur, INDIA, Dec. 11 – 15, 2012

defined as the angle between major axis of the ellipse and the horizontal axis. The polarization orientation angle shift is the angle of rotation about the line of sight. In Polarimetry, orientation angle can be used directly to compensate POLSAR data in rugged terrain areas. Due to presence of distributed targets each pixel of the polarimetric data has been incorporated with various orientation angles. Hence, deorientation is an obligation. For horizontal medium these shifts are zero and for objects which are not horizontal to the line of sight, the orientation angle produces some shift [10], [13]. These shifts reduce the size of the physical scattering area leading to error in the power gained by the receiver [6]. These shifts result in increased cross-polarized intensity. This leads to increase in the volume scattering power and wrong interpretation on the observed information [12], [33]. So to estimate the accurate scattering power of the target scatterers the compensation in polarization orientation shift before carrying out decomposition using coherency matrix plays key role. The estimation process includes polarization orientation angle compensation before decomposition and then modeling of the forest area with the help of Water Cloud Model (WCM). Previous studies on semi-empirical modeling approaches for SVD and AGB estimation were carried out with the help of single- or dual-pol SAR data which includes backscatter amount contributed by ground and vegetation. Retrieval of backscatter amount from ground and vegetation involves rigorous and iterative mathematical procedure with uncertain results. The present study intends to retrieve the SVD and AGB of forest area with the help of SAR polarimetry based semi-empirical modeling. Previous studies elaborate usage of different types of models for forest parameter retrieval such as the WCM [11], [29], [30], Interferometric Water Cloud Model (IWCM) [11], [20], [22], [24], Michigan Microwave Canopy Scattering (MIMICS) [8], [18], [19] etc. Amongst which WCM is found to be one of the most reliable and suitable model for forest biophysical parameter retrieval which has been so far implemented for single and dual polarized SAR data [26]. This paper is focused on implementing the semi-empirical model on the decomposition components of quad- pol SAR data. II. STUDY AREA Five scenes of ALOS PALSAR Quad-Pol data in SLC format comprising Gujarat and Dadra & Nagar Haveli Forest area has been used for present study. A. Dadra & Nagar Haveli Forest Area The Dadra & Nagar Haveli forest area has been covered in ALOS PALSAR scene having scene centre longitude 73.1335960 E and latitude 20.169129 N which is being acquired on 31st March 2010. There are in all 9 forest field plots which have been covered in this scene. Among these 4 plots are of Dry Deciduous Forest, 3 are of Moist Deciduous Forest and 2 are of Scrub Forest type. Circumference at breast height (CBH) and tree height was measured for all 9 field

plots encompassing total 383 trees of 51 species and 30 different families. The forest cover in Union Territory (UT), based on – screen visual interpretation of satellite data of November 2008, is 211 km2 which is around 42.97% of its total geographical area. Forest type mapping using the satellite data has been undertaken by the Survey of India with reference to Champion and Seth Classification (1968) [9]. As per this classification, the UT has four forest types which belong to two groups, namely Tropical Moist Deciduous and Tropical Dry Deciduous. Dadra & Nagar Haveli forest plays a major role in the business and economy of Dadra & Nagar Haveli. The forest of Dadra & Nagar Haveli is rich in economic resources and is furnished with a variety of species like: teak (Tectona grandis), khair (Acacia catechu), shisham (Dalbergia sissoo), mahua (Madhuca longifolia var. latifolia) and sadad (Terminalia crenulata). It is also rich in the underground growth of nirgudi (vitex negundo), ukshi, karamda (Carissa carandas), and bamboo (Dendrocalamus strictus). B. Gujarat Forest Area The Gujarat forest area has been covered in four ALOS PALSAR scenes out of which scene having scene centre longitude 73.5650020 E and latitude 20.6658920 N, which is being acquired on 29th April 2010 comprises of 3 forest plots. ALOS PALSAR scene having scene centre longitude 73.4577020 E and latitude 21.1626920 N, which is being acquired on 29th April 2010 comprises of 1 forest plot. ALOS PALSAR scene having scene centre longitude 73.3506520 E and latitude 20.2099080 N, which is being acquired on 16th May 2010 comprises of 7 forest plots and ALOS PALSAR scene having scene centre longitude 73.2476690 N and latitude 20.7074000 E, which is being acquired on 16th May 2010 comprises of 3 forest plots respectively. 2 field plots of Dry Deciduous forest, 1 field plot of Eucalyptus plantation and 1 field plot of Forest plantation has been covered in Valsad District. CBH and tree height was measured for all 4 field plots encompassing total 83 trees of 21 species and 17 different families. 2 field plots of Dry Deciduous forest, 1 field plot of Moist Deciduous Forest has been covered in Dang District. CBH and tree height were measured for all 3 field plots encompassing total 43 trees of 16 species and 10 different families. 7 field plots of Dry Deciduous forest has been covered in Surat District. CBH and tree height was measured for all 7 field plots encompassing total 223 trees of 31 species and 20 different families. Total geographical area of Gujarat is 1, 96,022 km2 (6.0% of country). Total Forest area is 19,113 km2 which is 9.75% state’s geographic area and 2.47% of country’s forest area. District-wise Surat comprises of 1,354 km2, The Dangs comprises of 1,417 km2 and Valsad comprises of 995 km2 forest area. The main forest types in the South Gujarat district viz., Surat, The Dangs, Valsad and Dadra & Nagar Haveli are: Moist teak forest, Southern moist mixed deciduous forest, Southern secondary moist mixed deciduous forest and scrub

129

8th International Conference on Microwaves, Antenna, Propagation & Remote Sensing ICMARS-2012, Jodhpur, INDIA, Dec. 11 – 15, 2012

forest (Champion & Seth, 1968). Owing to the arid climate, the flora and fauna of South Gujarat also possess a wide range of xerophytic vegetation [25]. This kind of vegetation includes Acacia Arabica, Acacia leucophloea, Capparis ophylla, Zizyphus mauratiana, etc. The main tree species are: teak (Tectona grandis), sadad (Terminalia crenulata), shisham (Dalbergia sissoo), khair (Acacia catechu), timru (Diospyros melanoxylon), mahuda (Madhuca longifolia var. latifolia), dhavdo (Anogeissus latifolia), khakhar (Butea monosperma), kalam (Mitragyna parvifolia), bondarao (Lagerstroemia parviflora), billi (Aegle marmelos), moina (Lannea coromandelica), etc [2], [3] III. DATA USED PALSAR was a joint project between Japan Aerospace Exploration Agency (JAXA) and the Japan Resources Observation System Organization (JAROS). ALOS was launched on January 24, 2006 from Tanegashima Space Center with 46 days repetivity. Finally, ALOS mission was completed in May 12, 2011. The ALOS carried three remotesensing instruments: the Panchromatic Remote-sensing Instrument for Stereo Mapping (PRISM), the Advanced Visible and Near Infrared Radiometer type 2 (AVNIR-2) for precise land coverage observation, and the Phased Array type L-band Synthetic Aperture Radar (PALSAR) for day-andnight and all-weather land observation. The Phased Array type L-band Synthetic Aperture Radar (PALSAR) is an active microwave sensor using L-band frequency to achieve cloudfree and day-and-night land observation. Modes of acquisitions were Fine, ScanSAR and Polarimetric. Among which data acquired in polarimetric mode is quad-pol which has been used for current study. Total 5 scenes of ALOS PALSAR data in SLC format of level 1.1, compressed in range and azimuth direction has been employed. This level of product gives separate image files for each polarization.

AGWB was converted into Aboveground Biomass (AGB) using the Biomass Expansion Factor (BEF) developed by Chhabra et al. (2002) [1] and further implemented by Shashi (2009). The two equations to calculate the BEF given are; BEF = exp {1.91-0.34*ln (GSVD)}, (for GSVD ≤ 200 m3/ha) - 4.2 BEF = 1, (for GSVD >200 m3/ha)

- 4.3

Where, BEF and GSVD represent Biomass Expansion Factor and growing Stock Volume Density respectively. GSVD is also known as SVD. Finally, BEF was multiplied by the AGWB to calculate AGB as follows AGB = AGWB * BEF

- 4.4

B. Satellite data processing

1st order Scattering Matrix has been generated which represents the scattered properties of target in HH, HV, VH and VV polarizations containing both amplitude and phase information. ALOS PALSAR follows Backscatter alignment co-ordinate convention. Hence, this matrix is known as Sinclair Matrix. The scattering matrix is the form in which the polarimetric information is preserved. The scattering matrix can be expressed in lexicographic basis or Pauli basis. One of the second order derivatives of this matrix is coherency matrix which is created by expressing the scattering matrix in Pauli format. The same coherency matrix before and after polarization orientation angle compensation are used for radar vegetation index (RVI) image generation which utilizes the Eigen values from the coherency matrix. The RVI follows the IV. METHODOLOGY Van Zyl range between 0 – 4/3 for ALOS PALSAR data [10], Field data processing and satellite data processing form two [17], [31]. The coherency matrix has been generated for the crucial phases for data generation and analysis. calculation of polarization orientation angle shift using matrix elements. The negative of the calculated orientation angle has A. Field Data Processing been used in a unitary rotation matrix for data compensation for orientation angle shift. After compensation the matrix is CBH and height of trees in 31.62m X 31.62m sampling transformed accordingly, to fetch the correct decomposition plots designed using stratified random sampling method has results [4], [14]. Deorientation results into the rotated matrix been collected from Forestry and Ecology Department of IIRS, which will confiscate the shift in orientation occurred due to Dehradun, India. The data is supplemented with the local and terrain variability yielding the same decomposition results botanical names for each tree which greatly helped for from identical targets. So, orientation angle compensation is species-wise stem volume calculation. Specific volume implemented before decomposing the coherency matrices [15]. equations collected from FSI 1996 were used to calculate Polarimetric decomposition is the process of the breaking Stem volume [7]. Aboveground woody biomass (AGWB) has down of the received signal information from numerous been calculated using SVD and specific gravity (SG) values scatterers covered in the radar cross-section into the collected for each species from Indian Wood, Volume 1 to 6. explainable and understandable information segregated into These specific gravity values have further converted into various components depending upon the target nature. The 2nd wood density values using density of pure water as a reference. order coherency matrix generated was used as an input for It is given as: decomposition of data into different scattering mechanism. Depending upon the distribution and orientation of scatterers AGWB = SVD * SG - 4.1 detected by the radar, the polarimetric response from a target

130

8th International Conference on Microwaves, Antenna, Propagation & Remote Sensing ICMARS-2012, Jodhpur, INDIA, Dec. 11 – 15, 2012

is measured in terms of surface scattering, double bounce or volume scattering. The comparison of the decomposition results produced before and after compensation for polarization orientation angle shift, the values for the field plots before and after compensation images were plotted for every component. Comparison has also been done for RVI before and after polarization orientation angle compensation but the value observed were unchanged. C. Water Cloud Modeling WCM was developed by Attema et al. (1978) [5] which portray the response of the upper vegetation layer to radar microwave radiation and it was suggested that the plant canopy can be modelled as water cloud by assuming plant canopy as cloud within which water droplets are arbitrarily distributed. The vital reason to model the plant canopy as water cloud was that the dielectric constant of dry vegetation (1.5) is much smaller than that of pure water but greater than the dielectric constant of air (1.0). Also, 99% volume of vegetation canopy is usually composed by air but the moisture available in plant canopy shows much larger dielectric constant than the dry canopy and air (Attema et al., 1978). For this incentive, canopy moisture is considered as water droplet and canopy as cloud. WCM considers the plant canopy as a homogeneous medium of water droplets and due to this constraint the model has been further developed [24] and the fissures in the canopy was also incorporated for forest mapping as forest is not a continuous layer but related to the scattering from the ground through the gaps in the canopy [21], [22] . In present study, WCM based on semi-empirical modeling approach has been implemented for SVD and AGB retrieval using field estimated data. The three unknown parameters of the model namely ζ (Empirically defined coefficient, representing two way transmmitivity), σ0vegetation (backscattered from vegetation) and σ0ground (backscattered from ground) has been estimated using the SAR polarimetric data and field measured data. For model training it is required to estimate ζ as in this study, the backscattered energy from ground has been retrieved from the single bounce or odd bounce scattering component and the backscattered energy from vegetation is retrieved using the aggregation of double bounce and volume scattering component obtained after decomposition of quad-pol SAR data. The decomposed components after polarization orientation angle compensation have been utilized as they are liberated from wrong estimates. For this purpose, out of total 22 plots, 10 plots have been used for training model. Input of these values give empirically defined 2-way transmmitivity coefficient for each plot taken into consideration for training. Using least square estimation method out of 10ζ values 1ζ value has been selected for running model which showed minimum least square error of around 0.001 in SVD and 0.0012 in AGB respectively. Plotwise forest SVD and AGB is retrieved by the inversion of the semi-empirical WCM. The inversion process is applied for remaining 12 plots which are not taken into consideration for

training. The selected ζ value is incorporated for estimating SVD and AGB based on semi-empirical model. Hence, obtaining model derived SVD and AGB. D. Performance evaluation of Model The accuracy of the of the results obtained using WCM modeling approach has been estimated using two statistical parameters as – • Coefficient of Determination – This method is based on regression analysis approach carried out using plotting of field based SVD and AGB against model derived SVD and AGB. • Root Mean Square Error – RMSE is a frequently used measure of the differences between values predicted by a model or an estimator and the values actually observed. It is a good measure of accuracy. It is calculated using field based SVD and AGB values with model derived SVD and AGB using formulae – RMSE = √1/n*∑n i=1 (V i model – V i measured)2

- 4.5

RMSE = √1/n*∑n i=1 (B i model – B i measured)2

- 4.6

Where, V i and B i are SVD and AGB of ith plot modelled by WCM V i measured and B i measured is the ground-truth SVD and AGB obtained from field measurements n is the total number of plots from where in-situ data have taken ( in present case number of plots is 22) V. RESULTS AND DISCUSSIONS A very high coefficient of determination 0.973 is obtained after regressing field retrieved SVD and AGB, which assures the reliability of the estimation of these biophysical parameters. According to Van Zyl (2006), RVI values for ALOS PALSAR data range between 0 to 4/3. The data for present study gave approximately around 0 - 1.18 as a whole image statistics for RVI. The RVI range for the plots taken into consideration for study is 0.03 – 0.70. The value range denotes very short range as compared to the range given by Van Zyl. Hence, the correlation of RVI with SVD and AGB is also found comparatively low. Agriculture, buildings, grasslands, waterbodies, etc show very low RVI as compared to the RVI values for the forest area. This may be due to the study area is majorly covered by dry deciduous forest and also the satellite imageries are of dry season which dictates the moderate range of RVI values. RVI values above 0.70 are very negligible in number as compared to the total number of scene pixels which is around less than 1% of total scene pixels. Hence, above 0.70 RVI values can also be speckles in the image. The generated coherency matrix has been utilized for the decomposition to analyse the nature of target. The decomposition has resulted into odd or single bounce which is majorly the surface scattering information, double bounce which is from the corner reflectors and volume scattering which is from distributed scatterers. The effect after orientation angle compensation shows that the volume

131

8th International Conference on Microwaves, Antenna, Propagation & Remote Sensing ICMARS-2012, Jodhpur, INDIA, Dec. 11 – 15, 2012

scattering mechanisms shows on an average considerable decrease for majority of plots while the double bounce information increases subsequently unchanging the odd bounce scattering values. Hence, the effect of orientation angle compensation on the decomposition results eradicates the error in the wrong estimate of the decomposed results. Figure 1 depicts the change in values of decomposition components after shift in orientation angle.

TABLE 1 COEFFICIENT OF DETERMINATION Regressions Field retrieved SVD Vs Field retrieved AGB Field retrieved SVD Vs RVI

R2 0.973 0.573

Field retrieved SVD Vs Modelled SVD Field retrieved SVD Vs Volume Scattering

0.213 0.101

Modelled SVD Vs Volume Scattering

0.382

Field retrieved AGB Vs RVI

0.617

Field retrieved AGB Vs Modelled SVD Field retrieved AGB Vs Volume Scattering

0.270 0.130

Modelled AGB Vs Volume Scattering

0.340

TABLE 2 WCM MODEL ESTIMATED RESULTS Outcome SVD AGB Zeta(ζ) -0.013 -0.022 RMSE 167.82 (m3/ha) 87.89 (t/ha) 2 R 0.21 0.27

C. Accuracy Assessment Accuracy obtained using RMSE approach for SVD is 167.82 (m3/ha) and for AGB is 87.89 (t/ha) after semi-empirical modeling. VI. CONCLUSIONS

Fig.1 Plotting of Volume scattering, Double bounce scattering, Odd bounce scattering before and after Polarization orientation angle compensation against plots

Model derived SVD and AGB have been retrieved from the scattering contributed by ground i.e. single bounce scattering, the scattering contributed by vegetation i.e. aggregation of volume scattering and double bounce scattering mechanism along with the field derived SVD and AGB estimates was carried out by using WCM. The Results obtained with the selected empirically defined coefficient ζ derived using WCM model are shown in Table 2 whereas Table 1 depicts the coefficient of determination obtained after regressing field retrieved SVD and AGB with polarimetric decomposition component, RVI and Model derived SVD and AGB.

The present study involves first estimation of SVD and AGB using field measured CBH and height of tree. A very high positive correlation R2 of 0.973 is obtained which concludes the reliability of the estimated SVD and AGB. The cross-polarized backscattered channel gives highest correlation of 0.174 and 0.170 with SVD and AGB respectively as compared to co-polarized channel values which concludes the potential of cross-polarized effect. A moderate correlation of biophysical parameters estimated with RVI is obtained which may be due to the satellite data acquired at dry season time due to which may be high leaf fall occurred and hence, RVI range falls between 0.03 to 0.70 which is comparatively moderate given by Van Zyl. The majority of forest cover type covered under the study area is dry and moist deciduous. Hence, to retrieve better correlation for the RVI along with the decomposition results with the forest stand parameters the SAR images of the wet seasons could prove advantageous. Volume scattering showed a dominant effect as compared to odd bounce and double bounce hence, concluding the scattering due to the attenuation from distributed scatterers. In Yamaguchi decomposition model 6 parameters from coherency are taken into consideration. Hence, a decomposition model should be developed, taking into account more than 6 elements of the coherency matrix which can promote revealing more biophysical traits of the forest vegetation. More than three parameter based semi-empirical model can be employed to draw better estimates for forest biophysical parameters. Effect of Orientation compensation showed demarcated effect by decrease in volume scattering mechanism and increase in

132

8th International Conference on Microwaves, Antenna, Propagation & Remote Sensing ICMARS-2012, Jodhpur, INDIA, Dec. 11 – 15, 2012

double bounce scattering mechanism with almost undeviated odd bounce scattering. High correlation is obtained using semi-empirical modeling approach between volume scattering and Modelled SVD as well as AGB as compared with empirical approach. Study undertaken experiences lack of collection of other parameters like Leaf Area Index and canopy moisture content (with higher precision in location) for achieving higher accuracy in SVD and AGB estimation along with height and CBH measurements for trees. Also, InSAR technique if integrated with PolSAR could have enhanced the probability and reliability of retrieving additional information regarding the forest stand parameters.

ACKNOWLEDGMENT The authors are immensely thankful to Director IIRS for supporting with all the requirements of present study in terms of forest field data and satellite data. Also, we are thankful to all the authors of whom the research articles were found helpful in gaining ideas, knowledge and proved as an accelerating factor for our research.

REFERENCES [1]

[2] [3]

[4]

[5] [6]

[7]

[8]

[9] [10] [11]

[12]

[13]

Abha Chhabra, S. Palria and V.K. Dadhwal, “Growing stock-based forest biomass estimate for India,” Biomass and Bioenergy, vol.22, pp. 187-194, 2002. Ambasta, S.P. 1986. The Useful Plants of India. Publication and Information Division, CSIR, New Delhi. Anonymous, 2006. Working Plan for Rajpipla East and Rajpipla West Divisions. Vol. II. Research and Working Plan Division, Rajpipla, District, Narmada, Forest Department of Gujarat, Gujarat, India. Charles D Richardson, “Polarimetric Scattering Model for Multilayered Vegetation in Tropical Forest,” M.Sc Thesis, International Training Centre, Enschede, Netherlands, 2011. E.P.W. Attema and F. T. Ulaby, “Vegetation modeled as a water cloud,” Radio Science, vo1. 3(2), pp. 357–364, 1978. F. Holecz, A. Freeman and J. Van Zyl, “Topographic effects on the antenna gain pattern correction,” in Proc. International Geoscience and Remote Sensing Symposium, IGARSS '95, 1995, Quantitative Remote Sensing for Science and Applications, Firenze, Italy, pp. 587-589. F.S.I. Volume Equations for Forests of India, Nepal and Bhutan. Forest Survey of India, Ministry of Environment and Forests, Govt. of India, Dehradun, 1996. F.T. Ulaby, K. McDonald, K. Sarabandi and M.C. Dobson, “Michigan Microwave Canopy Scattering Models (MIMICS),” in Proc. of IGARSS 88 Symposium, Winburgh, Scotland, 1988. H.G. Champion, and S.K. Seth, “A Revised Survey of Forest Types of India,” Government of India press, New Delhi, 1968. H. Kimura, “Radar Polarization Orientation Shifts in Built-Up Areas,” IEEE Geoscience and Remote Sensing Letters, vol. 5, pp. 217-221, 2008. J. Askne, P.B.G Dammert, P. Fransson, H. Israelsson and L.M.H Ulander, “Retrieval of forest parameters using intensity and repeat-pass interferometric SAR information,” in Proc. Retrieval of Bio- and Geophysical parameters from SAR datafor Land Applications, pp. 119129, 1995. J. Lee and T.L. Ainsworth, “The Effect of Orientation Angle Compensation on Coherency Matrix and Polarimetric Target Decompositions,” Geoscience and Remote Sensing, IEEE Transactions, pp. 1-12, 2010. J. Lee, D. Schuler, T. Ainsworth and W. Boerner, “Polarization orientation estimation and applications: a review,” IGARSS 2003, IEEE International Geoscience and Remote Sensing Symposium., Proceedings (IEEE Cat. No.03CH37477), Toulouse, France, pp. 428-430, 2003.

[14] Jong-Sen Lee and Thomas L. Ainsworth, “The Effect of Orientation Angle Compensation on Coherency Matrix and Polarimetric Target Decompositions,” IEEE Transactions On Geoscience And Remote Sensing, vol. 49(1), 2011. [15] Jong-Sen Lee, Dale L. Schuler, and Thomas L. Ainsworth, “A Review of Polarization Orientation Estimation from Polarimetric SAR Data,” Remote Sensing Division, Code 7263, Naval research Laboratory, Washington DC, USA. [16] Kristina R. Czuchlewski , Yunjin Kim and Jeffrey K. Weissel, “Assessing Natural Disaster Impacts And Recovery Using Multifrequency, Fully-Polarimetric Synthetic Aperture Radar (SAR),” in Proc. of the Workshop on POLinSAR - Applications of SAR Polarimetry and Polarimetric Interferometry (ESA SP-529), pp. 14-16 January 2003. [17] Ling, Z. Li, E. Chen and Q. Wang, “Comparison of ALOS PALSAR RVI and Landsat TM NDVI for Forest Area Mapping,” Asian and Pacific Conference on Synthetic Aperture Radar – APSAR, 2009. [18] M.C. Dobson, L. Pierce, K. Sarabandi, F.T. Ulaby and T. Sharik, “Preliminary analysis of ERS- 1 SAR for forest ecosystem studies,” IEEE Transactions on Geoscience and Remote Sensing, vol. 30(2): pp. 203-211, 1992. [19] M.C. Dobson, F.T. Ulaby, T. LeToan, A. Beaudoin, E.S. Kasischke and N. Christensen, “Dependence of radar backscatter on coniferous forest biomass,” IEEE Transactions on Geoscience and Remote Sensing, vol. 30(2), pp. 412-415, 1992. [20] M.L. Imhoff, “A theoretical analysis of the effect of forest structure on synthetic aperture radar backscatter and the remote sensing of biomass” IEEE Transactions on Geoscience and Remote Sensing, vol. 33(2), pp. 341-352, 1995. [21] M. Santoro, J. Askne and P.B.G. Dammert, “Tree height influence on ERS interferometric phase in boreal forest,” IEEE Transactions on Geoscience and Remote Sensing, vol. 43(2), pp. 207- 217, 2005. [22] M. Santoro, J. Askne, G. Smith and J.E.S. Fransson, “Stem volume retrieval in boreal forests from ERS-1/2 interferometry,” Remote Sensing of Environment, vol. 81, pp. 19– 35, 2002. [23] Nafiseh Ghasemi, Mahmod Reza Sahebi and Ali Mohammadzadeh, “A Review On Biomass Estimation Methods Using Synthetic Aperture Radar Data,” International Journal Of Geomatics And Geosciences, vol. 1(4), 2011. [24] P.B.G Dammert and J. Askne, “Interferometric tree height observations in boreal forests with SAR interferometry,” IEEE Transactions on Geoscience and Remote Sensing, pp. 1363-1336, 1998. [25] P.P. Patel, Ecoregions of Gujarat. Gujarat Ecology Commission. GERI Campus, Race Course Road, Vadodara, Gujarat, India, 1997. [26] Shashi Kumar, “Retrieval of forest parameters from Envisat ASAR data for biomass inventory in Dudhwa National Park, U.P., India,” M.Sc Thesis, International Training Centre, Enschede, Netherlands, 2009. [27] S.P.S. Kushwaha, R.S. Dwivedi and B.R.M. Rao, “Evaluation of various digital image processing techniques for detection of coastal wetlands using ERS-SAR data,” International Journal of Remote Sensing, vol. 21, pp. 565-579, 2000. [28] Uttara Pandey, S.P.S. Kushwaha, T.S. Kachhwaha, P. Kunwar and V.K. Dadhwal., “Potential of Envisat ASAR data for woody biomass assessment,” International Society for Tropical Ecology, Tropical Ecology, vol. 51(1), pp. 117-124, 2010. [29] U. Wegmuller and C. Werner, “Retrieval of Vegetation Parameters with SAR Interferometry,” IEEE Transactions on Geoscience and Remote Sensing, 35(1), pp. 18-24, 1997. [30] U. Wegmuller, M. Santoro and A. Wiesmann, “A novel methodology for parameter retrieval from multi-temporal data demonstrated for forest biomass retrieval from C-band SAR backscatter,” International Workshop on the Analysis of Multi-Temporal Remote Sensing Images, 2007. [31] Yun Shao, Huaze Gong, Guojun Wang, and Aimin Cai, “Applications of Polarimetric Decomposition technology in a dried up lake Evolution,” Geoscience and Remote Sensing Symposium (IGARSS), IEEE International, pp. 4499-4502, 2010.

133