Forest canopy height estimation using satellite laser ...

Appl Geomat DOI 10.1007/s12518-017-0190-2

ORIGINAL PAPER

Forest canopy height estimation using satellite laser altimetry: a case study in the Western Ghats, India S. M. Ghosh 1 & M. D. Behera 1

Received: 1 April 2016 / Accepted: 25 April 2017 # Società Italiana di Fotogrammetria e Topografia (SIFET) 2017

Abstract Canopy height is a crucial metric required to quantify the aboveground plant biomass accurately. The study explores the data derived using Light Detection and Ranging (LiDAR) technology from GeoScience Laser Altimeter System (GLAS) aboard Ice, Cloud, and Land Elevation satellite (ICESat) to derive canopy height estimate equations in the tropical forests of the Western Ghats, India. The interpretation of LiDAR waveforms for the purpose of estimating canopy heights is not straightforward, especially over sloping terrain where vegetation and ground are found at comparable heights. Canopy height models are developed using GLAS waveform extent and terrain index, derived from ASTER digital elevation, to counter the effect of topographic relief effects in canopy height estimates over steep terrain. The model was applied to calculate tree heights for whole of the Western Ghats. Results showed that the model can estimate tree heights within the specified height range with an accuracy of more than 90% while using percent overestimation/underestimation method of validation. This shows the effectiveness of the model, especially over steep slopes, also revealing that the models were able to successfully account for the pulse broadening effect. The study highlights the development of a LiDAR-based canopy height model for tropical forest and its ability to yield better canopy height estimates especially over steep slopes. Keywords Waveform extent . Terrain index . Slope . Pulse broadening effect . Vegetation . Canopy height model * S. M. Ghosh [email protected]

1

Centre for Oceans, Rivers, Atmosphere and Land Sciences, Indian Institute of Technology Kharagpur, Kharagpur, West Bengal 721302, India

Introduction Concern about global climate change have emphasized the importance of finding systematic ways of assessing terrestrial carbon stocks at regional, continental, and global scales (Boudreau et al. 2008). Forested areas contain almost half of these carbon stocks and that quantity is decreasing rapidly (Hayashi et al. 2013). As this reduction in forest carbon stock contributes to increase in carbon emission (Alamgir et al. 2016), estimating forest carbon became a primary concern for researchers worldwide. Quantifying forest carbon stock highly depends on reliable acquisition of forest structure information (Tian et al. 2015). Remote sensing technology has developed into a competent choice for the study of large-scale forested areas due to its synoptic coverage (Behera et al. 2015). Traditional optical remote sensing imagery is very useful in monitoring the horizontal distribution of forests. It is usually more difficult to acquire accurate information on the vertical profile of forests, which is needed for the measurement of forest biomass (Yu et al. 2015). Canopy height is an important factor in tree biomass estimation (Feldpausch et al. 2012; Lima et al. 2012), the inclusion of tree height in allometric equation increases the estimation accuracy in tropical forests across all regions (Chave et al. 2005). Satellite laser altimetry has the capability of measuring tree height with an unprecedented accuracy (Behera and Roy 2002; Kushwaha and Behera 2002). The Ice, Cloud, and Land Elevation satellite (ICESat)/ Geoscience Laser Altimeter System (GLAS) spaceborne Light Detection and Ranging (LiDAR) system was designed to obtain characteristics of the Earth’s surface features with unmatched precision and it is first of its kind (Nie et al. 2015). Ice elevation monitoring was the primary mission objective for GLAS. Measuring the height of vegetation canopies was one of the principal objectives of the mission

Appl Geomat

(Harding and Carabajal 2005). Forest observations for smaller areas using GLAS is not very convenient as its footprint extent is bigger than the convenient size needed for those kinds of observations (Pang et al. 2011). It is ideal for estimating forest parameters for larger areas, such as at country or global levels (Hayashi et al. 2013). Studies showed that GLAS data can be useful for obtaining detailed forest height variation (Tian et al. 2015). This is possible because GLAS recorded the changes in laser energy returned from the Earth’s surface as a waveform, which contains information on forest vertical structure. Reflection from ground surface and the objects above it produces the multiple energy peaks in a GLAS waveform data. Accurate detection of tree height depends on precisely finding the vegetation top and the ground return in the GLAS waveform (Yu et al. 2015). In direct canopy height estimation method, the waveform signal start is assumed to be the canopy top. Canopy height in this method is simply estimated based on the vertical difference between signal start and the ground peak (Chen 2010). However, over mountainous areas with large relief and complex terrain, the peaks from ground and surface objects can be broadened and mixed, making the identification of ground peak difficult. This event is known as pulse broadening effect (Lefsky et al. 2005; Tripathi and Behera 2013). As a result, ancillary topographic information like Digital Elevation Models (DEMs) are required to make accurate estimate of canopy height (Lefsky et al. 2005; Nie et al. 2015). The tropical forests of India, e.g., Eastern and Western Ghats, are expected to go through rapid and significant climate and vegetation changes over the next decades (Ravindranath et al. 2006). Though the Western Ghats has high potential to be a major carbon sink in the Indian Subcontinent region (Mushtaq and Malik 2014), the number of studies aimed at biomass estimation for the region is very few (Osuri et al. 2014; Mushtaq and Malik 2014; Kale et al. 2009; Shukla et al. 2015). None of those studies have used modern remote sensing technologies such as LiDAR, which is heavily used nowadays for biomass estimation (Babcock et al. 2015; Chen 2015; Li et al. 2015; Vaglio Laurin et al. 2014). Behera and Roy (2002) in their study suggested that LiDAR may provide useful information to detect changes in the aboveground carbon stores of tropical forests. The Western Ghats range, India, is topographically and biologically diverse. Terrain slope in the region varies widely and can go up to 70°. As steep slope causes more pulse broadening, we can expect overestimation with increase in terrain slope for the region. This study explores the data derived from GLAS aboard ICESat and Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) Global Digital Elevation Model, to estimate canopy heights for the Western Ghats region of India. Till now only a few researchers have developed statistical models to calculate canopy height using GLAS waveform and DEM (Hayashi et al. 2013; Rosette et al. 2008; Lefsky et al. 2005). None of those models

were made for tropical forests, which store approximately half of aboveground carbon sequestrated in global vegetation (Shukla et al. 2015). In this study, statistical models were developed for the Western Ghats to estimate tree canopy heights, and their efficiency was checked. Based on various canopy properties, six different forest types were selected for the study including dry deciduous, moist deciduous, evergreen, semi-evergreen, subtropical broad leaved, and sholas. Among the selected vegetation types canopy height variation for semi-evergreen, dry deciduous and moist deciduous trees are more, and for subtropical broad leaved, evergreen and sholas canopy heights are almost uniform. As a result, overestimation or underestimation may occur considerably for semi-evergreen, dry deciduous, and moist deciduous forests but overestimation or underestimation will be less for subtropical broad leaved, evergreen, and sholas.

Materials and methodology Study area The Western Ghats region of India primarily covers the states of Gujarat, Maharashtra, Karnataka, Kerala, Goa, and Tamil Nadu. This site was chosen owing to its wide range of vegetation and topographical features. Biogeographically, the hill chain of the Western Ghats constitutes the Malabar province of the Oriental realm, running parallel to the west coast of India from 8° N to 21° N latitudes, 73°E to 77° E longitudes for around 1600 km. Rising up from a relatively narrow strip of coast at its western border, the hills reach up to a height of 2800 m before they merge to the east with the Deccan plateau at an altitude of 500–600 m (Ramachandra and Suja 2006). The average width of this mountain range is about 100 km. The climatic and altitudinal gradient has resulted in a variety of vegetation types, from evergreen to semi-evergreen; from moist deciduous to dry deciduous formations. Major vegetation types which are considered for the study are shown in Fig. 1. Data products Figure 2 highlights the different stages of the process through which canopy heights were estimated from the GLAS/ICESat data over the study site. GLAS/ICESat Level-2 Global Land Surface Altimetry Data (GLA14), Version 33 (Zwally et al. 2012) was acquired from National Snow and Ice Data Centre, Boulder, CO, USA for this study. ASTER Global digital elevation model (GDEM) version 2, available at a resolution of 30 m, was utilized for calculating terrain index at each GLAS footprint location that fell within the study site. Terrain index was used in this study so that the canopy height models established for this study can properly quantify and remove the pulse broadening effect that plagues the

Appl Geomat Fig. 1 Location map of the study area showing state boundaries. Also forest vegetation distribution is shown (adapted from Roy et al. 2015)

accuracy of canopy height estimates at higher slopes. The validation report of ASTER GDEM2 (Tachikawa et al. 2011) states that GDEM2 has some positive bias in forested areas. As Hengl and Reuter (2011) noted in their study, this biasness in forested areas is common for other global DEM like SRTM also. Tighe and Chamberlin (2009) in their study showed that positive biasness for forested areas in SRTM and ASTER DEMs is similar. They even showed that biasness for ASTER is lower than SRTM for deciduous, evergreen and mixed forest types. We could not find any study where ASTER biasness was calculated over our study area, as a result we were unable to apply any correction to correct this biasness. In absence of a better alternative, one way to improve the result is to prepare a bald earth DEM after removing all nonterrain features like tree and buildings from the ordinary DEM, but that is beyond the scope of this work. In our study we used ASTER GDEM mainly to measure terrain index, which is defined as the difference between highest and lowest elevation in a 3 by 3 pixel window. It will be safe to assume within that small window all pixels are affected by almost same biasness. So it will not affect the calculation.

collection is a costly affair considering both financial constraints and time. Current study, being an independent one, does not have the fund to invest in field data collection. So for field tree height collection, a different approach has been taken. Roy et al. (2015) developed a vegetation type map of India with an accuracy of 90%. Using that map, LiDAR shots were filtered for the six selected forest types. Figure 1 shows the different forest types for the area. A literature survey is done to find the average height of all the forest types selected for this study and they are listed in the Table 1 below. Data processing GLA14 data product acquired in binary format was converted into ASCII format, utilizing the IDL code developed by the National Snow and Ice Data Center at University of Colorado Cooperative Institute for Research in Environmental Sciences (CIRES). All the relevant parameters, such as latitude (i_lat), longitude (i_lon), signal begin (i_SigBegOff), signal end (i_SigEndOff), centroid range increment (i_gpCntRngOff), and land range offset (i_ldRngOff), were extracted from the ASCII files for subsequent filtering and calculation.

Field data Estimation of canopy height LiDAR shots do not have wall to wall coverage, instead they cover some specific areas only. For a vast region like the Western Ghats, with challenging topography, precise field data

The simplest way for tree height estimation from LiDAR data is direct method, in which the canopy height is estimated

Appl Geomat

(2013) found this method as most accurate for predicting tree height using GLAS data.

Results Development of the tree height model for the Western Ghats

Fig. 2 Methodology flow diagram of the study

based on the vertical difference between the signal start point and the ground peak. The signal start point was provided in the GLA14 product (i_SigBegOff), the ground peak was identified using Gaussian decomposition of the GLAS waveform (Hayashi et al. 2013). However, over mountainous areas with large relief and complex terrain, the peaks from ground and surface objects can be broadened and mixed, making the identification of ground elevation difficult (Chen 2010). The effect of pulse broadening can be rectified by using topographical correction with the help of DEM. This method applies the property that the pulse is spread according to a terrain index, which is the vertical difference between the highest and lowest elevation within a GLAS footprint (Rosette et al. 2008). Elevation values were extracted for a 3 by 3 pixel window around the center of each GLAS footprint to find the terrain index. This area encompasses the GLAS footprint area which has a diameter of around 70 m. Terrain index is calculated by getting the difference between the highest elevation and the lowest elevation value within these 9 pixels. Hayashi et al.

Table 1 Tree height ranges collected from literature for present study

Forest class

Mean tree height

Tree height models were established by multiple-regression analysis. Canopy height, obtained by the literature survey, was used as dependent variable for the analysis. Independent variables were GLAS waveform signal extent and terrain index, calculated from DEM. As the Western Ghats covers a huge area, the points were chosen in such a way that they are distributed all over the study area. Hayashi et al. (2013) in their work found that for areas with steeper slope accuracy of height estimation decreases. We opted for the way they used in their study to counter this issue and developed separate equations depending on the slope. The equations were developed as follows: H ¼ 0:22  WE þ 0:61  TI þ 19:11 for slope ≤ 10Å

ð1Þ

H ¼ 0:21  WE−0:18  TI þ 23:95 for 10Å < slope ð2Þ Where H is the estimated canopy height; WE is the waveform extent from signal start to signal end, in meters; and TI is the terrain index, which represents the vertical difference, in meters, between the highest and the lowest elevations within a footprint. The models were established using 42 points and cross validated using 18 points. Canopy height value for all those points is collected from literature. The coefficient of determination (R2) for the model established for areas with less than 10° slope comes as 0.57 (Fig. 3) with root mean square error (RMSE) of 2.484 m. The canopy height model established for areas with slope more than 10° has a R2 value of 0.73 with

Assumed maximum height

Assumed minimum height

References

Evergreen

25–30 m

50 m

12.5 m

Parthasarathy et al. (2008)

Moist deciduous

10–20 m

5m

40 m

Utkarsh et al. (1998)

Dry deciduous

10–15 m

5m

30 m

Utkarsh et al. (1998)

Semi-evergreen

25–30 m

50 m

12.5 m

Davidar et al. (2007)

Sholas

25–30 m

50 m

12.5 m

Subtropical broad leaved

25–30 m

50 m

12.5 m

Bhat and Ramachandra (2012) Parthasarathy et al. (2008)

Appl Geomat

Fig. 3 Validation results of linear regression models

RMSE of 1.427 m. The coefficients of the equations are calculated with a significance level of 0.05. Overall performance of tree height models In general, canopy height model performance is measured based on the RMSE and coefficient of determination (R2) values generated from field observed canopy height versus predicted canopy height plot. Chen (2010) and Enßle et al. (2014) used the aforementioned method using airborne laser scanning data as field observed height. Hayashi et al. (2013) used field observed data to validate his results. In our work, coverage of study area is considerably larger than the study areas of studies mentioned earlier. As an independent research work, we did not have enough financial funding or time to collect field data for a vast area like the Western Ghats. We followed the steps used by Tripathi and Behera (2013) to validate our canopy height models. Tripathi and Behera (2013) in their study prepared the tree height map for north western India based on a direct method proposed by Los et al. (2012). They classified all LiDAR hits into three groups: overestimated points, underestimated points, and the points which are correctly estimated. Based on the number of LiDAR shots in each class, they assessed the efficiency of Table 2 Number of LiDAR shots for different slope ranges

Slope range 0–10 10–20 20–30 30–40 40–50 50–60 60–70

LIDAR shot count 9555 8667 3977 1236 300 69 13

their method. Tree heights depend mainly on tree species not only on forest type. As in our study, we considered only different forest types; it was difficult to select a perfect range of heights for chosen forest types. In this study, we assume that for a particular forest type the maximum canopy height that can be found is twice and minimum canopy height would be half of the mean canopy height, which is collected through literature. The exact number of LiDAR shots for different ranges of slope and types of vegetation considered for this study is shown in Tables 2 and 3. From Table 2 we can find that 98% of the points are in areas with slope below 40°. It is also clear that the number of LiDAR shots decreases with increase in slope range. Table 3 shows that moist deciduous is the most common vegetation type in the study area, it covers almost 50% of total LiDAR shots. It is followed by dry deciduous, evergreen, semievergreen forests, sholas, and subtropical broad leaved. Once the canopy heights for respective LiDAR shots were calculated, they were filtered, for shots over settlements, water bodies, and vegetation classes not included in the study. After the filtration, individual canopy heights were compared with species height ranges that was developed based on the previous studies held in the region (Table 1). Out of 24,652 total shots 299 were found to be overestimated, that is approximately 1.21% of total number of shots. Only 19 shots are found to be Table 3

Number of LiDAR shots for different kinds of vegetation

Type of vegetation

Number of LiDAR shots

Dry deciduous Evergreen Moist deciduous Semi-evergreen Sholas Subtropical broad leaved

4307 3522 13,034 3460 288 41

Appl Geomat Table 4 Number of overestimated LIDAR shots for different vegetation

Forest type

Within range estimation

Dry deciduous

Over estimation

Under estimation

Percentage of overestimated shots

Percentage of underestimated shots

4038

268

1

6.22

0.023

Evergreen Moist deciduous

3511 13,017

0 27

11 1

0 0.207

0.313 0.008

Semi-evergreen

3454

1

5

0.029

0.144

Sholas Subtropical broad leaved

284 41

3 0

1 0

1.04 0

0.355 0

This study establishes the tree canopy height model for the Western Ghats region by establishing relationship between

GLAS waveform extent, tree height data collected from literature and terrain index. The results from this study confirm that forest height for a sloped terrain can be estimated accurately using GLAS LiDAR data and terrain index. As we do not have exact height information of the trees for our study area, it is not possible to comment about the precise accuracy of the models. However, the canopy heights estimated by the models developed for current study agrees with the assumed range of tree heights for the study area. Interestingly, it works better with increase in slope. The models established for this study has two independent variables, waveform extent and terrain index. It is not necessary that tree height will depend on those two variables only. Factors like precipitation, sunlight, and nutrients affect tree growth so indirectly they control the tree height. A constant term was introduced in our height estimation equation to deal with the influence of the extra factors which were not considered for the current study. There are not many studies done for the Western Ghats region related to tree height determination. Tree height collection at field level was affected due to this scarcity of literature. Some of the tree heights collected from literature may not represent the true heights of related forest type considered for the study. This may be one of the reasons behind the overestimation of dry deciduous forest height. There are also some issues related to ASTER GDEM version 2 as it shows some positive biasness over the forested areas. Our study area comprises of only forests, as a result all parts of our study area will be equally affected by this biasness. We could not find any

Table 5 Overestimated shots for dry deciduous forest over different slope range

Table 6 Overestimated shots for semi-evergreen forest over different slope range

Slope

No. of shots

Overestimated shots

Underestimated shots

Slope

No. of shots

Overestimated shots

Underestimated shots

40°

2241 1383 504 139 22

173 76 16 3 0

0 0 0 0 1

40°

1031 1307 765 233 93

1 0 0 0 0

0 0 0 0 5

underestimated, that is only 0.077% of total shots. Table 4 shows the total number of overestimated shots for each vegetation type and its percentage with respect to total number of shots. The equations developed for the study correctly estimate the tree heights for all shots that fell on subtropical broad leaved. For other forest types also, there are very few points which are over or underestimated except dry deciduous. Tree heights for over 6% of total shots for dry deciduous are overestimated by the equations (Table 4). Two of the major forest types of the study area, dry deciduous and evergreen, are chosen to analyze the effectiveness of model in details. They are chosen out of the rest because in overall the model is somewhat most effective for semievergreen forests and least effective for dry deciduous forests. Examining Tables 5 and 6 we can see that the prediction accuracy of the models increases in areas with more than 20° slope for dry deciduous forests, with an estimation accuracy of 97.14%. In dry deciduous forest areas, models exhibit an accuracy of 93.13% for points with less than 20° slope. The height equations work excellently in semi-evergreen forest areas. There is no underestimation and only a single overestimation of tree heights for areas with less than 40° slope. Five points are found to be overestimated for areas with greater than 40° slope.

Discussion

Appl Geomat

earlier study which assesses the accuracy of ASTER GDEM over the study area. Global accuracy of ASTER GDEM as stated by Tachikawa et al. (2011) is 20 m, but for GDEM version 2, studies are done only for Contiguous United States (CONUS) and Japan. Accuracy of ASTER GDEM version 2 varies from 7.4 in Japan to 8 m in CONUS. Here we choose not to do any correction as a very small window, 3 by 3 pixel, is used for calculating terrain index. It will be safe to assume within this small window all pixels are affected by almost same biasness. So it does not affect the calculation.

Conclusions The Western Ghats region has a diverse topography, with steep slopes at many places. The pulse broadening effect of LiDAR shots progressively increases with increase in slope, which in turn create problems for proper estimation of tree heights (Harding and Carabajal 2005). This study shows that using terrain index and developing tree height equations for different slope ranges we can remove the pulse broadening effect and calculate the height very accurately for all slope range. The actual relationship between predictor variables and field height may not be always linear, it may be non-linear. In standard non-linear regression also we have to assume a relationship between the variables, which may or may not turn out as the actual relationship. To deal with this problem, the dataset can be used as input for non-parametric regression process to get a more appropriate relationship. Using a species-based model and usage of a digital terrain model, which is considered as a better representation of the bare earth surface, for the calculation of terrain index might further improve the accuracy of the estimates for canopy heights produced for this study. The field data used for establishing the model and validation of results in this study are all from previous studies. The results can be further improved by collecting field tree height data at exact GLAS footprints locations, and use them for developing models and result validation.

References Alamgir M, Campbell MJ, Turton SM, Pert PL, Edwards W, Laurance WF (2016) Degraded tropical rain forests possess valuable carbon storage opportunities in a complex, forested landscape. Sci Report 6: 30012. doi:10.1038/srep30012 Babcock C, Finley AO, Bradford JB, Kolka R, Birdsey R, Ryan MG (2015) LiDAR based prediction of forest biomass using hierarchical models with spatially varying coefficients. Remote Sens Environ 169:113–127. doi:10.1016/j.rse.2015.07.028 Behera MD, Roy PS (2002) Lidar remote sensing for forestry applications: the Indian context. Curr Sci 83(11):1320–1328 Behera M, Tripathi P, Mishra B, Kumar S, Chitale V, Behera SK (2015) Above-ground biomass and carbon estimates of Shorea robusta and

Tectona grandis forests using QuadPOL ALOS PALSAR data. Adv Space Res. doi:10.1016/j.asr.2015.11.010 Bhat SP, DSCM, Ramachandra TV (2012) Status of forests in Shimoga, Central Western Ghats 1–10 Boudreau J, Nelson RF, Margolis HA, Beaudoin A, Guindon L, Kimes DS (2008) Regional aboveground forest biomass using airborne and spaceborne LiDAR in Québec. Remote Sens Environ 112(10): 3876–3890. doi:10.1016/j.rse.2008.06.003 Chave J, Andalo C, Brown S, Cairns MA, Chambers JQ, Eamus D et al (2005) Tree allometry and improved estimation of carbon stocks and balance in tropical forests. Oecologia 145:87–99. doi:10.1007/ s00442-005-0100-x Chen Q (2010) Retrieving vegetation height of forests and woodlands over mountainous areas in the Pacific Coast region using satellite laser altimetry. Remote Sens Environ 114(7):1610–1627. doi:10. 1016/j.rse.2010.02.016 Chen Q (2015) Modeling aboveground tree woody biomass using national-scale allometric methods and airborne lidar. ISPRS J Photogramm Remote Sens 106:95–106. doi:10.1016/j.isprsjprs. 2015.05.007 Davidar P, Mohandass D, Vijayan SL, Ali S, History N (2007) Floristic inventory of woody plants in a tropical montane (shola) forest in the Palni hills of the Western Ghats, India. Trop Ecol 48(1):15–25 Enßle F, Heinzel J, Koch B (2014) Accuracy of vegetation height and terrain elevation derivedfrom ICESat/GLAS in forested areas. Int. J. Appl. Earth Obs. Geoinf 31(1):37–44. doi:10.1016/j.jag.2014.02. 009 Feldpausch TR, Lloyd J, Lewis SL, Brienen RJW, Gloor M, Monteagudo Mendoza A et al (2012) Tree height integrated into pantropical forest biomass estimates. Biogeosciences 9:3381–3403. doi:10.5194/ bg-9-3381-2012 Harding DJ, Carabajal CC (2005) ICESat waveform measurements of within-footprint topographic relief and vegetation vertical structure. Geophys Res Lett 32(21):1–4. doi:10.1029/2005GL023471 Hayashi M, Saigusa N, Oguma H, Yamagata Y (2013) Forest canopy height estimation using ICESat/GLAS data and error factor analysis in Hokkaido, Japan. ISPRS J Photogramm Remote Sens 81:12–18. doi:10.1016/j.isprsjprs.2013.04.004 Hengl T, Reuter H (2011) How accurate and usable is GDEM? A statistical assessment of GDEM using LiDAR data. Handbook of quantitative and theoretical geography or advances in quantitative and theoretical geography 000–046. Retrieved from http:// www.geomorphometry.org/HenglReuter2011 Kale MP, Ravan SA, Roy PS, Singh S (2009) Patterns of carbon sequestration in forests of Western Ghats and study of applicability of remote sensing in generating carbon credits through afforestation/ reforestation. J Indian Soc Remote Sens 37(3):457–471. doi:10. 1007/s12524-009-0035-5 Kushwaha SPS, Behera MD (2002) Lidar remote sensing and environment. Employment News (Government of India) XXVII(27):1–2 Lefsky MA, Harding DJ, Keller M, Cohen WB, Carabajal CC, Del Bom Espirito-Santo F, de Oliveira R (2005) Estimates of forest canopy height and aboveground biomass using ICESat. Geophys Res Lett 32(22):1–4. doi:10.1029/2005GL023971 Li W, Niu Z, Huang N, Wang C, Gao S, Wu C (2015) Airborne LiDAR technique for estimating biomass components of maize: a case study in Zhangye City, Northwest China. Ecol Indic 57:486–496. doi:10. 1016/j.ecolind.2015.04.016 Lima A, Suwa R, deMello Ribeiro G, Kajimoto T, dos Santos J, da Silva RP et al (2012) Allometric models for estimating above- and belowground biomass in Amazonian forests at São Gabriel da Cachoeira in the upper Rio Negro, Brazil. For Ecol Manag 277:163–172 Los SO, Rosette JAB, Kljun N, North PRJ, Chasmer L, Suárez JC, Hopkinson C, Hill RA et al (2012) Vegetation height and cover fraction between 60° S and 60° N from ICESat GLAS data. Geosci Model Dev 5(2):413–432

Appl Geomat Mushtaq H, Malik T (2014) Accounting carbon dioxide emission and stratification of carbon stock in Western Ghats, India. A geospatial approach. Int J Remote Sens Geosci 3(1):1–5 Nie S, Wang C, Zeng H, Xi X, Xia S (2015) A revised terrain correction method for forest canopy height estimation using ICESat/GLAS data. ISPRS J Photogramm Remote Sens 108:183–190. doi:10. 1016/j.isprsjprs.2015.07.008 Osuri AM, Kumar VS, Sankaran M (2014) Altered stand structure and tree allometry reduce carbon storage in evergreen forest fragments in India’s Western Ghats. For Ecol Manag 329:375–383. doi:10.1016/ j.foreco.2014.01.039 Pang Y, Lefsky M, Sun G, Ranson J (2011) Impact of footprint diameter and off-nadir pointing on the precision of canopy height estimates from spaceborne lidar. Remote Sens Environ 115(11):2798–2809. doi:10.1016/j.rse.2010.08.025 Parthasarathy N, Selwyn MA, Udayakumar M (2008) Tropical dry evergreen forests of peninsular India: ecology and conservation significance. Trop Conserv Sci 1(2):89–110 Ramachandra TV, Suja A (2006) "Sahyadri: Western Qhats Biodiversity Information System. Biodiversity in Indian Scenarios http://wgbis. ces.iisc.ernet.in/energy/water/paper/sahyadri/index.htm. Ravindranath NH, Joshi NV, Sukumar R, Saxena A (2006) Impact of climate change on forests in India. Curr Sci 90(3):354–361 Rosette JAB, North PRJ, Suárez JC (2008) Vegetation height estimates for a mixed temperate forest using satellite laser altimetry. International Journal of Remote Sensing, 29(5), 1475–1493. doi: 10.1080/01431160701736380 Roy PS, Behera MD, Murthy MSR, Roy A, Singh S, Kushwaha SPS, Ramachandran RM (2015) New vegetation type map of India prepared using satellite remote sensing: comparison with global vegetation maps and utilities. Int J Appl Earth Obs Geoinf 39:142–159. doi:10.1016/j.jag.2015.03.003

Shukla S, Jain SK, Singh J, Nanda S (2015) Geo-spatial technique for vegetation carbon pool assessment in Western Ghats of India. South Asian J Food Technol Environ 1(2):184–189 Tachikawa T, Kaku M, Iwasaki A, Gesch D, Oimoen M, Zhang Z, Carabajal C (2011) ASTER global digital elevation model version 2 – summary of validation results. NASA Land processes distributed active archive center, 27. Retrieved from https://lpdaacaster.cr.usgs. gov/GDEM/Summary_GDEM2_validation_report_final.pdf Tian J, Wang L, Li X (2015) Sub-footprint analysis to uncover tree height variation using ICESat/GLAS. Int J Appl Earth Obs Geoinf 35:284– 293. doi:10.1016/j.jag.2014.09.016 Tighe ML, Chamberlain D (2009) Accuracy comparison of the SRTM, ASTER, NED, NEXTMAP® USA digital terrain model over several USA study sites. Digital mapping from elevation to information ASPRS/MAPPS 2009 Conference, 12. Tripathi P, Behera MD (2013) Plant height profiling in western India using LiDAR data. Curr Sci 7(10):970–977 Utkarsh G, Joshi NV, Gadgil M (1998) On the patterns of tree diversity in the Western Ghats of India. Curr Sci 75(6):594–603 Vaglio Laurin G, Chen Q, Lindsell J a, Coomes DA, Frate FD, Guerriero L, Valentini R (2014) Above ground biomass estimation in an African tropical forest with lidar and hyperspectral data. ISPRS J Photogramm Remote Sens 89:49–58. doi:10.1016/j.isprsjprs.2014. 01.001 Yu Y, Yang X, Fan W (2015) Estimates of forest structure parameters from GLAS data and multi-angle imaging spectrometer data. Int J Appl Earth Obs Geoinf 38:65–71. doi:10.1016/j.jag.2014.12.013 Zwally HJ, Schutz R, Bentley C, Bufton J, Herring T, Minster J, Spinhirne J, Thomas R (2012) GLAS/ICESat L2 Antarctic and Greenland Ice Sheet Altimetry Data, version 33. NASA DAAC at the National Snow and Ice Data Center26, Boulder