is subject to errors when used as a potential biomass predictor. The seasonal
changes in leaf canopy induces changes in NDVI. The grasses and shrubs with
good amount of greenness can simultaneously affect NDVI values.
(Cammarano, Fitzgerald, Casa, & Basso, 2014) have used a suite of spectral indices to estimate
the N content in plant leaves.Even though the paper mentions the use of
vegetation indices for plant characterisation but it has not dealt with the
assessment of biophysical parameters through vegetation indices.
(Anaya, Chuvieco, & Palacios-Orueta, 2009) attempted to increase the details of biomass estimation
at regional scales by using MODIS products and field measurements in Columbia.
They classified the area into grassland, primary forests and secondary forests.
They used different MODIS products for different types of vegetation. However,
such estimation presupposed the classification of forests beforehand and no
distinction between different types of forests was made.
(Devagiri et al., 2013) used Remote sensing
and GIS based
approach for estimation
of above ground
biomass (AGB) and
carbon pool at
regional scale in
south western part
of Karnataka. He integrated field measured biomass with spectral responses of different bands
and indices of
MODIS 250 m
spatial resolution. Based
on relative forest
area within the
MODIS pixel, area
weighted biomass was
estimated for each
site using ground
measured plot biomass. They used the spectral modeling to estimate the AGB and
vegetation carbon pool and prepared a map to understand the geospatial
distribution in the region. However this study has been done on a relatively
less undulating area when compared with Sikkim.
(Ramachandran, Jayakumar, Haroon,
Bhaskaran, & Arockiasamy, 2007) conducted a study on
estimation of carbon stock in natural forests using geospatial technology in the Eastern Ghats of Tamil Nadu,
India. Expert classification technique was followed to prepare the forest cover
density map of the study area. Normalized Difference Vegetation Index (NDVI)
was prepared and recoded into four classes based on the density, viz. very
dense (>70%), dense (40–70%), open (10–40%) and degraded (<10%) 39. Digital elevation model (DEM) was prepared using the 20 m contours. They followed plot sampling technique to estimate the stand density in different forest types. In the hills the environmental gradients assume several forms. The elevation and slope aspect play a key role in determining the temperature regime of any sites. (Sharma, Gairola, Baduni, Ghildiyal, & Suyal, 2011) undertook a study in seven major forest types of temperate zone of Garhwal Himalaya to understand the effect of slope aspects on carbon (C) density. They assessed soil organic carbon (SOC) density, tree density, biomass and soil organic carbon (SOC) on four aspects, viz. north-east (NE), north-west (NW), south-east (SE) and south-west (SW), in forest stands dominated by Abies pindrow, Cedrus deodara, Pinus roxburghii, Cupressus torulosa, They found that SOC and TCD were significantly higher on northern aspects as compared with southern aspects. (Du et al., 2014) mapped forest biomass in China by combining the remote sensing and forest inventory. They have elaborated on the inherent limitation of the inability of the remote sensing pixel to capture the variation within the pixel and thereby the over or underestimation of biomass in the remote sensing methods. However the study was done on a very broad scale by dividing the forests in three classes namely coniferous forests, broadleaved forests and other lands. Moreover, the study was done at a broad scale by spatially downscaling the maps at 0.05° (~5600 m) resolution. Texture has long been used in the interpretation of RADAR images. A number of studies have reported improvement in estimates by using texture analysis for forest classification and biomass estimation. Most approaches to image texture analysis normally use a single band of spatial information to characterize texture but in a mixed Douglas fir (Pseudotsuga menziesii) and lodgepole pine (Pinus contorta var. latifolia) forest interspersed with stands of regenerated lodgepole pine in relatively flat to rolling topography in Interior Plateau region of British Columbia, Canada, (Coburn & Roberts, 2004b) used multiscale approach to image texture where first and second order statistical measures were derived from different sizes of processing windows and were used as additional information in a supervised classification. They observed an improvement up to a maximum of 40%. by using several bands of textural information processed with different window sizes. They compared a number of different statistical texture measures. Synthetic Aperture Radar (SAR) data were collected by (Mitchard et al., 2013) over the field site in 2007 and 2009 from the Phased Array L-band Synthetic Aperture Radar (PALSAR) sensor on the Advanced Land Observing Satellite (ALOS) . They created the map of AGBM by calibrating the Radar data with ground samples. They concluded that PALSAR data can be used to detect losses and gains in AGB in woodland ecosystems. They however said that further work is needed to precisely quantify the uncertainties in the change estimates, and the extent of false-positive and false -negative change detections that would result from using such a system. (Sarker & Nichol, 2011) have mentioned that most previous biomass estimation projects used Landsat TM data with 30 m resolution but texture is expected to be more effective with finer spatial resolution imagery since finer structural details can distinguished. Although texture measurement holds potential for biomass estimation it has not yet been fully investigated and results so far when applying texture to optical images have not exceeded 60% even in structurally simple temperate and boreal forests. (Safari & Sohrabi, 2016) used eight image textures per Grey Level Coherence Matrices ( GLCM) over four bands of Landsat 8 OLI with different offsets in the Oak Coppice forests of Zagros mountain and used stepwise regression to correlated the above ground biomass with various texture parameters. (Dube & Mutanga, 2015) were able to improve the accuracy of biomass assessment in medium-density plantation forest species in