Geospatial technology based diversity and above ground biomass assessment of woody species of West Kameng district of Arunachal Pradesh

ABSTRACT Comprehending the prominence of forest carbon in climate change, this study was piloted in different land use of West Kameng district, Arunachal Pradesh, India to record the floristic composition, community characteristics, and above ground biomass (AGB) carbon using random sampling and geospatial approach. Preliminary field survey was done in 2016. Altogether 45 quadrats (0.1 ha each) were laid. Total tree richness recorded was 164 species from 49 families. Dominance and frequency distribution pattern of species revealed heterogeneity in composition with majority species showing clumped distribution. Plantations showed highest tree density while mixed dense forest showed maximum basal area (58.89 m2 ha−1). Estimated AGB were 218.21 ton/ha for mixed dense forest, 84.94 ton/ha for abandoned forest, and 105.09 ton/ha for plantations. Total estimated carbon stocks were 120.01, 46.17, and 57.80 ton/ha for mixed dense, abandoned forest, and plantations, respectively. Predicted average AGB using Geographic Information System (GIS) techniques was 163.25 ton/ha. Field-based AGB was slightly greater than the values observed from satellite data. Findings of the study may be useful for calculating total biomass and carbon stored in the major land cover of the district in particular and region in general. It will also support in future studies for calculating the long-term data on biomass carbon sequestration.


Introduction
Forest ecosystem of earth surface accounts for 75% of the gross primary productivity of the earth's biosphere, and encompasses 80% of the plant biomass. They are the largest, complex, and self-regenerating natural resources. They are considered important for their unique role as major carbon sinks because of their capability to capture and store carbon dioxide from the atmosphere. Thus, maintaining forest biomass and sequestration potential is dependent on maintaining forests. However, structure and function of forest ecosystems are changing drastically due to various anthropogenic activities. Forest degradation and deforestation have direct impact on biomass and carbon pool. These activities have altered the concentration of greenhouse gases in the atmosphere thereby impacting climate and weather conditions, biodiversity, food production, and human health. Several studies have reported that carbon storage in forests are directly regulated by composition, age, site characteristics, succession, and climatic variation (Chen et al. 2005;Waring and Running 2007;Gough et al. 2008;Goward et al. 2008).
The amount of carbon sequestered by a forest can be estimated from the biomass accumulated and such database is important for many national developments planning pertaining to productivity and carbon budget. An increase of 0.65-1.06 C temperature was reported in the fifth assessment report of International Panel on Climate Change (IPCC). Hence, emphasis was given on control of emission and removal of carbon dioxide from the atmosphere to mitigate climate change. Remote sensing and GIS have been used for rapid and consistent assessments of above ground biomass (AGB) and carbon pool because of its coverage, systematic observation, and historical data archives. Recently, there have been tremendous efforts at national to global level on application of satellite and field-based data in assessing AGB and carbon pool (Chhabra et al. 2002;Zheng et al. 2004;Manhas et al. 2006;Gunlu et al. 2014;Wani et al. 2014;Salunkhee et al. 2016).
Himalayan ecosystem shows great variability in physiography and composition and housings over 51 million people. These ecosystems are dynamic to the ecological retreat of the Indian landmass, through forest cover, feeding perennial rivers, hydropower, safeguarding biodiversity, enriching soil and agriculture, and spectacular landscapes for sustainable tourism. The state of Arunachal Pradesh, India (Eastern Himalaya), where the study was carried out, has a geographical area of 83,743 km 2 . State biodiversity is very rich supporting 20% faunal species of the nation, 4500 flowering plants, 400 pteridophytes, 23 coniferous species, 35 variety of bamboos, 20 cane species, 106 Rhododendron taxa, and 500 orchid species. Degradation and poor forest management have the potential to reduce the carbon stock while sustainable management can increase the carbon stock. Like other parts, eastern Himalayan regions are also exposed to various levels of threat. Hence, there is a need to carry out the studies on spatial and temporal scale on diversity, degradations, management, etc. for their sustainability. However, the challenges remain to find a commonly agreed and scientifically sound methodological framework for accounting for carbon stock. The study emphasizes on plant diversity, community characterization, and stand biomass. Findings of the study could be helpful for developing suitable region-specific strategies by the land use planners. However, topography of the study area is mostly mountainous with steep terrain, hence CONTACT Om Prakash Tripathi tripathiom7@gmail.com, opt@nerist.ac.in limited plots could be established which otherwise would have given more accurate results.

Study site
Present study was undertaken in West Kameng district (26 54' to 28 01' N latitudes and 91 30' to 92 40' E longitudes) of Arunachal Pradesh (Figure 1). The district covers an area of 7422 km 2 accounting 8.86% of total geographical area of the state. A greater part of it falls within the higher mountainous zone, consisting of a mass of tangled peaks and valleys and its altitudinal variability ranges from 115 to 5780 m asl. The district shares an International border with Tibet in the north, Bhutan in west, Tawang district in northwest, and southern border with districts of Assam. The study area has three principle mountain chains, i.e. Sela, Bomdila, and Chaku ranges. The district is rich in biodiversity and is home to the Eaglenest Wild life Sanctuary covering an area of 217 km 2 and Sessa Orchid Wildlife sanctuary with an area of 100 km 2 . There are five major tribes inhabiting in the study area, namely Monpa, Miji, Sherdukpen, Aka, and Bugun. Monpas and majority of inhabitants follow Buddhist religion.

Data acquisition and analysis
Landsat operational land imager (OLI) satellite (Landsat 8) data was used (NASA's public domain). Landsat 8 collects data in nine spectral bands with 30 m spatial resolution except band-8 (15 m). Band-wise radiometric calibration was performed to remove spurious digital number present in the scene following the process laid down in Landsat 8 user handbook (2016). Images were re-projected to Universal Transverse Projection followed by layer stacking, subsetting, and extraction of the study area. Land use and land cover map was prepared using supervised classification in ERDAS Imagine 9.1 ( Figure 2). Based on the classified map, three major land use types, namely mixed dense forest, abandoned forest, and plantations were selected for detail studies.

Vegetation analysis
Total 45 quadrats of 30 m £ 30 m were laid in selected forest patch (100 m £ 100 m). All individuals (gbh > 30 cm) encountered in the quadrats were recorded with their height and girth at breast height (1.37 m). Coordinates and elevation of the sampling sites were recorded using GPS during the study. Collected specimens were identified with the help of regional floras and published literatures. Community characteristics such as frequency, density, basal area, spatial distribution pattern, and diversity indices were calculated according to Misra (1968), Mueller-Dombois and Ellengberg (1974), and Magurran (1988). Density-distribution, basal area, and biomass were studied by determining the number of individuals in different girth classes. Non-destructive approach of AGB estimation was adopted for estimating biomass and carbon.

Biomass and carbon estimation
Standing volume was estimated using diameter at breast height (dbh) and height of trees as an input in volumetric equations for existing tree species. In case of unidentified species and species with no species-specific equations, common equation for Arunachal Pradesh (Tirap) was used. Further, estimated volumes were converted into dry biomass by using specific gravity or wood density (FSI 1996). AGB carbon stock was estimated by assuming that the carbon content in wood is 55% of the total AGB (MacDicken 1997). Biomass was also estimated using remote sensed data through four widely used vegetation indices, namely normalized difference vegetation index (NDVI; Rouse et al. 1974), difference vegetation index (DVI; Tucker 1979), soil adjusted vegetation index (SAVI; Huete 1988), and red reflection band. Based on the results of linear regression analysis of different indices, best-fit model was used for spectral modeling of biomass of the study area. Various steps to study community characteristics, biomass, and carbon stock are presented in Figure 2.

Tree diversity and community characteristics
Altogether, 164 tree species were recorded from the study area. Maximum species richness (126 species) was recorded in mixed dense forest followed by abandoned (41 species) and plantations (33 species) (Table 1). Based on density, Castonopsis hystrix, Illicium griffithii, Duabanga grandiflora, Michelia champaka, Toona ciliata, Rhododendron sp., Macaranga denticulata, Altingia excela, Tectona grandis, Dipterocarpus macrocarpus, Citrus sinensis, Albiza sinus, Terminalia myriocarpa, and Schima wallichii were among the most dominant species. Shannon and Wiener diversity index was highest (1.38) in mixed dense forest and lowest (1.35) in plantation forest while reverse trend was observed in Simpson dominance index. Majority of the species (>90%) showed clumped distribution, whereas only few species showed random/regular distribution pattern. Sorenson's index of similarity was observed highest (24.32%) between abandoned forest and plantations followed by mixed dense forest and abandoned forest resulting marked difference in species composition between the sites. Raunkaier's frequency analysis revealed that most of the species (70%-90%) in different forest stands showed low frequency (<20%) distribution and species were completely absent in higher frequency classes signifying community is heterogeneous in composition. Forests' heterogeneity was further supported by occurrence of log-normal dominance distribution pattern of species. Such forest composition does not allow dominance of a species. However, in abandoned and plantations, it showed that a few species dominated the community which could be due to disproportionate sharing of resources among the plant species. Maximum tree density was observed in plantations followed by mixed dense and abandoned forests. Greater basal area was observed in mixed forest mainly due to presence of larger number of individuals having more girth (Table 1). Density-girth class distribution pattern showed a gradual decrease in number of individuals with increase in girth in all the selected areas (Figure 3(a)).
Above ground biomass and carbon AGB and carbon distribution in different girth class showed reverse trend to that of density distribution. Study showed that the lower girth class (<75 cm gbh) contributed to 64%-74% of the stand density while only 17% to biomass and carbon. However, 47% biomass was contributed by higher girth class (>151 cm gbh) although they represented 1%-8% of the total density (Figure 3(b)). Total AGB of selected land use showed maximum in mixed dense forest (218.21 ton/ha) followed by plantations (108 ton/ha) and abandoned forest (84.94 ton/ha). Distribution of AGB carbon also showed similar trend to that of AGB distribution (Table 1). Linear regression model exhibited significant correlation between biomass and basal area.
Plot-based biomass estimation using different vegetation indices of selected land use ranged between 0.53 and 0.76 for NDVI, 0.23 and 0.55 for SAVI, 0.10 and 0.30 for DVI, and 0.03 and 0.07 for red band. Linear regression analysis was carried out between field-based estimated AGB and satellitederived vegetation indices to understand their relationships (Table 2). R 2 values for NDVI, DVI, SAVI, and red band were 0.51, 0.51, 0.68, and 0.41, respectively, signifying that SAVI have resulted better relationship with biomass as compared to other indices which could be due to soil brightness factor. Hence, SAVI was considered the best-fit model and regression equation (Y = 192.77x ¡ 50.568, R 2 = 0.68) was used for AGB prediction in the present study. Average predicted AGB was 163.25 ton/ha. However, it was 192.74 ton/ ha (13.5-552.6 ton/ha) in mixed dense, 148.49 ton/ha (1.3-213.3 ton/ha) in abandoned forest, and 119.04 ton/ha (97.2-555.9 ton/ha) in plantation forest (Figure 4).

Discussion
Forests of Eastern Himalayan regions are exposed to various levels of anthropogenic threats such as shifting cultivation, habitat loss, fragmentation, and colonization of invasive species leading to serious ecological and environmental implications (Mishra et al. 2003;Sarma et al. 2008;Tripathi et al. 2010). Hengeveld (1996) argued that species composition is an important attribute of a natural community that influences functioning of an ecosystem. Higher species richness in a habitat is mainly due to the presence of synuisae in the forest (Richards 1996). Tree species richness of the study can be supported by the reported tree richness (76-103 species) from subtropical broad-leaved forests of Meghalaya (Jamir 2000;Tripathi and Tripathi 2011). Variation in species richness among the studied land cover could be due to either disturbance or management. Comparison of species richness among the habitats is quite perplexing and often leads to fallible conclusions mainly on account of wide variation in the sampling area studied by the researchers. Hence, it is extremely difficult to give plausible reasons for such a variation in the species richness.
Dominance is an imperative component of the community and such species may exert a controlling effect on associated species due to their competitive ability (Krebs 1994). They act as a key-species and have greater influence on structural and functional attributes (Janzen 1986;Krebs 1994). Dominance-distribution curves signify equitability and stability of the community. Log-normal distribution pattern signifies abundance of species that have transitional dominance values in the community and maturity and complexity of natural community (Magurran 1988). However, logarithmic or broken-stick distribution reflects that the community is primarily ordered with respect to one dominating factor (May 1975). Mixed forest was represented by older plants as is evident by the presence of larger number of trees having   Tripathi 2015, 2016). Majority of the species showed contagious distribution which could be attributed to inefficient mode of seed dispersal making community highly heterogeneous and patchy (Richards 1996;Tripathi et al. 2010;Yam and Tripathi 2016). Biomass is the function of tree density, age, girth, and height. It is also largely regulated by the habitats and species composition (Joshi and Ghose 2014). Large numbers of methodological literatures are available for modeling the spatial distribution of AGB using single value of biomass estimated from ground truth measurements to sophisticated methods that integrate different data sources. AGB observed in the study can be supported by the reported values of several researchers (Murphy and Lugo 1986;Devagiri et al. 2013;Yam and Tripathi 2016). NDVI-derived AGB showed higher R 2 value as compared to values (0.14 and 0.05) reported by Rahman et al. (2008) and Ren and Zhou (2014). However, Foody et al. (2003) reported lower R 2 value (0.01-0.08) from Thailand, Brazil and Malaysia using Landsat TM satellite data. Devagiri et al. (2013) reported R 2 value between 0.43 and 0.64 using temporal data for south western part of Karnataka. Maynard et al. (2007), Ullah et al. (2012), and Ren and Zhou (2014) have studied relationship between green biomass and SAVI and reported poorer correlation than the current study. Bragg (2011) found that models influence the estimates of AGB mainly due to composition and differences in volume equations. Therefore, using generalized methods may produce more reliable and accurate results. Among the indices, SAVI have resulted better relationship with biomass. Improved result of SAVI could be due to soil brightness effect and correction factor. Hence, best-fit SAVI derived regression equation (Y = 192.77x ¡ 50.568, R 2 = 0.68) was used for AGB prediction and average value was 163.25 ton/ha. Shen et al. (2016) reported 20-229.50 ton/ha from natural and plantation forest of Northern China. AGB of 92-268.49 ton/ha from tropical rain forest of Western Ghats (Bhat et al. 2003) and 230 ton/ha from evergreen forest of Karnataka (Devagiri et al. 2013) were also recorded. SAVI values varied across the vegetation types. In order to realize the relationship, regression analysis was performed between indices and area weighted biomass which revealed a significant positive correlation.

Conclusions
Findings of the study may be useful for calculating total biomass and carbon stored in the major land cover of West Kameng district. Based on the results, it was observed that tree diversity, density, basal area and biomass decrease from mixed forest to plantations. Therefore, to conserve tree diversity and maximize AGB, forest resource manager may minimize the influence of driving factors responsible for land use change. Further, protective buffer of edge species around the land cover may be created with reduced anthropogenic activities to ensure further degradation. SAVI derived average spatial AGB carbon was 89.79 ton/ha. Regression analysis between biomass and basal area showed positive relationship and suggested that AGB increases with increase in girth. It will also support in future studies for calculating the longterm data on biomass carbon sequestration.