Carbon Inventory Methods in Indian Forests-A Review

Under the United Nat ions Framework Convention on Climate Change (UNFCC), part icipating countries are required to report national inventory of greenhouse gas (GHG) emissions or uptake. The current challenge is to reduce the uncertainties in producing accurate and reliable act ivity data of Carbon (C) stock changes and emission factors essential for reporting national inventories. Improvements in above ground biomass estimation can also help account for changes in C stock in forest areas that may potentially participate in the Clean Development Mechanism (CDM), REDD plus and other initiat ives. The methods adopted for such estimations vary with respect to geography, objective of the study, available expertise, data and scientific excellence adopted. However the current objectives for such estimates need a unified approach which can be measurable, reportable, and verifiable. Th is might result to a geographically referenced biomass density database for tropical forests that would reduce uncertainties in estimat ing annual biomass increment and forest aboveground biomass. In the light of above requirements, this paper intends to present an overview of the methodologies adopted in India from local to country level estimates to assess C sequestration potential in d ifferent forest components. The paper also discusses remote sensing and Geographical Informat ion System (GIS) in itiat ives taken in this field and the possibility of adopting an integrated approach for reliable, accurate and cost effective estimates.


Introduction
At the first Conference of Part ies (COP), wh ich took place in Berlin in 1995, the parties agreed that the specific commit ments of the convention for the Annex I parties were not adequate because they were too vague and after two and a half year of intense negotiations, the Kyoto protocol was adopted at the third COP on 11 December 1997 in Japan. This protocol is the first international implementation of a cap and trade scheme. Kyoto Protocol in its Article 12 of Clean Develop ment Mechanism (CDM) allo ws Annex I countries to achieve 'additional' emissions and reduction in non-Annex I countries. Forests play a critical role [1] in stabilizing CO2 concentration for it acts as significant source of g lobal CO2 and also provides opportunities to act as sink through soil, vegetation and wood products.
Intergov ern mental Pan el on Cli mat e Change (IPCC) provides us the gu idelines for estimation of Carbon (C) inventory for land use change and forest sector [2] and for ag ricu ltu re, fo res t and oth er land uses [3] an d goo d practice guidance for land use, land use change and forestry (LULUCF) sector [4]. Adoption of C inventory methods and guidelines should lead to accurate, reliable and cost effective estimates of C stocks and changes for a given land use system and period [5]. Imp lementation of sustainable C forestry and C storage led forest management in India warrants for specific research support for status monitoring and technology generation. Considerable variat ions in terms of assumptions and estimates on C sequestration call for standardization of estimation of C estimation emissions for forest and other resources and land-use changes [6]. Worldwide nu merous ecological studies have been conducted to assess C stocks based on C density of vegetation and soils [7,8,9]. The results of these studies are not uniform and have wide variations and uncertainties probably due to aggregation of spatial and temporal heterogeneity and adaptation of different methodologies [10].
Five pools have been identified in Marrakech Accord viz. above ground biomass, below ground biomass, soil carbon, dead organic matter and litter. A mong these only those pools need to be measured and monitored under CDM which are most likely to be impacted by the pro ject activities. Various methods are available for estimation of carbon and flux in these pools but these methods vary on account of accuracy, precision, cost and scale of applicat ion. The broad categories of programmes requiring carbon inventory [5] are for national g reen house inventory, climate change mitigation pro jects, Clean Develop ment Mechanism projects, projects under the global environment facility and forest grassland and agroforestry development projects. Hence, it becomes extremely important for a developing country like India where there is an excellent opportunity of having 26 million degraded land as a potential storage for carbon to evolve and refine the methodology as per the objectives. The present paper gives an insight into the varied methodologies adopted by different workers as per their objectives of study for estimations at various levels related to carbon in India. India having diverse vegetation coupled with variat ion in climates, the inventory experts need to exp lore all sources of informat ion fro m all local sources and create data bases on the basis of inventory parameters (pools) and factors like growth rates, wood density etc. to improve the quality of carbon inventory.

Biomass Carbon
Bio mass is defined as the total quantity of live and inert or dead organic matter, above and below the ground, expressed in tones of dry matter per unit area, such as hectare. (Bio mass carbon = above ground bio mass carbon + below ground biomass carbon + dead organic matter). Above ground biomass is the most important visible and dominant C pool in forests and plantations, although not in grasslands and croplands5.

National Level Es timates
A study was conducted [10] to estimate contribution of India's forests from 1995 to 2005 towards C sink using secondary data of growing stock fro m various sources. Suitable bio mass increment values (expansion and conversion for calculating total tree above ground biomass) and the ratio of below and above ground biomass (for calculating total tree biomass above and below ground) as available in different studies covering a range of forest types of the country were used with a conservative value of C 40 % and 20 % of mo isture content on dry basis (mcdb) for realistic estimat ions [11,12].
Spatial data bases of climat ic, edaphic, and geomorpholo gic indices, and vegetation were used to estimate the potential carbon densities (without human impacts) in above and below ground biomass of forests in 1980. All data were p rocessed in GIS environ ment [13]. Land use data and carbon estimates for South and Southeast Asia were collected and analysed to help reduce the uncertainty associated with the release of C in the atmosphere caused by land use change. The database was developed in Lotus 1-2-3 TM using a sequential bookkeeping model. The source data were obtained fro m historical and geographical documents ( Fig. 1) [14]. The total amount of C sequestered in live vegetation of each ecological zone for 1880, 1920, 1950, 1970 and 1980 was calcu lated using the equation as , Where total C stock of vegetation at time i (TCi) is calculated based on Lji which is the total C (above and below ground) in vegetation type j at time i. Aji is the area in vegetation of type j at time i and n is the total number of land use categories within the zone.

Land Use Model
SPREADSHEET A Input: Land use data from all sources Output: T ime series (1880,1920,1950,1980) of areas in official land use categories for single district or division ⇓ SPREADSHEET A Input: sum of spreadsheet A output for all districts or division in single ecological zones + all information pertaining to evaluation of land use statistics in ecological terms. Output: T ime series of areas translated into ecological land use categories for single ecological zone.

Carbon Model
SPREADSHEET A Input: Output of Spreadsheet B for single ecological zone + all information needed to estimate maximum C stock per category and value of multipliers. Output: T ime series of C stocks for each category in single zone, estimated C release from live vegetation for each interval and total period for that zone. Accordingly [14], the actual C stock of a given vegetation class is calculated as the product of its potential maximu m C stock (M) and two fractional mult ipliers which quantify the estimated reduction of M by environmental limitat ions (E) and degradation (D). CPH = M × E × D. Similarly [15,16] a book keeping model was developed that tracks the C content of each hectare d isturbed by human activity.
In another study [17] estimated forest cover, growing stock and bio mass for the year 1984. This was done at state level for the entire country using information available fro m the vegetation maps, thematic maps and ground forest inventory collected by Forest Survey of India (FSI). For this purpose all the states and union territories were divided into grids of 2.5 0 × 2.5 0 . Data was collected for parameters related to growing stock fro m 170000 g rids. The growing stock of each state was estimated by calculating the nu mber of grids for each co mbination o f density and forest composition. The volume per ha (termed as wood volume factor) for a particu lar co mbination of density and forest composition was generated using data of forest inventory surveys. Three wood volume factors were calculated for each stratum and density class for each map sheet for each state. The estimated volume (or growing stock) was converted into biomass by using specific gravity [18,19] of dominant tree species in each grid and C stock was computer emp loying the formulae, A study was conducted [20] to estimate C flu x through litter fall in forest plantations in India. Data on 24 species fro m 82 stands was tabulated so as to cover the entire country. Mean litter fall (total and alone) fro m the plantation was computed. A C fraction of 0.45 was used for converting litter fall to C flu x. Above ground biomass was recorded at the site for shrubs and grasses whereas standard relationship was used to record tree biomass at the site in arid and semi arid areas of Rajasthan and 0.48 part of C was assumed in vegetation on dry weight basis [21].In another study CO 2 FIX a stand level simu lation model [22] was used to quantify the carbon storage and sequestration potential of selected tree species in India using published data on growth rate and biomass with a carbon factor of 50%.
Allo metric equations [23] (models) have been suggested for national level studies in estimating Above ground tree biomass (A GTB) developed [24] on the basis of climate and forest stand types. Bio mass stock densities are converted to carbon stock densities using the default carbon fraction [2] of 0.47. Furthermore root-to-shoot ratio value [25] of 1:5 was suggested to estimate below-ground biomass as 20% of above-ground tree biomass. Carbon sequestration projected 26] upto year 2050 has been calculated for forestry options under different land use scenarios in India fro m standing biomass, wood products and fossil-fuel use and the equation used is carbon = carbon in standing biomass + carbon in wood products + carbon in fossil fuel. Carbon in standing biomass is determined by mult iplying the area of each land use category by its average biomass and then mu ltiply ing the sum by the carbon content of bio mass, wh ich is assumed to be 0.5. If there is an insufficient amount of fuel wood in the project region, the model auto matically begins to burn fossil fuel wh ich results in increasing carbon emissions [26]. The model estimates the amount of carbon sequestered by approximating land use and relative b io mass changes in the landscape over time.

Regional Level Esti mates
Carbon mit igation potential [27] and cost effectiveness of different tree of med icinal importance of Haryana have been estimated for a period of 30 years (2008-2038) using spreadsheet model (PRO-COMAP), acrony m, Pro ject Based Co mprehensive Mitigation Assessment Process. The model uses data on selected carbon pools as collected from the field, viz. above ground biomass, below ground bio mass, soil carbon and woody litter along with data on costs and benefits. Above ground biomass was calculated by laying quadrates and Mean Annual Increment (MAI) was calculated using volume equations as given in Forest Survey of India publication [28]. Below ground bio mass was has been calculated using IPCC default value which is above ground biomass x 0.27 and carbon [2] sequestered was obtained after mult iplying the bio mass with 0.45). For soil organic carbon three samples were taken fro m each quadrant and samples were collected at the depths of 15 cm, 30 cm and 45 cm and Walkley's method was used for estimation of So il Organic Carbon (SOC). A similar study [29] was carried out to estimate elig ible carbon pools under CDM fo r med icinal trees of Haryana in which observations on growth (height, girth and crown cover) of selected plantation interventions was taken as per the structured data sheets. For calculation of Mean Annual Increment (MAI) of plantation intervention on private lands restricting to bund plantations, it was assumed that the farm size will be 0.25 ha (50 x 50 m) and 32 trees would be planted at a spacing of 7 m as co mmonly practiced in that area. Bio mass expansion factor and wood density have been used as per good practice guidelines by IPCC. Default value of 0.27 has been for below ground carbon and SOC has been calculated as per Walkley's method. Spreadsheet model PRO-COMAP was used for data analysis.

Local Level Es timates
A study [30] was conducted to evaluate C sequestration through commun ity based forest management in Sambalpur Forest Division Orissa. Two villages with total area of 200 ha were selected on the basis of number of years for which the allotted peripheral reserved forests have been protected. Quadrates were laid and observations were recorded for girth and height for each species of trees, shrubs and herbs. The data collected during 1997-98 was used for the estimation of gro wing stock and other indices. The growing stock was calculated using the regression equation [31] as (Standing Woody Bio mass (tonnes/ha) = -1.689 + 8.32 × BA).
Sequestration potential of natural forests in seven village forests of Ch indwara Forest Division o f Madhya Pradesh was estimated for different density classes using harvested method of stratified t ree technique. Quadrates were laid and sample t rees were felled and roots excavated for determination of above and below ground bio mass. The whole tree b io mass without foliage was recorded for different co mponents viz. t wigs, branches, bole and roots and presented on oven dry weight basis [32]. In a study carried out to find [33] C content of some forest tree species the plant samples of various parts were subjected to oven drying. Calciu m was estimated by flame photometer and C was carried out using Walkley and Black's rap id titrat ion method and regression equation method developed between Calciu m and C of various tree components. Ash content method was also used to estimate C. In another study [34] to access carbon sequestration potential under agroforestry in Rupnagar district o f Punjab PRO-COMAP (Project based Co mprehensive Mitigation Analysis Process) model was used for the period (2005-2030) as also suggested by Ravindranath [35] and five sample plots of 0.1 ha each were selected for measurements. Below ground bio mass was calculated as AGB × 0.26. Sequestered carbon was calculated in the model by mult iplying the dry b io mass with a default value of 0.45. In a study [36] to assess comparison between different methods for estimation of bio mass in a forest ecosystem it was concluded that stratified tree technique is the best but urged to develop estimat ing equations of wide applicability to obtain reliab le estimates of stand biomass without destructive sampling.
Carbon allocation in different parts of three year old agroforestry species was studied [37] adopting destructive method of sampling. Field measurements taken were fitted into regression equation with a general form factor of 0.5 regardless of the actual form or taper [38]. The carbon and nitrogen content percent in each plant component was estimated on CHNS analyser. Similarly destructive sampling [39] was adopted to assess carbon sequestration potential of selected bamboo species of Northeast India. Total dry bio mass of samp le co mponent was calculated by mu ltip lying weight of oven dry sample with total fresh weight of p lant component and divided it by fresh weight of plant sample co mponent taken. The total oven dry weight of each component was then multiplied by the total number of plants in that category. Carbon content was estimated by indirect method [40] using a factor of 0.48.

Remote Sensing and GIS Based Estimates
A study was conducted on biomass distribution of natural and plantation forests of humid tropics in northeast India using GIS and different types of forests were mapped using IRS LISS III imageries through supervised classification and a forest type map within the study area was prepared. Sampling of vegetation in the two forests was carried out by belt transect method. Because of high species richness in tropical forests, it is difficu lt to use species-specific regression models, as used in the temperate zone [41,42,43]. Therefore, mixed species tree biomass regression models (Table: 1) were used for A GB estimation of natural and plantation forests [44].
Geospatial technology [50] was used to estimate C stock in natural forests of Eastern Ghats Tamil Nadu. IRS 1D LISS III d igital data and Survey o f India topo sheets were used to prepare the fo rest cover density map and p lot sampling technique was followed to estimate the stand density. Volu me of about 1000 trees was estimated using Smalian's formu la by Chaturvedi and bi-variate equations were derived using calculated volu me, girth at breast height (gbh) and height for different girth class. Vo lu me was mu ltip lied with wood density to obtain biomass. C was obtained following standard methodology [51] by with 49.1 as the conversion factor.
On the basis of thematic maps prepared by FSI and survey done by FSI for forest inventory, growing stock, ground biomass and C stock was determined [52] for the assessment years 1979-1981 and 1994-1995 for a particu lar combination of density and forest composition in Ranchi district. Estimated volu me of growing stock was converted to biomass based on specific grav ity [18,19] of dominant tree species in each grid and dry b io mass was multip lied by the factor 0.48 for estimating Carbon [40]. Similarly satellite data was used in a study [53] to estimate carbon pool in Gov ind Wildlife Sanctuary and National Park to generate forest type and density maps by visual and digital interpretation methods. Field measurements of height and girth were taken to calculate volu me of sample plots of 0.1 ha using the site specific volu me equations provided by FSI. Vo lu me was mu ltip lied with specific grav ity to obtain biomass and later the results of bio mass were ext rapolated in the stratified fo rest type map. Carbon fro m b io mass was calculated and the min imu m value of 48 % was adopted as the conversion factor [51]. A ll above ground woody components have been assumed to have 47-50 % organic carbon [54]. Forest Bio mass and net assimilat ion of carbon of Rajaji National Park Uttar Pradesh (Now Uttarakhand) was mapped and assessed using IRS-1A and assessed using IRS-1A, LISS I digital data for the year 1988. The classified forest types were sub-classified into crown cover levels of 20 percent interval and calibrated through field checks. The crown cover for various forest types was related with the stand biomass (above ground) and the relationship was used mean bio mass was computed for each class which when mult iplied with the respective aerial extent gave total bio mass of the content. Belo w ground biomass was assumed [55] to be 23 % of the above ground biomass [56].
In a review work [57] bio mass distribution in a forest ecosystem was described as the function of vegetation type, its structure and site conditions. Phenology plays an important role in using satellite data for estimat ing qualitative and quantitative characters especially in deciduous vegetation as similarly reported [58] that SAR (Synthetic Aperture Radar) being sensitive to mo isture, temperature, branch architecture, b io mass, age classes, girth, canopy density etc. can provide us with species based forest stratification in areas with perpetual clouds. Spectral response modelling [59] was applied to estimate per unit biomass values of sample plots in homogenous vegetation strata. The results when ext rapolated to the entire area generated biomass map of the Madhav National Park. It was further reported that a combination of various forest parameters like trunk, branches, basal area, soil etc. show better relat ionship with b io mass coupled with merged data of optical (Landsat TM, and IRS LISS II and III) and SA R-X band sensors makes way for better enhancement techniques and mapping.

Soil Organic Carbon
After careful co mparison of the different international standards to be follo wed for forest carbon estimat ion, the carbon fraction (CF) 0.47 defau lt value [2] is proposed to convert the biomass value of standing trees into carbon stock [23]. A study [60] was carried out to estimate soil organic carbon store in different forests of India for which map sheets of all the states/UTs of the country were marked with 2.5' x 2.5' (lat itude and longitude) grids. Data on the extent of forest cover, forest stratum, density and volume per ha for each grid were collected. The major forest stratum in grid was marked using thematic maps prepared by FSI (forests of India have been stratified into 24 species strata). Grid volu me for a grid was calculated stratum wise. Map sheet wise addition of growing stock for all the map sheets falling in a part icular state/UT gives the total estimated growing stock of that state. Soil Organic Carbon (SOC) values under different forest species in various locations in India were collected fro m published literature in different journals, reports, books etc.
Six d ifferent eco zones were selected in arid and semi arid areas of Gu jarat and Rajasthan and data presented for only common access resources. Soil samp les were collected in triplicate fro m each type of land upto 75 cm depth divided into 0-25, 25-50 and 50-75 cm soil layers and analysed for SOC [21]. A study recommended [23] the collection of soil samples at 0-10, 10-20, and 20-30 cm depths and calculation of carbon stock density [62] [61] and methods vary in the choice of stratification, measuring carbon pools and values or factors of estimat ion (Table: 2).
In a study to assess carbon sequestration potential in Rupnagar district of Punjab [34] samples were drawn fro m each selected plantation and soil fro m within a depth of 30 cm and soil carbon was analysed by Walkley and Black rapid titration method [65]. In another study C sequestration potential in natural forests of Tamil Nadu [50] was studied using digital data and Survey of India topo sheets and adopted systematic samp ling technique to collect soil samples at pre-determined sampling points. Soil samples were co llected fro m three layers and after analysed using the equations as:  In a similar study [53] an integrated approach was used to assess carbon pool. The soil samp les (0-3-cm) within each clustered plot were collected and analysed for organic carbon and calculated with the same formu la. In another study [66] regarding soil organic carbon in different land use systems in Giri catchment of Himachal Pradesh soil samples were collected fro m all land uses by digging a pit of 30cm, 30 cm and 45 cm width, depth and length respectively. Bu lk density was calculated using standard core method [67]. Soil o rganic carbon was calcu lated by standard Walkley & Black method [68]. A ll the methods used in this study are in accordance with [10] Ravindranath & Ostwald (2008). The data for SOC pool was calculated by using the follo wing equation as suggested by IPCC Good Practice Guidance [3] for LULUCF: = � SOC Horizon 1 = � ⌊SOC × Bulk density × depth × (1 − C frag. ) 1

× 100 Horizon⌋
Where, SOC = Representative soil organic carbon content for the forest type and soil of interest, tonnes C(ha) -1 , SOC = Soil o rganic carbon content for a constituent soil horizon, tonnes C(ha) -1 , (SOC) = Concentration of SOC in a given soil mass obtained from analysis, g C (kg soil) -1 , Bulk Density = Soil mass per sample volu me, tonnes soil m -3 (equivalent to Mg m -3 ), Depth = Ho rizon depth or th ickness of soil layer, m, C frag ments = % volu me o f coarse frag ments/100.
In another study [69] to estimate soil organic carbon pool under different land uses in Champawat district of Uttarakhand the same methodology and equations were used. In an experiment to assess carbon sequestration potential in Himalayan region of Himachal Pradesh, split plot design [70] was adopted to assess carbon sequestration potential in Himalayan reg ion of H.P. using six land use systems viz. natural g rassland, Hortipastoral, Agriculture, agri-horticu lture and agri-hort i-silviculture each system replicat ing thrice. Agroforestry system fo rmed the main plot and soil sampling depth as sub plot. The soil organic pool expressed as Mega grams ha -1 for a specific depth was computed [71] by mu ltip lying the soil organic carbon (g kg -1 ) with bulk density (g cm -3 ) and depth (cm). A study [9] was carried out in India's forests for the assessment of forest carbon stocks using primary data for the soil carbon pool. The study covered a total of 571 samples in forest area and 101 addit ional samples in the nearby non-forest areas collected from a p it of 30 cm wide, 30 cm deep and 50 cm in length .Soil o rganic carbon was estimated by standard Walkley and Black method and bulk density was estimated using standard Clod method.

The Way Forward
Worldwide nu merous ecological studies have been conducted to assess carbon stocks based on carbon density of vegetation and soils [6,7,8]. The results of these studies are not uniform and have wide variations and uncertainties probably due to aggregation of spatial and temporal heterogeneity and adaptation of different methodologies [9]. A participatory approach for forest boundary delineation should be adopted by involving GIS experts, forest technicians, and members of co mmunity forest user groups (CFUGs ). High-resolution satellite images printed on a large scale can be to find the different land cover and natural boundaries and to trace individual forest blocks easily. For establishment of baseline scenario of the area local socio economic situation and regional economic trends should be taken into consideration. To create uniformity in estimat ions the IPCC Good Practice Gu idelines 1996, 2003 and 2006 must be adopted as per requirement; however an integrated approach to combine various methods is necessary for specific studies for which guidelines are not addressed. It is further reported [10] that IPCC guidelines till now do not provide guidance on certain methods and parameters. The use of GIS technology offers an approach to develop a biomass map of forests. It can be extended to areas in which data are not availab le because consistent patterns of biomass density frequently result fro m similar biophysical characteristics in the study area. A geographically referenced bio mass density database for tropical forests would reduce uncertainties in estimat ing annual biomass increment and forest aboveground biomass.