Estimation of Carbon Stock Using Remote Sensing: A Case Study of Indonesia

Estimation of Carbon Stock Using Remote Sensing: A Case Study of Indonesia

SHARE

Rokhmatuloh
World wide Fund for Nature-Pakistan
Department of Geography, University of Indonesia
Kampus UI Depok, Jl. Margonda Raya Depok 16424
Telp. +62-21-7270030, Facs. +62-21-7873067
E-mail: [email protected]

Abstract
Global climate is being affected by human activities that result from the emission of certain greenhouse gases (GHG) into the atmosphere. Different percentages of forest cover store different amounts of carbon and the changes in forest cover are used in the model to calculate the annual changes of carbon. This study describes an effort to estimate carbon stock using remote sensing for Indonesia. Measuring carbon stock particularly based on forest cover derived from remote sensing. This vegetation cover is then converted to carbon by multiplying with biomass-carbon conversion factors. The vegetation cover was accomplished using regression tree method. An advantage of the regression tree is its ability to effectively use proportional or continuous predictor data sets with different measurement scales. High carbon stock in Indonesia are distributed in Papua, Kalimantan (Borneo), Sumatra and Sulawesi (Celebes) islands, respectively. While, Java island is the lowest.

1. Introduction
Global climate is being affected by human activities that result from the emission of certain greenhouse gases (GHG) into the atmosphere. Carbon dioxide (CO2) in the atmosphere is a GHG that contributes considerably to global warming (IPCC, 1995). One possible strategy to reduce GHGs with great potential is to use forest to sequester CO2 (Prentice et al., 2000). Forests are relevant to climate change issues due to its function as a reservoir of carbon. Loss of forests is a significant contributing factor in climate change. On the other hand, the possibility of expanding carbon storage in forests has been identified as a potential measure to mitigate climate change (FAO, 2001, DeFries et al., 2000). Different percentages of forest cover store different amounts of carbon and the changes in forest cover, as expressed in a greenness factor to effectively surrogate biomass, are used in the model to calculate the annual changes of carbon (Houghton and Hackler, 2000; Myneni et al., 2001; Song and Woodcock, 2003).

Currently there is a tremendous amount and diversity being carried out related to forest and carbon accounting with a variety methods used for measurement. Remote sensing provides local/global estimates of carbon fluxes in forests. Remote sensing can fill in gaps where inventory information is unavailable. Remote sensing is most valuable applications in carrying out assessments of how climate change might be having an impact on forests by tracking major disturbances, changes in the growing season, and Net Primary Productivity (NPP). Carbon accounting is needed to support the objectives of international agreement to mitigate global climate change (UN, 1998). In conjunction with other spatial datasets such as climate, soil type, and tree height, the forest coverage is important in the carbon cycle model (DeFries et al., 2000).

The Kyoto Protocol sets a collective global target of reducing greenhouse gas emissions by about 5% of 1990 levels by the first commitment period of 2008 to 2012 (UN, 1998; DeFries et al., 2000). In order to enter into force, it needs to be ratified by at least 55 countries that are responsible for at least 55% of the countries’ carbon emissions (UN, 1998).

This study describes an effort to estimate carbon stock using remote sensing for Indonesia. Measuring carbon stock particularly based on vegetation (forest) cover derived from remote sensing. This vegetation cover is then converted to carbon by multiplying with biomass-carbon conversion factors. The vegetation cover was accomplished using regression tree method. The regression tree estimates a case’s target value in terms of its attribute values by constructing a model containing one or more rules, where each rule is a conjunction of conditions associated with a linear expression (Huang and Townshend, 2003; Breiman et al., 1984). An advantage of the regression tree is its ability to effectively use proportional or continuous predictor data sets with different measurement scales (Breiman et al., 1984).

2.Producing Vegetation Cover with Remote Sensing
Regression tree method was developed using training data obtained from QuickBird images and MODIS data. It involved two steps: selection for most relevant variables and preliminary regression tree modeling. The best-selected predictor variables were then used for constructing model by analyzing the relationships within the data and created an appropriate regression tree and rule set (Figure 1). The rules created from the developed model in Cubist (Cubist Tutorial, 1997) were then interpolated spatially to the entire 1-km MODIS data to produce a final vegetation cover map.


Figure 1. Splitting the data successively to form pure subsets (classes) in regression tree method.

Unsupervised clustering, supervised maximum-likelihood classification and on-screen digitizing were performed to obtain vegetation and non-vegetation classes from the training data. The percent vegetation cover is then calculated as the percent of vegetation class for each 1 x 1-km area corresponding to one MODIS pixel. The percent vegetation cover data as a predicted variable were then integrated to predictor variables data that were derived from MODIS and both sets of data were used in the regression tree method. Table 1 shows predictor variables derived from MODIS data. Vegetation cover created using regression tree method is illustrated in Figure 2.

Table 1. Predictor variables derived from MODIS that were used in regression tree method

Predictor Variables
1.Maximum NDVI
2.Average NDVI at three and seven highest NDVI
3.Minimum NDSI at three and seven highest NDVI
4.Average three minimum NDSI
5.Average SI at three and seven highest NDVI
6.Average Band 1-7 at three and seven highest NDVI
7.Minimum Band 1 reflectance
8.Maximum Band 2 reflectance
9.Minimum Band 3 reflectance
10.Maximum Band 4 reflectance


Figure 2. Vegetation cover created using regression tree method.


Figure 3. Vegetation cover produced in this study.

3.Estimation of Carbon Stock
Measuring carbon stock particularly based on vegetation cover derived from remote sensing. This vegetation cover is then converted to carbon by multiplying with biomass-carbon conversion factor. Table 2 shows conversion factors adopted in this study (CMFP, 2000). Based on these conversion factors, an estimated of carbon stock (terrestrial carbon) of Indonesia was then created (Figure 4). High carbon stock in Indonesia are distributed in Papua, Kalimantan (Borneo), Sumatra and Sulawesi (Celebes) islands, respectively. While, Java island is the lowest big island with carbon stock.

Table 2. Conversion factors from vegetation cover to carbon adopted in this study

Conversion factors
1. Total wood volume = vegetation cover x 1.454 x 0.396 (in m3)
2. Total dry matter biomass = wood volume x 0.43 (in tonnes)
3. Total carbon = dry matter biomass x 0.5


Figure 4. Estimated carbon (terrestrial carbon) of Indonesia created in this study.

4.References

  • Breiman, L., Friedman, J., Olshen, R. and Stone, C., 1984, Classification and Regression Trees (New York: Chapman and Hall), 358 pp.
  • Canada’s Model Forest Program, 2000, Carbon Budget Accounting at the Forest Management Unit Level: An overview of issues and methods, Canadian Forest Service, Ottawa, 13 pp.
  • Cubist Tutorial, 1997, An overview of Cubist, RuleQuest Research, Australia (from https://www.rulequest.com/cubist-win.html, last accessed: June 2006).
  • DeFries, R., Hansen, M., Townshen, J.R.G., Janetos, A.C. and Loveland, T.R., 2000, A new global 1km data set of percent tree cover derived from remote sensing, Global Change Biology, 6, pp. 247-254.
  • FAO (Food and Agriculture Organization), 2001, Global forest resources assessment 2000: Main report. FAO Forestry Paper 140, FAO, Rome, 479 pp.
  • Houghton, R.A. and Hackler, J.L., 2000, Changes in terrestrial carbon storage in the United States. I: The roles of agriculture and forestry. Global Ecology Biogeography, 9, pp. 125-144.
  • Huang, C. and Townshend, J.R.G., 2003, A stepwise regression tree for nonlinear approximation: applications to estimating sub-pixel land cover. International Journal of Remote Sensing, 24(1), pp. 75-90.
  • IPCC, Climate Change 1995: A report of the intergovernmental panel on climate change, IPCC report, Geneva, Switzerland, 64 pp.
  • Myneni, R.B., Dong, J., Tucker, C.J., Kaufmann, R.K., Kauppi, P.E., Liski, J., Zhou, L., Alexeyev, V. and Huhges, M.K., 2001, A large carbon sink in the woody biomass of Northern forests. Proceeding of the National Academic Science, 98(26), pp. 14784-14789.
  • Prentice, C., Heimann, M., and Sitch, S., 2000, The carbon balance of the terrestrial biosphere: Ecosystem models and atmospheric observations, Ecological Applications, 10(6), pp. 1553-1573.
  • UN (United Nations),1998, Kyoto Protocol to the United Nations Framework, Convention on Climate Change, 20 pp.