Supplement 1. Estimating carbon emissions and its associated uncertainty using the Monte Carlo simulation approach (taken from Lusiana et al., in prep.).
- Net carbon loss equation:
Lusianaet. al (in prep) estimated carbon emissions in Tanjabar using the net carbon loss approach with the following equation:
Δ Ct→t+1 = Σi(Ai,t+1 × Ci,t+1) - Σi(Ai,t × Ci,t) = Σi((Ai,t+1 - Ai,t) × Ci,t+1) + Σi(Ai,t ×( Ci,t+1 - Ci,t))...... (S1)
where Δ Ct→t+1 is net change in landscape carbon from time t to time t+1 (Mg), Ai,t+1is the area of land use i at time t+1(ha), Ai,tis the area of land use i at time t (ha), Ci,t+1 is carbon stock density of land use i at time t+1 (Mg.ha-1) and Ci,t is carbonstock density of land use i at time t (Mg.ha-1).
- Calculating carbon emissions
Carbon emissions (in CO2 equivalent) were derived by multiplying net carbon loss with a conversion factor of 3.67.
- Data source
Estimating carbon emission required data of land cover/use maps and plot-level carbon for each land cover/use category. Plot-level carbon data for Tanjabar were obtained from the carbon database compiled by and stored at the World Agroforestry Centre, which are based on various research studies that have been carried out by ICRAF and its partners[1]. We used land cover/land use maps for years 2000 and 2009 interpreted from Landsat-TM. The interpreted land cover mapshave 11 land cover categories and spatial resolution of 100 m.
- Methods to estimate uncertainty of carbon emission estimates
The methods to estimate uncertainty followed a Monte Carlo (MC) simulation analysis, which is a statistical based method using a probabilistic sampling. Figure Supp. 1 describes the basic steps involved in the MC analysis. The overall aim of the MC technique is to perturb (randomly vary) the land cover map with classification errors and the plot-level carbon value of each land cover category with its standard error of the meanand then to produce many realizations of carbon emission maps.Thus, the methods assumed that errors in land cover classification and variation in plot-level carbon measurements represent the uncertainty of carbon emission estimates. For the Tanjabar study case, land use classification errors were generated from users’ accuracy of the 2009 land cover map (Table Supp 1) and were assumed to follow uniform distribution. Plot-level carbon densities were generated following the normal distribution with variation based from standard error of the mean (Table Supp. 2).
- Outcome of the carbon emission uncertainty analysis in Tanjabar
The outcome of the uncertainty analysis was 10,000 maps of net landscape carbon loss between 2000 and 2009 perturbed with errors. Using these maps, the average netcarbonloss of Tanjabar district was estimated at 35.4 Mg.Cha-1over the period of 9 years which is equivalent to 129.8 Mg.CO2 eq.ha-1or 14.2 Mg CO2 eq.ha-1.year-1.
Figure Supp. 1. Steps in estimating net landscape carbon loss and its associated uncertainty using the Monte Carlo simulation approach.
Table Supp. 1-1. Error matrix: probability that pixel in classified land cover map of Jambi 2009 is classified as land use i when it is land use j on the ground.
Classified Map+ / Ground truth points+ / Users’ accuracy*Fu / Fd / Rf / Co / Cf / Ac / Rm / Op / Sh / Ag / Set
Fu / 0.78 / 0.17 / 0.00 / 0.00 / 0.00 / 0.00 / 0.00 / 0.00 / 0.06 / 0.00 / 0.00 / 0.78
Fd / 0.01 / 0.93 / 0.03 / 0.01 / 0.00 / 0.00 / 0.02 / 0.00 / 0.00 / 0.00 / 0.00 / 0.93
Rf / 0.01 / 0.03 / 0.73 / 0.01 / 0.01 / 0.01 / 0.08 / 0.09 / 0.01 / 0.01 / 0.04 / 0.73
Co / 0.00 / 0.00 / 0.00 / 0.25 / 0.00 / 0.00 / 0.38 / 0.38 / 0.00 / 0.00 / 0.00 / 0.25
Cf / 0.00 / 0.02 / 0.02 / 0.11 / 0.74 / 0.00 / 0.00 / 0.04 / 0.04 / 0.00 / 0.02 / 0.74
Ac / 0.00 / 0.00 / 0.00 / 0.00 / 0.00 / 0.33 / 0.44 / 0.11 / 0.00 / 0.00 / 0.11 / 0.33
Rm / 0.00 / 0.01 / 0.05 / 0.01 / 0.00 / 0.01 / 0.85 / 0.06 / 0.01 / 0.00 / 0.02 / 0.85
Op / 0.00 / 0.00 / 0.03 / 0.00 / 0.00 / 0.00 / 0.06 / 0.85 / 0.01 / 0.00 / 0.02 / 0.85
Sh / 0.04 / 0.00 / 0.02 / 0.06 / 0.02 / 0.00 / 0.02 / 0.04 / 0.75 / 0.00 / 0.06 / 0.75
Ag / 0.03 / 0.09 / 0.04 / 0.01 / 0.06 / 0.03 / 0.03 / 0.16 / 0.07 / 0.39 / 0.10 / 0.39
Set / 0.00 / 0.00 / 0.00 / 0.02 / 0.02 / 0.00 / 0.04 / 0.11 / 0.01 / 0.05 / 0.75 / 0.75
Note:
+ Fu: Undisturbed forest, Fd:Disturbed forest, Rf: Rubber agroforestry, Co: coconut/bettlenutagrofrestry, Cf: coffee agrofoerstry,
Ac: acacia plantation, Rm: rubber monoculture, Op: oil palm, Sh: shrub, Ag: agriculture; Set: settlement, cleared land, grass.
* User’ accuracy refers to probability that classified land cover map of Jambi 2009 matches gorundtruth data.
Table Supp. 1-2. Plot level carbon average versus time average carbon stocks
Land cover/ use systems / Average carbon stocks a (Mg.ha-1)Undisturbed forest / 214.2 / (17.6)
Disturbed forest / 151.0 / (9.8)
Rubber agroforestry / 91.2 / (9.8)
Coconut agroforestry / 64.2 / (24.3)
Coffee agroforestry / 31.0 / (4.2)
Acacia plantation / 38.2 / (8.3)
Rubber monoculture / 38.2 / (7.4)
Oil palm monoculture / 41.0 / (5.9)
Shrub / 47.7 / (6.6)
Agriculture / 3.5 / (0.7)
Settlement, cleared land, grass / 3.6 / (0.7)
aValue in brackets refers to standard error of the mean
[1]