RESULT AND DISCUSSION 1 AGB and BA estimates in the study area

Bogor, 21-22 October 2015 213 4. RESULT AND DISCUSSION 4.1 AGB and BA estimates in the study area We used airborne lidar for estimating forest metrics, i.e. AGB and BA, in tropical peat swamp forest of Central Kalimantan. In our study area, we estimated total AGB and BA of 24 million ton and 373 m -2 , respectively Table 4. The estimated mean AGB is 205 ton.ha -1 , with range between 15 ton.ha -1 and 787 ton.ha -1 . Table 5: Sumary of AGB and BA estimates in the study area Forest metrics Min Max Mean Standard Deviation Total AGB [ton] 15.516 787.21 205.05 150.2 24511230 BA [m2] 5.24 122.13 31.15 19.96 373.04 Figure 1 showed the AGB ton.ha -1 and BA m2.ha -1 distribution in the study area. The southern part or the study area, which was mostly degraded and deforested, stored less AGB than in the northern part, which was dominated by forests. In contrast to AGB maps derived from land cover maps Figure 2, we were still able to identify high variation of AGB in the the forested area. In this study, we claimed that lidar has ability to produce AGB maps in high accuracy. 4.2 AGB estimates using existing land cover maps In this study, we compared the AGB estimates derived from lidar analysis with the AGB estimates derived from landcover maps. We found that the AGB difference from MoF and KFCP with our estimate were 38 and 28 less, respectively Table 5. The KFCP landcover map had better conformity between the forest boundaries with our high AGB areas. The method in generating KFCP landcover map could be the reason of this better agreement. It was derived using object-based analysis of digital classification. While the MoF map, which was manually digitized, subjected to misclassification due to operator error. Tabel 5: Comparison of mean and total AGB estimates Approach mean AGB Total AGB Difference [] Lidar this study 205.05 24511230 MoF 126.67 15144378 -38.2 KFCP 146.44 17505100 -28.6 However, the KFCP AGB map failed to differentiate accurately the AGB variation within the forested area. In fact, the distribution of high density forest in the KFCP Figure 2 right map has no similarity at all to the distribution of high AGB in the lidar map Figure 1 left. The area with higher AGB mostly were classified as secondary forest, rather than in primary forest Figure 3b. Both maps were unable to distinguish the variation of AGB accurately in non forest area, as shown in Figure 3a and 3b, the standard deviation of the mean were relatively high. This finding suggested that the use of optical sensor-based satellite imageries for estimating AGB should be taken with care. In addition to that, the AGB density values from literatures used to derived total AGB stock must be examined for the applicability to the study area. The selection of the AGB values should not be only based on the land cover type classification, but also the similarity in vegetation densities as well as disturbance level and history. Bogor, 21-22 October 2015 214 Figure 2: Wall-to-wall AGB left and BA right maps of study area with 30-m resolution Figure 3: Landcover-derived AGB maps. The left is MoF map and the right is KFCP map Bogor, 21-22 October 2015 215 Figure 4: Mean AGB of each land cover types from a MoF map and b KFCP map. Error bars depict standard error of the mean Figure 5: Comparison of mean AGB estimates based on forest and non forest classes from KFCP and MoF maps 309.5 42.3 311.2 61.1 50 100 150 200 250 300 350 400 450 Forest Non Forest KFCP map Bogor, 21-22 October 2015 216 As can be seen from Figure 1 and Figure 2, the high AGB stock area and the forested area were strongly comparable. We further analysed the AGB estimates based on forest and non forest classes from both maps, which overlaid with our lidar-derived AGB map. We found that both maps have strong agreement on the mean AGB of forest and non forest classes Figure 4. The mean AGB of forest class from KFCP and MoF were almost similar with 309.5 ton.ha-1 and 311.2 ton.ha -1 , respectively. These numbers were higher compared to previous AGB studies in peat swamp forests using lidar Englhart, Jubanski, Siegert, 2013. The AGB estimate from KFCP forest class had an advantage of lower standard deviation than the MoF map forest class. Similarly, the non forest class from KFCP has lower standard deviation with mean AGB of 42.3 ton.ha -1 . The MoF map classified some high AGB area into non-forest class, thus resulting higher mean AGB with 61.1 ton.ha -1 . 5. CONCLUSION In this study we estimated AGB and BA of peat swamp forest using lidar data. Based-wall-to- wall map derived from lidar data, we estimated total AGB and BA of 24 million ton and 373 m -2 , respectively. The estimated mean AGB was 205 ton.ha -1 for the whole study area. The landcover classification map based on landsat imageries, should be used carefully to estimating AGB. The maps were unable to distinguish the variation of AGB accurately in non forest area, whereas the standard deviation of the mean were relatively high. However, we found that simple classification of forest and non-forest classes still produced similar estimates of AGB distribution between the maps. REFERENCES Ballhorn, U., Navratil, P., Jubanski, J., Siegert, F. 2014. Lidar survey of the Kalimantan Forests and Climate Partnership KFCP project site and EMRP area in Central Kalimantan, Indonesia. . Technical Working Paper. Kalimantan Forests and Climate Partnership. Englhart, S., Jubanski, J., Siegert, F. 2013. Quantifying Dynamics in Tropical Peat Swamp Forest Biomass with Multi- Temporal LiDAR Datasets. Remote Sensing, 55, 2368-2388. doi:10.3390rs5052368 Graham, L. L. B., Susanto, T. W., Xaveius, F., Eser, E., Didie, Salahuddin, . . . Applegate, G. 2014. KFCP Vegetation Monitoring: Rates of change for forest characteristics and the influence of environmental conditios n the KFCP study area. Kalimantan Forests and Climate Partnership Scientific Report. Hergoualch, K., Verchot, L. V. 2011. Stocks and fluxes of carbon associated with land use change in Southeast Asian tropical peatlands: A review. Global Biogeochemical Cycles, 252. doi:10.10292009gb003718 Hooijer, A., Silvius, M., Wosten, H., Page, S. 2006. PEAT-CO2, Assessment of CO2 emissions from drained peatlands in SE Asia. Delft Hydraulics report Q39432006. Retrieved from http:peat-co2.deltares.nl Ichsan, N., Vernimmen, R., Hooijer, A., Applegate, G. 2013. KFCP Hydrology and Peat Monitoring Methodology. Jubanski, J., Ballhorn, U., Kronseder, K., Franke, J., Siegert, F. 2012. Detection of large above ground biomass variability in lowland forest ecosystems by airborne LiDAR. Biogeosciences Discussions, 98, 11815-11842. Krisnawati, H., Adinugroho, C., Imanuddin, R., Hutabarat, S. 2014. Estimation of forest biomass for quantifying CO2 emissions in Central Kalimantan: A comprehesive approach in determining forest carbon emission factors. Research and Development Center for Conservation and Rehabilitation, Forestry Research and Development Agency, Bogor, Indonesia. Bogor, 21-22 October 2015 217 Kronseder, K., Ballhorn, U., Böhm, V., Siegert, F. 2012. Above ground biomass estimation across forest types at different degradation levels in Central Kalimantan using LiDAR data. International Journal of Applied Earth Observation and Geoinformation, 18, 37-48. Lefsky, M. A., Cohen, W. B., Parker, G. G., Harding, D. J. 2002. Lidar Remote Sensing for Ecosystem Studies Lidar, an emerging remote sensing technology that directly measures the three-dimensional distribution of plant canopies, can accurately estimate vegetation structural attributes and should be of particular interest to forest, landscape, and global ecologists. BioScience, 521, 19-30. Manuri, S., Andersen, H.-E., MacGaughey, R., Brack, C. Under Review. The influence of lidar density to the estimation of tree AGB in tropical peat swamp forest in Indonesia. McGaughey, R. 2014. FUSIONLDV: software for LIDAR data analysis and visualization. USDA Forest Service. Pacific Northwest Research Station. MoF. 2012. Improvement of Land Cover Change Estimate of Indonesia for Year 2011 In Indonesian. Directorate General of Forestry Planning, Ministry of Forestry, Jakarta, Indonesia, p. 40. Retrieved from http:appgis.dephut.go.idappgisdownload.aspx Page, S. E., Rieley, J. O., Banks, C. J. 2011. Global and regional importance of the tropical peatland carbon pool. Global Change Biology, 172, 798-818. PTSUI, P. T. S. U. I. 2011. Topographic mapping by airborne laser scanning August to October 2011. Kalimantan Forests and Climate Partnership Report. Roberts, G., Lewis, D., Soriano, L., Rosoman, G., Taufik, K., Wadji, F., . . . Tat, L. J. 2012. High Carbon Stock Forest Study Report: Defining and identifying high carbon stock forest areas for possible conservation. Golden Agri-Resources and SMART, The Forest Trust and Greenpeace. Rosenqvist, Å., Milne, A., Lucas, R., Imhoff, M., Dobson, C. 2003. A review of remote sensing technology in support of the Kyoto Protocol. Environmental Science Policy, 65, 441-455. Siegert, F., Navratil, P., Franke, J., Kronseder, K. 2013. Historical Land Cover Classification and Land Cover Change in the Kalimantan Forests and Climate Partnership KFCP site and the Kapuas Hulu District. Kalimantan Forests and Climate Partnership Report. Solichin, S., Lingenfelder, M., Steinmann, K. 2011. Tier 3 biomass assessment for baseline emission in Merang Peat Swamp Forest. Paper presented at the Workshop on Tropical Wetland Ecosystems of Indonesia: Science Needs To Address Climate Change Adaptation And Mitigation. Watilan Convention Center, Sanur Beach Hotel, Bali. Wulder, M. A., White, J. C., Nelson, R. F., Næsset, E., Ørka, H. O., Coops, N. C., . . . Gobakken, T. 2012. Lidar sampling for large-area forest characterization: A review. Remote Sensing of Environment, 121, 196-209. Bogor, 21-22 October 2015 218 PAPER B11 - Uncertainties of above ground biomass estimates in tropical peat swamp forest Solichin Manuri 1 , Shijo Joseph 2 , Christopher Martius 2 , Wiyono 2 , Cris Brack 1 1 The Australian National University, Linnaeus way 48 Acton, ACT 2601, Australia 2 Center for International Forestry Research, Center for International Forestry Research CIFOR, Jl. CIFOR, Situ Gede, Bogor 16115, Indonesia Corresponding Email: solichin.solichinanu.edu.au ABSTRACT Recently there has been renewed interest in accurate estimation of forest carbon stock in the context of climate change mitigation in the forestry sector. Many studies were carried out focusing on ground measurements and remote sensing techniques for the improvement of above ground biomass estimation at plot and landscape scales. However, most studies neglected the importance of above ground biomass AGB model selection. Our study aims at assessing existing models for estimating tree height and AGB in tropical peat land forest. We use destructive sampling data as reference value. In August 2014, ten trees from mixed species and maximum diameter of 94 cm were cut down and measured in a peat swamp forest of Central Kalimantan, Indonesia. We evaluated existing regional and local equations for tree height and AGB estimation. We found that all existing models showed mean absolute errors between 28 to 83 and 19 to 51 for tree height and AGB models, respectively. Tree height models tend to over- or under-estimate the reference values. Moreover, the use of tree height model into AGB model propagates these errors further into the AGB estimates. We also found that regional AGB model, which developed using datasets from Kalimantan and Sumatra, outperformed other local equations. These findings suggest that existing tree height and AGB models should only be used if cautiously validated through data, in particular for the tree height models which are used as a basis for the AGB. Keywords: allometric equation, tree height model, destructive sampling, logging damage, forest biomass 1. INTRODUCTION Reducing emissions from deforestation and forest degradation or enhancing forest carbon stocks REDD+ has been promoted to ensure participation of developing countries, especially those with substantial tropical forests, in reducing CO 2 emissions from forest related activities. Emissions from tropical deforestation and forest degradation in Indonesia account for 47 of national emissions. Emissions from peat swamp forest PSF alone contribute to 13 of national emissions MoE, 2010. This sector could be a potential contributor for the targets for emission reduction as set by the Indonesian Government at 26 and 41 of national emission by 2020, without and with support from international donors, respectively GoI, 2011. In order to meet the target, Indonesia implemented a moratorium for granting new licenses for forest utilization in natural primary forest or peat swamp forests in 2011. The policy was extended in 2015 and the moratorium areas were being enlarged. However, lack of law enforcement lead to pressures from timber demands to extract the timbers from τno man’s forests”Luttrell et al., 2011. During 2011-2013, deforestation rate in Indonesia increased in Bogor, 21-22 October 2015 219 spite of the policy Hansen Loveland, 2012; Margono, Potapov, Turubanova, Stolle, Hansen, 2014. Community logging in Kalimantan, Indonesia has a long history. Small scale community logging, small government program and NGO projects Fahmi, Zakaria, Kartodihardjo, Wahono, 2003; Ravenel, 2004 were promoted during the Soeharto era between 1970 and 1998. The first was often referred to as illegal logging Obidzinski, 2005. After decades of struggle with central government and large scale timber companies for benefit rights over forests, local governments granted many small scale logging permits for community logging during the beginning of decentralisation era in 1999. However, these smale scales activities lead to further depletion of forest resources Resosudarmo, 2004. In 2003, central government revoked the regulation and the permits due to the apparent failure of the small logging scheme. Small scale logging, nevertheless, continues especially in areas without timber concessions Englhart, Jubanski, Siegert, 2013; Luttrell et al., 2011. The quantification of the degradation level after selective logging using the current remote sensing technology is still problematic at large scales. Several studies make use of current technology based on satellite imageries and airborne LiDAR to asses forest degradation with promising results Asner et al., 2012; Boehm, Liesenberg, Limin, 2013; Franke, Navratil, Keuck, Peterson, Siegert, 2012. Validation of remote sensing data using site-specific ground measurements is still crucial. Remote sensing-based biomass assessments still rely on accurate ground measurement to develop the relationship between remote sensing derived parameters and biomass value. However, measurement of biomass directly is very expensive and time consuming, so it is common to rely on allometry and existing or derived model equations. Therefore the choice of appropriate AGB models is essential. The total residual error from individual trees is reduced at landscape level, because the errors from inaccurate equations were compensated each other Mavouroulou et al., 2014. Due to the variation in degradation level, tree species composition, site characteristics and the DBH-H relationship among different sites, a site specific relationship between remote sensing-derived parameters and measured AGB are required for accurate estimation. Previous studies confirmed that inclusion of H into AGB model increased the accuracy of the estimates Chave et al., 2005, which leads to a suggestion of improving estimates by incorporating the H model into AGB model Feldpausch et al., 2010. We carried out an assessment to evaluate and validate existing models for estimating AGB in peat swamp forest in Central Kalimantan. Most of previous studies failed to validate the models prior to use. Objectives of this study are 1 to select the best tree height model from existing models and 2 to validate existing biomass equations using destructive sampling dataset. 2. MATERIAL 2.1 Study site The study was conducted in peat swamp forest close to Mentaya River in Central Kalimantan, Indonesia. An average annual rainfall of 3287 mm was recorded from nearby Sampit airport from 1997 – 2010. The area is designated by law as production forest and was under selective logging regime since 1970s previously by large companies. Most of large dominant dipterocarp trees were harvested until 1990. In 1980s, ramin Gonystylus bancanus was economically interesting to be harvested leading to over exploitation of the species before it came under protection. Starting in 1990s, as the tree species composition altered, the locals started cutting Bogor, 21-22 October 2015 220 all large trees that were left, such as punak Tetramerista glabra and puri Dyospiros sp. Therefore the existing standing large dipterocarp trees found either have stem defects or hollow stems. Since 2000, the area is being managed by local communities for extracting smaller size and less commercial tree species. 2.2 Tree height measurements A total of 69 trees were selected for total height measurement Table 1. We measured tree heights using a Nikon Forestry laser hypsometer which has a design precision of better than 1. Only trees where the tree tops were visible were measured to allow good accuracy of total height measurements. We measured the tree height in the forest gaps in which felling recently took place to ensure better visibility of the base and the top of targeted trees. The height of harvested trees was measured after felling. Table 1: Data set used for this study Data type N DBH range cm H range m Tree height 69 10.1 – 94.0 8.7 - 38.3 AGB 15 22.0 – 94.0 19.0 – 38.3 2.3 Destructive sampling data We undertook destructive sampling for calculating volume and biomass of 15 crop trees. The sampled trees were fractioned into parts: stem; very large branches D 30; large branches 20 D 30; medium branches 10 D 20; small branches 3 D 10; twigs D 3 cm, leaves and fruits. All tree parts were weighed using digital scales, except for stem and branches with D 10 cm, which were measured only for diameter at both sides of each 1-m section and later calculated the volume using the Smalian formula. Hollows, if any, were measured at the base and end sections of each log. Samples were taken from each of the tree fractions. This included samples from lower and upper part of the stem and at various sizes of branches. Samples were weighted and labelled before being transported to Palangkaraya University Analytical Laboratory for wood density WD and dry massfresh mass ratio DMFM analysis. We calculated the fresh volume using water displacement method. Samples were saturated in water for 48 hours prior the measurement. After being oven-dried at 105 o C until reaching constant mass, all samples’ dry weight were recorded. Further the measured volume and fresh mass data were converted into AGB by multiplying with WD and DMFM ratio, respectively. 3. METHOD 3.1. Tree height model development and validation Using our dataset, we developed local tree height H models using the Weibull function model form as suggested by Feldpausch et al. 2012 see equation 1. 1 We examined the goodness of fit, RMSE and residual plot of the developed H model. We then compared the accuracy of the estimates with the accuracies of other existing H models. Several tree height H models see table 1 have been developed globally Chave et al., 2014, regionally Feldpausch et al., 2012; Rutishauser et al., 2013 and locally Boehm et al., 2013; Yamakura, 1986. We validated the performance of each models by comparing the estimated Bogor, 21-22 October 2015 221 tree heights with the tree heights measured in the field using laser hypsometer and destructive sampling method. Table 2: Existing H models used for validation H Model Equation n Source Glob H = exp0.893 – E i + 0.76 × lnD – 0.034 × lnD 2 . E i = -0.105396 Chave et al, 2014; pantropical dataset SEA H = 57.1221-exp-0.0332D 0.8468 2948 Feldpausch, 2012; South East Asia dataset Sum H = 56.7031-exp-0.0547D 0.739 4013 Rutishauser et al, 2013 Sumatra, Indonesia dataset Kal H = 1989.1441-exp-0.0018D 0.5306 3192 Rutishauser et al, 2013; Kalimantan, Indonesia dataset EastKal H = 111.757D+188.5 221 Yamakura et al, 1986; lowland dipterocarp forest; East Kalimantan dataset PeatCK H =-0.0007D 2 +0.2446D+5.5256 Boehm et al, 2013 Peat swamp forest Central Kalimantan, Indonesia dataset The best H model was selected based on the smallest percent true error PTE, mean percentage error MPE, the smallest mean absolute percentage error MAPE Sileshi, 2014, regression slope and intercept closest to 1 and 0, respectively Manuri et al., 2014. To calculate PTE, MRE and MAPE, we used equations 2, 3 and 4, where x p and x m are the predicted and measured values, respectively. ∑ ∑ ∑ 2 ∑ 3 ∑ | | 4 3.2 AGB models validation Due to limited harvested tree samples, we did not develop any new AGB equation. Instead, we validated existing PSF AGB equations using our destructive sampling data to select the best model. Several existing equations from peat swamp forests, developed regionally from western Indonesia Manuri et al., 2014 and locally from Central Kalimantan Dharmawan, Saharjo, Supriyanto, HS, Siregar, 2013; Jaya, Siregar, Daryono, Suhartana, 2007 were selected for this analysis see Table 3. As the study site has been intensively logged and is without silvicultural treatment, lower-quality trees and fast growing species are dominating the site. We selected the equations from Manuri et al. 2014 that represent similar tree and stand quality. We compared the PTE, MPE and MAPE of each model. Additionally, we compared the measured and estimated AGB values of the existing models by fitting linear regressions. The same statistical criterion for the selection of the best H model was used for the choice of the best AGB equation, i.e. PTE, MPE and MAPE. Bogor, 21-22 October 2015 222 Table 3: Existing AGB models used for comparison Model type Localities Equations Sources M D1 Peat swamp forest; Central Kalimantan 0.107D 2.468 Jaya et al., 2007 M D2 Peat swamp forest; Western Indonesia 0.136D 2.513 Manuri et al., 2014 M D3 Species group; peat swamp forest; Western Indonesia Dipterocarp: 0.108D 2.562 Manuri et al., 2014 Non-dipterocarp hardwood: 0.138D 2.537 Non-dipterocarp softwood: 0.149D 2.399 M DWH1 Peat swamp forest Central Kalimantan 0.061 D × W × H 1.464 Dharmawan, 2013 M DWH2 Peat swamp forest Central Kalimantan 0.040 D × W × H 1.524 Dharmawan, 2013 M DWH3 Peat swamp forest; Western Indonesia 0.15D 2.095 W 0.664 H 0.552 Manuri et al., 2014 M DWH4 Species group; peat swamp forest; Western Indonesia Dipt: 0.068D 1.662 W 0.352 H 1.230 Manuri et al., 2014 Non-dipterocarp hardwood: 0.077D 1.871 W 0.669 H 1.008 Non-dipterocarp softwood: 0.152D 2.099 W 0.427 H 0.369 Notes: D is diameter at breast height or above buttress, W is wood density and H is tree height

4. RESULT AND DISCUSSION