ESTIMATING ABOVEGROUND BIOMASS OF A TROPICAL FOREST IN NORTHERN BORNEO BASED ON INDIVIDUAL TREE CROWNS FROM IKONOS 2 DATA Mui-How PHUA a* , Zia-Yiing LING a , Wilson WONG a , Alexius KOROM a , Berhaman AHMAD a , Normah A. BESAR a , Satoshi TSUYUKI b , Keiko IOKI b , Keigo HOSHIMOTO b , Yasumasa HIRATA c , Hideki SAITO d , Gen TAKAO c a School of International Tropical Forestry, Universiti Malaysia Sabah, Jalan UMS, 88400 Kota Kinabalu, Sabah, Malaysia; Tel +6088-320765; E-mail: pmh@ums.edu.my b Graduate School of Agricultural and Life Sciences, The University of Tokyo, 1-1-1 Yayoi, Bunkyo-ku, Tokyo, Japan 113-8657 c Forestry and Forest Products Research Institute (FFPRI), Tsukuba 305-8687, Japan d Kyushu Research Center, FFPRI Kumamoto, Kumamoto 860–0862, Japan KEY WORDS: Biomass estimation, Tropical forest, Tree crown detection, IKONOS 2 Abstract: Estimation of AGB and carbon of tropical forests has been a crucial factor to ‘Reduction of carbon emissions from deforestation and forest degradation-plus’, a monetary mechanism for combating global warming. We examined the use of IKONOS 2 data for estimating aboveground biomass of a tropical forest in Long Mio, Sabah, Malaysia. The IKONOS 2 data was corrected for path radiance and topographic effect before segmentation using the watershed method in ArcGIS. Correlation analyses between the segment variables and forest variables revealed that crown area extracted from the IKONOS 2 data (CA s ) had the strongest correlation with the field-measured diameter at breast height (DBH f ). Linear regression analysis produced a model to estimate DBH f using CA s with R 2 of 0.71. Cross- validation using the remaining half of the data set also showed very consistent result. The CA s only represented upper canopy trees but estimated about 80% of the aboveground biomass. INTRODUCTION Tropical forest is important for carbon sink and source in global carbon cycling. Total 15 – 25 % of the global greenhouse emission per year comes from tropical forest (Houghton, 2005; Malhi and Grace, 2000). Estimation of aboveground biomass (AGB) of tropical forests has been an important input to ‘Reduction of carbon emissions from deforestation and forest degradation- plus’, a monetary mechanism for combating global warming. Accuracy of the estimation is important for reporting carbon stocks and carbon changes to the Kyoto Protocol (Muukkonen and Heiskanen, 2007). In fact, the AGB estimation methods must be verifiable, specific in time and space as well as covering large area at an acceptable cost. Remote sensing with combination of ground inventory data is a reliable approach for AGB estimation. Many tropical forest AGB estimation studies were carried out using medium resolution satellite image such as Landsat (Phua and Saito, 2003), ALOS-PALSAR (Morel et al., 2012) and Landsat ETM+ (Langner et al., 2012). Statistical models based on spectral band or vegetation index has been derived from medium resolution satellite image for AGB estimation. However, these models tend to underestimate tropical forest AGB due to dense canopy structure (Gibbs et al., 2007). Recently, tree crown detection using high resolution satellite images have been used to examine forest canopy structure (Pouliot and King, 2005, Wang et al., 2004 & Kubo and Muramoto, 2005). The high resolution remote sensing approach detects crown variables of upper canopy trees that comprise of most of the AGB. Significant relationships between ground measured AGB and crown variables derived from the high resolution satellite image have been found for temperate forests (Hirata et al., 2009; Wulder et al., 2002). This study examined the use of IKONOS 2 data for delineating individual tree crowns of tropical forest. We then explored the potential of the crown variable extracted from the IKONOS 2 data for constructing an AGB estimation model. 1