1. MAPPING OF CITES-LISTED ENDANGERED TROPICAL PEAT SWAMP FOREST TREE SPECIES USING AIRBORNE HYPERSPECTRAL SENSOR Khali Aziz Hamzah, Mohd Azahari Faidi and Hamdan Omar Forest Research Institute Malaysia (FRIM) ASIA GEOSPATIAL FORUM 17 – 19 SEPTEMBER 2012, HANOI, VIETNAM

2. 2 1. Introduction 2. The Project 3. Result and Discussion 4. Conclusion PRESENTATION OUTLINE

3. MALAYSIA – FOREST AREA 3 Malaysia total land area - about 328,300 km 2 About 59.5% (19.52 million ha) is still under forest cover RegionLand area (mil ha) Natural Forest Types Plantation forest Total Forested land % of total land Area Dry inland Swamp forest Mangrove forest Pen. Malaysia 13.165.400.300.100.085.8844.7 Sabah7.373.830.120.340.114.4059.7 Sarawak12.307.921.120.140.069.2475.1 Total (Malaysia) 32.8317.151.540.580.2519.5259.5

4. South China Sea Strait Of Malacca Sulu Sea MALAYSIA – FOREST MAP

5. 5 Ramin (Gonystylus bancanus) is an endangered peat swamp forest species Ramin (Gonystylus bancanus) is an endangered peat swamp forest species It has been listed in Appendix ll of the Convention on International Trade in Endangered Species of Wild Fauna and Flora (CITES) It has been listed in Appendix ll of the Convention on International Trade in Endangered Species of Wild Fauna and Flora (CITES) There is a need to identify the ramin population in the natural habitat for management purposes There is a need to identify the ramin population in the natural habitat for management purposes ISSUES

6. Scientific name: Gonystylus bancanus Family: Thymelaeaceae Local name: Ramin Brief description: large tree and can grow up to 40 m in height Uses: timber for high quality furniture Ecology: gregarious in peat swamp forests ABOUT THE SPECIES

7. fruits seedling flower timbers tree Ramin – Gonystylus bancanus

8. RESEARCH JUSTIFICATION 8 To identify and locate ramin trees in a highly mixed peat swamp forest is a challenging task. To identify and locate ramin trees in a highly mixed peat swamp forest is a challenging task. The ideal way is to inventories the whole population, but this will be very expensive to implement in the field. The ideal way is to inventories the whole population, but this will be very expensive to implement in the field. Opportunities on mapping using hyperspectral remote sensing technology Opportunities on mapping using hyperspectral remote sensing technology

9. Ramin is considered canopy layer treeRamin is considered canopy layer tree Advantage to use airborne hyperspectral dataAdvantage to use airborne hyperspectral data Forest profile of Plot E3: 39, Baccaurea bracteata; 54, Blumeodendron tokbrai; 14, 61, Calophyllum ferrugineum; 27, Calophyllum sclerophyllum; 47, Camnosperma coriaceum; 2, 33, 49, 50, Diospyros lanceifolia; 53, Diospyros maingayi; 18, 32, Durio carinatus; 3,9,17,24,Gonystylus bancanus ; 21, 29, 40, Koompassia malaccensis; 15, 30, Licania splendens; 43, 58, Litsea elliptica; 36, 38, 44, 56, 60, Litsea gracilipes; 51, Litsea grandis; 12, 16, 41, Lophopetalum floribundum; 45, Lophopetalum multinervium; 6, 7, 8, 11, 20, 55, 64, Neoscortechinia forbesii; 10, Palaquium ridleyi; 22, 37, 52, Parastemon urophyllus; 23, 48, Polyalthia glauca; 5, Polyalthia hypoleuca; 1, 57, Shorea platycarpa; 13, Syzygium cerinum; 19, 26, 31, 46, Syzygium inophyllum; 34, 62, Syzygium kiahii; 25, 35, Syzygium lineatum; 4, Tetractomia majus; 59, Xantophyllum ellipticum; 28, 42, Xylopia magna. RAMIN IN THE FOREST

10. THE PROJECT 10 Objective: To generate spatial distribution maps of ramin in peat swamp forest using hyperspectral technology. To generate spatial distribution maps of ramin in peat swamp forest using hyperspectral technology. Expected Output: Spectral library, spatial distribution maps and spatial database for ramin Spectral library, spatial distribution maps and spatial database for ramin

11. Peat Swamp Forest, Pekan, Pahang MALAYSIA PROJECT AREA

12. PEAT SWAMP FOREST DURING DRY SEASON

13. 13 PEAT SWAMP FOREST DURING WET SEASON

14. THE METHODOLOGY Spectral Library Field Survey Ramin Distribution Map Yes No Airborne Data Acquisition Data Pre-Processing Data Classification Discriminating Ramin Vectorisation Accuracy Assessment X,Y location height DBH (>20cm) crown width

15. SPECTRORADIOMETER DATA COLLECTION 15 Spectral Range 350 nm to 1050 nm Bandwidth interval 1.5 nm Technical Specifications:

16. AIRBORNE HYPERSPECTRAL DATA - Sensor : HySpex VNIR-1600 - Spatial resolution, 0.5m - Number of bands, 160 - Spectral Range : - Swath width,1km - Spectral Range : 0.4-1μm - Swath width,1km

17. AIRBORNE HYPERSPECTRAL DATA 17 Acquiring of airborne hyperspectral data in the study area HySpex V-NIR 1600 Hyperspectral system installations Aircraft (9M - PIH)

18. RESULT & DISCUSSION 18 StatisticGonystylus bancanus (%) Calopyllum ferrugineum (%) Minimum2.26.2 Maximum57.959.6 Mean26.931.7 SPECTRAL LIBRARY

19. RESULT & DISCUSSION 19 AIRBORNE HYPERSPECTRAL DATA

20. Spectral dimension Spatial dimension Pixel RESULT & DISCUSSION HySpex VNIR data cube HySpex VNIR data cube Ramin spectral signature Meranti Paya spectral signature Understory spectral signature Bintangor spectral signature

21. RESULT & DISCUSSION Spectral signature of selected features in hyperspectral imagesSpectral signature of selected features in hyperspectral images

22. RESULT & DISCUSSION Correlation (ramin reflectance) between hyperspectral data and spectroradiometer measurementCorrelation (ramin reflectance) between hyperspectral data and spectroradiometer measurement

23. RESULT & DISCUSSION 23 o Using Spectral Angle Mapper (SAM) classification technique the hyperspectral data can be used to map ramin distributions o It was found that the distribution of ramin within the study area is about 21 tree per ha, o Mapping accuracy of 86%

24. RESULT & DISCUSSION 24 Virgin Forest : Ramin – 491/25ha (19.64/ha) Ramin – 491/25ha (19.64/ha) Bintangor – 353/25ha (14.12/ha) Bintangor – 353/25ha (14.12/ha) Logged Over Forest: Ramin – 179/30ha (5.97/ha) Ramin – 179/30ha (5.97/ha) Bintangor – 50/30ha (1.67/ha) Bintangor – 50/30ha (1.67/ha) Ramin - Mean height, 34.70m - Mean height, 34.70m - DBH, 21.6 – 83.5cm - DBH, 21.6 – 83.5cm

25. 25 CONCLUSION Spectral library of the Ramin trees has been developed and can be used as reference spectral library for the future research project. Spectral library of the Ramin trees has been developed and can be used as reference spectral library for the future research project. Ramin trees can be identified using hyperspectral data with acceptable mapping accuracy (86%). Ramin trees can be identified using hyperspectral data with acceptable mapping accuracy (86%). Inventory of Ramin can be carried out faster. Inventory of Ramin can be carried out faster. The availability of accurate information on ramin population from this study can be used to assist in designing rehabilitation and conservation programs in order to conserve and sustainably manage this species in line with the CITES requirements. The availability of accurate information on ramin population from this study can be used to assist in designing rehabilitation and conservation programs in order to conserve and sustainably manage this species in line with the CITES requirements.

26. 26 This work was made possible by a grant from ITTO under its collaborative program with CITES and Malaysia to build capacity for implementing timber listings. Donors to this collaborative program include the EU (primary donor), the USA, Japan, Norway, New Zealand and Switzerland. Acknowledgements Thank You