1. Integrating Airborne LiDAR and Terrestrial Laser Scanner for Accurate Estimation of Above-ground Biomass/Carbon of Tropical Forests Accuracy Matters Muluken N. Bazezew *, Yousif A. Hussin, E. H. Kloosterman, Ismail M. Hasmadi 3 rd International Convention on Geosciences and Remote Sensing Oct. 19-20, 2018 Ottawa, Ontario, Canada Dilla University Ethiopia ITC, The Netherlands University Putra Malaysia 1

2. OUTLINES ___________________________________________________________________________________________________________________________________________________________ 3 rd International Convention on Geosciences and Remote Sensing  INTRODUCTION  OBJECTIVES  RESEARCH METHODS  RESULTS AND DISCUSSION  CONCLUSION AND RECOMMENDATIONS  ACKNOWLEDGMENTS 2

3. INTRODUCTION ___________________________________________________________________________________________________________________________________________________________ 3 rd International Convention on Geosciences and Remote Sensing  GHGs emissions; Impact on climate  UNFCCC, Kyoto Protocol (192 countries) 3

4. 3 rd International Convention on Geosciences and Remote Sensing 4 INTRODUCTION ___________________________________________________________________________________________________________________________________________________________

5. 3 rd International Convention on Geosciences and Remote Sensing The growing need of REDD+ MRV!! 5 INTRODUCTION ___________________________________________________________________________________________________________________________________________________________

6. 3 rd International Convention on Geosciences and Remote Sensing To meet the requirements of REDD+ MRV for Accurate Inventory !! 6 INTRODUCTION ___________________________________________________________________________________________________________________________________________________________

7. 3 rd International Convention on Geosciences and Remote Sensing ALS TLS  Integration of ALS and TLS for accurate topical forest monitoring  ALS- Upper canopy trees characterization  TLS- Lower canopy trees characterization 7 INTRODUCTION ___________________________________________________________________________________________________________________________________________________________ To meet the requirements of REDD+ MRV for Accurate Inventory !!

8. OBJECTIVES ___________________________________________________________________________________________________________________________________________________________ 3 rd International Convention on Geosciences and Remote Sensing The aim of this research is to develop an approach for estimating accurate AGB/Carbon of the tropical rainforests with integrating Airborne LiDAR Scanning (ALS) and Terrestrial Laser Scanning (TLS). Specifically, in the processing-chain: o Assess LiDAR-CHM in tree crown delineation of the tropical forests o Assess the accuracy of forest parameters (DBH, height) measurement with ALS and TLS o Compare the estimated AGB/Carbon between RS (ALS + TLS) and traditional field-based methods 8

9. STUDY SITE ___________________________________________________________________________________________________________________________________________________________ 3 rd International Convention on Geosciences and Remote Sensing  Located 3º0’0” to 3º2’0” lat. and 101º38’0” to 101º40’0” lon.  Elevation (15 to 233 m a.s.l).  Encompasses 430 plant species, and >60% of its emergent and middle canopy trees, and the rest understory trees and shrubs. 9

10. METHODS ___________________________________________________________________________________________________________________________________________________________ 3 rd International Convention on Geosciences and Remote Sensing  TLS data acquiring, processing and analyzing  ALS data acquiring, processing and analyzing  Data Integration 10 Various Image processing, and analysis software were used: ERDAS Imagine and ENVI for image processing, LAStool for ALS point cloud data processing, RiSCAN PRO for TLS point cloud data processing, eCognition for ALS-CHM segmentation, ArcGIS, R-studio.

11. TLS and Dataset Processing ___________________________________________________________________________________________________________________________________________________________ 3 rd International Convention on Geosciences and Remote Sensing RIEGL VZ-400 Terrestrial Laser Scanning sensor features Measurement range1.5 - 600 m Precision3 mm Accuracy5 mm Beam divergence0.35 mrad Footprint size at 100m30 mm Measurement (pulse) rate44 - 122 kHz Scan angle range (degree)100 o (+60 o / -40 o ) Laser wavelengthNear-infrared (1550 nm) GPS receiver Integrated, L1, with antenna Scanning mechanismRotating multi-facet mirror Scan speed3 - 120 lines/sec Weight9.6 kg Operating temperature0 to +40 o C; standard operation HumidityMax. 80%, non-condensing at +30 o C 11

12. 3 rd International Convention on Geosciences and Remote Sensing 12 TLS and Dataset Processing ___________________________________________________________________________________________________________________________________________________________

13. 3 rd International Convention on Geosciences and Remote Sensing 13 TLS and Dataset Processing ___________________________________________________________________________________________________________________________________________________________ Measuring with RiSCAN PRO

14. ALS and Dataset Processing ___________________________________________________________________________________________________________________________________________________________ 3 rd International Convention on Geosciences and Remote Sensing Sensor FeatureDescription Pulse rate Range between 70 kHz and 240 kHz Scan angle60° Scan patternRegular Beam divergence0.5 mrad Line/secMax. 160 A/c ground speed90 kts Target reflectivity 20 - 60% (vegetation 30%, cliff 60%) Flying height700 - 1000 m Laser points/m 2 5 to 6 points with 808 m to 1155 m swath width Spot diameter (laser)0.35 - 0.50 m Max (above ground level)1040 14

15. 3 rd International Convention on Geosciences and Remote Sensing 15 ALS and Dataset Processing ___________________________________________________________________________________________________________________________________________________________

16. 3 rd International Convention on Geosciences and Remote Sensing ALS-TLS Integration ___________________________________________________________________________________________________________________________________________________________ 16

17. RESULTS ___________________________________________________________________________________________________________________________________________________________ 3 rd International Convention on Geosciences and Remote Sensing TLS-based DBH Accuracy Tree parameter No. of ObservationsR2R2 r RMSE Bias (cm) (cm)(%) DBH7350.980.991.306.52– 0.52 17

18. 3 rd International Convention on Geosciences and Remote Sensing Reference polygons 1:1 matched polygons Over- segmentation Under- segmentationGoodness of fit (D) 1321030.240.300.27 Accuracy (%)7873 18 Segmentation Accuracy and canopy cross-sectioning RESULTS ___________________________________________________________________________________________________________________________________________________________

19. 3 rd International Convention on Geosciences and Remote Sensing The average success rate of treetops detection by LiDAR. Success rate displays the number of trees (%) that their treetops identified from ALS or TLS. 19 RESULTS ___________________________________________________________________________________________________________________________________________________________  Overall- ALS treetop detection potential is 57% of field identified trees, while TLS is 37%.

20. 3 rd International Convention on Geosciences and Remote Sensing LiDAR (ALS and TLS)-based trees height accuracy Tree ParameterNo. of ObservationsR2R2 r RMSE Bias (m) (m)(%) Upper canopy trees height4510.610.783.2420.18– 1.20 Lower canopy trees height2900.690.831.4514.770.42 20 RESULTS ___________________________________________________________________________________________________________________________________________________________

21. 3 rd International Convention on Geosciences and Remote Sensing LiDAR- Vs Field-based AGB/C Method No. of sampled plotsrInterceptSlopeP valueRMSERMSE (%) TLS270.930.49551.0399< 0.00011.08813.31 ALS270.951.14770.8418< 0.00010.96411.79 TLS-ALS Integration270.98-0.14440.909< 0.00010.6247.64 21 RESULTS ___________________________________________________________________________________________________________________________________________________________

22. 3 rd International Convention on Geosciences and Remote Sensing  There was a significant difference between field and various RS techniques- based AGB.  Traditional field-based methods underestimated the AGB.  Complementary use of TLS with ALS improved the accuracy of estimating AGB/C.  If the goal is to get highly accurate AGB/C, integration of TLS and ALS should be chosen.  If the simple parsimonious model is desired, then ALS-based AGB model could be chosen. Method No. of sampled plotsrInterceptSlopeP valueRMSERMSE (%) TLS270.930.49551.0399< 0.00011.08813.31 ALS270.951.14770.8418< 0.00010.96411.79 TLS-ALS Integration270.98-0.14440.909< 0.00010.6247.64 22 RESULTS ___________________________________________________________________________________________________________________________________________________________

23. 3 rd International Convention on Geosciences and Remote Sensing Statistics RS-based AGB (Mg) Field-based AGB (Mg) Combination of upper and lower canopies AGB (Mg) Upper canopy trees (ALS) Lower canopy trees (TLS) Upper canopy trees Lower canopy trees RS (ALS + TLS) methods Field- methods Mean/Plot8.4450110.7085607.4084480.767789.15368.1762 Std. Dev.3.2713950.3767382.9899250.415203.31463.0655 Min.3.4315900.0916333.2838450.024663.98073.9390 Max.17.891051.62106215.561551.9261518.231616.0022 Sum228.015319.13112200.028120.7302247.146220.758  Of a total AGB calculated from RS, 92% is the upper canopy trees AGB, and 8% is lower canopy trees AGB (168.8 Mg.ha -1, 14.2 Mg.ha -1, respectively).  It implies that it was able to capture an average of 14.2 Mg.ha -1 (0.71 Mg.plot -1 ) with the compliment of TLS.  Field methods underestimated AGB by 19.547 Mg.ha -1, accordingly for about 10.70%. 23 RESULTS ___________________________________________________________________________________________________________________________________________________________

24. CONCLUSION AND RECOMMENDATIONS ___________________________________________________________________________________________________________________________________________________________ 3 rd International Convention on Geosciences and Remote Sensing  Using ALS for detecting single tree in the complex biophysical structure of tropical rainforest have a potential to recognize not more than two-thirds of the trees.  The dense canopy and interlocking crown structure of the tropical forest results in a low accuracy of crown segments.  The TLS-based DBH measurements method used in this paper through distance function algorithm in the RiSCAN PRO offers more accurate result than the automatic detection of trees and determination of DBH.  Results from model evaluations based on ALS and TLS dataset proof that this approach can enhance the accuracy of predicted AGB or carbon stock than traditional field-based.  Integration enables to detect a comparable number of trees identified in the field. 24

25. 3 rd International Convention on Geosciences and Remote Sensing Main sources of errors  Field-based tree height measurements (with Leica laser DISTO D510)  Hunter et al. (2013) have reported that height error ranging from 3–20% of the total height results contribute to 5–6% of uncertainty in estimated biomass. 25 CONCLUSION AND RECOMMENDATIONS ___________________________________________________________________________________________________________________________________________________________

26. 3 rd International Convention on Geosciences and Remote Sensing  Although the number of plots acquired in this study was small, the range of canopy structures and the approach used for individual tree parameters measurement provide a clear indication of the potential of integrating ALS and TLS system.  However, the TLS data acquisition through observations of plots from multiple scanning viewpoints to reduce the occlusion effect requires more investigation.  The proposed approach can also be used to accurately predict forest structural variables other than AGB, such as stand density, basal area, and stand volume. 26 CONCLUSION AND RECOMMENDATIONS ___________________________________________________________________________________________________________________________________________________________

27. ACKNOWLEDGMENT ___________________________________________________________________________________________________________________________________________________________ 27 Organizers of 3 rd International Convention on Geosciences and Remote Sensing

28. 3 rd International Convention on Geosciences and Remote Sensing Oct. 19-20, 2018 Ottawa, Ontario, Canada Dilla University Ethiopia ITC, Netherlands University Putra Malaysia THE END THANK YOU FOR LISTENING!! Integrating ALS and TLS for Accurate Tropical Forest Monitoring 28

29. Email: mulukenn@du.edu.et Tel.: +251961013618 (M.N. Bazezew) Corresponding Author Address: Muluken N. Bazezew Dilla University College of Agriculture and Natural Resources P.O. Box: 419 Dilla, Ethiopia 29