Forest Nonforest FNF mapping for Viet Nam using

Forest / Non-forest (FNF) mapping for Viet Nam using PALSAR-2 time series images 2019/01/22 Truong Van Thinh – 201726070 Master’s Program in Environmental Sciences Supervisor: Prof. Kenlo Nishida Nasahara 1

The importances of forest Climate change (GHG’s emission) Ecosystem Biodiversity Forest Livelihood 2

Forest mapping and its issues Climate modeling Hydrological study Natural disasters study Input Forest map Survey mapping - Hight cost, - Limited area, - Laborious work Satellite images Aircraft photos -Low cost, -Large area coverage, -Less human labor intervention. 3

Satellite sensors and their advantage Optical satellite Landsat 8 (NASA) cloud cover Optical image Synthetic Aperture Radar (SAR) ALOS-2 (JAXA) No cloud SAR image 4

Forest / Non-Forest products JAXA (Japan Aerospace Exploration Agency) JAXA Forest / Non-forest maps Global Forest cover (Hansen et al. 2013) 5

The differences in JAXA FNF 2015_JAXA 2016_JAXA 2017_JAXA 6

Objectives 1. To produce forest / non -forest maps for Viet Nam: - with high accuracy - consecutive maps from 2014 to 2018 Forest monitoring and management in Viet Nam 2. To analyze forest cover change for Viet Nam during 2014 - 2018 7

Study area ● Located in Southeast Asia ● Total area: 332, 698 km 2 ● ~ 42 % of land area is covered by forest (GSO, 2017) Google terrain map of Vietnam 8

Method Satellite images Input data Algorithm Output map Training data 9

Method (cont. ) DATA SRTM Slope Scan. SAR time series MODIS Field GPS photos Convert DN to Backscatter NDVI Backscatter value (HH, HV, HH/HV, HH-HV) DATA PROCESSING Google Earth Training Data Validation Data Median filter CLASSIFICATION SACLASS Land cover maps OUTPUT MAPS Forest / Nonforest maps Validation 10

Satellite data Data Provider PALSAR-2 (Scan. SAR) JAXA MODISNDVI USGS SRTM Quantity (scene) 2014: 42 2015: 176 2016: 184 2017: 218 2018: 174 Time Resolution HH, HV 50 m 2014: 42 2014, 2015, (U. S. 2015: 176 2016, 2017, Geological 2016: 184 2018 survey) 2017: 218 2018: 174 NDVI 250 m USGS DEM 30 m 26 2014, 2015, 2016, 2017, 2018 Band 2002 11

Reference data Visual interpretation on Google earth 48, 957 training point The distribution of training data 21, 452 validation points The distribution of validation data 12

Results and Discussion Land cover map 2014 Land cover map 2015 Land cover map 2017 Land cover map 2016 Land cover map 2018 13

Integration of land cover maps into FNF maps Reclassification Land cover maps classes FNF classes 14

Results of Forest / Non-forest maps FNF map 2014 FNF map 2015 FNF map 2017 FNF map 2016 FNF map 2018 15

Forest area calculation based on FNF maps No. FNF map Number of forest pixels Forest area (ha) Total area (ha) Forest coverage (%) 1 2014 36, 697, 345 9, 174, 336 16, 440, 050 55. 8 2 2015 38, 649, 189 9, 662, 297 16, 440, 050 58. 8 3 2016 38, 145, 176 9, 536, 294 16, 440, 050 58. 0 4 2017 38, 387, 864 9, 596, 966 16, 440, 050 58. 4 5 2018 38, 965, 344 9, 741, 336 16, 440, 050 59. 3 16

Overall accuracy (OA) assessment No. Year OA of LULC (%) OA of FNF (%) 1 2014 64 86 2 2015 74 90 3 2016 73 91 4 2017 73 90 5 2018 73 90 17

Compare with JAXA FNF N 22 E 104_Scan. SAR N 22 E 104_JAXA 18

Future work 1. Continuing to make training data for the central region and southern Viet Nam by: + Visual interpretation + Field trip to Viet Nam on March 2019 2. Do FNF classification for the rest part of Viet Nam 3. Analyzing forest change and compare with statistical data of Viet Nam and other FNF global maps 19

THANK YOU FOR YOUR LISTENING! 21

Supplementary 22

23

24

25

26

27

Forest gain and lost between 2015 and 2018 28

2 9

30