Forest Dragon 3 Project Id

Transcription

1 Forest Dragon 3 Project Id Principle Investigator: Co-Investigator: Young Scientists: Prof. Li, Academy of Forest Sciences Prof. Schmullius, University of Jena Prof. Pang, Dr. Feilong, Dr. Santoro Wang Zhichao, Johannes Balling, Patrick Schratz, John Truckenbrodt

2 The seven objectives of the FOREST DRAGON 3 project are 1) the investigation of scaling effects in forest ecosystem mapping, 2) the long-term analysis of GSV and forest structure over NE China, 3) linking forest DRAGON products with existing land use, land cover and/or fire products and 4) the synergy of optical and radar data for mapping forest ecosystems, 5) adapt current forest mapping algorithms to Eastern Russia, 6) adapt current and develop new forest mapping algorithms in Continental Southeast Asia, and 7) use the Sentinels-1/-2 data for forest map updating.

3 First project year, KO / KO+12 I-i) Preparation of forest GSV maps and forest structure maps with different spatial resolutions. I-ii) Investigation of the scaling effects in mapping forest GSV and forest structure. I-iii) Further validation of the forest GSV- and forest structure and maps created within the Forest DRAGON 2 Project for Northeast China. I-iv) Preparation ERS and Envisat SAR data for Eastern Russia and Continental Southeast Asia.

4 BIOMASAR-II pan-boreal data products BIOMASAR-II Final Presentation April 22, 2013 Maurizio Santoro 1, Carsten Pathe 2,Julian Schwilk 2, Ina Burjack 2,Kerstin Traut 2 1 Gamma Remote Sensing, Gümligen, Switzerland 2 Department of Earth Observation, Friedrich-Schiller University, Jena, Germany

5 Forest DRAGON-2 GSV Change Map for Northeast China ( ) Continuous change values Derived from Envisat ASAR GSV maps based on BIOMASAR algorithm (Santoro et al. 2011) Total coverage of forest and shrubland: ~ km²

6

7 Coherence 1998 vs. GSV Classes 1998

8 Continuous GSV 2010 vs GSV Classes 1998.

9 Second project year, KO+12 / KO+24 II-i) Finalised Production of validated forest GSV change maps for Northeast China. II-ii) Finalised Production of a validated forest structure maps for Northeast China. II-iii) Publication of the validated 15-year change maps for Northeast China. II-iv) Using ERS and Envisat SAR data for forest GSV and coverage mapping in Eastern Russia and Continental Southeast Asia in 1995 and II-v) Publication of the forest changes detection and analysis for Eastern Russia and Continental Southeast Asia.

10 Major changes in GSV retrieval workflow Utilization of SRTM DEM at 1 arcsec resolution (ca. 30 m) Mapping of continuous GSV values instead of discrete classes Replacement of MODIS Vegetation Continuous Fields (VCF) product with Landsat VCF Tree Cover (30 m resolution) (Sexton et al. 2013) Fig. 1: Landsat 2000 Tree Cover example scene (UTM Zone 51N, WGS-84)

11 Comparison of old and new GSV maps: discrete classes Fig. 3: Old GSV classification. Projection: Albers Datum WGS-84 Spatial resolution: 50 m Fig. 4: New GSV classification. Projection: UTM Zone 51N Datum WGS-84 Spatial resolution: 30 m

12 Comparison of old and new GSV maps: continuous Fig. 5: Continuous approach for old SAR products Projection: Albers, Datum WGS-84, Spatial res: 50 m Fig. 6: Continuous approach for new SAR products Projection: UTM Zone 51N, Datum WGS-84.Res 30m

13 Third project year, KO+24 / KO+36 III-i) Investigation of synergies from using existing land cover and/or fire products for the and 2010 maps. III-ii) Investigation of synergies from using optical satellite data at different temporal and spatial resolutions for the and 2010 maps. III-iii) Development of forest GSV estimation and coverage mapping algorithms using Sentinel-1/2 data.

14 1. Determine a potential (!) Forest/Non-Forest threshold for the BIOMASAR GSV maps Non-forest < 80 m³/ha < Forest Why take a GSV of 80 m³/ha? Cartus et al Reiche et al [m³/ha] [m³/ha] [m³/ha] >80 [m³/ha] 4 classes BSc Thesis Schratz (2014): Forest/Non-Forest comparison of BIOMASAR GSV with Landsat & SPOT data resulted in best threshold of m³/ha

15 BIOMASAR change We assume 80 m³/ha as a threshold for Forest/Non-Forest Map shows areas which Feature a GSV value of 80m³/ha and more in 2005 (=Forest) Feature a GSV value of 80m³/ha and less in 2010 (=Non-Forest) Change from Forest (2005) to Non- Forest (2010)

16 Comparison to Global LC Products 1. Comparison to Forest/Non- Forest product from GLCF: Forest Cover Change (Hansen et al. 2013) 30 m Based on Landsat time series Forest loss was defined as a standreplacement disturbance or the complete removal of tree cover canopy at the Landsat pixel scale. Forest gain was defined as the inverse of loss, or the establishment of tree canopy from a nonforest state. (Hansen et al. 2013)

17 Intermediate step 1. Detect areas where only BIOMASAR shows change: 1. Areas of change (BIOMASAR) 2. Areas of change (GLCF) 3. Erase GLCF areas from BIOMASAR areas where only BIOMASAR shows a change Why this? To compare these areas with global LC products to explore potential patterns MODIS 2010 GlobCover 2009 CCI LC 2010 ( )

18 1. Comparison of BIOMASAR only change areas with 1. MODIS LC 2010 (500m) 2. GlobCover 2009 (300m) 3. CCI LC 2010 (300m) Let s look on the class statistics (next slide)

19 Class statistics: BIOMASAR only change areas (= BIOMASAR change GLCF Change) compared to MODIS LC Open shrublands 5% Grasslands 6% Woody savannas 7% Deciduous Broadleaf forest 2% Savannas 2% Cropland/Nat ural vegetation mosaic 25% km² Deciduous Needleleaf forest 13% Mixed forest 24% Croplands 15% 0 MODIS class

20 Some thoughts on these results (focusing on the 4 biggest classes) 1) 40 % of the areas are croplands (25% + 15%) 2) 37% of the areas are forest areas (24% + 13%) #1) Supports BIOMASAR map (change from Forest 2005 to croplands 2010) #2) supports GLCF FCC (no change; forest 2005 = forest 2010) unclear result, more research needed

21 1. Comparison of BIOMASAR only change areas with 1. MODIS LC 2010 (500m) 2. GlobCover 2009 (300m) 3. CCI LC 2010 (300m) Let s look on the class statistics (next slide)

22 Class statistics: BIOMASAR only change areas compared to CCI LC 2010 CCI LC class km² Cropland, irrigated or post-flooding 2% Sparse vegetation (tree, shrub, herbaceous cover) (<15%) 3% Mosaic tree and shrub (>50%) / herbaceous cover (<50%) 2% Cropland, rainfed 8% Tree cover, needleleaved, deciduous, closed to open (>15%) 37% 0 Tree cover, needleleaved, evergreen, closed to open (>15%) 4% Grassland 11% Tree cover, broadleaved, deciduous, closed to open (>15%) 25%

23 Comparison to LC CCI 2010 Some thoughts on the results (focusing on the 4 biggest classes) 1) 62 % of the areas are forest areas (37% + 25%) 2) 19% of the areas are grassland and cropland (11% + 8%) #1) supports GLCF FCC (no change; forest 2005 = forest 2010) #2) Supports BIOMASAR map (change from Forest 2005 to croplands 2010) supports the assumption of falsely detected forest change of the BIOMASAR map for threshold = 80m³/ha

24 Closer look on the GSV values for the selected areas of change 1. To what GSV values have the selected areas changed to? See statistics on the next slide Thoughts: Class m³/ha highly influenced by the variance of the BIOMASAR map production (multitemporal ASAR values) drop of GSV values can be very marginal can be caused by BIOMASAR product variance All other classes (60 and less) seem to have a (significant?) drop Explanation?

25 GSV values (2010) of areas with BIOMASAR 2005 > 80 and BIOMASAR 2010 < km² GSV value [m³/ha]

26

27 GSV in MODIS LC classes for 2005 and 10 Valu Label e 0 Water 1 Evergreen Needleleaf forest 2 Evergreen Broadleaf forest 3 Deciduous Needleleaf forest 4 Deciduous Boradleaf forest 5 Mixed forest 6 Closed shrublands 7 Open shrublands 8 Woody savannas 9 Savannas 10 Grasslands 11 Permanent wetlands 12 Croplands 13 Urban and built-up Cropland/Natural 14 vegetation mosaic 15 Snow and ice Barren or sparsely 16 vegetated

28 GSV in GlobCover classes, Value GlobCover legend 11 Post-flooding or irrigated cropla 14 Rainfed croplands Mosaic Cropland (50-70%)/Veget (20-50%) Mosaic Vegetation (50-70%)/Crop (20-50%) Closed to open (>15%) broadlea evergreen and/or semi-deciduo forest Closed (>40%) broadleaved decid forest Open (15-40%) broadleaved decid forest Closed (>40%) needleleaved everg forest Open (15-40%) needleleaved deciduous or evergreen fores Closed to open (>15%) mixed broadleaved and needleleaved fo Mosaic Forest/Shrubland (50-70 Grassland (20-50%) Mosaic Grassland (50-70%) / Forest/Shrubland (20-50%) 130 Closed to open (>15%) shrubla 140 Closed to open (>15%) grasslan 150 Sparse (>15%) vegetation Closed (>40%) broadleaved for regularly flooded - Fresh wate Closed (>40%) broadleaved sem deciduous and/or evergreen for regularly flooded - Saline wate Closed to open (>15%) vegetatio regularly flooded or waterlogged Fresh, brackish or saline wate Artificial surfaces and associated (urban areas >50%) 200 Bare areas 210 Water bodies 220 Permanent snow and ice

29

30 1.4 GlobCover 05 vs GlobCover 09

31

32

33 Fourth project year, KO+36 / KO+48 IV-i) Investigation of the suitability of fire emission data for forest GSV change estimation. IV-ii) Investigation of synergies from using Sentinel-1/2 data. IV-iii) Production of an ASAR GSV map for 2015 based on the BIOMASAR algorithm for Northeast China. IV-iv) Production of a forest GSV change map IV-v) Update forest GSV and coverage maps for Northeast and South of China, Eastern Russia and Continental Southeast Asia to 2015 using algorithms developed in III-iii.

34 DUE GlobBiomass Objective: Provide the user communities with a better characteristic of the distribution and changes, and an improved quantification of regional and global biomass User Consultation in Jena, October 2012: User Requirements from: Science: Carbon Cycle Science Community Policy: National Forest Inventory and REDD Forest Industry: timber production and certification ITT issue: Q KO: Q Budget: 1,500,000 Duration: 3 years Project Activities: 1. Improve above ground biomass maps (stock and changes) Better geometric resolution Improved accuracy Validation (discrepancy map and error statistics) 2. Platform for data sharing and validation 3. Better stratification of landscape (forest types/species) 4. Standardization of maps ESA UNCLASSIFIED For Official Use DLR ESA Consultation 5/7/2013 Slide 34

35 GlobBiomass Regional Maps Biomass stock and change maps with better spatial resolution than the global reference map ( m) and with a multi temporal approach comprising three epochs: 2000 or 2005, 2010 (reference year), and The regional maps will aim for an overall error of 20% or better: Poland: Temperate zone Sweden: Boreal zone Indonesia: Tropical zone Mexico Tropical-woodland transition Northern Congo forest-savanna mosaic ESA UNCLASSIFIED For Official Use GlobBiomass User Consultation 9-11/10/2012 Slide 35

36 GlobBiomass Global Map One of the key requirements is the generation of a global biomass map with specified requirements to spatial resolution ( m) and accuracy (expected 70%) and time frame = year ESA UNCLASSIFIED For Official Use GlobBiomass User Consultation 9-11/10/2012 Slide 36

37 Distribution of aboveground forest biomass (GEOCARBON 2013) Thank you for your attention!

38 ESA UNCLASSIFIED For Official Use GlobBiomass User Consultation 9-11/10/2012 Slide 38

39