Characterizing Forest Dynamics and Carbon
Biomass Assessment over Tropical Peatlands
using Multi Remote Sensing Approaches
Arief Wijaya
Center for International Forestry Research (CIFOR), Indonesia
Contributors: Ari Susanti, Oka Karyanto, Wahyu Wardhana, Lou Verchot, Daniel
Murdiyarso, Richard Gloaguen, Martin Herold, Ruandha Sugardiman, Budiharto,
Anna Tosiani, Prashanth Reddy Marpu and Veraldo Liesenberg
International Workshop on Forest Carbon Emissions
Technical Session 3: State of the Art Technology for Carbon Stock
Assessment and Monitoring
Jakarta, 3 – 5 March 2015
Project Background
 This work is part of CIFOR projects
– Global Comparative Study on REDD+ (GCS REDD) – work in 6
countries
– Sustainable Wetlands Adaptation and Mitigation Project (SWAMP) –
work in > 20 countries
 CIFOR is an international research organization working
based on three pillars – research, capacity building and
media outreach
2
Background
 The presentation focuses on mapping of tropical peatlands
in Indonesia using SAR and optical sensors
 Tested various classification approaches and SAR features
combined with reflectance of optical data to improve image
classification
3
Importance of Peatlands Ecosystem
 The GoI is preparing FREL submission to UNFCCC –
emissions from deforestation, peat decomposition and peat
fires
 Indonesia covers >80% (~20 Mha in 1990 out of 24 Mha) of
tropical peatlands in SE Asia
 1.1 Mha of intact peat swamp forests and 6.8 Mha of
secondary peatlands forest were deforested from 1990 –
2012
4
CO2 Emissions from Deforestation, Peat
Drainage and Peat Fires in Indonesia
Contributions of CO2 Emissions by Islands
Land Cover Classification System
7
Landuse/cover classification of Indonesia for the years 1990, 1996, 2000, 2003, 2006,
2009, 2011, 2012 and 2013.
Data source: LANDSAT satellite data (30 m resolution) (MOF, 2014)
No Classification
1 Primary Upland Forest
2 Secondary Upland Forest/Logged Forest
3 Primary Swamp Forest
4 Secondary Swamp Forest/Logged Area
5 Primary Mangrove Forest
6 Secondary Mangrove Forest/Logged
7 Crop Forest
8 Oil Palm and Estate Crops
9 Bushes/Shrubland
10 Swampy Bush
11 Savanna
12 Upland Farming
No Classification
13 Upland Farming Mixed with
Bush
14 Rice field
15 Cultured Fisheries/Fishpond
16 Settlement/Developed Land
17 Transmigration
18 Open Land
19 Mining/mines
20 Water Body
21 Swamp
22 Cloud
23 Airport/Harbor
 Characteristics: maps based on visual interpretation of
Landsat data, MMU 6.25 ha, need to assess the consistency
 Not yet included in any national reporting – FREL
submissions to UNFCCC during COP in Lima – issues of FD
definition, REDD activity degradation/carbon stock
enhancement
8
National Forest Degradation Mapping
Deforestation Drivers Analysis
9
What about drivers of forest degradation?
Saatchi biomass map
Baccini biomass map
Adjusted RS biomass measurement
Biomass map based on study by Baccini et al. (2012) including LIDAR shots data obtained
during Biomass mapping training at BIG
Carbon density by landcover type
Forest classes carbon (ton/ha) SD (ton/ha)
Primary dry forest (PF 2001) 179.9 16.9
Secondary dry/logged over forest (SF 2002) 173.7 15.2
Primary Swamp Forest (PSF 2005) 155.5 19.2
Secondary swamp forest(SSF 20051) 143.8 19.7
Primary mangrove forest (PMF 2004) 87.4 13.4
Secondary mangrove forest (SMF 20041) 62.6 8.9
Crop forest (CF 2006) 111.4 17.0
Non-forest classes (vegetated) carbon (ton/ha) SD (ton/ha)
Oil Palm and estate crops (PG 2010) 95.6 19.9
Bushes/Shrubland (B2007) 123.9 13.7
Swampy bush (SB 20071) 77.6 14.1
Savanna (S 3000) 63.1 11.3
Upland farming (UF 20091) 79.9 14.5
Upland farming mixed with bushes (Pc 20092) 115.2 17.2
Rice field (Sw 20093) 62.8 12.0
Carbon stocks change 2000 - 2009
Based on Multiply and Stratify approach. The figure shows only C stocks above ground.
2000
Carbon stocks change 2000 - 2009
2009
Landcover and carbon density
Landcover 2000 Landcover 2009
(a) (b)
(c) (d)
Degradation Mapping Exercise
17
Data
 Dual-polarimetry TerraSAR X data (2008)
 PLR data of ALOS Palsar (2007-2009)
 Landsat data
 Peatland maps from Wetland International
 Land use/land cover map from MoF
18
Peatlands under study
19
Class label Peat types Peat
thickness
Proportions
(%)
Bulk density
(gram/cc)
Carbon
contents (%)
Land
cover type
Mangrove
forest (MF)
- - - - - Mangrove
forest
Deep peat in
primary
swamp forest
(PDP)
Hermists/fibrists
(H3a)
2 – 4m (deep) 60/40 Hermists: 0.23
Fibrists: 0.13
Hermists:
36%
Fibrists: 43%
Mineral: 31%
Primary
forest
Shallow peat
in primary
swamp forest
(PSP)
Hermists/fibrists/
mineral (H1b)
0.5 – 1m
(shallow)
50/30/20 Hermists: 0.23
Fibrists: 0.13
Mineral: 0.32
Primary
forest
Very shallow
peat in sparse
forest (PVSp)
Hermists/mineral
(H1i)
<0.5m (very
shallow)
20/80 Hermists: 0.23
Mineral: 0.32
Sparse
forest
Shallow peat
in secondary
swamp forest
(PSS)
Hermists/fibrists/
mineral (H1b)
0.5–1m
(shallow)
50/30/20 Hermists: 0.23
Fibrists: 0.13
Mineral: 0.32
Secondary
forest
SAR Data Decomposition
21
SAR Backscatter Responses
22
Alpha Entropy Plane
23
Land Cover Map
26
PLR SAR Features
28
Polarimetric features: alpha angle (a), entropy (b) and anisotropy (c). Two
additional polarimetric features were also calculated, PolSAR random
volume over ground volume ratio (RVOG_mv) based on polarimetric data
inversion and accumulation of polarimetric backscatter (span in decibel /
span_db)
Alpha Entropy Plane
29
1
2
3
4
5
6
7
8
9
Initial SAR Classification
30
Technical Challenges/Opportunities
 Needs to upgrade technical competence in the country –
ground station is available
 Access to data might not be major concern – various
donors/bilateral cooperations continuously comes – JICA,
EU, USAID, Norway
 Methods for merging SAR and optical need good knowledge
of RS data pre-processing
 Relatively good IT infrastructure and facilities
31

Characterizing Forest Dynamics and Carbon Biomass Assessment over Tropical Peatlands using Multi Remote Sensing Approaches

  • 1.
    Characterizing Forest Dynamicsand Carbon Biomass Assessment over Tropical Peatlands using Multi Remote Sensing Approaches Arief Wijaya Center for International Forestry Research (CIFOR), Indonesia Contributors: Ari Susanti, Oka Karyanto, Wahyu Wardhana, Lou Verchot, Daniel Murdiyarso, Richard Gloaguen, Martin Herold, Ruandha Sugardiman, Budiharto, Anna Tosiani, Prashanth Reddy Marpu and Veraldo Liesenberg International Workshop on Forest Carbon Emissions Technical Session 3: State of the Art Technology for Carbon Stock Assessment and Monitoring Jakarta, 3 – 5 March 2015
  • 2.
    Project Background  Thiswork is part of CIFOR projects – Global Comparative Study on REDD+ (GCS REDD) – work in 6 countries – Sustainable Wetlands Adaptation and Mitigation Project (SWAMP) – work in > 20 countries  CIFOR is an international research organization working based on three pillars – research, capacity building and media outreach 2
  • 3.
    Background  The presentationfocuses on mapping of tropical peatlands in Indonesia using SAR and optical sensors  Tested various classification approaches and SAR features combined with reflectance of optical data to improve image classification 3
  • 4.
    Importance of PeatlandsEcosystem  The GoI is preparing FREL submission to UNFCCC – emissions from deforestation, peat decomposition and peat fires  Indonesia covers >80% (~20 Mha in 1990 out of 24 Mha) of tropical peatlands in SE Asia  1.1 Mha of intact peat swamp forests and 6.8 Mha of secondary peatlands forest were deforested from 1990 – 2012 4
  • 5.
    CO2 Emissions fromDeforestation, Peat Drainage and Peat Fires in Indonesia
  • 6.
    Contributions of CO2Emissions by Islands
  • 7.
    Land Cover ClassificationSystem 7 Landuse/cover classification of Indonesia for the years 1990, 1996, 2000, 2003, 2006, 2009, 2011, 2012 and 2013. Data source: LANDSAT satellite data (30 m resolution) (MOF, 2014) No Classification 1 Primary Upland Forest 2 Secondary Upland Forest/Logged Forest 3 Primary Swamp Forest 4 Secondary Swamp Forest/Logged Area 5 Primary Mangrove Forest 6 Secondary Mangrove Forest/Logged 7 Crop Forest 8 Oil Palm and Estate Crops 9 Bushes/Shrubland 10 Swampy Bush 11 Savanna 12 Upland Farming No Classification 13 Upland Farming Mixed with Bush 14 Rice field 15 Cultured Fisheries/Fishpond 16 Settlement/Developed Land 17 Transmigration 18 Open Land 19 Mining/mines 20 Water Body 21 Swamp 22 Cloud 23 Airport/Harbor
  • 8.
     Characteristics: mapsbased on visual interpretation of Landsat data, MMU 6.25 ha, need to assess the consistency  Not yet included in any national reporting – FREL submissions to UNFCCC during COP in Lima – issues of FD definition, REDD activity degradation/carbon stock enhancement 8 National Forest Degradation Mapping
  • 9.
    Deforestation Drivers Analysis 9 Whatabout drivers of forest degradation?
  • 10.
  • 11.
  • 12.
    Adjusted RS biomassmeasurement Biomass map based on study by Baccini et al. (2012) including LIDAR shots data obtained during Biomass mapping training at BIG
  • 13.
    Carbon density bylandcover type Forest classes carbon (ton/ha) SD (ton/ha) Primary dry forest (PF 2001) 179.9 16.9 Secondary dry/logged over forest (SF 2002) 173.7 15.2 Primary Swamp Forest (PSF 2005) 155.5 19.2 Secondary swamp forest(SSF 20051) 143.8 19.7 Primary mangrove forest (PMF 2004) 87.4 13.4 Secondary mangrove forest (SMF 20041) 62.6 8.9 Crop forest (CF 2006) 111.4 17.0 Non-forest classes (vegetated) carbon (ton/ha) SD (ton/ha) Oil Palm and estate crops (PG 2010) 95.6 19.9 Bushes/Shrubland (B2007) 123.9 13.7 Swampy bush (SB 20071) 77.6 14.1 Savanna (S 3000) 63.1 11.3 Upland farming (UF 20091) 79.9 14.5 Upland farming mixed with bushes (Pc 20092) 115.2 17.2 Rice field (Sw 20093) 62.8 12.0
  • 14.
    Carbon stocks change2000 - 2009 Based on Multiply and Stratify approach. The figure shows only C stocks above ground. 2000
  • 15.
    Carbon stocks change2000 - 2009 2009
  • 16.
    Landcover and carbondensity Landcover 2000 Landcover 2009 (a) (b) (c) (d)
  • 17.
  • 18.
    Data  Dual-polarimetry TerraSARX data (2008)  PLR data of ALOS Palsar (2007-2009)  Landsat data  Peatland maps from Wetland International  Land use/land cover map from MoF 18
  • 19.
    Peatlands under study 19 Classlabel Peat types Peat thickness Proportions (%) Bulk density (gram/cc) Carbon contents (%) Land cover type Mangrove forest (MF) - - - - - Mangrove forest Deep peat in primary swamp forest (PDP) Hermists/fibrists (H3a) 2 – 4m (deep) 60/40 Hermists: 0.23 Fibrists: 0.13 Hermists: 36% Fibrists: 43% Mineral: 31% Primary forest Shallow peat in primary swamp forest (PSP) Hermists/fibrists/ mineral (H1b) 0.5 – 1m (shallow) 50/30/20 Hermists: 0.23 Fibrists: 0.13 Mineral: 0.32 Primary forest Very shallow peat in sparse forest (PVSp) Hermists/mineral (H1i) <0.5m (very shallow) 20/80 Hermists: 0.23 Mineral: 0.32 Sparse forest Shallow peat in secondary swamp forest (PSS) Hermists/fibrists/ mineral (H1b) 0.5–1m (shallow) 50/30/20 Hermists: 0.23 Fibrists: 0.13 Mineral: 0.32 Secondary forest
  • 20.
  • 21.
  • 22.
  • 23.
  • 24.
    PLR SAR Features 28 Polarimetricfeatures: alpha angle (a), entropy (b) and anisotropy (c). Two additional polarimetric features were also calculated, PolSAR random volume over ground volume ratio (RVOG_mv) based on polarimetric data inversion and accumulation of polarimetric backscatter (span in decibel / span_db)
  • 25.
  • 26.
  • 27.
    Technical Challenges/Opportunities  Needsto upgrade technical competence in the country – ground station is available  Access to data might not be major concern – various donors/bilateral cooperations continuously comes – JICA, EU, USAID, Norway  Methods for merging SAR and optical need good knowledge of RS data pre-processing  Relatively good IT infrastructure and facilities 31

Editor's Notes

  • #19 Coverage of ALOS and Landsat data is national/global, TerraSAR X more on sub-national coverage Temporal resolution reolution of Landsat 16-18 days, ALOS Palsar has gone but replaced with ALOS 2 Source of ground data – we use high spatial resolution for study in Kalimantan, and additional field data and national land cover map for study in Sumatera Conventional Confusion matrices approach is used to validate the resulted maps This method, in terms of R&D needs good competence in RS data analysis, especially to handle preprocessing of SAR data which is normally not straight forward as the optical data The approach will complement national estimate on forest degradation with more accurate result. Jurisdictional approach of REDD project should find better and more accurate methods to map forest cover change and/or forest degradation and eventually come up with better predictions of carbon emissions