Indonesias National Carbon Accounting System- Land Cover Change - - PowerPoint PPT Presentation

Indonesia’s National Carbon Accounting System- Land Cover Change Analysis Program: A national system for monitoring forest changes

Orbita Roswintiarti

Indonesian National Institute of Aeronautics and Space (LAPAN)

Presented at “the 7th GEOSS Asia-Pacific Symposium Benefits for Society from GEOSS Evolution Toward Addressing Sustainable Development Goals” Tokyo, Japan, 26-28 May 2014

Background

• The development of the national forest carbon

Measurement, Reporting, and Verification (MRV) systems in Indonesia among others is done through the Indonesia’s National Carbon Accounting (INCAS) program.

• INCAS is designed to provide a credible and sustainable

system in Indonesia for greenhouse gas accounting and reporting for Indonesia’s land sector, with full national coverage.

• INCAS was commenced in 2009 under the Indonesia-

Australia Forest Carbon Partnership (IAFCP). The program consists of two major technical components:

• a. The remote sensing component, Land Cover Change

Analysis (LCCA), provides spatially detailed monitoring for the whole country of changes in forest area over time using satellite remote sensing imagery.

Background

• b. the biomass component includes forest biomass

measurement and modeling, forest disturbance mapping, and carbon stock estimation to produced carbon accounts.

• In both components, the initial approach was to transfer and

adapt knowledge and experience from Australia’s national system to build operational systems and capacity in Indonesia based on the Indonesia’s requirements and conditions.

LCCA objectives

The initial objectives of LCCA are to:

• undertake wall-to-wall of cloud-free mosaic images and forest

extent and change for 2000-2009 using Landsat data using nationally consistent methodology.

• undertake the feasibility of integrating other data sources, such

as radar or a variety of non-Landsat optical sensors, into the

• perational program.
• develop methods for detecting deforestation and forest

degradation.

• provide inputs for carbon accounting activities.

Recent updates:

• Add years (2010-2012) for the annual cloud-free mosaic images

and forest extent and change mapping currently being processed and on schedule for completion by June 2014.

• Developing methodology to integrate Landsat-8 data for 2013

and future updates.

Remote sensing data utilization

Landsat data (2000-2009)

2000 106 2001 113 2002 123 2003 61 2004 116 2005 119 2006 126 2007 94 2008 76 2009 96 Total 1030 2000 61 2001 76 2002 95 2003 40 2004 70 2005 81 2006 81 2007 63 2008 57 2009 113 Total 737 2000 55 2001 116 2002 78 2003 68 2004 102 2005 93 2006 95 2007 84 2008 73 2009 104 Total 868 2000 42 2001 84 2002 73 2003 60 2004 93 2005 100 2006 101 2007 81 2008 96 2009 98 Total 828 2000 66 2001 90 2002 80 2003 67 2004 94 2005 99 2006 94 2007 82 2008 71 2009 85 Total 828

Sumatera Kalimantan Sulawesi Papua Others

Remote sensing data utilization

High resolution data

Kalimantan: 53 Quickbird, Ikonos, World View-2, and Geo Eye-1 (2001 to 2010, mostly 2007 to 2009). Sumatera: 77 Quickbird, Ikonos, World View-2 (2006 to 2009). Papua: 62 Quickbird, Ikonos, World View-2, and GeoEye (2006 to 2011). Sulawesi: 79 Quickbird and World View-2 (2006 to 2011).

Forest cover

• perational processing stream

Scene selection Registration and Callibration Cloud Masking and Mosaicing Thresholding to map forest extent Multi-temporal processing to monitor change Attribution for purpose Processing of other products Quality Assurance Quality Assurance

Cloud and shadow masking

• In Indonesia, manual cloud masking isn’t an option.
• We have developed a semi-automated method based around
• brightness (albedo) and bare land indices with certain and maybe

thresholds.

• shadow and water indices with thresholds.
• thermal band with threshold.
• sun angle and height considerations (cloud/shadow usually comes

in pairs).

• growing regions to include mixed edge pixels.
• manual ‘add’ and ‘delete’ vectors to fix the automated algorithm

errors.

Cloud and shadow masking

RGB 453 of Landsat-7 (7 Oct 2000) Shadow masking Cloud masking

Cloud and shadow masking

Detecting cloud and shadow using cloud classification (temperature and brightness), shadow classification (band 4 and 5), and spatial correlation between cloud and shadow.

Cloud Shadow RGB 542 Landsat-7

Automatic mosaiking of masked images

Landsat-5/7:2002 Landsat-5/7:2003 Landsat-5/7:2004 Landsat-5/7:2005 Landsat-5/7:2006 Landsat-5/7:2007 Landsat-5/7:2008 Landsat-5/7:2009 Landsat-7:2010 Landsat-7:2011 Landsat-7:2012 Landsat-8:2013

Forest extent and change mapping

Stratification zones

High resolution satellite imageries are used in stratification and analysis, and then to optimise a classifier based on locally optimal indices and thresholds.

Base forest probability image

2008 Landsat TM image mosaic Forest classification (green) over Landsat image (grey) Quickbird multi- spectral image (27 Sept 2008) Quickbird panchromatic image (27 Sept 2008)

2008 forest extent ‘base’ maps

(Kalimantan, Sumatera, Papua, and Sulawesi)

The 2008 base maps of forest-non forest probabilities resulted from multi-temporal classification. Areas of missing data due to cloud in 2008 are predicted using other dates.

Forest extent and change

(Central and South Kalimantan, 2000-2009)

The region of central and southern Kalimantan is one of the transmigration regions, where extensive areas of wetland forest have been cleared and drained since 1996 for resettlement and agricultural activities. The products show where and when clearing has occurred since 2000, and also identify areas and timing of reforestation events. Green areas indicate constant forest cover; non-Green colours indicate clearing (hot colours) and reforestation (cool colours) at different dates within the decade; bright blue indicates clearing following by reforestation.

Forest extent and change

(Riau Province, 2000-2009)

The Riau Province of Sumatra has also been a transmigration destination. More recently, extensive areas of natural forest have been cleared for oil palm plantations. As well as the forest carbon implications, this forest loss has biodiversity and habitat impacts. Green areas indicate constant forest cover; non- green colours indicate clearing (hot colours) and reforestation (cool colours) at different dates within the decade; bright blue indicates clearing following by reforestation.

Forest extent and change

(Central and South Kalimantan and Riau provinces, 2000-2012)

Forest monitoring products for carbon accounting input

The forest monitoring products have been used as one of the inputs in the GHG emissions accounting process.

Land cover mapping using radar data

• SAR data:
• L-band: ALOS Palsar

Fine Beam strip data (FBD: HH+HV, FBS:HH, 50m resolution)

• C-band: Radarsat-2 SLC
• Year: 2009 and 2010
• Data for calibration and

validation :

• Landsat data
• Optical high resolution

data

• Aerial photo flights

Tiles of radar data stacks over Borneo (year 2009-2010)

Aerial photo flight: 5.5 hours, 6500 photos

Methodology

Operational processing chain has been developed for systematic mapping

• f:
• Forest/Non-forest
• Land cover

using ALOS PALSAR FBS and FBD strip data.

Forest from radar in 2009

Forest/Non-Forest map

(Kalimantan, 2009, source: ALOS Palsar)

F F

and/or

NF NF

Closing remarks

• While the main purpose of this activity is the ability

to develop a reliable and operational forest monitoring system in Indonesia for its national carbon accounting system, it has wider benefits in supporting Indonesia’s greenhouse gas inventories, MRV institution, and national communications.

• The INCAS cloud-screened mosaic images are also

being used for land use planning and forest management.

• At local and regional levels, reliable and consistent

information on historic forest changes are required for REDD baselines, and for assessing policy impacts into the future.

Acknowledgements

We wish to thank INCAS LLCA team (LAPAN), Suzanne Furby (CSIRO), Jeremy Wallace (CSIRO), Thomas Harvey (IAFCP), Radar team (LAPAN), and Dirk Hoekman (Wageningen University) for their valuable contribution.