GOFCGOLD Efforts in Harmonization and Validation of Land Cover Datasets

1
GOFC-GOLD Efforts in Harmonization and Validation
of Land Cover Datasets

• Chris Schmullius Martin Herold
• ESA GOFC-GOLD Landcover Implementation Project
  Office
• IGOL, 13-15 Sept. 2004

2
GOFC GOLD introduction

• Multifaceted international strategy to bring the
  Earths land cover under continuous operational
  observation
• Vision to share data, information and knowledge,
  leading to informed action and decision support
• Network of participants implementing coordinated
  research for global monitoring of terrestrial
  resources, global assessments and the study of
  global change
• Establishing international standards and
  protocols especially with respect to issues like
  validation

3
GOFC GOLDs role

• A Panel of the Global Terrestrial Observing
  System (GTOS)
• A long term process of building an improved match
  between Observations, Data Products and User Needs

Space agencies Data producer
Science community
GOFC GOLD
Data users FAO/UNEP etc.
4
Implementation priorities

• 2nd GOFC GOLD land cover implementation team w/s,
  Jena, March 2004
• Harmonization of global and regional land cover
  products
• Validation of global earth observation products
• Adequacy and advocacy of earth observation in
  serving international conventions
• Towards operational global observations of land
• Support of new IGOL-theme
• Strengthen international cooperations (e.g.
  regional networks)

5
Why harmonization?

• Land mapping uses different standards
• national maps, application specific, spectral
  classes
• Compatibility and comparability
• Within and between countries
• Within and between applications and disciplines
• From local to global scales
• Required for successful
• Improvement of mapping products
• Validation of global datasets
• Foster application land change dynamics

6
Dataset heterogeneity

• Syntactic hetero. e.g. logical data models
• Schematic hetero. e.g. spatial reference system
• Semantic hetero. e.g. cognitive
  conceptualizations
• Problems for harmonizing land cover semantics
• Confusion between classification/legend (mixed
  units)
• Mix of land use/cover terms
• Internal unbalance/inconsistency

7
Framing Harmonization

• Union of similarities in existing definitions
• Harmonization - Bottom up process
• from an existing divergence to a state of
  comparability/compatibility
• does not necessarily eliminate all
  inconsistencies
• Standardization - Top down process
• introduces a new, common definition or standard
• standards should eliminate all inconsistencies

8
Harmonization strategy

• Originally strong push for single lc/lu legend
• Too much standardization reduces application
  relevance of lc/lu maps
• Standardizing terminology rather than categories
• W/S FAO/UNEP 1994, GOFC GOLD 2004 (Jena/Rome)
• Implementation strategy
• Two working groups cover and use
• Find common language (LCCS)
• Need for case studies
• Impact on operational lc/lu data collection

9
Harmonization mechanisms

• Harmonization resources
• Web-based resources (LCCS)
• Raise awareness and foster use of harmonized
  products
• Harmonization experiences for existing datasets
• Develop legend translation protocols/case studies
• Compatibility/Comparability of datasets
• Harmonization in future mapping products
• Standardized legend generation (GLOBCOVER)
• Consider inconsistencies in previous maps
• Harmonization and Validation of land cover
  products

10
FAO Land cover classification system (LCCS)

• FAO standards common in harmonization
• Harmonization in UNCEDs Agenda 21 (Chapter 10),
  for which the FAO is task manager within the UN
  system
• LCCS
• Pure Land Cover classification system
• Not based on predefined list of names, rather set
  of simple elements (classifiers)
• High flexibility (ability to describe/classify
  L.C. features at any scale or level of detail)
• Systematic class definition and description for
  the user to build up its own legend
• Cartographic rules fully standardized and
  embedded system

11
LCCS Status and Use

• LCCS-2 Software completed and ready for download
  September 2004
• Manual available September 2004, translations in
  French, Spanish and Arabic scheduled in the year
  2005
• GOFC GOLD endorses LCCS
• Feedback to improve LCCS
• Translation of legends (start _at_ Rome w/s)

12
IGBP legend translation

• Rome workshop
• translations
• IGBP DIS
• MODIS LC
• MODIS Cont.Flds.
• CORINE
• GLC2000
• - already in LCCS

MODIS LC
Workshop discussions
LCCS Software
13
Forest areas in global land cover maps
Forest definitions IGBP legend percent tree
cover 60 / tree height 2m GLC2000 legend
percent tree cover 15 / tree height 3m
14
Differences for urban areas
Shanghai
Sao Paolo
Sources of urban land IGBP DIS Digital chart
of the world MODIS MODIS(2000) DMSP
(1994/95) GLC2000 DMSP (1994/95)
15
Forest cover from continuous field products
Credit M. Hansen/University of Maryland
0
100
tree cover threshold
16
VCF can support harmonization of land cover data
Credit M. Hansen/University of Maryland
0
100
tree cover threshold
17
Agreement between datasets
Agreement match between all three datasets (in
)
18
GLOBCOVER

• Partnership ESA - JRC
• Programs GOFC/GOLD - IGBP
• Users FAO, UNEP, UN Conventions (CCD, CB)
• Objectives Global Land Cover Map in 2005 by
  using
• ENVISAT MERIS data at 300m (15 bands)
• Schedule Definition phase Spring 2004
• Invitation To Tender July 2004
• KO December 2004
• First Monthly Mosaic September 2005
• GLOBCOVER V1 September 2006
• GLOBCOVER V2 September 2007
• Validation September 2008
• Implementation One European consortium
  industry/companies

19
GLC2000 versus GLOBCOVER
GLC-2000 map at 1km res.
LC map from MERIS at 300m res.
20
Validation of land cover datasets

• GOFC GOLD fosters international effort to
  validate all existing global land cover products
• Consistent with Best Practices identified by
  CEOS Land Cover Validation Working Group
• Create living dataset for future land cover
  products (e.g. GLOBCOVER)
• Use of GOFC GOLD regional networks
• Harmonization experiences are essential
• LCCS provides the framework to translate the
  individual in situ interpretations
• Frustration with lack of funding for validation