LCCS legend translation and harmonization appoaches

LCCS legend translation and
harmonization appoaches
Martin Herold
GOFC-GOLD project office
Friedrich Schiller University Jena, Germany
www.fao.org/gtos/gofc-gold
www.gofc-gold.uni-jena.de
Global Observations of Forest Cover and Land Dynamics
Starting point in 2004
• National/international land mapping initiatives:
¾ reflect different interests, requirements and methodologies
• Current maps exist as independent datasets:
¾ limited compatibility, validation, and no or irregular updates
• Low efficiency and applications (change, scaling)
• Heterogeneity of land cover information users
• GIScience/information age requires standardization
• Requirements for consistency and continuity in land
assessment:
¾
¾
¾
¾
UNFCCC (GCOS implementation plan)
GEO(SS) workplan and global land cover task (DA-07-02)
Global change community
Need for operational land observations
Integrated land cover observations
high
IN-SITU (+ IKONOS type)
periodically (usually 1-10 yrs)
Spatial detail
Assuming observation
continuity and
consistency
M
at
O
e
DI
hi
S
(in /M
gh
tra E
pa -)a RIS
Lo tte nn
ng rn ua
l
-te
tre rm
nd
s
LANDSAT/SPOT – type
inter-annual (1-5 yrs)
rt
fo
rf
re
qu
en
t
up
d
Vegetation physiognomy
Ef
fo
low
Detailed physiognomy
Floristics and species distribution
Land use: i.e. crop type/rotation
Calibration and validation
Land change
processes
Land type/
Phenology
Thematic detail
high
Harmonization steps
1. Standardize terminology, i.e. LCCS classifiers
2. Accepted terminology should be proposed and related
resources distributed, i.e. within the scientific community
or for capacity building (i.e. in developing countries)
3. Existing land cover legends should be translated to
explore (in)consistencies in current approaches for land
cover characterization,
4. Agreement and synergy among existing land cover maps
should be studied and applied in the context of specific
applications
5. An evolving and consolidated standard terminology for
land cover should be applied in operational observing
programs
LCCS translation of existing legends
100
UMd
(14 classes)
100
92
IGBP
(17 classes)
97
59
CLC
(44 classes)
67
42
ACS
(37 classes)
Consistency of class
description
83
Confidence in translation
0
25
50
75
100
Evaluation score (%)
Evaluation scores for legend consistency and translation confidence
Common LCCS classifiers
• Vegetation life form: trees, shrubs, herbaceous vegetation
(maybe separated into grasslands and agricultural crops),
lichen and mosses, non-vegetated
• Leaf type (needle-leaf, broad-leaf) and leaf longevity
(deciduous, evergreen) for the different vegetation life
forms
• Non-vegetated cover types (bare soil or bare rock, built up
areas, snow, ice, open water)
• Density of life form and leaf characteristics in percent cover
• Terrestrial areas versus aquatic/regularly flooded
• Artificiality of cover and land use
• ….
Harmonized land cover characterization
Existing global land
cover datasets
Common land cover classifiers (LCCS)
Cover type/ life form
Trees
Shrubs
Herbaceous
Bare
Snow & Ice
Artificial
Evergreen
Leaf type Deciduous
Needle-leaved
Aquatic/
flooded
Leaf
longevity
Broadleaved
Cultivated
/
managed
“Living” validation database for
comparative assessment
Categories in existing global datasets
Terminology: land cover classifiers (LCCS)
Trees
Shrubs
Translation
Evergreen
Deciduous
Herbaceous
Bare
Cultivated/
managed
Snow & Ice
Built up
Needle-leaved
Broadleaved
Aquatic/
flooded
Land cover/use in Thuringia
CORINE
(CLC2000)
Landuse
classification
(Landsat)
30 m
250 m
GlobCover
300 m
1 km
MODIS LC
(MOD12)
1 km
GLC2000
comparisation concept
products
CORINE
trees
Generic classes
Grasses
agriculture
crops
Built up
GlobCover
GLC2000
MODIS LC
(MOD12)
treecover in thuringia
GlobCover V2
CORINE
GlobCover
GLC2000
MODIS LC (MOD12)
Accuracy Assessment
Legende
Sample
Design
1,5 x 1,5
1 x 1 km
500 x
300 x
250 x
Thematic standards
Reference
database (GLC2000)
Comparative
validation & assessment
SYNMAP – for carbon cycle modeling
SYNMAP – a global synthesis product of existing global land cover maps
to provide a targeted and improved land cover map for carbon cycle
modelling purposes; here shown as life form assemblages (Source: M.
Jung et al. 2006, Remote Sensing of Environment).
GLOBCOVER (2005/6)
Dataset release: October 2008
Further needs
• Further international consensus discussions on the
adoption of LCCS concept
– Definition of generic classes/legend
• Existing legends should be translated/ revisited
• Land cover legends should be developed using LCCS
and the common set of classifiers.
– more thematic detail can be specified that meet the mapping
requirements without losing compatibility on a broader level
• Explore how a harmonized land cover product can link
to, or benefit from existing mapping initiatives on both
finer and coarser scales (e.g. in situ and global) and
vice versa