The EAGLE concept – a paradigm shift in European - SIA

EIONET Action Group on
Land Monitoring in Europe
The EAGLE concept – a paradigm shift in European land monitoring
Move from CORINE Land Cover towards a new harmonized approach of object-oriented landscape mapping
Barbara Kosztra Institute of Geodesy, Cartography and Remote Sensing (FÖMI) Hungary
Stephan Arnold Federal Statistical Office (Destatis) Germany
Elise Järvenpää & Markus Törmä Finnish Environment Institute (SYKE) Finland
4a Characterizing bread
7 EAGLE KEY MESSAGES
1 Given situation
• Many applications lead to numerous classification systems
• Most classification systems contain mixture of LC and LU
information
• Redundant data collection on EU and national level
• Lack of compatibility between European and national datasets
• Harmonization between top-down and bottom-up approches is
needed to meet...
different data collection methods,
different scales,
tailored-to-purpose class definitions,
lack of completeness for either LC or LU information
Classification
•
Possible applications of the EAGLE model:
Bread ≠ Bread
Characterization
• Analytical decomposition of class
definitions
Outer Appearence
• weight
• size
• shape
Inner details
• grainsize
• density
Ingredients
• salt
• wheat / rye
• water
• yiest
• E 510, …
Color
• light
• dark
Other characteristics
• Bio-certified
• GMO-free
Use
• Food, Alimentation
• Semantic translation between different
classification systems
• Collection and mapping of land related
information
Conceptual basis for harmonized future
European Land Monitoring Framework
4b Characterizing grassland
8 LandIntegration
monitoring
vision
Scheme of Europeanframework
Land Monitoring Framework
2 Seed questions
• How to strictly separate LC and LU information in a model?
Classifications
•
European Level
C
O
P
E
R
Urban
CLC
Atlas
N
I
C
HR
Layers
• What information can be captured by remote sensing?
• How to apply object-oriented data modelling for European and
national land monitoring?
U
S
BioPhys
Par
I A C S
LUCAS
LPIS
Characterization
Growth structure
• homogenous
• heterogenous
Growth density
• closed
• sparse
Moisture
• Wet soil
• Surface water
Use
• agriculture
• conservation
• sports
Management
• mowing
• grazing
Ecosystem type
• Inland marsh
EAGLE
Concept
• How to decompose and recombine LC and LU information?
National Level
National (A)
Land Monitoring
3 Criteria for data model
Sub-National Level
•
Object-oriented description instead of classification
•
Separation of LC and LU information
•
Describe land cover in a mutually exclusive and
comprehensive way
•
Scale independence
•
Semantic translation between classification systems
•
INSPIRE compliance
•
Support both bottom-up and top-down initiatives
National
CLC
Regional (a)
Land Monitoring
National (B)
Land Monitoring
Regional (b)
Land Monitoring
9 Structure of EAGLE data model
5 Decomposing objects
Structure of the
EAGLE
data of
model
Structure
the
Land Cover
Dataset
(1..* LCUs)
Describing an object by decomposition into
components and further characterization
EAGLE data model
LUA
Land Use Attribute
(HILUCS
EAGLE-extended)
CH
Storage of parameterized data
•
Handling of multi-dimensional /-temporal aspects
•
Flexibility for extension by new model elements
+
Characteristics
ABIOTIC LCCs
Artificial/Natural
•
LCU
Land Cover Unit
(1..* LCCs)
+
+
Grassland
≠ pasture
≠ lawn
≠ natural grassland
CH
+ -- ......
- ...
BIOTIC LCCs
Vegetation
CH
+
-
WATER LCCs
...
...
...
...
CH
+
-
...
...
...
...
Land Cover Unit (LCU) is formed by one or several Land Cover
Components (LCC), further described with Characteristics
(CH), and with one or many Land Use Attribute (LUA).
© Ursus Wehrli
6 Structure of EAGLE matrix
III. Landscape Characteristics
II. Land Use Attributes
I. Land Cover Components
6a Decomposing built-up area
6b Decomposing agricultural land
6c Decomposing wetland
I. Land cover components (LCC):
- conventional buildings
- broadleaved trees
- herbaceous plants
- open sealed surfaces
I. Land cover components (LCC):
- regular graminaceous
- broadleaved trees
- bushes
I. Land cover components (LCC):
- inland water body
- reeds
- regular shrubs
II. Land use attributes (LUA):
- permanent residential
- agriculture for commercial / for own
consumption
- road network
II. Land use attributes (LUA):
- agriculture for commercial production
- forestry
II. Land use attributes (LUA):
- nature conservation, protected site
III. Characteristics (CH):
- soil sealing degree = 35%
- spatial pattern = discont. single houses
Photo © György Büttner
III. Characteristics (CH):
- ecosystem type = inland marshes
- salinity = fresh water
- water regime = permanent
- spatial pattern = mosaic
Photo © Barbara Kosztra
III. Characteristics (CH):
- linear landscape pattern = hedgerows
- cultivation pattern = crop rotation
Photo © György Büttner
Use case I. Enhancement of CLC nomenclature guidelines
TASK
APPROACH
Mixture of LC, LU, other characteristics and thresholds in
CLC definitions results in difficulties to match national
features types to European classes in a bottom-up CLC
production process.
Defining elements in CLC classes need to be identified and
categorized through semantic analysis.
The EAGLE matrix is used to decompose CLC classes to LCC,
LUA, CH and to reveal inconsistencies like semantic gaps and
overlaps in the class definitions.
Example: CLC 111 Continuous urban fabric
“Areas mainly occupied by dwellings and buildings used by
administrative/public utilities. Most of the land is covered by
structures and the transport network. Buildings, roads and artificially
surfaced areas cover more than 80 % of the total surface. Nonlinear areas of vegetation and bare soil are exceptional. ”
LC / LU / CH / Threshold
RESULT
New structure of CLC class definitions:
 „This heading is applicable / not applicable for“  expressing the applicable landscape situations and land uses
 „This heading includes / excludes“  expressing the relevant land cover components
Use case II. Extractable LC information from Sentinel-2
Satellite data from the up-coming Sentinel-2 mission are useful for land monitoring purposes. By visual interpretation
or automatic analysis the imagery can be used to detect and monitor the following landscape properties of EAGLE
data model:
Abiotic
• Artificial/Sealed: GIO
Land HRL Soil
sealing
• Size of the buildings
or nature type using
visual interpretation
• Natural/Consolidated:
Bare rock and
Hardpan could be
possible to classify
Vegetation / Biotic
• Woody / Herbaceous / Lichen&Moss / Succulent -classification
should be possible
• Woody/Trees: Coniferous & Broadleaf
• SpeciesType: would require spectrometer
• CrownCoverDensity: HRL Forest
• ForestHistoryType & ClearCut: multiyear time series
• Start & EndGrowingSeason: seasonal time series of e.g. NDVI
• Wetness: Tasselled Cap-transformation
• Irrigation & Fertilizing: visual interpretation
• AgriculturalCultivationType/PermanentGrassland: HRL
Grassland
Water
• Liquid: HRL Water
• Further division of Liquid
is possible using visual
interpretation
• Solid: PermanentSnow &
IceAndGlacier
Other
• Spatial pattern: Partly
possible to automatize,
but generally requires
visual interpretation
Conclusion
Sentinel-2 kind of satellite images are not able to provide all information for EAGLE data model, so other information
sources like existing GIS databases, in-situ measurements or interpretation of VHR-images are also needed.
Results presented in the poster contain contribution from all members of the voluntary EAGLE group:
Gebhard Banko, Christoph Perger (Austria), Tomas Soukup (Czech Republic), Markus Törmä , Elise Järvenpää (Finland), Stephan Arnold, Michael Bock, Stefan Kleeschulte,
Andreas Littkopf (Germany), Barbara Kosztra, Gergely Maucha (Hungary), Gerard Hazeu (The Netherlands), Geir-Harald Strand (Norway), Julián Delgado Hernández, Roger
Milego, César Martínez Izquierdo, Alejandro Simon Colina, Nuria Valcarcel Sanz (Spain), Charlotte Steinmeier (Switzerland), Geoff Smith (UK)
Contact
EAGLE web:
http://sia.eionet.europa.eu/EAGLE
EAGLE e-mail:
[email protected]
Authors:
[email protected], [email protected],
[email protected], [email protected]