Final Report of the Semantic Testing of the EAGLE Matrix

Contract SC55998 based on the restricted procedure No
EEA/MDI/14/012 following a call for expression of interest
EEA/SES/13/005-CEI
Final Report
Assistance to the EEA in the production of
the new CORINE Land Cover (CLC) inventory,
including the support to the harmonisation of
national monitoring for integration at
pan-European level
–
Task 2.2 - Semantic testing
Prepared by:
Stephan Arnold,
Geoff Smith, Gerard Hazeu, Mario Caetano, Barbara Kosztra,
Stefan Kleeschulte, Charlotte Steinmeier, Julian Delgado
31 July 2015
Version 2.1
Service Contract 55998 following a call for expression of interest EEA/SES/13/005-CEI
Legal notice
The contents of this publication were created by space4environment under EEA service contract
No. 3436/B2014/R0-GIO/EEA.55998 and do not necessarily reflect the official opinions of the
European Commission or other institutions of the European Union. Neither the European
Environment Agency nor any person or company acting on behalf of the Agency is responsible for
the use that may be made of the information contained in this report.
Copyright notice
© EEA, Copenhagen, 2015
Reproduction is authorised, provided the source is acknowledged, save where otherwise stated.
i
TABLE OF CONTENTS
1
Introduction ............................................................................ 3
2
The approach .......................................................................... 5
3
4
5
2.1
General issues on semantics ................................................................... 5
2.2
EAGLE Matrix Bar Coding Method ............................................................. 6
2.3
Semantic analysis of class definitions ...................................................... 12
2.4
Matrix filtering for class comparison ........................................................ 13
Results .................................................................................. 14
3.1
Feedback on matrix content: general thematic issues ................................ 14
3.2
Bar Coded Classes ................................................................................ 16
Lessons learnt and open issues ............................................. 40
4.1
Explicit definition versus logic interpretation ............................................. 40
4.2
Merging of Bar Code Values ................................................................... 41
4.3
Richness of class definitions ................................................................... 41
4.4
Matrix re-coding for practical class interpretation ...................................... 42
4.5
Eligible elements & Borders of bar code function within the matrix .............. 42
4.6
Machine readability of grouped matrix cells .............................................. 45
4.7
Further matrix/model fine tuning ............................................................ 46
4.8
Input for task 3.1 (matrix population and comparison tool) ........................ 46
Conclusions ........................................................................... 47
Abbreviations and acronyms ........................................................ 50
ii
1 INTRODUCTION
The aim of this activity is to test the use of the EAGLE matrix for a transfer of textual
class definitions of a given nomenclature or a classification system into a standardized
format which makes them machine readable and allows a comparison between class
definitions. The primary aim of this testing exercise was to check if the EAGLE matrix (by
its thematic content) is capable of describing some of the most common Land
Cover/Land Use (LC/LU) nomenclatures. Together with this, the secondary aim was to
see if the current bar coding method is capable of providing unambiguous class
descriptions. The long-term goal of semantic testing is an automated comparison
between class definitions, but for the moment it is out of scope. Further, the focus of this
task lies only on the semantic dimension, not on a mapping activity.
The testing of the EAGLE matrix content is executed by applying it to the analyzing of
class definitions and by expressing them in terms of the EAGLE matrix. That is to
describe the class by using the matrix elements: Land cover components (LCC), Land use
attributes (LUA) and further characteristics (CH). These matrix elements are represented
as columns in an excel sheet.
The procedure of using the matrix as a tool to decompose class definitions starts by
selecting one or more land cover components from the LCC block, continues with
attaching land use information from the LUA block, and is finalized by adding further
characteristic information on landscape. In theory, the LCC form a complete and mutually
exclusive collection of matrix elements that can describe 100% of the Earth´s surface in
the sense of land cover at any scale.
This report is structured in the following manner: After the project background and this
introduction (chapter 1) to the task, approach explanation follows (chapter 2) where the
applied bar coding methodology and some general reflections are described. In
connection with that, Annex 1 contains the EAGLE Matrix Bar Coding Manual as a
separate document. Chapter 3 contains the results of the feedback from contributors on
the content of the EAGLE matrix used for the presented semantic testing exercise,
including a list of all bar coded classes.
The actual bar code results of the semantic translation of classes, with all their bar code
values, is compiled in Annex 2 (T2.2_Annex2_BarCodingResults.xlsx), which is added as
a separate file to this report. The most common Land Cover/Land Use (LC/LU)
nomenclatures tested that can be found in the referred annex are
•
•
•
•
•
•
•
•
•
•
•
•
•
CLC (Corine land cover),
LUCAS (Land Use and land Cover Area Frame Survey),
Urban Atlas,
EUNIS (European Nature Information System),
LPIS (Land Parcel Information System,
Cross-border nomenclature between Bulgaria and Romania,:
MePAR (national high-resolution Land Parcel Identification System (LPIS) of
Hungary),
BBG (Bestand Bodemgebruik) and LGN (Landelijk Grondgebruiksbestand) of the
Netherlands,
SIOSE (Sistema de ocupacion del suelo de Espana),
DLM-DE (Digitales Landbedeckungsmodell Deutschland),
LISA (Land Information System Austria),
BAP BH (UK Biodiversity Action Plan Broad Habitats),
Arealstatistik of Switzerland.
The lessons learnt from the methodological point of view of applying the EAGLE concept
to various European and national nomenclatures are presented in chapter 4. Some open
issues as a result of the testing are also described.
3
In the final chapter 5, the conclusions from this semantic testing exercise are drawn in a
Strengths, Weaknesses Opportunities and Threats (SWOT) schema.
4
2 THE APPROACH
In this chapter the approach of the exercise is explained briefly. The general idea of the
EAGLE concept is to decompose and semantically analyse the class definitions of any
given classification system. The applied method is call “Bar Coding Method”. More
detailed facts about the bar coding method are noted further down. The matrix bar
coding manual is briefly commented, followed by general methodological aspects and
thematic aspects, before the bar coded classes from different nomenclatures are listed.
(For a better understanding of the meaning of bar code values mentioned in this report
please refer to the Bar Coding Manual (Annex 1) and the Bar Coded Results (Annex 2) to
this report.)
2.1 GENERAL ISSUES ON SEMANTICS
From the conceptual idea of a classification system to the usage of information content,
several phases or perspectives can be identified when applying a classification system to
landscape´s reality:
1) conceptual design of nomenclature: the targeted “ideal” meaning of the class,
the purpose behind the creators´ design of a classification system,
2) determination of definitions: the implementation of the idea into the definition
of a class, mostly given by textual description or by just a name,
3) application of definitions: the way the class definitions are applied and
interpreted by the data producers and editors, and
4) interpretation of content: the way the data users themselves understand the
meaning of classes and assume the actual content of information due to their
interpretation of the class definitions.
In the best case phase (1) and (2) are strongly congruent. The challenge is to keep
aspects (3) and (4) as close as possible to aspect (2). Given the fact that a theoretical
description can hardly embrace the entire reality of landscape where exceptions confirm
the rules, compromises have to be made for particularities, borderline cases or situations
that have not been foreseen at all when designing the classes.
As for mapping activities and data capture procedures phases (3) and (4) are practically
inevitable, this also became visible during the semantic testing exercise of this task 2.2.
To minimize the interpretative effect of phase (3) and (4) above, the contributors for this
task were instructed to stick as close as possible to the strict textual class definitions (i.e.
aspect (2)) and try to avoid their own interpretations of the class definitions. This strict
approach is necessary to be able to compare the plain textual class definitions with each
other. Otherwise, if the logic interpretative approach was allowed, it would be more
difficult to draw a line between text-based and interpretation-based bar coding and to
avoid differences coming from subjectivity and different levels of expertise of testers
Besides the obvious to identify compulsory elements in a class definition, it is an
important question to which level of detail or abstraction the landscape situations in the
‘real world’ are represented in the class definitions. The answer can be of three
variations:
a) Typical examples, which represent the essential content of the class.
b) Particularities that might occur and explain the application of the class in not so
typical and rare situations.
c) Land unit types that normally would be classified with another class, but due to
their own area size below the minimum mapping unit (MMU) they are generalized
under the given class.
5
All these three variation may occur in combinations. Depending on how elaborate a class
is explained with examples of the definition according to the above mentioned variations
(b: particularities) and (c: areas smaller than MMU), the more matrix elements may be
candidate to be assigned to a bar code. This leads to the situation where two class
definitions from different nomenclatures, which are meant to address in principle the
same type of land cover or land use, may have different bar coding results, one
containing more elements in total than the other. The mandatory LCC, LUA and CH will
most likely be the same. This issue needs to be considered for this task´s exercise and is
also addressed under the lessons learnt in chapter 4 further down.
2.2 EAGLE MATRIX BAR CODING METHOD
The first version of bar-coding method had been developed in the initial phase of the
EAGLE concept development and described already in the first Explanatory
documentation of the concept (published on EAGLE website1 in June 2013). The
explanatory documentation had then been consolidated within the frames of EEA service
contract No. 3436/B2014/R0-GIO/EEA.55828 in 20142. The method has been further
discussed during the EAGLE meeting in Frankfurt in first week of March 2015 as part of
the present task. The core team of task 2.2 contributors has revised the EAGLE matrix
bar coding method and a bar-coding manual has been compiled. Compared to the initial
guidelines of “How to use the matrix” (as part of the Explanatory Documentation (from
Task 1.1 under restricted procedure EEA/MDI/14/009) the manual has been revised and
extended from the “single-line approach” (still applicable) to a “multi-line approach”. The
bar coding manual has been then applied by the task contributors to a range of European
and national nomenclatures. During this semantic testing exercise contributors also gave
feedback on the method and the applicability of bar code values and the logic of their
usage. The Bar Coding Manual itself is part of this report and is attached to the report in
a separate file as Annex I (T2.2_Annex1_BarCodeManual.docx).
To recall the overall structure of the EAGLE matrix – as it was used for this semantic
testing exercise - is shown in Figure 1 and shall be briefly explained. The Matrix
represents a cross table between the three main matrix block for Land Cover
Components (LCC), Land Use Attributes (LUA) and further Characteristics (CH). They
contain the matrix elements in the form of matrix columns. All the matrix elements are
structured hierarchically from the lowest level to the top level, and therefore are group
into groups of elements (= segments). The classes to be codified are arranged in lines in
the matrix table. In the figure, the CH block is only displayed party because it is even
broader in reality (indicated by dotted line).
When applying the bar coding method, each matrix cell (table field) along a class line is
filled with a coded value called. It indicates the presence/absence/relevance of the matrix
element for the chosen class and the relations between the elements.
The whole sequence of coded values of the class in a single row results in the so called
“bar code”, analogously with a scannable product digit code.
In this chapter, this method is further explained in detail, with the meaning of the bar
code values, how to apply the multi-line approach, and some examples.
1
At the time of writing: http://sia.eionet.europa.eu/EAGLE
2
http://sia.eionet.europa.eu/EAGLE/EEA_contracts/%2301/T1_A1_EAGLE_ConceptExplDocumentati
on_v2.3.1.pdf
6
Figure 1: Overview of the structure of the EAGLE matrix, with the three main blocks of I. Land
Cover Components (LCC, in red-green-blue), II. Land Use Attributes (LUA, in purple) and III.
Characteristics (CH , in light yellow). The codified classes are arranged in rows as matrix lines.
2.2.1 Bar code values
The bar code values express the relationship between the elements of the EAGLE matrix
and actual class definition (e.g. presence / absence etc.) For a better understanding of
the following paragraphs, the bar code values are cited here as also explained in the
Manual:
7
Value Explanation
n/a
x
0
Example
This matrix element is not applicable to the
class. It does not occur because of logic
considerations and therefore can be excluded.
Buildings have no Leaf Type.
This matrix element must not occur, because it
is explicitly excluded by definition from the
class content.
This matrix element is not mentioned in the
class definition, though in some cases a certain
element or character would be logic to occur.
However, the class definition does not allow
any precise conclusion on its existence.
Definition of natural grassland
excludes fertilization or use of
pesticides
Fertilization as a cultivation
measure is probable for arable
land, but is not mentioned in
the class definition.
OR: The element is not mentioned in the
A few single trees or buildings
definition because it is insignificant for the class inside vast arable land parcels
and rather unlikely to occur, but still not
illogical. In case it is present in landscape, it is
lost in generalization due to its negligible size
and does not have any influence on the
assigning of class.
1
2
3
4
This matrix element can be expected in the
class. It is mentioned in the class definition. Its
presence can be typical but has a minor role in
the class definition. It may occur that these
elements are mentioned in the class definition
as examples (in a non-exhaustive list). The
scale and the applied minimum mapping unit of
the class has to be taken into account for this
bar code value.
Any kind of vegetation inside
densely built-up continuous
urban fabric, specific buildings
like a small church inside
suburban areas dominated by
residential use
This matrix element is mentioned explicitly in
the class definition as a defining element. It is
a selective mandatory element with an ORrelation to other elements with the same coded
value. It means that only ONE of these
elements must be present in the land unit
assigned to this class.
Woody vegetation can consist
of trees or bushes, either LCC
trees OR LCC bushes MUST be
present.
This matrix element is mentioned explicitly in
Both broad-leaved trees and
the class definition, it is compulsory element. It needle-leaved trees must be
is a cumulative mandatory element with an
present in a mixed forest.
AND-relation to other defining elements (value
2, 3, 4). It means that ALL of the value 3coded elements must be present in the land
unit assigned to this class.
These matrix elements are mentioned explicitly
in the class definition. They belong to a group
of defining mandatory element of the class.
Such value 4-coded elements cannot occur as
stand-alone elements. They are alternatively
combined mandatory elements with EITHERAND-OR-relation. It means that at least two or
more (but not necessarily all) out of the group
of elements marked with this code must be
present.
The presence of EITHER
grassland AND arable land, OR
grassland AND permanent
crops, OR arable land AND
permanent crops, OR ALL of
them lead to the assigning of
CLC class 242 complex
cultivation pattern.
8
Under chapter 4.1 an open issue on the meaning of bar code value is addressed.
2.2.2 Matrix single-line approach
Initially, as published in the “Explanatory Documentation of the EAGLE concept”3 the bar
coding method was foreseen to be applied as a single-line approach to express class
definitions in EAGLE matrix terms. All matrix elements that are relevant for a given class
definition had to be bar coded in only one line for each class. However, as within one
single line several land cover components can occur, and connected to them several
more land use attributes and characteristics, it is not always clear, which coded character
belongs to which land cover component or land use type. Some combinations can be
excluded logically, but some others allow various connection combinations by just
reading the bar code results.
2.2.3 Matrix multi-line approach
To overcome the constraint of the single-line approach, the bar coding method was
extended to the “multi-line” approach. Some complex classes that contain several kinds
of different land cover are difficult to describe within a single matrix line. The different
land cover types embraced by such a complex class shall be called “semantic partitions”.
With this approach, complex classes that embrace several semantic land cover partitions
can be described in a more distinct manner. The multi-line approach foresees a main line
and one to several sub-lines. The main-line continues to contain all relevant bar code
values for the given class at once. Each semantic partition of a more complex class can
be described in a separate matrix sub-line. This results in three or more matrix lines in
total (= main-line plus at least 2 sub-lines) for each codified class.
A single line may of course still be sufficient to describe a simply structured class
definition, so it is not in all cases necessary or obligatory to create sub-lines.
The bar code values function both in horizontal and in vertical direction. It means that
they express the defining elements and their relation to each other within a matrix (main
or sub-) line (horizontal), and also the relation between the sub-lines (vertical). The field
for the vertical bar code in the matrix is placed in the beginning of the line between the
class name/code and the bar code value fields on the left side of matrix. (See also Figure
2 or Figure 3 further down)
When applying the multi-line approach, differences in bar code values can occur between
the main-line and the sub-lines. These differences represent no error or uncertainty, but
are both necessary and logical. Depending on where an element is coded (main-line or
sub-line), the bar code value for that specific element may change due to a change in
relation to other elements.
This means in detail, that two elements (or matrix segments=unified elements on higher
semantic level) that stand beside each other in the main-line with a selective mandatory
relation (OR-relation = value 2: one of them must be present) can change their meaning
in their sub-lines to a cumulative mandatory element (bar code value 3) that must be
present without any alternative. Each element (or group of elements) for which a subline has been opened, is a stand-alone compulsory element within the subline. The
selective mandatory relation that was expressed with value 2 for two neighboring
3
http://sia.eionet.europa.eu/EAGLE/EEA_contracts/%2301/T1_A1_EAGLE_ConceptExplDocumentati
on_v2.3.1.pdf
9
elements in the main-line is then expressed with the vertical bar code between the sublines.
The EAGLE matrix bar coding manual (Annex 1) contains a more detailed explanation of
the modified bar coding method.
2.2.4 Bar coding exercise examples
In this section examples of bar code results are given for a better understanding of the
bar coding method, in particular the multi-line approach. For better readability, the
matrix elements (columns), which are not relevant for the codified classes, have been set
invisible in the figures below. The examples show the chosen classes as lines, and the
matrix elements (LCC, LUA, CH) as columns. The horizontal bar codes have no
background color (white), the vertical bar codes that put the sub-lines into relation are
placed at the beginning of each line and are highlighted in yellow. The main lines are
labeled with the coded number of the classes, the sub-lines have an appendix to the
class code.
Figure 2 shows an example of a bar coding result for the CLC class 313 Mixed Forest. For
this class, a single line would already be sufficient, because there are only the two
cumulative mandatory land cover components (coded with 3) of broad-leaved trees and
coniferous-trees in the class definition. However, when applying the multi-line approach,
it is easy to explain how this approach works. The two sub-lines (313_1 and 313_2) have
only one difference in the Leaf Form. The yellow highlighted vertical code puts them in
the compulsory AND-relation with code 3. In addition, in the two sub-lines it is possible
to indicate a percentage area share of 75% that none should exceed, otherwise it would
not be a mixed forest anymore but a plain broad-leaved forest, respectively a coniferous
forest (according to CLC definition).
Figure 2: EAGLE matrix multi-line bar coding approach, 1st example: CLC 313 (mixed forest).
Main-line is labelled 313, two sub-lines are labelled 313_1 and 313_2. Bar code values with white
background express relations between matrix elements, bar codes with yellow coloured background
express vertical relations between the sub-lines.
Figure 3 and Figure 4 show an example of the CLC class 242 Complex cultivation
patterns, where Figure 4 is the continuation of Figure 3 towards the right side into the
Characteristics block. Here too, the non-relevant matrix columns have been removed,
still two figures (Part A and B) are necessary to show all elements used for the
description of class 242 with the main line and its four sub-lines. The four semantic
partitions which the class 242 basically consists of are (optional) scattered houses
(expressed in sub-line 242_1), permanent crops (242_2), permanent grassland
(pasture/meadow) (242_3) and arable land (242_4).
In the main-line all these four semantic partitions are assembled together and
respectively bar coded. All relevant LCCs (partly grouped) have bar code value 4, which
10
means two or more of them must occur for this class. Only the conventional buildings are
optional in this constellation. In the main-line, the relation between the elements is
expressed by the horizontal neighbouring bar code values. In the sub-lines (which
represent now each semantic partition embedded in the main-line), the relation between
them is now expressed by the vertical bar codes highlighted in yellow at the beginning of
the line.
Figure 3: Second EAGLE matrix multi-line bar coding approach, 2nd example: CLC 242 (complex
cultivation pattern) – Part A. Main-line is labelled 242, four sub-lines are labelled 242_1 – 242_4.
Bar code values with white background express relations between matrix elements, bar codes with
yellow coloured background express vertical relations between the sub-lines.
Figure 4: EAGLE matrix multi-line bar coding approach, 2nd example (continued): CLC 242
(complex cultivation pattern) – Part B. Main-line is labelled 242, four sub-lines are labelled 242_1
– 242_4. Bar code values with white background express relations between matrix elements, bar
codes with yellow coloured background express vertical relations between the sub-lines.
11
2.2.5 Conceptual connection between multi-line approach and UML
model
From the conceptual point of view- and to make a connecting bridge between the EAGLE
matrix and the EAGLE UML model - the principle of the multi-line approach as used for
the matrix bar coding method is the equivalent to the constellation of LandCoverUnit,
AbstractLandCoverComponent and SpatialComposition in the UML data model.
The main-line in the matrix, where all relevant matrix elements are tick-marked and
coded, corresponds to the LandCoverUnit (LCU) in the UML model. The sub-lines under
each matrix main-line correspond to the AbstractLandCoverComponent in the UML
model. The single matrix elements (or segments on higher semantic level) correspond to
the LandCoverComponents in the UML model (see Figure 5).
In the UML model, characteristics can be attached to the different levels of abstraction,
either directly to the basic singular LCC, or to the entire LCU that can contain several
LCCs. For example spatial pattern for an entire LCU can be labelled differently from the
spatial pattern of a single LCC within a LCU. It is also possible that a LCU consists only of
one homogenous LCC.
Figure 5: simplified structure of the EAGLE UML model
2.3 SEMANTIC ANALYSIS OF CLASS DEFINITIONS
A demonstration of how to analyze the class definitions and to extract terms of land
cover, land use and further characteristics and parameters from the text is shown below
on the example of two British habitat classes. The definitions are contained in the
“Guidance on the interpretation of the Biodiversity Broad Habitat Classification (terrestrial
and freshwater types): Definitions and the relationship with other classifications (2000)”.
12
Examples for semantic analysis:
(Color Legend: Land Cover Component: Biotic; Land Cover Component: Abiotic; Land Cover
Component: Water; Land Use Attribute; Characteristics, excluded by definition)
“2.6 Neutral grassland
This broad habitat type is characterised by vegetation dominated by grasses and herbs on a
range of neutral soils usually with a pH of between 4.5 and 6.5. It includes enclosed dry hay
meadows and pastures, together with a range of grasslands which are periodically inundated
with water or permanently moist.”
“2.14 Rivers and streams
The 'Rivers and streams' broad habitat type covers rivers and streams from bank top to bank
top, or where there are no distinctive banks or banks are never overtopped, it includes the
extent of the mean annual flood. This includes the open channel (which may contain
submerged, free-floating or floating-leaved vegetation) water fringe vegetation and exposed
sediments and shingle banks. Adjacent semi-natural wetland habitats such as unimproved
floodplain grasslands, marshy grassland, wet heath, fens, bogs, flushes, swamps and wet
woodland, although intimately linked with the river, are covered in other broad habitat types.”
2.4 MATRIX FILTERING FOR CLASS COMPARISON
For the comparison of the decomposed classes as they result from the bar coding
exercise, the filter function in Excel can be used.
Firstly, all bar coded classes need to be collated in one single matrix sheet underneath
each other. A constraint of Excel is that the filter function only works in vertical direction
for columns.
In the standard horizontal orientation of the matrix, the matrix elements are columns
and the codified classes are represented in lines. In this matrix from, filters can be
applied to show all codified classes that contain a certain bar ode values for a specific
matrix element, e.g. to show all classes that contain trees as a mandatory element.
For an Excel-filtering in the orientation of class lines, a complementary matrix has to be
generated, where all matrix elements are displayed in rows and the codified classes are
shown in columns (change columns against rows). In such a switched matrix, the classes
can be filtered for a query according to certain matrix elements, e.g. to show all relevant
matrix elements that are mandatory for a certain class.
The EAGLE matrix bar coding results can be filtered in two forms, by using the bar code
values as criteria in both ways.
However, this part of exercise towards comparison between bar coding results (beyond
the scope if this task) is basically working in a testbed, but would need further
elaboration to prove it as functional.
13
3 RESULTS
Under this task, the semantic testing of the EAGLE concept has been carried out by
applying the EALGE matrix on five European and eight national nomenclatures, out of
which around 600 classes have been bar coded. The current version used for this task
was the file T2_A1_EAGLE_MATRIX_20141230_v02.xlsx (as outcome of the Task 1.2
“EAGLE Matrix fine tuning” under EEA service contract No. 3436/B2014/R0-GIO/EEA.55828,
restricted procedure EEA/MDI/14/009).
This chapter is subdivided in two parts, where firstly the contributors´ feedback on the
applied method is collated, and secondly the chosen classes from the nomenclatures are
listed.
The feedback from task contributors was related to two guiding questions:
1.) Methodology: Does the bar coding method itself work, is the meaning of the bar
code values logical and is the method description understandable?
Reflections and feedback on this question are collected and addressed in the
Lessons learnt and open issues (see chapter 4).
2.) Content: Is the thematic content of the EAGLE matrix suitable for the semantic
description the chosen classes; are there any modifications necessary to the
matrix?
Regarding the thematic content, comments from the contributors’ feedback
identified as of generic relevance are mentioned under the sub-chapter on
content-related general thematic issues (see chapter 3.1).
Country-specific or nomenclature-related issues are mentioned directly under the
subchapters of bar coded classes (see chapter 3.2). If no further country-specific
comments can be found, they were considered to be relevant for general thematic
issues.
3.1 FEEDBACK ON MATRIX CONTENT: GENERAL THEMATIC ISSUES
From the contributors’ feedback on the semantic testing, the most essential issues on the
matrix thematic content are listed. They can be subdivided into two sub-headings, a)
modifications of the existing content of the matrix and b) proposals to add elements to
the matrix.
3.1.1 Modification of matrix content
- Crop Types / Plot Activity
In the current matrix version, crop types are handled with the so called plot activity
according to the EU regulation (EC) 1200/2009 Annex II chapter 2, as it is used for farm
structure surveys or for IACS (Integrated Administration and Control System). For
practical reasons it was decided to integrate these plot activities, because it is a
classification of crop types which is already in use and available as a library for data
models. From the EAGLE concept point of view however it is not mutually exclusive, the
plot activities contain partly overlapping crop species because it includes usage of crops
or manner of planting (e.g. industrial crops, energy crops, fodder crops, production of
grain, harvested green, kitchen garden). Tropical fruits like coffee, cacao, coco, banana,
pineapple etc. are not part of the (European) plot activity list, but can be added
independently to the data model.
14
- Leaf character (under vegetation characteristics)
The current matrix version contains “leaf character” under vegetation characteristics with
the subtypes “sclerophyte” or “regular”. For future modifications of the matrix other leaf
characters could be addressed (like spines, or leafless). Also the name “leaf character”
could be changed in a more fitting term like “leaf type” or “leaf morphology”.
- Alpine skiing sport infrastructure
So far, ski pistes are elements in the matrix as a subtype under the LUA sport
infrastructures. However, the impact of alpine sports is not restricted to the ski tracks
and pistes. Also ski lifts and water reservoir for artificial snow making could be included
in a renamed “alpine sports infrastructure”.
- Coastal / Transitional / Marine Waters
In the tested nomenclatures, water surfaces that have the character of a transitional
zone between inland areas and open sea like estuarine and intertidal areas are
categorized differently. In particular this is the case for estuaries and lagoons. The EAGLE
model sees all water surfaces with no direct connection to sea water as “inland water”.
All water surfaces in the zone seaward of the average high tide line belong to marine
waters. During the testing exercise it was identified as sometimes difficult to categorize
any water surface beyond high tide line as marine. It means that an estuary area, as
long as it is under tidal influence, would be seen as “marine”.
In the current matrix version, a lagoon disconnected from the open sea is “inland water”,
the salinity can be described in the characteristics. A lagoon with connection to the sea is
“marine”. The estuary of a river can be described as “inland water course” under “tidal
influence”.
The question is – with regards to an aggregated point of view to sum up water areas on a
higher semantic level - if this is sufficiently solved with the current matrix/model (with
inland and marine), or if it would be better to have a new component like “transitional
waters” between inland water bodies and open sea/marine waters, that allows to clearly
differentiate between those three water surface types. The reason for this ambiguous
point of view of water surfaces is that several criteria appear to be relevant, namely the
geographical location, the genesis, the connection to the sea, and the salinity of water
bodies.
- Ecosystem types / Habitat types / Bio-geographical regions
Within the matrix, the European nature information system (EUNIS) classification is
integrated as a Characteristic to describe the ecosystem types. It is an existing
classification that is somehow commonly used but still some critiques from the user side
remains. Here the scope is not to evaluate EUNIS but see it if can serve as a standard
nomenclature to be integrated in the matrix. In this case, it is a special situation because
some of the EUNIS classes (forest areas) have been selected to be bar coded while
themselves being part of the EAGLE matrix under the Characteristics block as “Ecosystem
Types”. It is not decided yet finally, if the EUNIS classes will remain part of the matrix to
indicate the ecosystem type of a land unit, or if another code list is integrated to the
matrix, which is more consistent within its own hierarchy.
The question remains, if also some generic terms to address bio-geographic regions (e.g.
boreal, continental, Mediterranean, Pannonian etc.) should be included in the matrix as a
character, which are also partly used for the EUNIS definitions.
3.1.2 New matrix element proposals
- Minimum Mapping Unit
It was suggested to integrate a character for the minimum mapping unit (MMU) that is
applied to a class. It can help to indicate the scale of class application.
15
- Soil types
For habitat or ecosystem description and evaluation it is important to have some
information on the soil type and soil condition such as soil pH, mineralogy, nutrients,
presence of peat and its depth, or also drainage of the soil. As soil is an INSPIRE annex
theme, there exist already data specifications on soils. When some more information
about soils is integrated in the EAGLE model, consideration of the INSPIRE specification
should be included. In the best case, some details from the soil specifications can be
taken over in the form of code lists.
- Protection from natural hazards
A new matrix segment for “protection from natural hazards” (with elements like
avalanche protection, flood protection, forest fire protection etc.)is proposed to be added
under the LUA block, possibly integrated in the position of “Other uses” (from
Hierarchical INSPIRE Land Use Classification System (HILUCS)) or separately and in
connection with the INSPIRE theme ‘Natural Risks Zones.
- Continuous parameter values and range
Typical parameters for land cover types are the soil sealing degree, the crown cover
density or the general proportion of a land cover component within a land unit. In the
current matrix version, there is a matrix cell to place a percentage value for each of
those characters. In addition to that it would be helpful to also allow ranges (minimum
and maximum) of parameter values, like e.g. 30-80% soil sealing degree. This would
help to capture thresholds like often used in class definitions.
- Generic landscape type, spatial context
For class definitions that address a broader variety within generic land cover or land use
types (with terms like “urban areas” or “forest areas” or “agricultural land”), a larger
number of matrix elements can apply, many of which do not have to be mandatory
elements but still are somehow typical for such broad classes. It is more the general
surrounding that is relevant than the single LCCs.
For such class definitions with little precise information about land cover or land use, the
character of “spatial context” has been added to the Characteristics block of the matrix
with elements like “urban, rural, coastal, mountainous”. For further diversification the
term “forested areas” could be added (not only addressing tree covered areas, but in
general areas which are dominated by forests). Multiple selections of these characters
are possible in the model/matrix.
3.2 BAR CODED CLASSES
The chosen nomenclatures (respectively their selected classes) have been expressed in
terms of the EAGLE matrix by using the bar coding method. The extent of semantic
testing has been limited to a reasonable depth of hierarchy and/or a representative
selection of major land cover sub-themes. The following subchapters include brief
paragraphs for each nomenclature on its data producer, scope, purpose and application
context. The classes listed here are only the ones which have been bar coded in the
semantic testing exercise. The following lists therefore do NOT contain in all cases the
entire and complete nomenclatures.
A list of all class names as well as the bar coding results in matrix format are attached as
Annex 2 (T2.2_Annex2 _BarCodingResults.xlsx). Each contribution for each nomenclature
is contained in a separate excel sheet. An additional sheet ”ALL_compilation” contains the
bar coded classes from all addressed nomenclature unified in one single sheet. Like that,
it is now possible to directly compare the bar coded classes among each other. The
source column (A) indicates the abbreviation of the source nomenclature.
16
Between the matrix element lines and the coded class line, a filter has been established
for each matrix column, i.e. the LCCs, LUAs and CHs. With this filter it is possible to
customize the portrayal of coded classes by given bar code values. For example, to find
out all classes that describe “broad-leaved timber forest”, the filter can be set to show all
classes where the presence of ‘trees’ is a mandatory element (bar code value 2, 3 or 4)
of class definitions. The filters on LCCs can be combined with LUA and CH, for example
with the land use types ‘forestry’ under the LUA block, and under the CH block,
vegetation characteristics, the ‘leaf form’ can be set on ‘broad leaved’.
3.2.1 EU, EEA: CORINE Land Cover
•
Dataset details:
CLC is the primary land cover / land use (LC/LU) dataset for the EEA and is used in
reporting, indicator development and environmental assessment. Since its launch in 1986
it has become a quasi-standard for LC/LU mapping in Europe. The CLC2012 projects
represents the third update cycle in the history of CLC datasets, with precursors
CLC1990, CLC2000 and CLC2006. All update phases have included the mapping of
changes between the two reference dates. The CLC2012 project is implemented within
the framework of GMES Intial Operations continental component (GIO land). The
coordination of the project is delegated to EEA. The CLC datasets (CLC2012 and CLCChanges2006-2012) are created by national teams of the participating 39 European
countries.
The basic methodology of CLC production is described in detail in the CLC2006 Technical
Guidelines. According to this, CLC-Change2006-2012 is the primary and most important
product of the CLC2012 project. CLC-Change2006-2012 is an individual product having a
smaller MMU (5 ha) than CLC2006 and CLC2012 (25 ha). The overall thematic accuracy
of the change database shall be >85 % (similarly to CLC2006).
The land cover changes are mapped by interpretation of visually detectable land cover
differences on images from 2006 and 2012. As ancillary data, topographic maps,
orthophotos, HRL, Google Earth imagery are used most frequently.
The Majority of countries still apply computer aided visual photointerpretation in the CLC
production, however an increasing number of countries is changing to the deriving of CLC
data from national datasets with semi-automated methods in order to ensure synergy
with national land monitoring initiatives.
CLC Illustrated Guide URL: http://sia.eionet.europa.eu/CLC2000/classes
Enhanced CLC nomenclature guidelines, as outcome of ETC-SIA IP2013 Task 261_2:
URL: http://forum.eionet.europa.eu/nrc_land_covers/library/gio-land/corine-land-coverclc/technical-guidelines/enhanced-clc-nomenclature-guidelines
•
CLC Level 3 classes
1.1.1.
Continuous urban fabric
1.1.2.
Discontinuous urban fabric
1.2.1.
Industrial, commercial, public and private units
1.2.2.
Road and rail networks and associated land
1.2.3.
Port areas
1.2.4.
Airports
1.3.1.
Mineral extraction sites
1.3.2.
Dump sites
1.3.3.
Construction sites
1.4.1.
Green urban areas
17
1.4.2.
Sport and leisure facilities
2.1.1.
Non-irrigated arable land
2.1.2.
Permanently irrigated land
2.1.3.
Rice fields
2.2.1.
Vineyards
2.2.2.
Fruit trees and berry plantations
2.2.3.
Olive groves
2.3.1.
Pastures
2.4.1.
Annual crops associated with permanent crops
2.4.2.
Complex cultivation patterns
2.4.3.
Land principally occupied by agriculture with significant areas of natural vegetation
2.4.4.
Agro-forestry areas
3.1.1.
Broad-leaved forest
3.1.2.
Coniferous forest
3.1.3.
Mixed forest
3.2.1.
Natural grassland
3.2.2.
Moors and heathland
3.2.3.
Sclerophyllous vegetation
3.2.4.
Transitional woodland / shrub
3.3.1.
Beaches, dunes and sand planes
3.3.2.
Bare rock
3.3.3.
Sparsely vegetated areas
3.3.4.
Burnt and damaged by disaster areas
3.3.5.
Glaciers and perpetual snow
4.1.1.
Inland marshes
4.1.2.
Peat bogs
4.2.1.
Salt marshes
4.2.2.
Salines
4.2.3.
Intertidal flats
5.1.1.
Water courses
5.1.2.
Water bodies
5.2.1.
Coastal lagoons
5.2.2.
Estuaries
5.2.3.
Sea and ocean
•
Comments:
Regarding the codification of CLC it once more turned out that especially the mixed
classes like CLC 242 or 243 have been designed to “handle” a possible combination of
different (and often unrelated) small scaled land cover types (below MMU) within one
spatial unit, delineated by the editor in the form of a cartographic object (CLC polygon).
The spatial distribution and spatial allocation of the different land elements is mostly
arbitrary and polygon-related, not class-specific. Even if the appearance of all mixed
class polygons may look different, it is still possible with the EAGLE concept to
disaggregate the thematic composition of a mixed class in separated semantic sub-units
(by the multi-line bar coding approach), characterize them, and then bring them into a
more generic “mosaic” spatial pattern.
18
With a view to a database implementation, in combination with a possible future larger
mapping scale, this means that the concept of “mosaic class” should be managed at the
level of the database rather than at the level of the classification system. This is not
applicable for mixed classes that refer to inherent fine mixed pattern, i.e. classes 241
and 244.
3.2.2 EU, Eurostat: LUCAS
•
Dataset details:
LUCAS stands for Land Use and Cover Area Frame Survey. Since 2006, EUROSTAT
carries out a survey on the state and the dynamics of changes in land use and cover in
the European Union called in short the LUCAS survey. It is a sample point-based survey
(not a wall to wall mapping initiative). The sample point are randomly chosen from a 2x2
km grid, after a pre-selection of major land cover types and weighted total number of
potential sampling points out of all grid points. For creation of statistical time lines,
Eurostat tries to keep the same points in the sample from survey to survey. However,
due to modifications of the criteria for the total pool of potential sample points (exclusion
above 1000/1500 meters height, distance to road, accessibility) the number and
distribution of field sample points has partly altered.
Since 2006, LUCAS is executed every three years as an in-situ survey on the ground all
over the EU. The latest finalized LUCAS survey (2012) covers all the then 27 EU countries
and observations on more than 270 000 points. Currently in 2015 between March and
October, EUROSTAT is carrying out the LUCAS 2015, where surveyors from 28 Member
States will visit a total of 273.401 points. The surveyors also capture linear landscape
features along a transect connected to the point in the field to measure the
fragmentation of landscape.
The LUCAS points can fall on all land cover types (cropland, grassland, forest, built-up
areas, transport network, etc.). The surveyors examine land cover and land use,
irrigation management and structural elements in the landscape. In addition, a top-soil
sample is taken in one out of 10 points. The sample is analysed in a laboratory and used
for purposes related to assessing environmental factors, such as updating European soil
maps, validating soil models, and measuring the quantity of organic carbon in the soil
which is an important factor influencing the climate change.
From LUCAS survey three types of information are obtained: micro data (LC/LU and
environmental parameters associated to the single surveyed points), point and landscape
photos in the four cardinal directions. From the micro data, statistical tables with
aggregated results by land cover, land use are estimated and weighted for the use on
European level.
Data and information on land cover/use status and changes is according to Eurostat used
on European scale for: nature protection, forest and water management, urban and
transport planning, agricultural policy, natural hazards prevention and mitigation, soil
protection and mapping, monitoring climate change, biodiversity, etc.
URL: http://epp.eurostat.ec.europa.eu/statistics_explained/index.php/LUCAS__Land_use_and_land_cover_survey
19
•
LUCAS land cover classes (as specified by Eurostat for the 2012 survey)
A10 BUILT-UP AREAS
A00 ARTIFICIAL LAND
A20 ARTIFICIAL NON-BUILT UP
AREAS
A11 Buildings with one to three
floors
A12 Buildings with more than three
floors
A13 Greenhouses
A21 Non built-up area features
A22 Non built-up linear features
B11 Common wheat
B12 Durum wheat
B13 Barley
B14 Rye
B10 CEREALS
B15 Oats
B16 Maize
B17 Rice
B18 Triticale
B19 Other cereals
B21 Potatoes
B20 ROOT CROPS
B22 Sugar beet
B23 Other root crops
B31 Sunflower
B32 Rape and turnip rape
B33 Soya
B30 NON-PERMANENT INDUSTRIAL
CROPS
B00 CROPLAND
B34 Cotton
B35 Other fibre and oleaginous
crops
B36 Tobacco
B37 Other non-permanent industrial
crops
B41 Dry pulses
B42 Tomatoes
B40 DRY PULSES, VEGETABLES AND
FLOWERS
B43 Other fresh vegetables
B44 Floriculture and ornamental
plants
B45 Strawberries
B51 Clovers
B50 FODDER CROPS (mainly
leguminous)
B52 Lucerne
B53 Other Leguminous and
mixtures for fodder
B54 Mix of cereals
B55 Temporary grassland
B71 Apple fruit
B72 Pear fruit
B70 PERMANENT CROPS: FRUIT
TREES
B73 Cherry fruit
B74 Nuts trees
B75 Other fruit trees and berries
B76 Oranges
20
B77 Other citrus fruit
B81 Olive groves
B80 OTHER PERMANENT CROPS
B82 Vineyards
B83 Nurseries
B84 Permanent industrial crops
C10 Broadleaved woodland
C20 Coniferous woodland
C00 WOODLAND
C30 Mixed woodland
D00 SHRUBLAND
C21 Spruce dominated coniferous
woodland
C22 Pine dominated coniferous
woodland
C23 Other coniferous woodland
C31 Spruce dominated mixed
woodland
C32 Pine dominated mixed
woodland
C33 Other mixed woodland
D10 Shrubland with sparse tree cover
D20 Shrubland without tree cover
E10 Grassland with sparse tree/shrub cover
E00 GRASSLAND
E20 Grassland without tree/shrub cover
E30 Spontaneously re-vegetated surfaces
F10 Rocks and Stones
F00 BARE LAND AND
LICHENS/MOSS
F20 Sand
F30 Lichens and Moss
F40 Other bare soil
G10 Inland water bodies
G00 WATER AREAS
G20 Inland running water
G30 Coastal water bodies
G50 Glaciers, permanent snow
H10 Inland wetlands
H00 WETLANDS
H11 Inland marshes
H12 Peatbogs
H21 Salt marshes
H20 Coastal wetlands
H22 Salines
H23 Intertidal flats
•
Comments:
As within the LUCAS nomenclature the themes of land cover and land use are handled
with two separate class lists, the focus of the bar coding exercise lies only on the
targeted theme, i.e. LUCAS LC classes are only codified with Land Cover Components
(and Characteristics), LUCAS LU classes focus only on Land Use Attributes.
3.2.3 EU, DG Regio: Urban Atlas 2012
•
Dataset details:
The European Urban Atlas is part of the local component of the Copernicus land
monitoring services. It provides high-resolution land use maps for 700 Functional Urban
21
Areas(FUA) (in 2009: 305 Large Urban Zones (LUZ)) and their surroundings (more than
100.000 inhabitants as defined by the Urban Audit) for the reference year 2012.
The Urban Atlas was created to fill a gap in the knowledge about land use in European
cities, focusing on inter-comparable land use data in urban areas, to compare land use
patterns amongst major European cities and to benchmarking cities in Europe. It is
based on satellite imagery.
The legend of Urban Atlas is designed to capture urban land use, including low density
urban fabric at a scale of around 1:10.000. The Urban Atlas also aims at providing a
picture of urban sprawl in the fringe of urban zones.
The Urban Atlas is a joint initiative of the European Commission Directorate-General for
Regional Policy and the Directorate-General for Enterprise and Industry with the support
of the European Space Agency and the European Environment Agency.
URL:
•
http://land.copernicus.eu/local/urban-atlas/urban-atlas-2012
Level 4 classes of Urban Atlas.
Classes in grey font have not been encoded, but for completeness reasons are listed as
part of artificial surfaces.
111 - Continuous
Urban Fabric
11 - Urban
fabric
1 - Artificial
surfaces
12 Industrial,
commercial,
public,
military,
private and
transport
units
112 - Discontinuous
Urban Fabric
113 - Isolated
Structures
121 - Industrial,
commercial, public,
military and private
units
122 - Road and rail
network and
associated land
123 - Port areas
124 - Airports
131 - Mineral
extraction and dump
13 - Mine,
sites
dump and
133 - Construction
construction
sites
sites
134 - Land without
current use
14 141 - Green urban
Artificial
areas
1110 - Continuous Urban Fabric (S.L. >
80%)
1121 - Discontinuous Dense Urban
Fabric (S.L. : 50% - 80%)
1122 - Discontinuous Medium Density
Urban Fabric (S.L. : 30% - 50%)
1123 - Discontinuous Low Density
Urban Fabric (S.L. : 10% - 30%)
1124 - Discontinuous Very Low Density
Urban Fabric (S.L. < 10%)
1130 - Isolated Structures
1210 - Industrial, commercial, public,
military and private units
1221 - Fast transit roads and associated
land
1222 - Other roads and associated land
1223 - Railways and associated land
1230 - Port areas
1240 - Airports
1310 - Mineral extraction and dump
sites
1330 - Construction sites
1340 - Land without current use
1410 - Green urban areas
22
nonagricultural
vegetated
areas
2Agricultural
areas,
seminatural
areas and
wetlands
142 - Sports and
leisure facilities
21 - Arable land
22 - Permanent crops
23 - Pastures
24 - Complex and mixed cultivation
patterns
25 - Orchards
31 - Forests
3 - Forests,
herbaceous
vegetation
and open
spaces
32 - Herbaceous vegetation
associations
33 - Open spaces with little or no
vegetation
4 - Wetland
5 - Water bodies
1420 - Sports and leisure facilities
21000 Arable land (annual crops)
22000 Permanent crops (vineyards,
fruit trees, olive groves)
23000 Pastures
24000 Complex and mixed cultivation
patterns
25000 Orchards
31000 Forests
32000 Herbaceous vegetation
associations (natural grassland,
moors…)
33000 Open spaces with little or no
vegetation
40000 wetland
50000 Water bodies
3.2.4 EU, DG Env: EUNIS habitat classification
•
Dataset details:
EUNIS stands for European Nature Information System. It brings together European data
from several databases and organisations into three interlinked modules on sites, species
and habitat types. The EUNIS information system is part of the European Biodiversity
data centre (BDC) and it is a contribution to the knowledge base for implementing the EU
and global biodiversity strategies and the 7th Environmental Action Programme.
The EUNIS database includes information about:
-
Data on species, habitat types and designated sites compiled in the framework of
Natura 2000 (EU Habitats and Birds Directives);
-
The EUNIS habitat classification;
-
Data from material compiled by the European Topic Centre of Biological Diversity;
-
Information on species, habitat types and designated sites mentioned in relevant
international conventions and in the IUCN Red Lists;
-
Specific data collected in the framework of the EEA's reporting activities, which
also constitute a core set of data to be updated periodically, e.g. Eionet priority
dataflow
-
Nationally designated areas (CDDA).
The purpose of EUNIS is to provide a reference information system for data users
working in ecology and conservation. It is also used for assistance to the Natura 2000
process (EU Birds and Habitats Directives) and coordinated with the related EMERALD
Network of the Bern Convention, the development of indicators (EEA Core Set) and also
environmental reporting connected to EEA reporting activities.
In this task the forest classes of the EUNIS habitat classification are used for bar coding.
http://biodiversity.europa.eu/maes/typology-of-ecosystems
23
•
EUNIS Forest areas
G - Woodland, Forest and other
wooded land
•
G1 - Broadleaved deciduous woodland
G2 - Broadleaved evergreen woodland
G3 - Coniferous woodland
G4 - Mixed deciduous and coniferous woodland
G5 - Lines of trees, small anthropogenic woodlands, recently felled
woodland, early-stage woodland and coppice
Comments:
The EUNIS coding has been restricted to the forest part.
The options to describe dwarf trees affected by limited growth due to climate conditions
could be optimized within the matrix.
Going deeper into the sub-classes of EUNIS they become too species-oriented, and
therefore are not addressed in this exercise.
The further subtypes of EUNIS classes (beyond class G4) partly are related to
biogeographical regions. When simply following the EUNIS hierarchy, it is not necessary
to change the matrix because the information is given within these very same EUNIS
classes as they are at the moment integrated in the matrix.
The question remains, if also some generic terms to address bio-geographic regions (e.g.
boreal, continental, Mediterranean, Pannonian etc.) should be included in the matrix as a
character, which are also partly used for the EUNIS definitions.
3.2.5 EU, JRC: Pan-European list of agriculture LPIS land cover
types
(as used within the scope of the annual LPIS Quality Assessment)
•
Dataset details:
The list given below represents a set of possible agriculture land cover types found in
Europe that can be eligible for direct aids within the scope of the EU Common Agriculture
Policy (EU CAP). This nomenclature was designed and compiled by the Joint Research
Centre of the European Commission (DG JRC) for the annual quality assessment of the
Land Parcel Identification System (LPIS). Its purpose is to provide a common starting
point for the EU Member States to produce a complete list of their agriculture land cover
types (arable land, natural grassland, permanent crops, etc...) with a quantification of
their agriculture potential. As such quantification had to be based purely on the
physiognomic-structural characteristics of the land cover, independent from actual land
use or socio-economic factors, the land cover types modelled by JRC were based on the
3-dimensonal Tegon concept, described with the semantics of the FAO Land Cover
Classification System (v.2).
This list was made part of the so-called “eligibility profile”, which is the conversion table
allowing mapped land cover features to be expressed in eligibility terms. Eligibility is
relevant for the calculation of the maximum eligible area and for counting the presence
of ineligible features. It provides the correct quantitative determination within a single
methodology by joining the common pan-European qualitative aspect of the land cover
features with the national rules and supporting schemes applied to the measured areas.
In other words, it converts the results of the land cover mapping into “eligible hectares or
features found”.
JRC WikiCAP, LPIS Quality Assessment, Executable Test Suite, v. 5.3, Annex III
URL:
ftp://mars.jrc.ec.europa.eu/lpis/documents/v53_May2014/5_3_Annex_III_LC_concept_e
ligibility_20140507.pdf
24
•
Agriculture land cover classes:
1
Arable Land (general)
2
Arable Land (rainfed with fallow system)
3
5
Arable Land (temporary resting)
Arable Land with Patches of Trees (up to 15% of the
surface)
Arable Land with Patches of Scattered Trees (up to 4% of
the surface)
6
Agriculture with Cultivated Trees (intercropping)
7
Permanent pasture (self-seed or sown)
8
Permanent pasture (self-seed)
9
Permanent pasture (sown)
4
10
Permanent pasture (self-seed with shrubs)
11
Permanent pasture (self-seed with sparse trees)
12
Kitchen Gardens
13
Permanent crops (tree type)
14
Permanent crops (orchards)
15
Permanent crops (plantation)
16
Permanent crops (olive trees)
17
18
Permanent crops (shrub type)
Permanent crops
(soft fruits)
19
Permanent crops (vineyards)
20
Permanent crops (hops)
21
Cotton Fields
22
Rice fields
23
Starch Potatoe
24
Sugar beet
25
Pulses and Vegetables
26
Greenhouses
27
28
29
Waterlogged natural or semi-natural vegetation developed
on former agriculture land due to the implementation of
Art 34 (2i and 2iii) of 73/2009
Afforested former agriculture land - implementation of Art
34 (2ii) of 73/2009
Short Rotation Coppice – broadleaved deciduous (definition
2013, revised)
25
30
Short Rotation Coppice – broadleaved evergreen
(definition 2013, revised)
3.2.6 Bulgaria & Romania: Cross-border LC nomenclature
•
Dataset details:
The common CBC land cover nomenclature was designed to enable the elaboration of
reference land cover for the cross-border area between Bulgarian and Romania. This is a
key dataset as part of the common information resources at regional level developed
within the Cross-Border Cooperation (CBC) Project “Common Strategy for Sustainable
Territorial Development of the cross-border area Romania-Bulgaria – project SPATIAL”,
funded by the European Regional Development Fund. The elaboration of the reference
land cover dataset took into account the abundant information on land cover already
captured and stored in various products elaborated in the scope of different panEuropean and national initiatives, such as CORINE Land cover (CLC), FAO TCP/BUL/8922,
Bulgarian Spatial Data Infrastructure/BULCOVER-2009-2010 (BSDI) and FAO
TCP/ROM/2801. The requirements for this reference land cover were more stringent and
demanding than many other specifications of classical land cover mapping. What was
needed was a management system to effectively monitor the land changes and to
support impact assessment of sectoral policy interventions and EU funds expenditures at
regional level. Thus, the land cover nomenclature is generated following the
recommendations given in Annexes F and G of INSPIRE Data specifications on Land
Cover. It has no multi-level classification, as all classes are on the same level of semantic
description and are considered mutually exclusive and exhaustive. Nomenclature is scale
and product (map) independent and all classes in the legend represent either single or
“pure” land cover type, or a functional mix. It doesn’t include cartography-related
mixtures between different land cover types, as classes. It is purely based on land cover
semantics and all class entries are conceptually modelled in the semantic apparatus of
LCML
Bulgarian National BulCover, Bulgarian reference layer BulCover: URL: http://bsdi.asdebg.org/lccs.php;
CBC Project SPATIAL – MIS ETC 171 (SPATIAL), URL: http://www.cbc171.asdebg.org/project_results.php?wp=3
LCCS-based common cross-border nomenclature; Project – MIS ETC 171 (SPATIAL),
URL: http://www.cbc171.asde-bg.org/docs/pp9-wp3CBC_Nomenclature_Samples_final.pdf
PP9-[WP3] – LCML modelling of classes for the CBC - Reference Land Cover Layer in
Bulgaria,
http://cbc171.asde-bg.org/docs/pp9-wp3-CBC_Legend_LCML_Codes_Ref_LC.pdf
Project FAO TCP/ROM/2801, URL: http://rosa.ro/index.php/en/cercetare/proiecteinternationale/105-un/122-fao-tcp-rom-2801-land-cover-land-use-inventory-by-remotesensing-for-the-agricultural-reform
•
CBC Land Cover Classes
1
Arable land
2
Managed Grassland
3
Paddy Rice Field
4
Natural Grassland
5
Tree Crop
26
6
Tree Plantation
7
Broadleaf Deciduous Forest
8
Broadleaf Evergreen Forest
9
Coniferous Forest
10
Mixed Broadleaf and Coniferous Forest
11
Woodland
12
Waterlogged Forest
13
Urban Vegetated Areas
14
Association of herbaceous and woody crops
15
Association of crops and natural trees
16
Shrub Crop
17
Scrubland
18
Waterlogged Vegetation
19
Artificial non-build up surface
20
Continuous urban fabric
21
Discontinuous urban fabric
22
Fragmentary urban fabric
23
Artificial build up surface
24
Artificial build up network
25
Consolidated bare surface
26
Unconsolidated bare surface
27
Rivers
28
Channels
29
Lakes and impoundments
30
Reservoirs
31
Covered Agriculture land (Greenhouse)
3.2.7 Hungary: national high-resolution Land Parcel Identification
System (LPIS), MePAR
•
Dataset details:
The Institute of Geodesy, Cartography and Remote Sensing created and started
operating the Land Parcel Identification System (LPIS; in Hungarian: MePAR) in 2004, in
accordance with the relevant Community legislation. MePAR has several functions:
MePAR is the exclusive reference, land parcel identification and spatial information
system of the agricultural and rural development subsidies, financed by EU and national
sources (regulation 115/2003 of the Ministry of Agriculture and Rural Development on
the Land Parcel Identification System).
MePAR is a reference system, based on which the farmers as clients of the Agriculture
and Rural Development Agency can specify their area related grant applications and data
services by unique identifiers out of the data set of the MePAR land parcel blocks.
27
MePAR is a land identification system, because the paying agency and the contributory
agricultural management organizations can use the land parcel identification system for
area-based grants management and administrative and physical (conventional on-site
and remote sensing) controls, investigating the justness of claims.
MePAR is a GIS too, which provides up-to-date information for each participant about
each agricultural table, indicating which areas are eligible, which payment entitlements
can be requested, and what restrictions should be obeyed during the farming.
The basic data of MePAR are the boundaries of physical blocks and (for payment)
ineligible areas, the unique block identifier, and the area data. The additional data of
MePAR, the thematic layers provide support to the implementation of certain support
schemes, designating the areas eligible for application or defining the territorial scope of
certain obligations.
•
1.
1.1.
1.1.1.
1.1.2.
1.2.
1.2.1.
1.2.2.
1.2.3.
1.2.4.
1.3.
1.3.1.
1.3.2.
1.3.3.
1.3.4.
1.3.5.
1.4.
1.4.1.
1.4.2.
1.4.3.
1.5.
1.5.1.
1.5.2.
1.5.3.
1.5.4.
1.6.
1.6.1
MePAR classes:
Subsidizable Areas
Arable land
Arable land, glass greenhouses and plastic greenhouses,
floriculture and ornamental horticulture, nurseries
Arable land split up by trenches, dikes and channels, rice
fields.
Grassland, meadow, pasture, hayfield
Pure grassland (without woody vegetation)
Bushy-woody grassland with trenches and/or other
infrastructure (part of animal husbandry)
Corral, sheepfold
Wooded meadow and pasture (agro-forestry)
Plantations
Fruit tree plantation
Vineyards
Energy crop plantation (not registered in the woody energy
crop layer)
Plantation (berries. mixed or type unidentified)
Energy plantations registered in the woody energy crop
layer
Areas with complex cultivation structure
Agricultural areas with fragmented parcels and mixed
agricultural use.
Agricultural areas with fragmented parcels and mixed
agricultural use, with scattered infrastructure, and/or other
not subsidized area
Kitchen gardens
NMUV(Non-Cultivated Areas)
Non-cultivated areas on arable land
Non-cultivated areas on grassland
Non-cultivated areas on plantation
Non-cultivated areas on complex cultivation structure
Special subsidizable area
Wooded-shrubby areas on subsidizable “kunhalom”
(ancient burial mounds)
111
112
121
122
321
322
131
132
133
134
135
141
142
143
151
152
153
154
28
2.
2.1.
2.1.1.
2.1.3.
2.1.4.
2.1.5.
2.2.
2.2.1.
2.2.2.
2.2.3.
2.2.4.
2.2.5.
2.3.
2.3.1.
2.3.2.
2.3.3
2.3.4
2.3.6.
2.4.
2.4.1.
2.4.2.
2.4.3.
2.5.
2.5.1.
2.5.2.
2.5.3
2.5.4.
2.5.5.
•
Non-subsidizable Areas
Forest
Forest
Forest patches, group of trees or shrubs
NVT Afforestation of agricultural areas (NVT-MGTE)
EMVA- Afforestation of agricultural areas (EMVA-MGTE)
Linear facilities
Dirt road, sealed road (not yet separated to 224 and 225)
Railway
Dike, dam
Sealed road
Permanent dirt road, unsealed road
Infrastructure
Settlements, built-up areas, industrial areas
Military area masked on orthophoto
Homestead / solitary farmstead
Permanent bale and manure depot
Miscoded military area
Waters
Lake, fishpond
River, streamlet, creek, canal
Reed (Fields registered within the AKG-Cane coverage)
Natural vegetation
Parcel edges, bounds, tree lines, narrow shelterbelts not
subdivided yet to 254 and 255
Marshes, morasses, wetlands, bogs which are not AKGfields
NMUV non-subsidized areas (that undergone irreversible
changes not allowing cultivation any more)
Parcel edges with grass / herbaceous vegetation cover
Parcel edges with woody vegetation cover
211
213
214
215
221
222
223
421
422
231
232
233
234
432
241
242
243
251
252
253
451
452
Comments:
Some classes address areas under agricultural usage with presence of cultivation
measures like irrigation, fertilization etc. In such cases, the classes can include several
types of crop parcels that not all are managed necessarily in the same way. In theses
cases, the Characteristics are assumed in a categorical way. It then more expresses the
probability of certain cultivation measurements.
3.2.8 Netherlands: National Land Use Datasets (Bestand
Bodemgebruik BBG) and Land Cover Database (LGN)
•
Dataset details National land use datasets (BBG)
Land use database produced by the Office of Statistics (CBS) with Top10vector/Top10NL
database as geometrical basis. The BBG recognises 38 different land uses (on basis of
functionality) with special focus on urban area (CBS 2002). The MMU for most land use
classes is 1 ha. In the near future the 2012 version will be published. Other versions are
the ones indicating the land use with reference years 1989, 1993, 1996, 2000, 2003,
2006, 2008 and 2010 showing an increase in update frequency. The social/economic use
29
of land is the main criteria to classify land into different categories. Recent aerial photos
and other ancillary data are used for regular updates of Dutch land use. Trend in land use
changes can be monitored.
http://geodata.nationaalgeoregister.nl/bestandbodemgebruik2010/atom/bestandbodemg
ebruik2010.xml
Hazeu, G.W., Bregt, A.K., de Wit, A.J.W and J.G.P.W. Clevers, 2011. A Dutch multi-date
land use database: Identification of real and methodological changes. International
Journal of Applied Earth Observation and Geoinformation 13, 682–689
•
BBG classes
20
21
22
23
24
40
41
42
50
51
60
•
Residential
Retail trade, hotel and catering
Public institutions
Socio-cultural facility
Industrial area and offices
Park and public garden
Sports ground (incl. car parks)
Allotment garden
Greenhouses
Other agricultural usage
Woodland
Dataset details National Land Cover Database (LGN)
The LGN database is a gridded database of 25 by 25 m cells reflecting the LC for 39 welldefined LC classes including urban areas, forest types, water, infrastructure, crop types
and several ecological classes. Since 1986 at regular intervals of 3-5 years a version of
LGN was released presenting the LU/LC for the Netherlands for a specific year (Hazeu et
al., 2011; Hazeu, 2014). The dataset is produced by Alterra – Wageningen University
and Research Centre. From the beginning the nomenclature of the LGN database had its
focus on the agricultural domain. From 1995 onwards also the nature areas were further
subdivided into various LC/LU classes. Recently the 7th version of LGN (LGN7) was
produced reflecting the Dutch LU/LC for the year 2012 (Hazeu et al., 2014). The LGN6
methodology as described in Hazeu (2014) and Hazeu et al. (2011) was used in a slightly
adapted form for the production of LGN7 (Hazeu et al., 2014). The two main processing
steps are 1) an object-oriented classification in which national databases are integrated
and a LGN classes is assigned to the polygons and 2) a pixel-based classification in which
the homogeneous polygons are enriched with more spatial information. In the first step
also the LC changes are detected between 8 main LC groups.
URL: www.lgn.nl;
http://www.wageningenur.nl/en/Expertise-Services/Research-Institutes/alterra/FacilitiesProducts/Land-use-database-of-the-Netherlands.htm
Hazeu, G.W., 2014. Operational land cover and land use mapping in the Netherlands. In:
Manakos, I & Braun, M. (Eds.): Land Use and Land Cover Mapping in Europe. Practices &
Trends. Series: Remote Sensing and Digital Image Processing, pp. 282-296. Springer,
ISBN: 978-94-007-7968-6/3
Hazeu, G.W., C. Schuiling, G.J. Dorland, G.J. Roerink, H.S.D. Naeff and R.A. Smidt,
2014. Landelijk Grondgebruiksbestand Nederland versie 7 (LGN7); Vervaardiging,
nauwkeurigheid en gebruik. Wageningen, Alterra Wageningen UR (University & Research
Centre), Alterra-rapport 2548. 86 blz.; 16 fig.; 12 tab.; 15 ref.
CBS (2002). Productomschrijving Bestand Bodemgebruik. Den Haag, the Netherlands,
Centraal Bureau voor de Statistiek (CBS).
30
•
LGN classes
1
2
3
4
5
6
8
9
10
11
12
18
19
20
23
28
61
62
•
Pasture/grassland
Maize
Potato
Sugar beet
Cereals
Other crops
Greenhouses
Orchards
Flower bulbs
Broadleaved forest
Coniferous forest
Housing in urban areas
Housing in semi-urban areas
Forest in urban areas
Grass in urban areas
Grass in semi-urban areas
Tree nurseries
Fruit nurseries
Comments:
The bar coding for the Netherlands was only applied to a limited number of classes from
two databases: the national land cover database (LGN) and the national land use
database (BBG). The focus of the bar coding exercise was on urban, forest and
agricultural classes.
In the Dutch case, land characteristics are often not described in the definitions of the
classes. Crop type and agricultural management (limited) are often indicated in the
definitions. Spatial pattern is indirectly defined – not directly named in the definitions.
The same counts for ecosystem types and biophysical characteristics. For that reason
biophysical characteristics and ecosystem types have been coded with bar code 1
(optional can occur). Furthermore, from class definitions it is not in every case clear
whether the classes fulfil all criteria of a specific ecosystem type (as described by EUNIS
definitions).
In NL it is difficult to describe classes with multiple lines. The class description is
condense and not easy to link to different LUA or LCC. The bar code vale “x” was not
used because in the Dutch class definitions nothing is explicitly excluded.
The bar coding for some LGN classes is difficult as the classes are based on the
combination of different datasets. To do the proper bar coding you have to take into
account the definitions of all datasets used to come up to the definition of the LGN class.
3.2.9 Spain: Land Use Information System / Sistema de
Información sobre Ocupación del Suelo en España (SIOSE)
•
Dataset details:
SIOSE is the Spanish national database for land cover and land use. It is produced within
the responsibility of the regions and is compiled at the National Geographic Institute
(IGN) to a complete national dataset. The data set is updated on a regular bases. It is
used for national purposes but also to derive the Spanish contribution to CORINE Land
Cover through semi-automated generalization processes. The updating is mainly based
31
on photointerpretation of aerial and satellite imagery, combined with ground-truthing in
the fields, if needed. The MMU in SIOSE varies from 0,5 to 2 ha.
SIOSE has 40 simple classes and 45 composed classed, which means that in those classes
both LC and LU terms occur. Also simple classes can be described with attributes, a total
of 20 attributes are present in SIOSE.
Sistema de Información sobre Ocupación del Suelo en España (SIOSE). Documentation
at (in Spanish): http://www.siose.es/web/guest/documentacion SIOSE classes
701
Dehesa
844
Campsite
702
Olive grove-vineyard
860
Urban park
703
Agricultural settlement
842
Hotel resort
704
Home orchard
881
Road network
811
Town centre
882
Rail network
812
Urban expansion area
884
Airport
813
Discontinuous
883
Port
831
Farming
897
Pipeline
832
Logging
911
Water treatment
833
Mining
912
Water supply network
834
Fish farming
913
Desalination plant
821
Planned industrial park
921
Dump site
822
Non-planned industrial park
922
Waste treatment plant
823
Isolated industry
101
894
Electric
104
893
Nuclear
895
131
Thermal
Buildings
Roads, parking lots and other artificial
surfaces
Extraction and dumping
891
103
Wind power
Artificial water body
892
111
Solar
Other constructions
121
Non-built soil
102
Green urban area
211
Rice
212
Herbaceous other than rice
222
Citrus fruit trees
223
Non-citrus fruit trees
231
Vineyard
232
Olive grove
241
Other woody crops
290
Pastures
300
Grassland
312
Deciduous broad-leaved
313
Evergreen broad-leaved
316
Coniferous
896
Hydroelectric
841
Commercial and business park
900
Telecommunications
851
Government
855
Penitentiary
854
Educational
852
Health
856
Religious
853
Cemetery
857
Cultural
843
Amusement park
858
Sports
859
Golf course
32
320
Shrub
412
Peat bogs
333
Bare soil
413
Inland salines
334
Burnt areas
421
Coastal marshes
335
Glaciers and perpetual snow
422
Coastal salines
351
Sea cliff
511
Water courses
352
Outcrop
513
Lakes and lagoons
353
Scree
514
Reservoir
354
Quaternary lava flow
521
Coastal lagoons
331
Beaches, dunes and sand plains
522
Estuaries
336
Dry riverbed
523
Sea and ocean
411
Marshes
• Comments:
The bar coding has been applied to 85 elements. Usually, the bar code values 1 (element
can occur), 2 (selectable mandatory element) and 3 (accumulative mandatory element)
were applied, but rarely bar code value 4 (minimum two out of value-4-coded elements
mandatory). Value ‘x’ was barely used because the definitions are in most cases mutually
exclusively defined and work without explicitly mentioning excluded elements in the
definitions.
In the case of SIOSE, the approach of unified matrix cells was not applied, but each
matrix element (column) has been coded separately. In some cases, the matrix columns
where applied for LCCs, but others for LC classes.
3.2.10
Germany: Land Cover Model for Germany LBM-DE
(formerly DLM-DE)
•
Dataset details:
The Land Cover Model for Germany (LBM-DE4, formerly DLM-DE5) is a polygon vector
data set with a MMU of 1 ha and a 3-year update cycle. It is based on input data from
the national topographic reference data ATKIS Basis-DLM, which updated with remote
sensing methods to a specific reference year (2009, 2012). For the 2009 production, the
CLC class definitions were applied as consistently as possible (with respect to the much
smaller scale) to the DLM-DE. During the update phase of LBM-DE2012 again ATKIS data
is used as input data (including parts of the DLM-DE2009 results). A new semantic
approach was adopted, namely a nomenclature that contains 36 LC classes (level 2) and
17 LU classes (flat hierarchy). For each polygon, both the LC and LU information is
captured separately. Based on these LC and LU classes, CLC classes are semantically
derived.
Bundesamt für Kartographie und Geodäsie (BKG), last updated 30.03.2012, product
description (in German), URL http://www.geodatenzentrum.de/docpdf/dlm-de2009.pdf
URL: http://www.ufz.de/export/data/36/54055_Feigenspan_CLC-DLMDE_2013_11_05.pdf, Slide 13, (English documentation exists but not yet publically
provided by BKG)
4
5
LBM-DE – Landbedeckungsmodell Deutschland
DLM-DE – Digitales Landbedeckungsmodell Deutschland
33
•
LBM-DE2012 Land Cover Classes
B111
Houses (sealing > 80%)
B112
Houses (sealing 50% - 80%)
B113
Houses (sealing 30% - 50%)
B121
Halls and other facilities
B122
Sealed areas without buildings
B133
Unsealed areas without buildings
B211
Arable land
B221
Winegrowing
B222
Orchards and soft fruit populations
B224
Hops
B231
Homogeneous grassland
B321
Inhomogeneous grassland
B233
Grassland with trees (< 50%)
B421
Salt marshes (coast)
B242
Mixed areas (regular structure)
B322
Dwarf shrubs (heath)
B324
Bushes, shrubs, young trees
B311
Deciduous trees
B312
Coniferous trees
B313
Deciduous and coniferous trees
B333
Sparse vegetation
B331
Sand plains
B332
Rock
B334
Burnt areas
B335
Snow (permanent) and ice
B336
Loose rock
B411
Swamp
B412
Bog
B413
Swamp with bushes/trees < 50%
B414
Bog with bushes/trees < 50%
B423
Mud flat
B511
Watercourse
B512
Waterbodies
B521
Lagoon
B522
Estuary
B523
Open sea
34
•
LBM-DE2012 Land Use Classes
N112
N120
N121
N123
N131
N132
N134
N122
N124
N142
N141
N211
N214
N215
N311
N133
N999
•
Housing
Production
Public
Port
Mining areas
Dumpsites
Stock
Road and railway traffic
Air services
Sports and leisure
Urban green area
Agriculture (intensively)
Extensive use
Fallow land
Forestry
Construction site
Not relevant
Comments:
As within the LBM-DE nomenclature the themes of land cover and land use are handled
with two separate class lists, the focus of the bar coding exercise lies only on the
targeted theme, i.e. LBM-DE LC classes are only codified with Land Cover Components,
LBM-DE LU classes focus only on And Use Attributes, even when some LBM-DE classes
partly contain semantic of the respectively other theme.
The design of the LBM-DE classes follows the principle of separation between LC and LU,
but still is oriented quite close to the terms of CLC.
3.2.11
•
Austria: Land Information System Austria (LISA)
Dataset details:
LISA stands for Land Information System Austria. It has been developed throughout a
range of national research projects and is currently implemented by the national mapping
agency. Standardized products (land cover using basic LISA nomenclature) will be
available from 2017 onwards with a 3-year update cycle (according to orthophotoproduction). Pilot projects are currently carried out on more than 1/3 of the permanently
inhabitable (“Dauersiedlungsraum”) area financed by ESA. It is foreseen to enlarge the
classes concerning their thematic and temporal characteristics using the SENTINEL data.
LISA core project: http://www.landinformationsystem.at/en-us/lisa/overview.aspx
LISA extension: http://www.landinformationsystem.at/enus/cadasterenvcurrent/overview.aspx
35
•
LISA Land Cover Classes
built-up
abiotic
non-built
up
water
woody
biotic
herbaceous
•
1
2
3
4
5
6
7
8
9
10
11
12
13
building
other constructed area
bare soil
screes
bare rock
surface water
snow
ice
trees
bushes
dwarf shrubs
herbaceous vegetation
reeds
Comments:
As LISA is based on a feature-oriented approach in a very large scale with a MMU of 25
m2, the componential way of decomposing landscape with the LCC is straightforward.
In the current status of LISA only land cover is operationally implemented, therefore land
use classes have not been included in this exercise.
3.2.12
United Kingdom: Biodiversity Action Plan Broad
Habitats (BAP BH)
This nomenclature has been used in the UK Countryside Survey and other surveys and a
subset was the basis for the nomenclature used for EO-based national land cover
mapping in 2000 and 2007
•
Dataset details:
The Broad Habitats are described in Jackson D.L., (2000): Guidance on the interpretation
of the Biodiversity Broad Habitat Classification (terrestrial and freshwater types):
Definitions and the relationship with other classifications, JNCC Report 307, 73 pages,
ISSN 0963 8091
BAP BH is a framework classification for 37 habitat types across the whole of the UK to
inform national and local policy and action. The selection of broad habitats types should
represent a limited number of habitat types and the definitions should be simple and
easily understood by a broad range of people. In addition they aimed to provide:
- a comprehensive framework for surveillance of the UK countryside and
surrounding seas which is compatible with other widely-used habitat and landcover classifications, particularly Phase 1 mapping and the Countryside Survey.
- a means of setting priority habitats in context and a system for identifying gaps
and dealing with emerging new priorities in the list of priority habitats; and
- a means of characterising patterns and mosaics upon which wide-ranging species
are dependent.
36
•
•
UK Habitat Types:
1
Broadleaved, mixed and yew woodland
2
Coniferous woodland
3
Boundary and linear features
4
Arable and horticulture
5
Improved grassland
6
Neutral grassland
7
Calcareous grassland
8
Acid grassland
9
Bracken
10
Dwarf shrub heath
11
Fen, marsh and swamp
12
Bogs
13
Standing open water and canals
14
Rivers and streams
15
Montane habitats
16
Inland rock
17
Built-up areas and gardens
Comments:
The structure of the definition of the Broad Habitats was in some cases a bit difficult to
use in straightforward manner for bar coding, because some BH definitions contain
information about other broad habitats which is not in their own definitions. The UK
Biodiversity Action Plan focuses on soils and indicator species as the main discriminators.
Therefore, distinct land cover information is in some cases not part of the habitat type
definition itself. In a few cases the wording of habitat definitions needed some
interpretation to match to EAGLE terms. Urban areas and land use is not within the scope
of UK BAP BH
3.2.13
•
Switzerland: Areal Statistik
Dataset details:
The Swiss Land Use Statistics lies in the responsibility of the Swiss Federal Statistical
Office. It captures information on land use and land cover in Switzerland based on aerial
photographs from the Federal Office of Topography at intervals of nine to twelve years.
In addition to statistical tables, the land use statistics provides also spatial data with 1 ha
MMU for Geographic Information Systems (GIS). In addition, it provides input into
national programs (space observation Swiss, Swiss biodiversity monitoring, hydrological
study areas) and indicators systems. It contains 46 categories of land use, and 27
categories of land cover. The Swiss Land Use Statistics is a point sample survey based on
aerial photographs with 4.1 million sampling points at intervals of 100x100m. The data
are available since 1990, new data re provided no later than two years after the aerial
photo survey.
Standard LCLU nomenclature:
http://www.bfs.admin.ch/bfs/portal/de/index/infothek/nomenklaturen/blank/blank/noas0
4/01.html
Land Cover nomenclature:
http://www.bfs.admin.ch/bfs/portal/de/index/infothek/nomenklaturen/blank/blank/noasl
c04/02.html
37
Land Use nomenclature:
http://www.bfs.admin.ch/bfs/portal/de/index/infothek/nomenklaturen/blank/blank/nolu0
4/01.html
Settlement and
urban areas
Building areas
Recreational areas and
cemeteries
Agricultural
areas
Orchard, vineyard and
horticulture areas
Intensive Orchards
Field fruit trees
Vineyards
Horticulture
Arable land
Meadow, farm pastures
Arable land
Meadows
Farm Pasture
Brush meadows and farm pasture
Alpine meadows
Favorable alpine pastures
Brush alpine pastures
Rocky alpine pastures
Sheep pastures
Alpine agricultural
areas
Wooded areas
Forest
Brush forest
Woods
Unproductive
one- and two-family houses
Surroundings of one- and two-family houses
Terraces houses
Surroundings of terraced houses
Blocks of flats
Surroundings of blocks of flats
Public buildings
Surroundings of public buildings
Agricultural buildings
Surroundings of agricultural buildings
Unspecified buildings
Surroundings of unspecified buildings
Public Parks
Sports facilities
Golf courses
Camping areas
Garden allotments
Cemeteries
Unproductive
Normal dense forest
Forest strips
Afforestations
Felling areas
Damaged forest areas
Open forest (on agricultural areas)
Open forest (on unproductive areas)
Brush forest
Groves, hedges
Clusters of trees (on agricultural areas)
Clusters of trees (on unproductive areas)
Scrub vegetation
38
areas
•
vegetation
Unproductive grass and shrubs
Avalanche and rock fall barriers
Wetlands
Alpine sports facilities
Comments:
In the matrix there is no element for “normal forest areas” in general that could be used
as general spatial context or as “excluded by definition” for classes in Switzerland that
explicitly exclude forest. (This issues has been taken into account for matrix
modification.)
Horticulture was uncertain to assign to a LUA under Agriculture (“commercial crop
production” or “production for own consumption”). Horticulture (German: Gartenbau) can
in theory be both, commercial and for subsistence. When it comes to mapping
operations, it has to be decided how to apply the attribute. For the moment it is assigned
to “commercial crop production”.
In some cases the assigning of ecosystem types (from the EUNIS classes) was uncertain.
39
4 LESSONS LEARNT AND OPEN ISSUES
After the introduction and explanation of the bar coding approach and the list of bar
coded classes during the semantic testing exercise, the following paragraphs assemble
the lessons learnt from the exercise. The following subchapters address some
methodological issues and questions that occurred during the application of the EAGLE
matrix bar coding.
4.1 EXPLICIT DEFINITION VERSUS LOGIC INTERPRETATION
An important issue for the activity of bar coding a class is the difference between
encoding the class by just following the explicit textual definition or also allowing logical
and broader interpretation of the definitions.
The question here is how strictly the textual definition is taken as the basis for the bar
coding, and if obvious or implicitly logical but not explicitly mentioned details of a class
definition are integrated in the bar coding or not. This can be seen differently by the data
producer/editor and the user. In this task however, contributors were asked to stick a
close as possible to the strict and explicit textual definitions, otherwise a later
comparability between bar coding results cannot be guaranteed.
The semantic testing exercise led to three specific findings regarding applicability of the
approach:
a) The structure and format of class definitions and the criteria they are built upon
differ between classification systems. It can happen that a detailed textual
description of a class is not available. In such cases the orientation for the
selection of matrix elements is very vague or must come from another source like
visual illustrations. For such cases the border between strict definition and
interpretation of classes can hardly be drawn. Sometimes class definitions are not
given in textual form, but as specific codes (LCCS or LCML6); in such cases the
coded format of the class definition is already close to conceptual perspective of
the EAGLE matrix. All this can have an influence on how classes are bar coded.
b) The more specific and precise the semantic details are outlined in a class
definition, the easier it is to express them distinctly with bar coded matrix
elements.
Differences in the hierarchical structure of two nomenclatures (e.g. specific
grouping of classes on a higher level) do not necessarily imply a problem for
semantic transformation between the systems, as long as the source and targeted
classes are designed in sufficient semantic detail on both sides. Therefore it is
always best to go as far as possible into hierarchical detail in barcoding the
classes of a nomenclature, to be able to extract comparable LC/LU information.
c) The bar coding result and expression of thematic content can only be as good
as the textual definition. Information content that is not given by class definitions
of course cannot be replaced by any means (besides allowing interpretation of the
definition).
d) The thematic consistency of the nomenclatures has an influence on the
applicability of the matrix. Depending on how closely a nomenclature follows the
principle of separation of land cover and land use, the easier it is to assign the
matching matrix elements to each class. By consequence, this precondition helps
to translate the bar coded information from the source to another targeted
classification system through the EAGLE concept.
The semantic testing exercise of this task was decided to follow strictly the explicit
textual class definitions. During the exercise it turned out to be very important to have
such a basic encoding at first that reflects the plain class definitions without any logic
6
LCCS – FAO´s Land Cover Classification System; LCML – Land Cover Meta Language
40
interpretation of them. Only then a comparison can be made between class definitions
and identify commonalities and differences as a next step, after having the results of
both the strict and the logic interpretive bar coding.
4.2 MERGING OF BAR CODE VALUES
The bar code values attempt to indicate, somewhat categorically, features and concepts
which may be fuzzy in the class definitions and within the perceptions of those
interpreting these definitions. Uncertainties on behalf of the contributors have arisen in
the differentiation between the meanings of the bar code values which were defined as:
n/a = ‘Element is not applicable in class, logically does not make sense’, and
0 = ’Element is not relevant for class’,
as well as in the differentiation between
0 = ’Element is not relevant for class’, and
1 = ‘Element can be expected and is typical for the class but not mandatory’.
It has turned out that difference between these values is not always as clear as the other
bar code values.It is therefore to be discussed, whether these three bar code values
(n/a, 0, 1) should be kept as they are or should be re-thought in their meaning. The
decision has been made that for application in the matrix population tool (from Task 3.1),
eventually the codes ‘n/a’ and ‘0’ are not needed (they are expressed differently than
with coding) and are therefore not applied. Instead, these meanings can be solved with
tool-internal rules that allow/forbid certain combinations of model elements.
The question about how to make use of bar code value 0 and 1 is further discussed in the
next chapter.
4.3 RICHNESS OF CLASS DEFINITIONS
Based on the work and experience in this task, a final decision needs to be made, how to
handle “foreign” fractions or particularities of class definitions with the view on how to
use bar codes 0 and 1. The term “foreign” is here used for land cover / use types that
appear in association with the given class but would be assigned to another class in the
nomenclature, if they were big enough in area size (MMU). They may occur within the
spatial extent of a geometric unit of an assigned class. Sometimes they are explicitly
mentioned in the class definitions as examples of sub-units which are smaller than the
determined MMU and only therefore are included in the class. This scenario mainly exists
with nomenclatures that work with relatively large MMU where generalizations and
inclusions of smaller “foreign”-class elements is common practice (typically like in CLC).
For example, cemetery, sports ground, commercial centre with an area size smaller than
the MMU are meant to be included within the delineation of residential areas. These
examples help the interpreter, data editor, data user to better understand the content of
the class, but the intended thematic content of the class in principle does not necessarily
get a bigger or richer content through the mentioning of more examples.
As a consequence of sticking strictly to the explicit textual class definitions when bar
coding the classes, the bar code results may appear richer in thematic content. During a
comparison of the definitions of two classes that by their logical meaning address the
same kind of land cover/use, one being described poorly and the other being described
richer, the result may imply a difference in content of the classes which is not existent in
their application reality. The impact would be that explicitly mentioned examples for a
certain class suggest that this class has a larger content (numerous value-1 coded
optional elements), than another class that in principle addresses the same thematic
content, but the definition only contains the compulsory elements. Comparison and clear
distinction between the explicit bar coding results with the logic interpretation results are
only possible if classes are coded in both but separate manners.
41
An option could be to include all mentioned examples and inclusions of “foreign”-class
elements smaller than the MMU with bar code value 1. In that case, these elements could
be flagged as such by either attaching a matrix element that indicates them as being
smaller than the MMU of the applied classification system, or by inventing a separate bar
code value to indicate that they do not fall into the above-mentined “foreign” fraction
category, but are real ‘optional elements’. The user of the bar coding results then can
decide, if he wants to work with the entire coding, or just the matrix elements that are
foreseen above the MMU. It is clear that this idea is difficult to implement and has to be
carefully reflected before integrating it as a rule for bar coding method, because many
features can occur with a broad range on sizes depending on the "basic" spatial scale of
the landscape.
The other direction of taking into account only the compulsory elements for comparison
and generally leaving out all optional (value 1-coded) elements seems not feasible,
because some classes would loose typical parts. In consequence, when comparing class
definitions, this would result in a higher degree of similarity between classes and
nomenclature.
This question becomes obsolete when the EAGLE concept is applied to distinct mapping
procedures, as all inclusions then can be described for and attached to each individual
spatial unit and don’t have to be categorically abstracted for an entire class.
4.4 MATRIX RE-CODING FOR PRACTICAL CLASS INTERPRETATION
As already addressed in the report above (sections 2.1 and 4.1), the selected classes
have been bar coded according to their explicit textual definitions, and other implicit
meanings of class definitions that do not appear in the text and may still be of
importance for the practical assigning of the class to land units, have not been included
in the bar coding exercise. Therefore, an additional exercise might be necessary at a later
step to also investigate the effect of the inclusion of implicit class meanings. In that case,
the exercise shall not only be based on the strict textual definitions but also on the
pragmatic use of the classes to classify landscape. The focus would lie on the CLC
classes.
4.5 ELIGIBLE ELEMENTS & BORDERS OF BAR CODE FUNCTION WITHIN THE
MATRIX
The bar codes, which are used to assign the matrix elements (represented as matrix
columns) for each codified class (represented as lines in the matrix) not only express the
presence or absence of matrix elements in the class definitions, but also the bar codes
express a relationship between the assigned elements (inclusive AND- / exclusive ORrelations) (see section 2.2.1).
In this regard it needs to be determined, what is the extent of the OR/AND bar code
function along a matrix line, in other words how far reaches the effect of the relation
between bar codes within a matrix line. When assigning the bar codes to the elements,
the structural difference between the three matrix blocks needs to be taken into account,
as explained in the following paragraphs. (To recall: Figure 1shows the overall structure
of the matrix with the three blocks of Land Cover Components (LCC), Land Use Attributes
(LUA), and Characteristics (CH).)
The LCC and LUA blocks are hierarchically structured, every hierarchical level has
thematic content, it means it represents on the lowest level a single LCC (single column
cells), or on upper levels superior LCCs (matrix segments) that unify elements from
lower levels.
42
The LCCs aim at being mutually exclusive and exhaustive, trying to cover all kinds of
land cover. The LUA block also tries to offer a land use type for any possible use, but
does not claim to be completely exhaustive; categories like “other uses” in theory help to
do so. From top to bottom the LCCs and LUAs start with a generic overall category and
are further subdivided into more distinct subtypes (see Figure 6 and Figure 7). All LCCs
or LUAs on every hierarchical level can be assigned to a bar code by unifying the
neighboring matrix cells (within the codified line) that reach the same broadness as the
segment in the blocks. It means that every hierarchical level of LCC or LUA can be bar
coded as a relevant matrix element or segment.
Figure 6: Matrix extract: Land Cover Component (LCC) block, Vegetation segment with subtypes
(sub-elements)
The borders of the function of bar codes include the entire LCC matrix block, the relation
of bar coded LCC is therefore related to any other coded LCC within the matrix block.
Every LCC (element or segment) on any hierarchical level is eligible to be assigned to a
bar code.
43
Figure 7: Matrix extract: Land Use Attribute (LUA) block, Industrial Use (Secondary Sector)
segment with subtypes (sub-elements)
Also for the LUA block it is the same setting like for the LCC matrix block, all bar code
values are related to all other land uses and their effect reaches throughout the entire
LUA matrix block. Every LUA (element or segment) on any hierarchical level is eligible to
be assigned to a bar code. So far, no conflict to this approach has been identified.
The Characteristics (CH) as the third matrix block is structured differently from LCCs and
LUAs. It is also hierarchically structured, but not as an entire unit but segment-wise.Not
every marix “segment” within the CH block has thematic content but is a heading for the
subsequent elements below and therefore cannot be connected to a bar code. From top
to bottom the CH block starts with headings that are further subdivided along the
columns. Though in principle it is structured hierarchically, the levels are not comparable
throughout the CH block. In most cases, only the elements on the lowest level of CH
(single column cells) are eligible to be bar-coded. All other matrix segments above are
just headings to group the characters by their themes (see Figure 8). In the extracted
example of the matrix block it would not make sense to unify matrix cells with the
broadness of the segment “cultivation patterns” in the class line, and give value 2 for
“must be under cultivation practice”, because if such information is available / written in
the class definition, it would already covered by the LUA “agricultural use”. Only when
more detailed specifications are known additional to the general agricultural land use
(bar coded under the LUA block), then a certain type of cultivation practice (e.g. crop
rotation) would be eligible to be bar coded with say value 2.
44
Figure 8: Matrix extract: CH block, Agricultural Management segment, with sub-headings and
character elements on lowest level
Regarding the borders of the relational effect of bar code value within the CH matrix
block the setting is different from the LCC and LUA block. Unlike the LCCs, the CHs are
not mutually exclusive, but can exists beside and independent from each. In
consequence, the borders of the extent of bar code functions are not the entire CH
matrix block but many CH segments have their own borders, within which the bar code
values unfold their relational effect to neighboring characteristics. Bar codes which are
assigned respectively in neighboring CH matrix segments mostly have no relation to each
other.
For example, three out of the cultivation practices (e.g. plantation, agroforestry,
intercropping) are coded with value 2 (=one of those three must be present). Further,
under the cultivation measures, also some elements (e.g. irrigation, fertilization, liming)
are bar coded with value 2. Unlike in the LCC or the LUA matrix block, these selective
mandatory OR-relation cannot be established between all CH elements, but only in a
separate way: the OR-relation within the cultivation practices segment PLUS the ORrelations within the cultivation measures segment. This situation also applies for the
other CH segments like spatial patterns, status, ecosystem types, etc.
The boundaries to determine the range of relation of bar code values inside the CH block
are to be revised for matrix population tool. Besides, inside the functionality of the tool –
together with the EAGLE database - it will be technically clearer which CH belongs to LCC
and LUA.
4.6 MACHINE READABILITY OF GROUPED MATRIX CELLS
As stated also in the Matrix Bar Coding Manual, the meaning of bar code value 2
(mandatory selective elements7) and value 3 (mandatory cumulative elements8) can in
general be used on different hierarchical scales: a) to describe relations of single matrix
columns (single land cover components, single land use attributes or single
characteristics/parameters) among each other, or b) to express the relations between
groups or segments of LCCs, LUAs or CHs.
7
one of the value 2-coded matrix elements is compulsory in the class definition
ALL of the value 3-coded matrix elements are compulsory in the class definition,
combinations with selective mandatory elements are possible
8
45
For example, a forest area has to have trees, so the LCC “trees” is bar coded with value
2 or 3. For another class, vegetation can be a mandatory part of the definition, no matter
what kind of vegetation, besides other LCCs like artificial areas or water areas. In this
case, the entire Vegetation LCC segment in the matrix is grouped and given the value 3.
At this point it is not yet completely solved whether it is possible to automatically
distinguish between singular coded matrix elements and grouped matrix elements and to
extract this information automatically from the Excel spreadsheet with a macro. For the
time being, the grouping of matrix cells will help the visual comparison between bar code
classes. Further development will take place under task 3.1 (development of matrix
population tool).
Once the EAGLE Matrix Population and Comparison Tool is available, the bar coding of
matrix cells will be supported with a user interface that optimizes the matrix population
procedure. Also, the storage of information about the relations (mandatory, optional)
between matrix elements or matrix segments on different hierarchical levels (and their
combinations) within the matrix will be supported.
4.7 FURTHER MATRIX/MODEL FINE TUNING
The semantic testing task had two aims:
a) to test the barcoding method itself with its bar coding values and their
applicability, and
b) to test the content of the matrix elements.
These two actions went on in parallel.
As for part b), some proposals to modify and fine-tune the EAGLE matrix content have
been noted. They are not entirely addressed in the report of this task, only a selection of
them has been listed under section 3.1. However, they are internally well documented in
the comment resolution table and will be tackled in detail together with others by the
time open issues are dealt within in task 5.3.
If modifications are implemented in the matrix/model, it will also be noted accordingly in
the explanatory documentation (outcome of task 1.1). Besides, some minor editorial
corrections will also be made regarding the mirror content between matrix/model and
documentation. It will be tackled in the maintenance task (T 5.2).
4.8 INPUT FOR TASK 3.1 (MATRIX POPULATION AND COMPARISON TOOL)
In this exercise of semantic testing, the matrix in its current form of an excel sheet was
used to describe land cover and land use classes by decomposing them into their
separate components. From its basic form in the beginning of development, the EAGLE
matrix (and the model) has reached a more holistic and complex form for a broader
range of applications. It was designed for the semantic analyses of single classes but also
for semantic comparison between class definitions. With the matrix, the visual
comparison between class definitions by their bar codes is possible by listing the classes
underneath each other in the matrix, and compare them by using the filter function of
Excel. However, this method does not give any formalized information n on the semantic
closeness of classes.
To ease the handling of the matrix content and foster a semi-automated approach, an
EAGLE matrix population and comparison tool (EMPACT, Task 3.1) is under development
that will help to populate the matrix elements with the bar coding method. Like that, the
classes can be described easier and semantic comparison between class definitions will
be possible.
46
The feedback collected during the semantic testing, especially the issues regarding the
meaning and relation of bar code values, will be integrated as input in the development
of this tool.
5 CONCLUSIONS
Based on this testing exercise, the content of the EAGLE matrix is well suitable to
describe the content of LC/LU nomenclatures’ class definitions. With the help of the
contributors of this task, a number of issues were identified that need to be addressed
and solved to further optimize the applicability of the EAGLE matrix and the concept
behind.
From the semantic testing exercise of this task, some conclusions for the future
development can be drawn. To give this a structured frame, a SWOT analysis related to
the future operational implementation of the EAGLE matrix has been collated.
SWOT analysis
Aim of SWOT analysis: Operational usage of EAGLE matrix & concept for the purpose of
semantic transformations of LC/LU information between classification systems.
Strengths
Weaknesses
+ idea of concept is convincing and
consistent
- bi-dimensional structure of matrix is
constraint for complex relationships
between concepts
+ aim of covering the theme LC and LU
completely(at least for a European
scenario) is completed
+ multi-line approach can solve the coding
of complex classes with complex relations
between matrix elements
+ INSPIRE conformant (data model)
+ results based on consensus and
voluntary collaboration between different
national institutions responsible for land
monitoring
+ matrix in the form of an excel sheet does
not require complex software nor expert
knowledge for use
+ matrix population tool in combination
with the data base can help to codify real
data
- unified matrix cells not machine readable
but only single filled out cells
- complex structure and extent of matrix
content not easy to overview
- versioning of matrix is difficult in Excel.
After a version is modified and columns are
added or removed, it is difficult to compare
already exported bar code sequences with
newer sequences after the matrix
modification.
- open issues of whether the functions of
bar coding are fully clear, understood and
needed as they are currently designed
- matrix content in theory is never
complete to explain any possible situation
on earth
+ through merging of matrix cells it is
possible to bar code matrix
elements/segments on every hierarchical
level
47
Opportunities
Threats
+ after operational implementation as a
basis for LC/LU mapping activities the data
will be better comparable
- concept is not yet established as a
standard, has to compete with already
existing standards like LCML, LCCS
+ implementation of EAGLE concept by EEA
(e.g. in CLC2018 production) would be a
signal for MS to also go into same direction
- pressure to deliver tangible results in
time for the preparation of CLC2018.
However, CLC2018 could be business as
usual.
+ standardization and establishment of
EAGLE in the direction of ISO concepts,
INSPIRE, CEN could be also addressed with
more focus + from the current matrix
status, two matrix versions can be
developed: one basic version with
rudimentary content for application on
simple generic nomenclatures; one fully
extended version which is open for further
user requirements;
- matrix / model gets more and more
complicated and difficult to maintain /
manage / understand if taking into account
the cross relations with other standards
and compatibility with them in high detail
this includes the option to have thematic
modules of the matrix according to
thematic fields of work (like agriculture,
forestry, habitats, urban areas)
+ matrix population tool will solve the
machine readable coding of merged matrix
cells / segments representing higher
hierarchical levels and its transfer into
database
+ first result of pilot beta version of matrix
population tool shows promising approach,
first developed functionalities are designed
according the ideas and suggestions of
contributors, recent brainstorming session
for further elaboration go in right direction:
user friendly and easy to understand
interface, excellent web-based
performance. further functions like
load/save/export functions for bar code
results, user registry are under
development.
+once the matrix population tool is
elaborated enough to be operational, it can
take the place of the matrix
The overall conclusion can be summed up as follows:
The two main questions connected to this semantic testing exercise were
1.) Does the EAGLE matrix bar coding method work? Is the concept logic and
understandable for users?
Based on the test results, the bar coding method seems logical in principle. The main
important bar code values that have mandatory character, namely value 2 (=
48
exclusive OR-relation, selective mandatory elements), 3 (= inclusive AND-relation,
cumulative mandatory elements) and 4 (alternatively combined mandatory elements,
EITHER-AND-OR-relation) are clear how to apply.
The multi-line approach - compared to the initial single line idea – is a step forward to
feasibility of the matrix bar coding method itself. It can solve many of the former
uncertainties in the relations between bar coded elements.
In some special cases there can be more than one solution during decomposing
classes when using the values 0 (element is not relevant for class), 1 (element can be
expected but has a minor role). The meaning of those bar codes will be reflected on
once more. The Bar Coding Manual will be enhanced accordingly in contract 5.
2.) Is the thematic content of EAGLE matrix applicable for semantic class
decomposition?
The thematic content of the EAGLE matrix is already very comprehensive and is
capable of describing the majority of tested classes. In some detail, the content could
be modified to further optimize its applicability. Details on this can be found in section
3.1.
The semantic precision is different for each nomenclature, depending on the richness /
semantic detail in the definition of the classes. To assess the discrepancies between the
strict orientation on explicit textual definitions and logical interpretation of class
meanings it is necessary to repeat this exercise for some classes (preferably CLC classes)
at a later step and to compare it with the explicit result.
The concept is well developed, but the application needs to be a bit more clarified. The
application towards working with physical databases was not part of this task, but it is
within reach after accomplishing this semantic testing exercise, after the tasks 3.1. and
4.1 will be accomplished.
An open issue is how effectively the matrix population tool can be connected to the
EAGLE data base for data load and export functions. This connection has two strings:
-
The conceptual string goes from the EAGLE matrix to the UML data model to the
physical data base.
-
For the practical string, the matrix tool should reproduce/deal/manage/read the
registry tables of code lists inside the EAGLE database. These registry tables store
the bar coded LCC, LUA, CH that represent the real data.
Regarding the cross connection with Task 4.1 (Geometric test case) and Task 4.2 (Grid
approach), the bar coding results as delivered in the Annex 2 (Semantic testing results of
bar coding in excel format) can be uploaded into the matrix population (Task 3.1), which
is still under development. Through that tool, the bar coding results and semantic
decomposition could then be transferred to the EAGLE data base, where physical data will
be stored.
49
ABBREVIATIONS AND ACRONYMS
BAP BH
-
Biodiversity Action Plan Broad Habitats (in the UK)
BBG
-
Bestand Bodemgebruik (Dutch national Land Use Database)
CH
-
Characteristics
EAGLE
-
EIONET Action Group on Land Monitoring in Europe
EEA
-
European Environment Agency
EIONET
-
Environment Information and Observation Network
EUNIS
-
European Nature Information System
Eurostat
-
European Statistical Office
FAO
-
United Nations (UN) Food and Agricultural Organization
FUA
-
Functional Urban Areas
IACS
-
Integrated Administration and Control System
INSPIRE
-
Infrastructure for Spatial Information in the European Community
JRC
-
LBM-DE
-
Joint Research Centre
Digitales Landbedeckungsmodell Deutschland (Land Cover Model
Germany)
Landbedeckungsmodell für Deutschland (renamed DLM-DE)
LC
-
Land Cover
LCC
-
Land Cover Component
LCCS
-
FAO´s Land Cover Classification System
LCML
-
DLM-DE
-
LISA
-
Land Cover Meta Language
Landelijk Grondgebruiksbestand Nederland (Dutch national Land Cover
Database)
Land Information System Austria
LU
-
Land Use
LUA
-
Land Use Attribute
LUCAS
-
Land Use / Cover Area statistical frame Survey
LUZ
-
Large Urban Zones
MMU
-
Minimum Mapping Unit
SIOSE
-
Sistema de ocupacion del suelo de Espana (Spanish Land Use System)
LGN
-
50