andskap.no
]uires continuous deliv-
vegian Forest and Land-
orway. The statistics are
ination of different na-
tistics are calculated for
and glacier, mountain-
en source software. The
-files that are published
handle geometric opera-
ribution within the three
best soil quality are de-
r classes (Tenge, 2011).
ed ARSTAT, which is a
tion that is relevant and
e, but not intended for
CES
national land capability
set that describes land
ity for agriculture and
the productive part of
or above the tree line is
; the best information
ow the tree line suitable
'he dataset classifies the
te class, tree species and
104 classes based . on
inimum mapping unit I$
in outfield. The accuracy
1 2 meters.
completed in 1995. The
ipdate was completed in
iodically revised using
)-50 cm) and maintained
celo, 2006).
yographic 1:50 000 map
ined by the Norwegian
kartverk, 20122). N50
Jasses, transport, place
names, buildings, borders and restrictive areas. However, the
land cover classes do not distinguish between different kind of
forest or arable land. This dataset is suitable in scales from 1:25
000 to 1:100 000. The accuracy is +2-50 metres depending on
the feature type.
Some classes in N50 are updated once or twice a year, while
others are updated more seldom depending on access
to/available source data.
122 ARmountain: The before mentioned CLC project
speeded up the mapping of the mountain areas in Norway which
are about 43 % of Norway's mainland, see Figure 1. In these
sparsely populated arcas with low-intensity land use, only
topographic maps (e.g. N50) provided by NMA with a coarse
division of land cover classes were originally available.
Skog og landskap started some years ago with manual
interpretation of satellite images covering these areas. But this
approach soon turned out to be too time-consuming. However,
the development of the capacity of computers, the improvement
of the relevant software and the CLC2000-project, gave Skog
og landskap the opportunity to make *ARmountain map in 1:50
000" semi-automatically and in this way obtain a more
homogenous and detailed dataset above the tree line.
ARmountain (Norwegian: AR-FJELL) is a dataset that covers
mountains and unproductive areas above the tree line and
includes a wide spectre of resource-types. It contains land cover
ranging from unproductive boulder fields to high-productive
meadows. Five classes are produced: abiotic, sparse vegetation,
lichen, moderately fresh vegetation and vigorous vegetation.
ARmountain was produced using a semi-automatic method de-
veloped at Skog og landskap. The mapping was carried out us-
ing à two-step segmentation and classification of satellite imag-
es (using SPOT4, Landsat 7 and 5 scenes from 1999-2006)
followed by a manual correction of the results. The input da-
tasets were the satellite mountain area mask generated from the
N50 dataset and a snow and cloud mask.
The resulting ARmountain dataset is not published as a separate
product, but is used to complete other land resource databases,
ensuring that these products have complete national coverage
(Gjertsen, Angeloff, Strand, 2011).
Figure 1.
Typical mountain area in Norway. Photographer:
Michael Angeloff, Skog og landskap.
2.2.3 — ARS0: To offer a homogeneous dataset covering the
whole mainland of Norway (below and above the tree line),
Skog og landskap produces a land cover map in 1:50 000 called
AR50. This map is created through a generalization of ARS
below the tree line and ARmountain (interpretation of satellite
images) above the tree line. In addition, N50 is used where
neither ARS nor ARmountain is available (Skog og landskap,
2011). ARSO is suitable for cartographic presentation in scale
1:20 000 to 1:100 000 and is used on a regional and
municipality level.
2.2.4 Area figures for land and freshwater: The official
acreage figures for each municipality in Norway are provided
by NMA. These numbers are downloaded from the internet
(Statens kartverk, 2012b) as an XLS-file and converted and
stored as a table in PostgreSQL (PostgreSQL, 2012a). The
administrative borders and numbers have been unchanged since
2008 (no municipalities have been merged together). In 2012
the numbers will change slightly due to the new coastline
(Statens kartverk, 2012c).
3. SOFTWARE
The majority of the software used by Skog og landskap has
been proprietary; the data are stored in Oracle relational data-
base with ESRI's ArcSDE extension. ArcGIS is used for the
production of paper maps, carrying out all kind of analyses and
processing data. However, the software has proven to fail when
processing large vector datasets (and Skog og landskap are to a
large extend dealing with national datasets) and many of the op-
erations are surprisingly and disappointingly very time-
consuming. In addition, you have the cost of purchase and sup-
port of the software.
The use of open source software is expanding, and since 2005,
Skog og landskap has gradually been moving from proprietary
and licensed software to open source software solutions for a lot
of tasks. Datasets are stored in PostgreSQL relational database
with PostGIS extension (PostGIS, 2012) and GeoTools for
different GIS tasks. The internet application for viewing and
downloading maps is entirely based on open source software
like GeoServer, MapServer, OpenLayers and GeoExt.
The production of land resource statistics is based on
PostgreSQL and PostGIS.
To automate the production, the calculation is run by the use of
batch files (Wikipedia, 2012a). Even though batch-files may
sound a little bit “old-fashioned”, it is still a very easy, simple
and convenient way of automating processes when running
Windows. Every Windows system is capable of creating and
using batch files.
4. STATISTICS
4.1 Producing ARSTAT
ARSTAT is stored in PostgreSQL and created using the
geometric functionality of the PostGIS extension. The final
database contains 14 million polygon objects, with a total of
one billion points.
ARSTAT is made of data from different sources to produce the
"best possible” land resource statistics. The minimum mapping
unit, classification systems, mapping methods, accuracy and
