update regimes of the data sources varies greatly, as seen in
Figure 2, and ARSTAT is not suited for cartographic
presentation.
As mentioned, the amount of data and insufficient tools for
geometry processing made the production cumbersome and
tedious. Data had to be exported to shape-datasets in small
chunks in order to overcome time constraints and errors in
geometric operations, and the results imported back to the
database. The migration to PostGIS made it possible to do the
processing on ARSTAT stored as one table in the database
(likewise for the source datasets). Most operations could be
executed as single SQL-queries on the national datasets. Still,
some of the heavy geometric operations were implemented as
functions and run on chunks. However, the convenience of all
data resident in the same database and the ability to use SQL,
together with the improved speed and near error-free
performance of geometry-operations, was a major
improvement.
The land cover and capability of the objects is characterized by
seven attributes, inherited from AR5 and ARSO. There are
several thousand allowable combinations of these attributes,
which are used to compute the classes needed for the statistics.
The alignment of classes across the different classification
schemes (briefly described below) is developed in cooperation
with the domain experts and the end users.
3 ) : m dm HE San
Figure 2. The dataset ARSTAT is not homogeneous with
respect to MMU and classification.
4.2 Land resource classes
Acreage numbers for different land resources on municipality,
county and country level is calculated automatically.
The final results contain total area (land and fresh water, see
2.2.4), acreage figures for land resource classes based on ARS
(see 2.1.1) and ARSO (ie. ARSTAT, see 4.1). Further,
mountainous areas are dived into five sub-classes and forest is
distinguished based on productivity classes and tree species.
Land type classes from ARS: Fully cultivated land, Surface
cultivated land, Pasture, Productive forest, Unproductive forest,
Open peat bog, Open land with soil-cover, Open land with little
or no soils, Area that is not mapped (unknown classification)
Land type classes from ARS50: Fully cultivated land, Surface
cultivated land, Pasture, Forest, Open peat bog, Mountainous
arcas, Freshwater, Glacier and permanent snow, Built-up area
and Transport network area.
Forest classes from ARS50: Very high productivity, High
productivity, Medium productivity, Low productivity, Non-
productive and Unknown classification for Coniferous forest,
Mixed forest, Deciduous forest and Forested peat bog.
Mountainous areas classes from ARS0: Vigorous vegetation,
Moderately fresh vegetation, Sparse vegetation, Bare rock and
Unknown classification.
4.3 Calculating and saving acreage figures
The workflow is the same for computing numbers for 430
municipalities, 19 counties or for the whole country.
The calculation of the statistics is done automatically by
running a BAT-file which executes a PSQL-command
(PostgreSQL, 2012b) which for each statistical level starts an
SQL-file. The SQL-file contains SQL-queries to different tables
stored in PostgreSQL. The resulting acreage figures for the
different classes mentioned in 4.2, is written to an XML-file.
Computing time for one municipality is about 25 seconds, 65-
75 minutes for one county and about 23 hours for the whole
country.
4.4 Publishing the statistics
The resulting XML-files are available on Skog og landskap's
home page on the internet. In addition to the acreage figures,
some additional information about the statistics, the involved
datasets and the land types is included.
To make the XML-files more easy to read and nice to look at,
XSL T-files have been created. This ensures a nice layout of the
statistics in different web browsers. To save work, one CSS-file
(Cascading Style Sheets) is being created. CSS defines how
HTML elements are to be displayed. Styles saved in this CSS-
file enable us to change the appearance and layout of all the
pages (XML-files) in a web site, just by editing this single file
(WC3, 2012).
Statistics are produced every second year, and acreage figures
from 2006, 2008 and 2010 are available. People may look at,
print and/or download these land resource statistics sheets as
HTML-files. An example, land resource statistics for Norway:
http://kart2.skogoglandskap.no/xml filer/2010/Norge arstat 20
10.xml
4.5 Synergy effects
The dataset ARSTAT has a national coverage, but is inhomoge-
neous with respect to MMU, classification, and update frequen-
cy. In order to quantify changes of specific importance more
accurately Skog og landskap also conduct sample-based moni-
toring programs for different strata.
One of these is the 3Q-programme for monitoring of the agri-
cultural landscape. The 3Q-programme (Dramstad et al, 2003)
computes indicators for various landscape phenomena based on
intensive monitoring of 1000 sample sites (1x1 km squares) by
stereo interpretation of true colour aerial photos. Each square i
mapped repeatedly every fifth year to record changes. Four 1n-
dicator themes have been in focus: landscape spatial structuré,
biological diversity, cultural heritage and accessibility.
The specific m
combined with
erage area reso
possible to ana
Further, the pr
successful use
application wa:
ing several toa
exploit the GI
*translating" th
in complexity a
The GIS functi
cation (ISO, 2€
spatial informal
cators using rel
The statistics ai
detailed classifi
the topographi
images in mou
ARS0 is revisec
is accordingly p
dataset may the
two years after t
A copy of ARS
January, e.g. ye:
at the beginning
These changes v
produced Januar
The changes ii
municipalities a
the landscape a
one year. Althoi
efficient, there a
scale. In the fi
changes, the stat
Total acreage n
from the official
Map projection,
Lambert azimut
azimuthal origin
for their dataset
for all municipal
hand, are given i;
Àn additional ch.
quality version o
October 2011. W
Correct our data
this introduced a
OK. Norway ha
Canada). This o
exiting functions
The open softwar
*àSy to use and to
