Republic of Ireland Corine 2006 Import

What is this all about

The Corine Dataset intended for import

It is intended to import Corine Land Cover data from the 2006 Corine data set into OSM in the area of the Republic of Ireland. This data set provides much richer land cover data than can be manually surveyed, and forms a good basis for more detailed local area mapping.

The licence under which Corine data is published requires attribution only and has been verified to be OSM-compatible in other countries in which Corine Imports have taken place or are planned. These countries include France, Romania and Poland.

About land cover

The Corine Land Cover project (CLC for short), provides, free of charge and with no licensing restrictions data about land cover in ESRI shapefile format for most of Europe. The data set consists essentially of tagged polygons. Polygons are split into about 45 categories, not all of which are relevant to Ireland.

Data processing

Corine data are made available as ESRI shapefiles, one all-Europe file per land cover type. These have been cut down to the Ireland bounding box and merged into a single shapefile, then converted to .osm format using ogr2osm. The output of this conversion contains a large number of duplicate nodes (about 6,000) and a smaller number of overlapping ways. These have been corrected using JOSM and an import candidate file prepared.

The mapping of Corine layers to OSM tags was performed using the translation facility of ogr2osm, and is largely consistent with the Proposed Corine Tagging Scheme, but with some layers given provisional tagging not part of that proposal. A full tag mapping is visible at http://maps.openstreetmap.ie/~mackerski/corine/corine.py

The import will be performed using the Upload.py suite of scripts. These allow optimal ordering of the uploaded data so that, for instance, nodes are quickly used to make ways rather than left unused until the tail end of the import process. Since the import is slow (nearly 3 days on the test server), the risk is too high that the nodes unused for such a long time would be "cleaned up" by well-meaning mappers or bots.

Accuracy

The land cover data accuracy is, in subjective terms, of higher quality than can be traced from Landsat imagery but much less detailed than can be managed using high-resolution imagery or ground survey. This means that no fixed assumption can be made as to how best to deal with cases where an imported land cover polygon overlaps one already in OSM.

Land Cover already in OSM

Currently in OSM, land cover is very sparsely mapped, meaning that the Corine import will represent a large improvement in the Irish map. The current proposal is to perform the import without attempting to avoid overlap with existing elements - such elements are fairly localised and can, over time, be de-duplicated manually. The table below summarises the main types of elements currently in OSM that will become overlapped with Corine data, along with comments as to whether Corine or existing data are likely to be better for both alignment and detail:

Feature type	Alignment	Detail	Comments	How to fix overlap
Residential areas	OSM better	OSM better	OSM data likely to be from ground survey or High res imagery. Also tends to be broken down by estate	Cut holes in Corine polygons to accommodate existing (usually smaller) OSM data
Forests, woods	Depends	Depends	Anything from landsat is worse than Corine. Anything from imagery is more detailed and accurate, but will lack tree type.	Review aerial imagery and decide which cover data is most accurate. Keep tree type information from Corine if missing in OSM.
Lakes, Riverbanks	Corine usually better	Corine often better	Many lakes and rivers already in OSM are traced from uncalibrated landsat and therefore badly offset.	Review aerial imagery and decide which version more accurate. Where there is a big offset, consider that Corine is likely better positioned. In such cases you might keep a more detailed traced lake but move it to the Corine location.

Corine Land Cover classes

1-Artificial surfaces

	CLC code	CLC description	OSM tags	Count	Notes
	1.1	Urban fabric
	1.1.1	Continuous urban fabric	landuse=residential residential=continuous	34
	1.1.2	Discontinuous urban fabric	landuse=residential residential=discontinuous	759
	1.2	Industrial, commercial and transport units
	1.2.1	Industrial or commercial units	landuse=industrial;retail FIXME=Corine Import. Retag to just industrial or retail or split between the two	125
	1.2.2	Road and rail networks and associated land	landuse=transport FIXME=Corine Import. This is a road or rail corridor. Consider whether this area is useful for anything	30
	1.2.3	Port areas	landuse=harbour	14	Can be replaced by leisure=marina after survey
	1.2.4	Airports	aeroway=aerodrome FIXME=Corine Import. Please arbitrate between this and any existing airport boundary	13
	1.3	Mine, dump and construction sites
	1.3.1	Mineral extraction sites	landuse=quarry	189
	1.3.2	Dump sites	landuse=landfill	12
	1.3.3	Construction sites	landuse=construction	45
	1.4	Artificial, non-agricultural vegetated areas
	1.4.1	Green urban areas	landuse=green_urban FIXME=Corine Green Urban Area. Please retag as something specific	40	Skipped in France, imported as "park" in Romania
	1.4.2	Sport and leisure facilities	landuse=sport_misc FIXME=Corine Sport and Leisure Facilities. Please retag as something specific	308

2-Agricultural areas

	CLC code	CLC description	OSM tags	Count	Notes
	2.1	Arable land
	2.1.1	Non-irrigated arable land	landuse=farm irrigated=no note=Corine Non-irrigated arable land	4062
	2.1.2	Permanently irrigated land	landuse=farm irrigated=yes note=Corine Permanently Irrigated arable land	0
	2.1.3	Rice fields	landuse=farm crop=rice note=Corine Rice Fields	0
	2.2	Permanent crops
	2.2.1	Vineyards	landuse=vineyard	0
	2.2.2	Fruit trees and berry plantations	landuse=orchard note=Corine Fruit trees and berry plantations	0
	2.2.3	Olive groves	landuse=orchard trees=olive_tree note=Corine Olive Grove	0
	2.3	Pastures
	2.3.1	Pastures	landuse=farm farm=pasture note=Corine Pastures	2104	Tagging improvements under discussion in talk-ie mailing-list
	2.4	Heterogeneous agricultural areas
	2.4.1	Annual crops associated with permanent crops	landuse=farm note=Corine Annual crops associated with permanent crops	0	Skipped in France
	2.4.2	Complex cultivation patterns	landuse=farm note=Corine Complex cultivation patterns	1381	Skipped in France
	2.4.3	Land principally occupied by agriculture, with significant areas of natural vegetation	landuse=farm farm=natural_mix FIXME=Corine Land principally occupied by agriculture, with significant areas of natural vegetation	2413	Skipped in France
	2.4.4	Agro-forestry areas	landuse=forest forest=agro note=Corine Agro-forestry areas	0	Skipped in France

3-Forest and seminatural areas

	CLC code	CLC description	OSM tags	Count	Notes
	3.1	Forests
	3.1.1	Broad-leaved forest	landuse=forest wood=deciduous note=Corine Broad-leaved forest	461
	3.1.2	Coniferous forest	landuse=forest wood=coniferous note=Corine Coniferous forest	2286
	3.1.3	Mixed forest	landuse=forest wood=mixed note=Corine Mixed forest	293
	3.2	Scrub and/or herbaceous vegetation associations
	3.2.1	Natural grasslands	natural=grassland note=Corine Natural grasslands	456
	3.2.2	Moors and heathland	natural=heath note=Corine Moors and heathland	144
	3.2.3	Sclerophyllous vegetation	natural=scrub note=Corine Sclerophyllous vegetation	0
	3.2.4	Transitional woodland-shrub	natural=wood wood=mixed note=Corine Transitional woodland-shrub	4202
	3.3	Open spaces with little or no vegetation
	3.3.1	Beaches, dunes, sands	natural=beach note=Corine Beaches, dunes, sands	101
	3.3.2	Bare rocks	natural=rock note=Corine Bare rocks	34
	3.3.3	Sparsely vegetated areas	natural=scrub note=Corine Sparsely vegetated areas	132
	3.3.4	Burnt areas	natural=burnt FIXME=Corine Burnt areas	0	Skipped in France, "rock" used in Romania
	3.3.5	Glaciers and perpetual snow	natural=glacier note=Corine Glaciers and perpetual snow	0

4-Wetlands

	CLC code	CLC description	OSM tags	Count	Notes
	4.1	Inland wetlands
	4.1.1	Inland marshes	natural=wetland wetland=marsh note=Corine Inland Marshes	222
	4.1.2	Peat bogs	natural=wetland wetland=bog note=Corine Peat bogs	2023	Corine does not distinguish between blanket and raised bogs, nor between exploited and natural bogs. Other tags have been discussed for these distinctions, but since they cannot be applied during import there is no opportunity to save work. It should be noted that peat bog cover is included in Corine in some unlikely places, so prepare for some cleanup. Tagging improvements under discussion in talk-ie mailing-list
	4.2	Maritime wetlands
	4.2.1	Salt marshes	natural=wetland wetland=saltmarsh note=Corine Salt Marshes	56
	4.2.2	Salines	landuse=salt_pond note=Corine Salines	0
	4.2.3	Intertidal flats	tidal=yes note=Corine Intertidal flats	222	(See Proposed_features/Water_cover)

5-Water bodies

	CLC code	CLC description	OSM tags	Count	Notes
	5.1	Inland waters
	5.1.1	Water courses	waterway=riverbank note=Corine Water courses	28
	5.1.2	Water bodies	natural=water note=Corine Water Bodies	539
	5.2	Marine waters
	5.2.1	Coastal lagoons	natural=water note=Corine Coastal Lagoons	10
	5.2.2	Estuaries	natural=water_estuary note=Corine Estuaries	36
	5.2.3	Sea and ocean	natural=water_sea note=Corine Sea and ocean	N/A	Not Imported

Northern Ireland and the Border

An annoying feature of the import is that Corine data for the UK are not yet processed and available for download. This means that this import will be of land cover only in the Republic of Ireland, making it likely that a future import will have to be careful of overlap along the border. Our import data set seems (with one exception, see below) to follow the border pretty accurately, so it is hoped that the UK data set, when published, will fit into place rather nicely. A side effect of the missing data is that we can adjust the (currently very rough) border in OSM to match the Corine boundary - keep in mind, though, that those border road crossings for which we have ground surveys are more accurate than Corine and these should be considered definitive.

Missing data in Monaghan-almost-enclave

The Border follows a very unusual route in an area south-west of Clones, giving rise to an almost-enclave of County Monaghan sticking into Fermanagh. In this area, the road from Clones to Belturbet crosses the border many times, alternating number between N54 and A3. This almost-enclave is missing from the Corine data set, and will, presumably, be provided at the same time as the UK data set. For this reason, it is important that we not adjust the border in OSM at this location.

State of Play

To end-to-end test the import process and to allow for users to play at editing a map containing Corine data, a test instance of the OpenStreetMap site (complete with slippy map, API and Potlatch) has been set up:

http://apidev2.openstreetmap.ie

A test import of the data has been performed on this instance. The slippy map displayed helps to visualise what the rendered map will look like, though it should be noted that the map tiles it is serving are actually from another box, so any edits you make on this test instance will not be reflected on the map.

Anybody wishing to make test edits on apidev2 may do so, but will need to register first with the usual OSM signup process, but carried out on apidev2. It is not possible to sign up using a username already in use on the real OSM (not even your own), so pick a new ID for testing purposes. Also, although the server is secured, it is not recommended that you choose an important password that you use elsewhere.

Test Editing

An important thing to understand before doing the real import is how everyday mapping will be affected by the addition of so much new data to the map. Many parts of Ireland are very sparsely mapped, and it is certainly easier to map in empty countryside than in areas with lots of nodes and ways. New or occasional mappers could become intimidated by the need to work around the land cover data. In addition, the imported data polygons are implemented as multipolygons rather than simple closed ways. Although there are good reasons for this, many mappers are not experienced with multipolygons and should consider how they will maintain such data, how to update it or correct it based on their local knowledge.

So while test editing, try to consider answers to the following questions - and if you have concerns, discuss them on the mailing list before the import so the community can help address them:

• Are you happy with the quality of what we intend to import?

• Are you happy that you can continue mapping as effectively with the extra data on the map?

• Do you know how to use your normal OSM editor to edit the imported multipolygons? (splitting, changing shape, merging etc.)

• If you have previously mapped land cover - forests, farmland, industrial areas, resdidential areas - this import will most likely create duplicate data overlapping yours. This will require a process of de-duplication, selecting the best of what was done by hand and what has been imported. Do you know how to go about doing this? If not, others can help, but please identify the areas where this is required.

To edit the test data using Potlatch 1 or 2, simply log into the test server and use Potlatch as normal. To use JOSM, Merkaartor or another editor that can be configured to talk to a particular API URL, set the API URL to: http://apidev2.openstreetmap.ie/api or http://apidev.openstreetmap.ie/api

Map Previews

There are other zoomable slippy maps showing:

• OSM with Corine Preview (optional hill-shading).
  • This should give a close represenation of what the main OSM map will look like after the CORINE import.
• Corine Layer prepared for import (optional hill-shading).
  • Shows only the CORINE data we plan to import using OSM map styles.
• Original EEA Corine Shapefiles for Ireland
  • Shows CORINE landcover using EEA supplied shapefiles rendered with the default CORINE colour scheme.

Rogue Polygon

A setback to the preparation occurred due to poor Mapnik performance. This was traced to a single pasture land cover area covering most of Ireland. Removing this fixed performance problems with PostGIS import and with the Mapnik rendering. We have now split this large area into 1552 more reasonably-sized pieces.