Geodata Repository

The notes on the Talk:Geodata_Repository Talk page for this page describe the background to this effort

A full list of suggestions for public domain data sets that are nice-to-haves is maintained at Geodata Discovery Working Group.

= Getting involved =

The geodata repository has a dedicated blade server at telascience - cf SAC Service Status. Right now this is geodata.telascience.org ; we need to get an osgeo.org domain pointed to it. The plan is to have something very much up, running and demoable by FOSS4G.

How you can help

 * Offer feedback on the Geodata Metadata Requirements - if you have a dataset you would like to contribute, does this model express it adequately?
 * Write or contribute an example Simple Catalog Interface
 * If you would like to install software and/or data on the repository machine, please talk to the good people at SAC or visit #osgeo and #telascience on irc.freenode.net to find out how to get an account and get started

Who is involved now

 * Jo Walsh is working on the backend data store and machine level metadata interfaces (also presenting this material at FOSS4G 2006)
 * Chris Schmidt is helping with an OpenLayers-based browsing interface
 * Schuyler Erle is helping with web service / application glue
 * Sean Gillies' OWSLib is, half-beknownst to him, a big part of all these applications - see the OWSLib API user stories at http://trac.gispython.org/projects/PCL/wiki/OwsCapabilitiesUserStory
 * Markus Neteler is contributing data processing code and resources as part of the GRASS data packaging effort
 * Norm Vine is providing a point of contact for administering the system and general reality-checking to all involved.
 * Martin Spott maintains the content of the PostGIS-repository on 'zuluviz'.

= Interface Design =

In order to be useful to people, a geodata repository needs a way for users to become quickly + easily informed as to what kind of data they're getting: metadata about the data, and the quality/quantity of data involved.

For Raster images, one way to do this is to include an extent-wide image of the data -- http://openlayers.org/gallery/ has screenshots which show some of what I mean where, although more directed towards applications. By having a whole-extent screenshot overview, users can quickly and easily see what they're getting. Additionally, assuming this data is to be made available as WMS, setting up an OpenLayers instance to allow users to browse the rasters and see at a more detailed level would be beneficial.

Vector data/attribute data would also need to be described, in text or some other way. http://freemap.in/world/ is an example of a browsable map which displays attribute data -- clicking on a country fills the sidebar with data. This interface was set up in about 20 minutes, and given decent map files, this kind of set up could be largely automated, allowing users to (again) get an overview of the data they're looking at before they download it or set it up on their own servers.

= Data sources =

Sources of public geodata for initial setup in a repository.

PostGIS serving vector data
There's a PostGIS server (thanks to John Graham for hosting this site at TelaScience) which stores different flavours of vector data; also known as "Landcover-DB". Current datasets include: VMap0, VMap1, AptNav, GSHHS, PGS, SWBD, MGRS, TIGER, StatsCan, OSM, GeoNames, CountryCodes and FGSODB - see below for further explanation.

Access - how to get to it !?

 * Graphical MapServer interface
 * Download Shapefiles
 * WMS + WFS + TileCache services - see below

The web mapping colour schema has been applied in accordance with the Corine Land Cover project.


 * Direct read access to the database is available to users of the 'geodata.telascience.org' system. Retrieval in SQL syntax as "SELECT asText(wkb_geometry) FROM  WHERE  LIKE ''".
 * Collaborative editing via FeatureServer is being prepared.
 * Dump into Shapefiles provided occasionally or upon request; regular schedule possible if required.
 * Other data access upon request, depending on the purpose and use.

CAVEAT We regret that, due to technical reasons, direct PostGIS database access is currently unavailable.

On Offer !
Short explanation of available datasets (to be extended - the number of datasets as well as their explanation ;-):

Many thanks to Norman Vine for running the HCRA, the "Human Communications Relay Agent" :-)

How the data gets in
Many datasets are readable with the GDAL/OGR(/OGDI) toolbox. Notably these are the
 * VMap datasets,
 * GSHHS,
 * PGS,
 * SWBD and
 * MGRS.

The 'ogr2ogr' command is used here, hidden behind a somewhat complex contruct of shell scripts which automates the whole thing. Writing to the DB is accomplished by the PostGIS driver provided by 'ogr2ogr'.

Some other datasets are automatically or manually (for those sets which are not expected to change often) transformed into OGC-compilant SQL scripts and run through the respective SQL monitor.

Few datasets are being imported directly into the DB with Perl/DBI. AptNav, for example, is parsed from a text file with a home-grown parser in Perl and written to the database table with DBI.

Everything that is meant to represent a geometry is stored in the DB using OCG-compilant geometry types POLYGON, LINESTRING and POINT. A POINT, for example, is written into the DB as

INSERT INTO  ([...]) VALUES ([...] PointFromText('POINT( )', 4326) [...])

The data are stored internally in the respective, geospatially searchable geometry data type.

Phase 1

 * Import data from many different sources and shape it into a unified format - first results available (see above).
 * Retrieve exact locations of major river dams and waterfalls (keywords: St. Lawrence, Niagara, Bosporus, Gibraltar, ....).
 * Build joins between VMap0/1 urban areas and GeoNames populated places (via names and geographic vincinity).

Phase 2

 * Generation of static per-country and per-region shapefiles; distribution via HTTP and via geotorrent.org.

Phase 3

 * Design and implement a storage-/data-model for road data that is capable of serving the needs of OpenStreetMap while remaining conformant to OGC-standards. Merge the ideas explained in the OSM New Data Model paper as well as Schuler's OSM on PostGIS initiative - in other words: Try "squaring the circle" ;-)
 * Now that OpenStreetMap has (also) imported TIGER, convince OSM to imcorporate an equivalent of VMap0 at places that are currently not covered, finally to create a global road network of maximum detail and accuracy.
 * Merge landcover- and stream-layers from VMap1, TIGER/OSM and Landsat7-classification into the foundation of VMap0 to create a global landcover dataset of maximum detail and accuracy.

Sequence undefined

 * Add proper metadata according to OSGeo metadata recommendation/standard.
 * Find suitable desktop client to FeatureServer.
 * Add polygon data that's been automagically retrieved from Landsat7 images at the World Custom Scenery Project - procedure available.
 * Add polygon data that's been manually digitized from Landsat7 images.

What you can do

 * Help is very much appreciated to add a reasonable colour schema. Work is currently on its way to convert the map display over to the CORINE colour values but this schema doesn't cover all our needs.

DONE

 * Generation of shapefiles on-the-fly for user-defined region and layer(s). Use 'pgsql2shp' and consider the 2 GByte-per-file limit: Landcover Database Shapefile Download
 * Rebuild TIGER layers from 2006se release according to these instructions - mostly finished.
 * Limit the "viewing angle" of WMS/WFS-services in order to save the DB-servers' life .... Solved by setting MAXSCALE for some layers.

Old status

 * see Geodata processing at telascience.org

Blue Marble NG

 * Mapserver as WMS

Status

 * By bittorrent: http://data.freemap.in/view/DataSet/9

Status

 * SRTM30 data is available now by WMS/WCS: http://data.freemap.in/view/DataSet/11

Landsat-7

 * Mapserver as WMS
 * Telascience already serving the raw data - http://onearth.telascience.org/

Status

 * Waiting for disk space to finish unpacking the raw data on zuluviz.sdsu.edu
 * Interim plan is to write a simple WMS wrapper script that generates a GDAL VRT to assemble composites on the fly

Status

 * Create mechanism to have multi-resolution seamless "best of" imagery server Multi-resolution seamless imagery

WMS + WFS + TileCache services
Many layers (see list above) are served as WMS/WFS at http://mapserver.flightgear.org

WMS: http://mapserver.flightgear.org/ms?Service=WMS&Version=1.1.1&request=GetCapabilities WFS: http://mapserver.flightgear.org/ms?Service=WFS&Version=1.0.0&request=GetCapabilities TileCache: http://mapserver.flightgear.org/tc (EPSG:900913 !!)

TelaScience / OSGeo / FlightGear Landcover Database Shapefile Download
See http://mapserver.flightgear.org/download.psp

City of Vienna, Austria
OGD Wien offers data under CC-BY. This data can be imported to OSM. Supported interface standard include: Contact on IRC at #ogdwien.
 * WMS
 * WFS GML
 * JSON
 * KML geplant

= Metadata =


 * Geodata Metadata Requirements - a baseline for what would be asked for / expected of data sets in an OSGeo repository
 * A simple metadata-in-GeoRSS webservice that is also as FGDC compliant as possible/desirable.
 * Simple Catalog Interface

= See Also =


 * OSGeo's Geodata Discovery Working Group - half-dormant
 * http://grasswiki.osgeo.org/wiki/Global_datasets
 * MapDex offers a metadata search facility
 * GeoNames is a gazetteer and naming collection
 * http://www.fao.org/geonetwork/srv/en/main.search
 * http://geoportail-geoportal.ainc-inac.gc.ca/metaindex_e.asp
 * http://en.wikipedia.org/wiki/List_of_towns
 * http://mapas.mma.gov.br/i3geo/menutemas/servicoswms.php WMS servers collection in RSS format
 * http://mapas.mma.gov.br/i3geo/wscliente.htm Web service client