wiki:UserGuidelines/GIS/Data

Version 238 (modified by Fran Boon, 10 years ago) ( diff )

--

User Guidelines for GIS Data

Table of Contents

  1. Data Sources
  2. Import Data
    1. UUIDs
    2. CSV
    3. Shapefiles
    4. GDB
    5. KML
    6. Geonames
    7. OpenStreetMap
    8. WFS
    9. Yahoo
  3. Display Data
    1. GeoServer
      1. Import Shapefiles
      2. Configure GeoServer
      3. Configure GeoWebCache
      4. WFS
      5. Add WMS Layer to Sahana Eden
    2. WMS Reprojection
    3. Grid
    4. Administrative Areas
      1. GADM
    5. Population Density
      1. GPWv3
      2. GRUMPv1
      3. US Census
    6. Topographic Maps
      1. WMS
      2. OpenStreetMap
    7. Old Printed Maps
  4. OpenStreetMap
  5. PostgreSQL management
    1. PostGIS functions

Assumes installation of the relevant tools

Data Sources

See GIS/Data

Import Data

UUIDs

See: UUID

CSV

Locations can be imported using the normal import procedure:

The CSV needs to have specific columns:

  • WKT column if we have polygon info (or Lat and Lon for Points, if not)
  • For L1, we need these columns: Country, L1 (& WKT)
  • For L2, we need these columns: L1, L2 (& WKT) [Country can also be used to help separate duplicates]
  • For L3, we need these columns: L2, L3 (& WKT) [L1 and Country can also be used to help separate duplicates]
  • For L4, we need these columns: L3, L4 (& WKT) [L2, L1 and Country can also be used to help separate duplicates]
  • For specific locations, we need these columns: Lx (for appropriate parent level of hierarchy) Name (& Lat/Lon) [L2, L1 and Country can also be used to help separate duplicates]
  • Key/Value columns are used, if-present
    • e.g.: "L1 KV:GADM" adds a record in gis_location_tag for every L1 record with a key of "GADM" & a value of whatever is specified in the cell.
  • Population & Elevation columns are read, if-present
  • Ensure that names are consistent between Levels
  • The PROPER() spreadsheet function is useful to get the names in the correct format (then Paste as Text).
  • The VLOOKUP() spreadsheet function is useful if the different levels of the hierarchy are in different sheets & linked via a code instead of the name (as we need):
    • If the lookup sheet is called 'lookup' and the lookup table is in B2-C87 then lookup code in C2 to column 2 of the lookup table: =VLOOKUP(C2; lookup.B$2:C$87; 2; 0)
  • To remove duplicate rows, can create a new column with =IF(A1=A2;1;0), Paste Special, Sort & Delete 1s

Basic Hierarchy can often be found from Wikipedia (although currently there's no easy way to download this - a student project to enhance Wikipedia for this would be much appreciated! e.g. Using ​WikiData).

For the Polygon data, it is normal to get this from Shapefiles, such as ​GADM or ​UN CODS.

Shapefiles

Inspect the data using ​qGIS.

Use ​ogr2ogr to convert the data to CSV: 

ogr2ogr -select ISO,NAME_1,NAME_2 -f CSV CSV USA_adm2.shp -lco GEOMETRY=AS_WKT
ogr2ogr -f CSV CSV TM_WORLD_BORDERS-0.3.shp -lco GEOMETRY=AS_WKT
ogr2ogr -f geojson TM_WORLD_BORDERS-0.3.json TM_WORLD_BORDERS-0.3.shp

If needing to reproject (e.g. for the ​Haiti Departements): 

ogr2ogr -f CSV haiti_departments Haiti_departementes_edited_01132010.shp -s_srs EPSG:32618 -t_srs EPSG:4326 -lco GEOMETRY=AS_WKT

NB AS_WKT requires OGR v1.6+

If the data is Admin Boundaries then it can be imported into the gis_location table via ​http://host.domain/eden/gis/location/import

Otherwise the data can be imported into the gis_theme_data table via ​http://host.domain/eden/gis/theme_data/import

  • you will first need to define the layer and activate in the config(s) of your choice

An alternative way to deal with Shapefiles is to upload to GeoServer & serve as WMS/WFS from there...

Dissolving Polygons (e.g. recreate L0 from L1s): 

ogr2ogr output.shp input.shp -dialect sqlite -sql "SELECT ST_Union(geometry), dissolve_field FROM input GROUP BY dissolve_field"

GDB

ESRI's File Geodatabase format.

Can use this online service:

or use GDAL:

NB It isn't included in GDAL by default. Windows users can use ​osgeo4w to install the gdal-filegdb driver.

KML

Can convert a KML to CSV using the attached script: python KML2WKT.py <filename>.kml

This can then be imported into Sahana by editing the column headers & using the Importer

qGIS can be used to convert this into a Shapefile (uses ogr2ogr so can also do using the CLI, if you prefer): give it column headers with 'WKT' for the WKT column name.

  • This is the easiest way to load into PostGIS (using PGAdmin III's Shapefile Importer plugin) to allow GeoServer to serve as WMS

An alternate approach is to use this XSL:

Geonames

There is an import_geonames() function in S3GIS which downloads/unzips the country file (a TAB-separated list) from ​http://download.geonames.org/export/dump/

It should be run for the different levels of hierarchy that you wish to import (generally just the lowest level as Geonames just has Point data, so it's best to use other sources for the Polygons 1st, that way the Geonames importer can locate these Points within the correct Polygons of the hierarchy)

NB It takes some time to do this import! Pakistan imports 95000 locations!

Update: Geonames schema 2.2 supports parentADM(1-4): ​http://geonames.wordpress.com/2010/09/29/geonames-ontology-2-2/

  • will be good for when we only have hierarchy, not polygons
  • need to check whether much data has this populated though.

Python 2.5 doesn't support Zipfile.extract() & Zipfile.read() isn't unicode-safe. Until this is fixed, download the file manually 1st: 

cd ~web2py/applications/eden/cache
wget http://download.geonames.org/export/dump/PK.zip
unzip PK.zip

In Web2py CLI: 

gis.import_geonames('PK', 'L5')
db.commit()

Alternate approach:

  1. Transform each line in this file into XML by regular expression: 
    ^(\d*)\t([^\t]*)\t([^\t]*)\t([^\t]*)\t([0-9\.]*)\t([0-9\.]*)\t[^\t]*\t([A-Z]*).*

into:

<location>
       <id>$1</id>
       <name>$2</name>
       <asciiName>$3</asciiName>
       <localNames>$4</localNames>
       <lat>$5</lat>
       <lon>$6</lon>
       <featureClass>$7</featureClass>
</location>

This can be done using an RE-capable editor (e.g. Kate), Perl or even Python. Note: Need to replace & with &amp; and to remove any invalid characters

  1. Transform into S3XRC-XML using XSLT, stylesheet is available at

OpenStreetMap

See GIS/OpenStreetMap#Import

WFS

It is possible to use the WFS Plugin to get data into qGIS & thence export into other formats.

May need to use a Custom CRS (in Settings menu - remember to Save!) such as:

Can then go to the Layer Properties & Specify CRS to this User Defined Coordinate System.

Can then Save As and change the CRS to something like the standard WGS84.

Yahoo

Display Data

GeoServer

GeoServer can provide geospatial data in Raster (WMS) or Vector (WFS/KML) formats.

Once you have installed in Linux or Windows, then login:

  • l: admin
  • p: geoserver

PostGIS is recommended as the main data store.

  • Allow Read access to GeoServer: 
    su postgres
psql
CREATE USER geoserver WITH PASSWORD 'geoserver';
\c sahana
GRANT CONNECT ON DATABASE sahana to geoserver;
GRANT SELECT ON gis_location to geoserver;
GRANT SELECT ON geometry_columns to geoserver;
GRANT SELECT ON spatial_ref_sys to geoserver;
GRANT SELECT ON stats_demographic to geoserver;
GRANT SELECT ON stats_demographic_data to geoserver;
GRANT SELECT ON stats_demographic_aggregate to geoserver;

Configure:

Import Shapefiles

e.g. Country Outlines:

These can be loaded direct into GeoServer, however there will be better performance by importing into PostGIS:
(can also use pgAdmin III GUI's Shapefile loader on plugins menu) 

su postgres
shp2pgsql -s 4326 -I TM_WORLD_BORDERS-0.3.shp public.countries | psql -d gis

To reproject the data into 900913 for a slight performance advantage: 

drop constraint srid;
update table set geomcolumn=transform(geomcolumn,900913);

Configure GeoServer

Colours:

Zoom Levels (in Spherical Mercator):

Zoom Level(s) MinScale MaxScale
1 250000000 n/a
2 100000000 250000000
3 50000000 100000000
4 25000000 50000000
5 10000000 25000000
6 5000000 10000000
7 2500000 5000000
8 2000000 2500000
9 1000000 2000000
10 500000 1000000
11 250000 500000
12 100000 250000
13 50000 100000
14-22 n/a 50000

Configure GeoWebCache

The raw WMS server will be slow, so once you've chosen your style, then you should serve via GWC. This caches pre-rendered tiles & also does MetaTiling so that the WMS has less separate requests (at a cost of increased RAM requirements)

The version embedded within GeoServer is great for providing a zeroconfig of the common options, however there are cases where you need to define a layer manually:

  • Want a Background Colour (bgcolor)
  • Want to specify an alternate style (& you'd rather not republish the layer on the WMS)
  • Want to render a set of layers into a single tileset (so that clients don't need to download them separately & merge locally)

Tips for optimal usage:

Example geowebcache.xml in the GADM section:

WFS

If you are displaying a complex dataset at zoomed-out resolutions, then you will want to have simplified views.

e.g. Hospitals aren't shown at all at low zooms, are shown as Points at medium zooms & shown as Polygons at high zooms.

Scale-dependent styling using SLD in GeoServer:

Scale-dependent styling in OpenLayers:

Simplifying Polygons in PostGIS:

Showing the different layers at different zooms using GeoServer:

Add WMS Layer to Sahana Eden

  • tbc

WMS Reprojection

  • Have a remote WMS source that you want to access?
  • Have a desire to keep OpenStreetMap/Google/Bing layers?
  • WMS source server doesn't support the 900913 projection?

e.g. TRMM Rainfall Monitoring

Solution: MapProxy

Grid

We have a 'Coordinate Grid' Layer available by default.

Other options:

Administrative Areas

GADM

GADM is the best source of global Administrative Boundaries:

There are often better local sources for specific countries, although getting hold of these can be difficult. Note that some countries have boundaries which change frequently and so datasets can often be a little out of date.

To import into Sahana Eden's gis_location table (for consistency of naming/boundaries across basemap & dynamic data):

  • Install latest Python GDAL bindings
  • Open a web2py CLI: 
    python web2py.py -S eden -M
  • Optionally, define a filter for which countries you wish to import data, e.g. for Asia-Pacific (without TL, as that will be imported from UN CODS): 
    countries = [ "AF", "AU", "BD", "BN", "CK", "CN", "FJ", "FM", "HK", "ID", "IN", "JP", "KH", "KI", "KP", "KR", "LA", "MH", "MM", "MN", "MV", "MY", "NP", "NZ", "PG", "PH", "PK", "PW", "SB", "SG", "SL", "TH", "TO", "TV", "TW", "VN", "VU", "WS"]
  • Import: 
    gis.import_admin_areas(countries=countries)

This can then be served as separate WMS layers using GeoServer & GeoWebCache.
You can use GeoServer's ​SQLView feature.
SLD files are attached:

  • L0 Base
    • SQL View: SELECT id, name, area, the_geom FROM gis_location WHERE level='L0'
  • L0 Overlay
    • SQL View: SELECT id, name, area, the_geom FROM gis_location WHERE level='L0'
  • L1 Overlay
    • SQL View: SELECT id, name, area, the_geom FROM gis_location WHERE level='L1'
  • L2 Overlay
    • SQL View: SELECT id, name, area, the_geom FROM gis_location WHERE level='L2'

Example geowebcache.xml for GADM attached which provides 3 layers:

  • L0 Base
  • L0-L2 Base (merged)
  • L0-L2 Overlay (merged) 
    cp geowebcache.xml /var/gis/geoserver_data/gwc
/etc/init.d/tomcat6 restart

Population Density

GPWv3

Gridded Population of the World, version 3 (GPWv3) is the standard global dataset for both measured & projected population densities.

Download gl_gpwfe_pdens_10_wrk_25.zip from (requires registration): 

mkdir /home/data/GPWv3
cd /home/data/GPWv3
unzip gl_gpwfe_pdens_10_wrk_25.zip
/usr/local/bin/gdal_translate -of GTiff glfedens10/glds10ag/hdr.adf glds10ag.tif
ln -s /home/data/GPWv3 /var/gis/geoserver_data/coverages/GPWv3

Add new GeoTIFF Store to GeoServer to serve as WMS: 

URL: file:coverages/GPWv3/glds10ag.tif

Style:

GRUMPv1

Global Rural-Urban Mapping Project, version 1 (GRUMPv1) is a newer dataset which combines satellite with census data to locate people within settlements rather than just administrative areas:

US Census

For the US, data is available down to block level:

​http://factfinder2.census.gov/faces/nav/jsf/pages/download_center.xhtml Dicennial Census 2010 SF1 100% Data

Convert data from NAD83 to WGS84 (e.g. using qGIS) 

su postgres
cd /data/Census2010BlockGroup
shp2pgsql -s 4326 -I tl_2010_06037_bg10_WGS84.shp public.Census2010BlockGroup | psql -d gis
psql
\c gis
ALTER TABLE census2010blockgroup ADD COLUMN population integer;
ALTER TABLE census2010blockgroup ADD COLUMN population_density integer;
\q
exit

w2p
%autoindent
pop_dict = {}
input = os.path.join("/", "home", "data", "Census2010BlockGroup", "DEC_10_SF1_P1_with_ann.csv")
inputFile = open(input, "r")
header = 2
for line in inputFile:
    if header:
        header -= 1
        continue
    parts = line.split(',', 6)
    geoid = parts[1]
    pop = int(parts[6].strip())
    pop_dict[geoid] = pop

inputFile.close()
from __future__ import division
db_string = "postgres://gis:GIS@localhost:5432/gis"
db2 = DAL(db_string, migrate_enabled = False)
table = db2.define_table("census2010blockgroup", Field("gid", "id"), Field("geoid10"), Field("aland10", "integer"), Field("population", "integer"), Field("population_density", "integer"), migrate=False)
rows = db2().select(table.geoid10, table.aland10)
for row in rows:
    data = {}
    geoid10 = row.geoid10
    population = pop_dict.get(geoid10)
    if population:
        data["population"] = population
        aland10 = row.aland10
        if aland10:
            area = aland10 / 2589988
            population_density = population / area
            data["population_density"] = population_density
        db2(table.geoid10==geoid10).update(**data)

db2.commit()

Can also have the Tract-level data for lower zoom (9-15) & create a Layer Group to serve the 2 layers with a style which shows only 1 level at each zoom

Topographic Maps

WMS

Topography can be rendered as WMS using e.g. GeoServer

Download the SRTMs in GeoTIFF format from ​ftp://xftp.jrc.it/pub/srtmV4/tiff

  • get.sh can assist with this
  • Beware this is a *lot* of data! (At least 60Gb for a global dataset)
    • The GeoTIFFs are already well ​compressed
    • Partial datasets are, of course, possible (@ToDo: A script to select an area & download just that area) 
mkdir /tmp/SRTMv4
cd /tmp/SRTMv4
sh get.sh

Unzip the data into a folder called 'SRTMv4' in the GeoServer 'coverages' folder (or use a symlink): 

cd /var/lib/tomcat6/webapps/geoserver/data/coverages
mkdir SRTMv4
unzip -o /tmp/SRTMv4/\*.zip
rm /var/lib/tomcat6/webapps/geoserver/data/coverages/SRTMv4/*.hdr
rm /var/lib/tomcat6/webapps/geoserver/data/coverages/SRTMv4/*.tfw
rm /var/lib/tomcat6/webapps/geoserver/data/coverages/SRTMv4/readme.txt

Give Tomcat permission to the folder: 

chown tomcat6 /var/lib/tomcat6/webapps/geoserver/data/coverages/SRTMv4

Configure GeoServer by adding a new Store using the ​Image mosaicking plugin

  • URL: file:coverages/SRTMv4

Publish (defaults OK)

Styling:

Test out using the direct WMS URL:

Once happy then start using the GeoCache URL (but don't do this too early as otherwise you have to invalidate the cache to see your changes):

OpenStreetMap

Contours can be rendered using OSM tools:

​http://de.wikipedia.org/wiki/Benutzer:Alexrk2/SRTM-Reliefs

  • suggests using GIMP's emboss filter! Azimuth = 135, height = 50, depth = 10

Old Printed Maps

Old Printed Maps can be 'Rectified' to be overlaid on the base maps:

OpenStreetMap

PostgreSQL management

PostGIS functions

  • Centroids 
    SELECT name, iso2, asText(ST_Transform(ST_Centroid(the_geom), 4326)) AS centroid FROM countries;

GIS

Attachments (12)

Download all attachments as: .zip

Note: See TracWiki for help on using the wiki.