Merging statistics and geospatial information, 2017 projects - Norway

GEOSTAT house and health project; 2017 project; final report December 2019

This article forms part of Eurostat’s statistical report on the Integration of statistical and geospatial information.

Full article

Problem

The project was in two parts, one for houses and one for health (and fire) services.

Light detection and ranging (LiDAR) data are available for many countries and their use for statistics should be investigated, for example for price statistics for houses.
Concerning the part on health (and fire) services, accessibility statistics (driving distance and time) were of interest for a range of services.

Objectives

The initial plan for the part on houses was to establish routines for handling three-dimensional (3D) data as an input into price models for houses. The amount of data and the number of algorithms needed were limiting factors. Furthermore, the data still lack coverage and quality checks need to be done. In practice, the project became mainly a literature study that focussed on the possible uses of the third dimension (3D) in public statistics, many of which are relevant for house price models.

Concerning the part on health, the aim was to produce new official statistics describing the average travel time and distance from a patient’s residence to their GP’s office and the nationwide coverage of a selection of emergency services.

Method

Houses

The project focused on LiDAR (light detection and ranging) data and its possible use for making official statistics related to population and housing. LiDAR data can be used to generate detailed digital elevation and terrain models (DEM and DTM), in other words 3D models: these can be used to describe the macro landscape as well as buildings and infrastructure.

Health (and fire) services

A lot of geodata are available within Statistics Norway, but crucial data had to be acquired from other sources and then structured and georeferenced.

The geographic position of maternity wards (in fact maternity wards and delivery rooms) was scraped from a website administered by the Norwegian Directorate for eHealth (helsenorge.no). The collected addresses were then joined to a georeferenced version of the cadastre. An alternative would have been to source the addresses directly from the Norwegian Directorate of Health; this should be used in future as the web-scraping exercise was rendered obsolete by changes to the website.
No source of information on which of the hospitals had emergency rooms was found. Data were gathered directly though e-mails sent to every health trust known – from their website – to have an emergency room. Once collected, these data were joined to a georeferenced version of the cadastre.
The Norwegian Directorate for Civil Protection (DCP) provided data on the geographic position of fire stations. These are now published on their own geoportal and the national geoportal (geonorge.no).
In sparsely populated municipalities, the location of emergency clinics (a 24-hour service to either treat or transfer patients in need of immediate medical attention) can change as general practitioners (GPs) take turns to provide the service. Because of the inherent difficulties of identifying the alternating locations, this service was excluded from the project.
Residential addresses are georeferenced points derived from the Norwegian population register. The data are available from Statistics Norway and 99.8 % of the population is georeferenced.
The national road network is documented in the national road database (NVDB) which belongs to the national road authority and is updated continuously. Statistics Norway downloads the dataset once a year and makes it available for internal use through its geodatabase.
Data on administrative regions (counties and municipalities) are produced annually by the National Mapping Authority and made available by Statistics Norway for internal use through its geodatabase.
Polygons were constructed representing hospital trust district and health regions by merging the polygons representing the municipalities included in the respective entities (one municipality was treated as an exception).
Polygons representing the fire department areas and the 110-central areas (dispatch areas) were attained from the DCP’s geoportal.
Polygons were constructed representing service areas (see below).

A common production method was applied for calculating driving time and distance between residential addresses and each service location; the calculation can be used either to or from a residential address.

The lowest cost path through the network can be calculated using ArcGIS origin-destination cost matrix function (a multiple-origin, multiple-destination algorithm). The information on the road network takes account of a variety of factors, such as road type, speed limit and length of the road segment. A cost factor was also applied to each road segment, with different factors for different road types. Approximately 0.1 % of addresses were excluded, for example because they were on islands without a mainland connection or because they were more than 5 km from the nearest road. In order to remove the complexity of multiple-origins and multiple-destinations, the calculation was performed in two stages.

Separately for each facility (maternity wards, emergency rooms and fire stations), service areas were calculated to identify the addresses closest (in time) to them. This was done in such a way that the service areas (for any particular type of facility) did not overlap.
Once the service areas had been established, the calculations (time and distance) were performed for residences within a single service area.

Figure 1: Isolating facilities and residential addresses for iterative calculations

Results were aggregated to six types of areas. Data were produced for the country as a whole, for counties and for municipalities for the three types of facilities. For maternity wards and emergency rooms, results were also calculated for (modified) service areas and for health regions. For fire stations, results were also published for 110-central areas (dispatch areas); data were also compiled for service areas and published as tables, but not as maps.

Results

Houses

The review of geographic analyses focused on an area of Kongsvinger (80 km north-east of Oslo). The area was 1.25 km (east–west) x 1.50 km² (north–south) and contained a mixture of land use and of land cover.

Slope and aspect: slope is the angle of inclination to the horizontal and aspect is the compass direction that a slope faces. Useful in the calculation of solar energy potential of surfaces for solar panels.
Hill shade: using solar path algorithms, the total potential number of sun-hours may be estimated for any point.
Roughness: the degree of irregularity of the surface; calculated by the largest inter-cell difference of a central pixel and its surrounding cell. Roughness plays a role in the analysis of terrain elevation data and is useful for calculations of river morphology, in climatology and physical geography in general.
Topographic position index: the difference between the elevation of a central pixel and the mean elevation of its surrounding cells.

Figure 2: DTM Slope Degrees

Regional statistics for land cover, land use and place typologies.

Ecological and land classification: provides physical and biological information about a given landscape to help with sustainable management. Could be used for statistics on sustainability and biological diversity.
Coastlines: map coastlines and identify their structure, helping coastline management. Combined with information on species, can help analyse the impact on the eco-system of coastline changes and use of coastal areas.
Watershed and water courses: watershed areas can be identified and water courses (streamlines) delineated.

Sun conditions / residential lighting: light/shade and temperature of buildings are impacted by many factors, such as the location, rotation and layout of buildings, the terrain, the height and distance from surrounding buildings. Sun conditions and other physical neighbourhood variables may be correlated with house prices.

Viewshed analysis: what can be seen from a particular point and what is blocked. A viewshed analysis could contribute to an attractivity index for property at a location.

Construction: information (location, volume, area and roof types) can be gathered about the stock, construction and removal (such as demolition) of buildings. This can be used for planning residential and non-residential construction work.

Forestry statistics: information about stocks, changes in stocks and quality. These are of relevance for forest management, forest fire management, biodiversity and carbon absorption.

Agriculture: information on field (soil) and crop conditions with implications for sowing, fertiliser use and yields.

Transport: data for use in planning new roads/tracks as well as accurately mapping existing networks.

Floods and landslides: data on the nature and position of riverbanks can be used to predict the possibility of flooding as well as mitigation plans. More generally, advanced topographical information can help to create effective flood relief simulation plans. Information can be used to determine the nature and strength of landslides and also to mitigate their impact. These data could be used to assess where people and property are at risk of floods or landslides.

Pollution: through differential absorption LiDAR (DIAL), information such as the composition of atmospheric gases and other atmospheric constituents can be calculated precisely. Together with a building or terrain model, these data can be used to observe pollutant build-up in a given area and develop mitigation plans and when combined with population data can assess health risks.

Solar energy: information for planning solar panel installation, concerning aspect and minimum area.

Urban utility planning: electric grid mapping, sagging of electric wires and terrain surveillance.