Merging statistics and geospatial information, 2016 projects - the Netherlands

Paving of public and private domains in the Netherlands; analyses of spatial and temporal patterns during the last decades; 2016 project; final report December 2018

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

Full article

Problem

Paved surfaces (for example, land covered by concrete, bricks or buildings) prevent infiltration of precipitation, leading to increased flood frequency and intensity. The urban heat effect is directly related to the extent and concentration of paved areas. Analysing temporal and spatial patterns of paving can contribute to ecosystem accounts.

Terrestrial ecosystems are the focus of several policy-related initiatives, such as the sustainable development goals (SDGs). In the Netherlands, the urban heat effect and urban flooding have both been recognised as potential threats for human health and the economy. No known data are available to systematically study where, when, and to what purpose land areas have been paved in the Netherlands during the past few decades.

Objectives

This study aims to map the development of paved surfaces in the public and private domains in the Netherlands over the past few decades for a number of time periods.

The project should lead to results that can be directly used in ecosystem accounts as well as contributing to policy-related discussions. It provides an excellent example of the integration of geographical and statistical information.

Method

The study mapped the development of paved surfaces in public and private domains in the Netherlands over the past few decades. The temporal and spatial patterns have been analysed.

The municipality of Eindhoven was used as a case study to show how the paving in the residential domain – such as the gardens of dwellings – or yards of industrial areas have evolved. A method combining (available) topographical data and a classification of green unpaved areas using aerial photography was used. This was done for 2013 and 2016. The surface area, location, and direction of change was calculated. An estimation of the accuracy was made at several locations in the study area.

There are some land use/cover data from topographical sources which can, with high certainty, be classified as either paved or unpaved.

Data on buildings are extracted from a register – Basic Registration Addresses and Buildings (Basisregistratie Adressen en Gebouwen; BAG)) – which contains the polygon shapes of each building, including private homes and even sheds. The construction dates attributed to each building can be used to identify buildings in a particular year. Errors due to demolition or replacement are estimated to be negligible.
Data on roads are taken from the Top10NL dataset, where there are both polygon and polyline datasets available for each year. Lane width (and therefore overall road width) was estimated.
A polyline railroad dataset is available. The number of tracks was used to estimate an overall width.
For some agricultural fields, vector data are available from the General Register Agricultural Lots (Basisregistratie Percelen; BRP).
Water bodies are also found in the Top10NL dataset. A polygon dataset includes lakes, ponds and wide canals, while a polyline dataset includes the small canals for which widths were estimated.

Surface areas which cannot be accurately classified using topographical data were classified from aerial photographs using remote sensing techniques. Aerial photographs are available with visible and near-infrared (NIR) bands for Eindhoven in 2013 and 2016. These data were used to explore the possibilities and limitations of estimating the paved surface using a near difference vegetation indicator (NDVI). The absorption and reflection patterns in the red and near-infrared spectrum are very specific for vegetation surfaces. The NDVI is a method of calculating the ratio between the red and the near-infrared reflection of the spectrum.

The topographic data can be assumed to be very certain. The unpaved class of the aerial photography also has a relatively high certainty. A lack of vegetation does not mean a paved surface (for example, bare agricultural fields), yet this is assumed when classifying areas based on the NDVI of the aerial photography and so the classification of these areas is relatively uncertain. Areas close to buildings (within 5 metres) also have a high degree of classification uncertainty. The result of this is a single map (for each year) with the different relative degrees of uncertainty.

A new classification of each cell was calculated to identify the direction of change for each cell combining the binary classification of paved and unpaved surfaces for each of the two years.

A second study was done for the municipality of Weert. This used a similar methodology to that for Eindhoven, but with a somewhat more detailed classification (buildings, roads, (other) paved, water and (other) unpaved surfaces.

Results

Results for the case study of Eindhoven, comparing 2013 and 2016.

Approximately 40 % of the area (in both years) was classified using topographical sources. As a share of the total area (paved, unpaved and unclassified), the paved area increased 0.2 % while the unpaved area decreased 1.3 % and the unclassified area increased 1.1 %.
Using aerial photography, it was possible to classify all areas. As a share of the total area (paved and unpaved), the paved area decreased 9.4 % while the unpaved area increased 9.5 %.
Combining both (with preference for topographical data), it was possible to classify all areas. As a share of the total area (paved and unpaved), the paved area decreased 4.4 % while the unpaved area increased 4.5 %.
2.7 % of the total area changed from paved to unpaved, while 7.1 % changed from unpaved to paved. 41.0 % was paved in both years, while 49.2 % was unpaved in both years.

Figure 1: Change in paved and unpaved surface areas in Eindhoven using combined classification methods between 2013 and 2016

For 2016, an analysis of the certainty of classification within Eindhoven shows that 39.4 % of the area was classified with high certainty and 43.9 % with some certainty. By contrast, 1.4 % was classified with some uncertainty and 15.4 % with high uncertainty. The uncertain areas were mainly areas classified as paved using aerial photography or areas close to buildings. These were often areas of particular interest, between private buildings.

Figure 2: Relative certainty of the pavement classification of Eindhoven in 2016

Policy-related analyses

The population and dwelling stock of Eindhoven increased between 2013 and 2016 and so the increase in the paved area can, in part, be related to increased population density.
The problems related to heat stress and flooding due to heavy rainfall may partly be associated with the increased paving of gardens. While the map for Eindhoven clearly shows an increase in paving between some buildings, this is often in areas where the certainty of the classification is low; the apparent increase in paved areas may be due to significant shadows or other issues related to the photography.

Figure 3: Land use map municipality of Weert in 2017