European Forum for Geography and Statistics Vienna Conference

European Forum for Geography and Statistics Vienna Conference

Vienna, 10 – 12 November 2015

Name (s) of author(s)

(1) Raffaella Chiocchini – (1) Stefano Mugnoli – (2) Luca Congedo – (2) Michele Munafò

Organization

(1) ISTAT (Italian National Institute of Statistics)

(2) ISPRA (Italian National Institute for Environmental Protection and Research)

IT Geostat Population Grid 2011

(1) R. Chiocchini, (1) S. Mugnoli, (2) L. Congedo, (2) M. Munafò (1) National Institute of Statistics (ISTAT), Rome, Italy

(2) Italian National Institute for Environmental Protection and Research (ISPRA), Rome, Italy {rachiocc, mugnoli} @istat.it

[email protected] [email protected]

Abstract

ISTAT presentation has the aim to describe the activities that have brought at the new Italian 1K Population Grid. The final result represents just the last step of a very worthwhile collaboration between ISTAT (Italian National Institute of Statistics) and ISPRA (Italian National Institute for Environmental Protection and Research).

The template used was the Italian portion of new GEOSTAT Grid 2011. Many different geographic layers were used. The most important were: - ISTAT 2011 Census cartography and data; - Copernicus Degree of Imperviousness HR Layer at 20m of resolution; - ISTAT Statistics Synthetic Map, and Regional Land/Cover Map to mask the uninhabited areas on the

Imperviousness HR Layer; In brief: the Copernicus Degree of Imperviousness is a raster layer that represents the percentage of soil

sealing per pixel (i.e. cell), which was assumed to be related to population distribution; but the Degree of Imperviousness layer does not distinguish between residential and non-residential areas, therefore a detailed masking work has been first implemented in order to exclude non-residential pixels using ancillary data. Population data collected for each enumeration area during Census 2011 survey, has been proportionally distributed basing on the sum of Degree of Imperviousness, obtaining the number of resident population per pixel.

Then, the resident population for each GEOSTAT cell has been calculated as the sum of the single pixel values of the raster layer that are included in that cell.

Many GIS software tools were used but the main algorithm, used to calculate the population for each 1Kmq cell was ‘Zonal statistics as table’ of the ARCGIS 10.1 ESRI© software.

This algorithm calculates the entire pool of spatial statistics of a reference layer on the base of a ‘value’ raster layer. The output is a table in which there is a single statistic value for each input polygon.

Used SW: ARCGIS 10.1 ESRI© for desktop (ArcInfo license), ERDAS IMAGINE 2013 INTERGRAPH©.

Keywords: ISTAT, ISPRA, Enumeration areas, Census project, ARCGIS 10.1, ERDAS IMAGINE, Imperviousness HR Layer, Copernicus, Population grid

Introduction

This paper has the purpose to describe briefly the principal activities carried out by ISTAT and ISPRA with the aim to realize the 1 km2 population grid for the entire Italian territory.

The final product represents very worthwhile collaboration between two Italian National Research Institutes that permitted a successful synthesis among many geographic datasets, in particular two of them:

- Copernicus Degree of Imperviousness HR Layer at 20m of resolution; - ISTAT 2011 Census cartography and data;

This new grid can be surely considered a big improvement compared to the one that can be produced by a simple ‘disaggregation techniques’. In fact starting from the bond between ISTAT geographic datasets and ISPRA Imperviousness Layer, we could estimate with a really good approximation which are the residential zones. Moreover, the resolution of the Copernicus satellite data ensures very precise estimations.

Brief description of the main layers used

The European Environmental Agency (EEA), in the frame of Copernicus initiative, has developed several High Resolution Layers (HRLs), referred to the year 2012. These HRLs have the main purpose of monitoring land cover of European countries with a high level of detail (20m resolution); in particular, the following environmental issues are represented in a land cover map: Degree of Imperviousness, Forest, Grassland, Wetland, and Water Bodies.

In this work, the HRL Degree of Imperviousness was used as main layer for the definition of built-up areas. In fact, the Degree of Imperviousness describes the percentage of soil sealing inside the pixel area (i.e. 400m2), which is related to the irreversible process of the degradation and removal of the soil surface (Munafò & Tombolini, 2014).

The HRL Degree of Imperviousness is the result of a classification methodology, developed by EEA, based on the multispectral classification and object-oriented classification of remote sensing images and the calculation of vegetation indices (e.g. NDVI) and biophysical parameters for improving the

identification of vegetated and non-vegetated areas; also, ancillary data (i.e. land use/cover maps) were used if available, and interactive editing was performed on classification results if needed (EEA, 2012). It is worth mentioning that the HRL Degree of Imperviousness is a land cover raster of the built-up area (without distinction of the land use as shown in Fig.1) that includes the following elements as described by EEA (2012):

• Housing areas • Traffic areas (airports, harbours, railway yards, parking lots) • Industrial, commercial areas, factories • Amusement parks (excluding the pure green areas associated with them) • Construction sites with discernible evolving built-up structures • Single (farm) houses (where possible to identify) • Other sealed surfaces that are part of fuzzy categories, such as e.g. allotment gardens, cemeteries,

sport areas (visible infrastructure), camp sites (roads and infrastructure, possibly influenced by caravans), excluding green areas associated with them.

• Roads and railways associated to other impervious surfaces • Water edges with paved borders

Fig. 1: Example of impervious surfaces (in red) from the Degree of Imperviousness 2012

ISTAT Census Enumeration Areas vector layer represents the base to analyze the Italian territory as regards statistical data. In fact, all the data collected during Census surveys are linked to each of the over

400.000 enumeration areas ‘drawn’ upon Italy. This dense plot helps us to describe Italian Territory in a very detailed way overall in urban areas. It is also important to remember that Italy is planning continuous population Census, that should start in about two years; these activities will be absolutely fundamental to have a very detailed reference cartography and statistical data. So, the Census geographical datasets are essentially used for classifying and characterizing Italian Territory in relation to resident population, buildings, services and industry, and this fact can be considered precisely the starting point to realize a reliable population distribution grid. Furthermore it is very important to remember that several attributes are linked to each enumeration areas; this data were very useful overall during ‘masking activities’. In fig.2 an example of ISTAT enumeration areas (in red); in yellow the administrative boundaries.

Fig.2 – Example of ISTAT enumeration areas layer

Project PROJCS['WGS_1984_UTM_Zone_32N',

GEOGCS['GCS_WGS_1984',

DATUM['D_WGS_1984',

SPHEROID['WGS_1984',6378137.0,298.257223563]],

PRIMEM['Greenwich',0.0],

UNIT['Degree',0.0174532925199433]],

PROJECTION['Transverse_Mercator'],

PARAMETER['False_Easting',500000.0],

PARAMETER['False_Northing',0.0],

PARAMETER['Central_Meridian',9.0],

PARAMETER['Scale_Factor',0.9996],

PARAMETER['Latitude_Of_Origin',0.0],

UNIT['Meter',1.0]]

ETRS_1989_To_WGS_1984

PROJCS['ETRS_1989_LAEA',

GEOGCS['GCS_ETRS_1989',

DATUM['D_ETRS_1989',

SPHEROID['GRS_1980',6378137.0,298.257222101]],

PRIMEM['Greenwich',0.0],

UNIT['Degree',0.0174532925199433]],

PROJECTION['Lambert_Azimuthal_Equal_Area'],

PARAMETER['False_Easting',4321000.0],

PARAMETER['False_Northing',3210000.0],

PARAMETER['Central_Meridian',10.0],

PARAMETER['Latitude_Of_Origin',52.0],

UNIT['Meter',1.0]]

Methodology in brief

European grid layer re-projection ISTAT enumeration areas layer and ISPRA imperviousness

layer are absolutely congruent regarding their geographic reference system. In fact these were both realized in WGS84 / UTM zone 32N (EPSG Projection 32632). So, even if European reference population grid is in ETRS89/ETRS LAEA (Lambert Azimuth Equal Area - EPSG Projection 3035), the re-projection of this has become very simple given that both of them are based on GRS80 Datum.

So we decided to re-project European Grid in WGS84/UTM Zone 32N using conventional ARCGIS 10.1 algorithm. This because a projected coordinate system such UTM and LAEA requires a definition for the projection transform. This transform is used to translate between linear positions (e.g. meter) and angular longitude/latitude positions. In the sidebar all the transformation parameters are shown.

At the end of the re-projection process, ISPRA layer results completely covered by 310.981 European grid cells. This number is entirely in keeping with the Italian territory extent that can be estimated in 302.070,8 Km2.

Zonal statistics calculation After the acquisition of

European grid projection, the next step involved the calculation of zonal statistics. So using Zonal statistics ARCGIS 10.1 tool, we were able to have the sum of the resident population for each European grid cell.

In fact, with the Zonal Statistics tool, statistics are calculated for each zone defined by a zone dataset, based on values from another dataset (a value raster). A single output value

Box 2 – Example inputs and output from zonal statistics

is computed for every zone in the input zone dataset. The Zonal Statistics as Table tool calculates all, a subset or a single statistic that is valid for the specific

input but returns the result as a table instead of an output raster. A zone is all the cells in a raster that have the same value, whether or not these are contiguous. The

input zone layer defines the shape, values, and locations of the zones. An integer field in the zone input is specified to define the zones. A string field can also be used. Both raster and feature datasets can be used as the zone dataset. The input value raster contains the input values used in calculating the output statistic for each zone.

In the following illustration (Box 2), the zone layer is described as an input raster that defines the zones. The Value layer contains the input for which a statistic is to be calculated per zone. In this example, the maximum of the value input is to be identified for each zone.

Before calculating resident population for the european grid, a zonal statistics as a table algorhytm has been used to evaluate ISPRA data at municipality scale. So using ISPRA layer as value layer, we calculated resident population for each Italian municipality. The result are shown in the excel file attached where the meaning of the fild are:

- COD_REG: Code of the Region in which the Municipality is located; - COD_PRO: Code of the Province in which the Municipality is located; - PRO_COM: ISTAT code of the Municipality; - DEN: Municipality denomination; - PIXEL_COUNT: N. of the DEM pixels (20*20 m) that are included inside the perimeter of the

Municipality; - ISPRA_AREA: Area of the Municipality calculated starting from ISPRA data (PIXEL_COUNT*400).

Area is expressed in m2; - ISTAT_AREA: Area of the Municipality extract from official ISTAT municipality shapefile. Area is

expressed in m2; - POP_ISPRA: Resident population of the Municipality calculated by ISPRA; - POP_ISTAT_2011: Resident population of the Municipality calculated by ISTAT during Census 2011

survey; - DIFF_AREA: Difference between ISTAT_AREA and ISPRA_AREA (ISTAT_AREA-

ISPRA_AREA); - DIFF_POP: Difference between POP_ISTAT_2011 and POP_ISPRA (POP_ISTAT_2011-

POP_ISPRA); - DIFF_AREA/ISTAT_AREA (%): Percentage value calculated as (DIFF_AREA/ISTAT_AREA)*100;

- DIFF_POP/POP_ISTAT_2011 (%): Percentage value calculated as (DIFF_POP/POP_ISTAT_2011)*100;

The ‘Masking work’ As already mentioned in the abstract, the Degree of Imperviousness layer is very suitable to evaluate the

percentage of soil sealing per pixel, however it does not discriminate the residential part of sealed soil.

Therefore, it was necessary to plan a strategy to isolate as many uninhabited areas as possible, in order to distribute population data just to imperviousness layer pixels that represents residential zones.

Thematic digital cartographies (i.e. Land cover and use Maps, Road Networks, etc.) were used as a mask filtering definitely uninhabited zones: it means that all the imperviousness layer pixels that are included inside the polygons, even if they are sealed, were labeled as ‘not-residential’.

Examples of these sealed but uninhabited areas are: Streets and Roads, Airport, Railway stations and network, etc. (see fig.3); in yellow the administrative boundaries.

Fig.3 – Airport enumeration areas before and after masking work.

A summary of the process is reported in box 3 describing how the value of each pixel of the imperviusness layer is proportionally calculated depending on resident population of each enumeration area (in red in Box 3a). Then, in Box 3b, for each grid 1km2 is calculated the sum of the resident population by zonal statistics algorithm and masking work. In the end, in Box 3c, the final result where the square grid is classified depending on the sum of their resident population (in light grey the most populated).

Box 3a) – Merge between enumeration areas and Box 3b) – Merge Imperviousness HR Layer and the grid Imperviousness HR Layer

Box 3c) – The final result. In light grey the most populated grid cells.

Use, Future developments and Conclusion This work can be consider the first step to produce important statistical information about population.

For example, starting from a grid with a pixel wide 400 m2, as the Imperviousness HR Layer really is, it is possible to estimate resident population not only inside ‘conventional’ boundaries (i.e. administrative ones), but inside areas that cannot be obtained grouping enumeration areas too. About this, ISTAT and ISPRA, using Imperviousness HR Layer data, held an experimentation to calculate resident population inside nature conservation areas.

Another very important consideration: starting from ISTAT and ISPRA datasets it is possible to draw population nets with small mesh sizes, best suited for urban areas.

It is also very easy the grid upgrade, if new data will be available in the future; in fact, the process to distribute resident population inside Geostat grid is computerized into a Python language algorithm that is very easy to change.

Moreover, data collected at this resolution can also be used to upgrade DEGURBA (Urbanization degree) algorithm.

In order to improve the final product, it could be taken into account the third dimension. In particular: if it is known the exact height and volume of the building, resident population could be distributed in proportion not only on the base of the HRL Imperviousness classification, but also in relation to buildings characteristics. This objective can be achieved, for example, processing both census buildings data and radar images. This is a difficult process if we consider that Italian building census data are georeferenced just for major cities and not for the entire Italian territory. Furthermore nowadays census buildings data have only a geocoding address.

Nevertheless, starting from the principal cities data it is possible to draw geo-statistical procedures that automatically associate the building height and then calculate volume for all ‘residential’ pixels.

Another important perspective is the estimation of other relevant environmental statistic data such as those related to biomass in forest or agriculture. These quantitative data are very different from population ones because they come from sample surveys. Some interesting results on these topics could be reached only by integrating a lot of different sources but it is not so easy with current available data.

It is worth mentioning that the Copernicus HRL Degree of Imperviousness is planned to be updated every three years, with the next production referred to the year 2015. This frequent production process will be very useful for keeping the Degree of Imperviousness aligned with census data and therefore update the estimation of population distribution.

Finally, on the next page IT Geostat Population Grid is shown (fig.4).

Fig.4 – IT Geostat Population Grid

REFERENCES

- ISTAT Geographic datasets downloaded from: http://www.istat.it/it/archivio/104317#accordions;

- Ines Marinosci ‘I servizi Copernicus/GMES per la valutazione del consumo del suolo’ – Convention:

‘Il Consumo di Suolo: lo stato, le cause e gli impatti’ – Rome, feb 5th 2013

http://www.isprambiente.gov.it/files/eventi/2013/convegno-consumo-del-suolo-2013/Marinosci.pdf;

- S. Mugnoli, R. Chiocchini, R. Molinaro – ‘IT Geostat Population Grid 2011: Passato, Presente, Futuro’ –

FORUM PA 2015 – Rome, May 28th 2015; http://www.forumpa.it/forum-pa-2015/geolocalizzazione;

- EEA, 2012. Guidelines For Verification Of High-Resolution Layers Produced Under Gmes/Copernicus

Initial Operations (Gio) Land Monitoring 2011 – 2013 version 4.

- Freire S., Halkia M. ‘GHSL application in Europe: Towards new population grids’ – EFGS Krakow

Convention – Krakow, Poland Oct 22-24th 2014;

http://www.efgs.info/workshops/efgs-2014-krakow-poland/efgs-2014-conference-1/ghsl-application-in-

europe-towards-new-population-grids;

- Goerlich F., Canatrino I. – ‘Comparing bottom-up and top-down population density grids: The Spanish

Census 2011’ - EFGS Krakow Convention – Krakow, Poland Oct 22-24th 2014;

http://www.efgs.info/workshops/efgs-2014-krakow-poland/8_efgs-2014_abstract-goerlich

- Munafò, M. & Tombolini, I., 2014. Il consumo di suolo in Italia, ISPRA rapporti 195/2014.