View EDGAR spatial Maps

Historical maps (1890 - 1970/1990) for greenhouse gases (CO2, N2O and CH4), ozone precursors (NOx, NMVOC and CO) and acidifying compounds (NH3 and SO2) are   constructed using the EDGAR/HYDE 1.3 sectoral gridded emissions. For the most recent years (1970 and later for greenhouse gases and 1990 for all other gases) we used the EDGARV32 maps. In the table below we specify the datasets used to contstuct the emission maps.

The EDGARV32 maps comprise all anthropogenic sources, including international air traffic and international shipping:

• Fossil-fuel production, transmission, transformation (e.g. coke production, oil refineries) and combustion (F category)
• Biofuel production, transformation (charcoal production) and combustion (B category)
• Industrial production and consumption processes (including solvent use) (I category)
• Agricultural activities (L category)
• Biomass burning (L category)
• Waste handling (W category)     

At present, several gridded population map are publicly available, such as the 1x1 maps from Harvard/Logan (Logan, 1993, pers. comm.), NASA-GISS, two maps at 10'x10' NGCIA (2000), and the GEIA map on 1ox1o by Li (1996). In EDGAR 3 we use the Li map (see Fig. A.1) since it is the only map available at 1x1 degree which has the following qualifications:

• a uniform spatial quality for all countries - in contrast with the NGCIA maps, which are based on sub-national census data that have in many cases a very high spatial resolution, but also include several countries - e.g. the Russian Federation - of which the smallest units are much larger then a 1x1 grid cell;
• a detailed rural population distribution due to the inclusion of small towns to the   size of 10,000 inhabitants and compiled for a recent year - in contrast with the NASA-GISS and Harvard/Logan maps, which were compiled for an older year and using less details for the rural area;
• it locates population in the proper grid cell - whereas allows taking account of border cells, which include areas of more than one country/sea area.  

For these reasons the map compiled by Li was selected by GEIA as the default GEIA population map for new inventories. Therefore, we also decided to use this map, although the EDGAR software presently does use the multi-country/sea feature provided in the dataset. We combined the Li map with the NASA-GISS one grid cell-to-one-country relation table for distributing national total emissions of a particular source to the grid cells allocated to the countries. Moreover, due to the more detailed spatial resolution in the rural areas, the Li map is better suited for splitting into urban and rural sub-maps.

Construction of separate urban and rural population maps

The reason for compiling separate maps for urban and rural population from the total human population map of Li at 1ox1o was to be able to restrict emissions from large scale activities, e.g. from industry and power generation, to the urban population areas when no source-specific map is available. In this way we can avoid that a fraction of the emissions of these sources is also allocated to distant, rural areas. Although that share would have been small (say up to 10-20% in most cases) and sometimes distributed over many cells, in absolute levels it may be a substantial amount, thus increasing rural emissions significantly. This effect has now largely been eliminated from the previous spatial distribution used e.g. in EDGAR 2. We believe that for some atmospheric model applications this may prove to be a sensitive issue.

Construction of urban and rural population maps for EDGAR 3 from the GEIA/Li total population map

There is no generally agreed definition of 'rural' and 'urban' population or 'rural'  and 'urban' area (Clark and Rind, 1992). Sometimes the distinction is between settlements smaller or larger than 10,000 inhabitants living in a built-up area. However, then the question is what is a built-up area: are these households living all adjacent to each other, or are they (partly) distributed over a large area with distances between houses of 100s or 1000s of metres? Also the definitions that countries use in reporting their national fraction of urban and rural population to the UN Population Statistics Division differ in practice, often to an unknown extent. 
Therefore we decided to use a pragmatic approach in separating out the area and population in the country that is living in smaller units, e.g. smaller than 10,000. Since the grid cells at 1x1 can be as large as the order of 100x100 km near the equator, many rural communities may live in one grid cell. So taking a cut-off of 10,000 persons per grid cell would be a too simple approach. Instead we made the following argument:make the spatial distinction into rural and urban areas within a country only for the larger ones, since for the smaller ones the spatial redistribution effects will be much smaller;

• use a default urban population density (in pers/km2) for continuously built-up urban areas based on data for a selection of European type of cities;
• if the population density of a grid cell is higher than is this urban cut-off value, we assume the cell to be 100% urban;
• if the population density is less than a selected rural cut-off density, we assume the cell to be 100% rural;
• also if the total population of a grid cell is lower is than 10,000, we assume the cell to be 100% rural;
• for all intermediate cells, i.e. grid cells with a population density between the urban and rural cut-off values, we ranked the grid cells per country in order of decreasing population density and applied an algorithm for allocating fractions of urban/rural population per grid cell in such a way that the national urban and rural population fractions were equal to the fractions published in the 1990 population statistics of the UN (1999).  

The resulting maps have been compared with high-resolution urban and rural population maps for China, where urban population was based on a cut-off density per high resolution grid cell. Both visual inspection of the spatial pattern as well as the shape of the urban and rural population density distributions (number of grid cells ranked according to decreasing densities) showed good agreement.

EDGAR/HYDE 1.3

J.A. van Aardenne, F.J. Dentener, J.G.J. Olivier, C.G.M. Klein Goldewijk, and J. Lelieveld, A 1° x 1° resolution dataset of historical anthropogenic trace gas emissions for the period 1890-1990, Global Biogeochemical Cycles, Vol. 15, No. 4, pages 909-928, December 2001

EDGARV32

Olivier, J.G.J., J.J.M. Berdowski, J.A.H.W. Peters, J. Bakker, A.J.H. Visschedijk en J.-P.J. Bloos (2001b) Applications of EDGAR. Including a description of EDGAR 3.0: reference database with trend data for 1970-1995. RIVM, Bilthoven. RIVM report no. 773301 001/ NOP report no. 410200 051.

Population