Global Population Distribution Database

United Nations Environment Programme
Environment for Development

Global Resource Information Database - Sioux Falls

Methodology

The ideal way to produce global population data is to obtain detailed census data associated with specific regions within each country. This, however, is both time-consuming and costly, and hence not realistic in the short term. Environment Canada has therefore developed a new methodology in order to produce satisfactory global population data within a reasonable time period. First, a basic global population dataset with 1° X 1° degree latitude/longitude resolution was compiled. Next population data for certain countries in the basic dataset were replaced by more accurate data. The process will be continued as more detailed census data become available.

The two phases of the methodology are outlined in detail below:

Phase I. Creation of the Basic Dataset:

The first phase of the project consisted in creating a global population base map with 1° X 1° degree latitude/longitude resolution by gridding population data from Rand McNally (ref. 2) for more than 6,000 cities according to their latitudes and longitudes. The remaining population of each country was distributed over its habitable areas. This phase was carried out in four stages:

1) Compilation of the Global City Population in 1990

In order to allocate the urban population of each country, we used Rand McNally World Atlas (Ref. 2) population of around 6,000 cities worldwide with population greater than 50,000 inhabitants. The population figures were taken either from censuses or official estimates, other than for a few cities for which only unofficial estimates were available. Population figures were given for different years, ranging from 1967 to 1990, and sometimes varies even for cities in the same country. Unfortunately, average annual urban growth rates are available only from 1985 (ref. 4). So, in order to obtain 1990 population estimates, for cities with data from earlier than 1985, first we applied a growth rate of 0.63 for those in developed countries, and 2.13 for those in less developed countries (ref. 3) in order to obtain 1985 populations for each city. Then the average annual urban growth rates (1985-1990) (ref. 4) were used to obtain 1990 estimates from city populations for the years 1985-1989.

2) Compilation of the Global Rural Population Distribution Factor for 1990 with 1° X 1° degree latitude/longitude resolution

Approximately 90,000 cities and towns (20,500 major cities and 69,000 minor cities, see ref. 5) from 223 countries were allocated to 12,200 grid cells, and population factors were assigned to each city: 2 for each major city, and 1 for each minor city. The assigned population factors in each grid cell were summed up for each country, and a rural population distribution factor dataset for 1990 with 1° X 1° degree latitude/longitude resolution was obtained.

3) Distribution of the rural population data⁽¹⁾

The total rural population was obtained by subtracting the global city population in 1990 from the total of national populations for 1990 (Refs. 6 and 7). This total rural population was allocated to grid cells according to the formula:

total rural equation

This method is based on the assumption that rural population density is in proportion to the number of cities and towns within each cell. This is generally true since small villages tend to be concentrated around cities and towns. Note the larger the country, or the bigger of the rural population in comparison to the total country population, the less accurate the data will be.

D) Finalization of Global 1990 Population Basic Dataset with 1° X 1° degree latitude/longitude resolution

After combining the city and rural populations in each grid cell for each country, a global 1990 population basic dataset with 1° X 1° degree latitude/longitude resolution was created. The dataset was set up in a flexible way so that data from each country can be easily updated. In fact, the population data for United States, Canada, China (including Taiwan), and former Soviet Union have already been replaced by better datasets in Phase II of the project, as described below.

Phase II. Refinement of the Dataset:

1) United States Census Data

Two 1990 census population datasets for approximately 23,400 USA cities and 3,141 counties (Refs. 8 and 9) were obtained. A USA rural population distribution factor dataset was developed for each county using a method similar to one described above for global rural populations. The population factor assigned to each city is given in Table 1.

Table 1: USA rural population distribution factors.

Population Factor
City Population

1
< 1,000

3
1,000-10,000

4
10,000-100,000

5
> 100,000

The rural population for each county was obtained by subtracting the total city population from the county from the total county population, then allocating the remainder to each cell according to the formula:

total rural equation

2) Canadian Data

A gridded Canadian population dataset for 1991 with a 1/6 latitude and 1/4 longitude grid resolution was obtained from the Canadian National Pollutant Release Inventory (ref. 10), and scaled to 1990 for each grid by a factor of 26,647/27,297 (Canadian population for 1990 is 26,647,000, for 1991 is 27,297,000).

3) Chinese Population Data (including Taiwan)

The Chinese 1990 census data, which contains population data for 2,405 administration units (cities and counties), were allocated over the cities and county capitals (ref. 11). For the big cities, like Beijing, Shanghai, Tianjin, the city population (agglomeration data, including the surrounding rural population) were carefully allocated among the residential areas of the city.

4) Former Soviet Union Data

The 1990 population data for 15 former USSR republics were obtained from Dr. Alexey G. Ryaboshapko, Institute of Global Climate and Ecology, Moscow, Russia. City and town population figures were allocated to each cell according to its latitude and longitude. The rural population was determined separately for each county (raion) in Russia, and for each province (oblast) in other former Soviet Republics.

After we replaced the old the population data by the census data for United States, Canada, China, and a better dataset for former Soviet, the total number of populated grid cells was reduced. This is why the file PopFact.dbf contains more grid cells than the file GlPop90.dbf.

⁽¹⁾ For the purpose of this project, we have used an unusually wide definition of "rural population" which encompasses urban populations smaller than 50,000 inhabitants.

[ Results and Discussion || Table of Contents || Mail to: Yin-Fan Li ]

Sitemap

Contact