Global population distribution pdf

South Korea. Saudi Arabia. North Korea. Sri Lanka. Burkina Faso. South Sudan. Dominican Republic. Czech Republic Czechia. United Arab Emirates. Papua New Guinea. Sierra Leone. Hong Kong. El Salvador. State of Palestine. Costa Rica. Central African Republic. New Zealand. Bosnia and Herzegovina. Puerto Rico.

North Macedonia. Equatorial Guinea. Trinidad and Tobago. Solomon Islands. Western Sahara. Cabo Verde. French Guiana. New Caledonia.

French Polynesia. Saint Lucia. Channel Islands. Virgin Islands. The process for generating raster population surface is illustrated in Figure 2. PFLAreai is the proportion of the building floor area of a residential land parcel to the total SubPFLArea j building floor area of all residential land parcels in the subdistrict. Figure 2 about here The above calculations disaggregate subdistrict population to residential land parcels based on their proportions of residential building floor area in the total residential building floor area of the subdistrict.

Our approach is built upon more detailed residential building floor areas and is not limited to the resolution of satellite imagery. It is more flexible and we can choose the resolution of the vector grid optionally. After this operation, each vector grid cell has a population counts attribute. Using the original census data, Figure 3 presents population density distribution by subdistrict, and Table 1 provides the summary statistics of the subdistricts, with most of the subdistricts falling totally within the study area.

Figure 3, a general map of population distribution in Nanjing, shows high-density districts in the central city within the old city wall. From Figure 3, we cannot observe any nodes of population distribution in the suburban areas in the south of Yangtze River, which, as we will show in the next sections, is misleading.

Figure 3 and Table 1 about here With our disaggregated population data, Figure 4 shows the raster population surface of Nanjing with m resolutions.

It can be clearly observed that the raster population surface provides more detailed information than the vector-based population density representation by subdistrict shown in Figure 3. Figure 4 about here The uneven distribution of urban population within subdistricts can be further illustrated by surface models of individual subdistricts.

Figure 5 compares the representation of population in the CBD by raster surface with the representation by vector-based subdistricts.

Coefficient of variance CV of population density by subdistricts was only 0. In contrast to the homogenous population distribution by vector representation within individual subdistrict, the raster population surface reveals high heterogeneity of population in the CBD, as revealed by the map and the high CV value. Figures 5 and 6 about here 4. Urban Population Density and Spatial Structure The generated population surface provides us a valuable platform for studying urban population distribution in Nanjing.

From the raster population surface Figure 4 , we can see that in general, population density declines from the central city to suburban areas. Though Chinese cities including Nanjing have been experiencing population suburbanization Feng and Zhou, ; Zhou and Ma, since the mids, the central city of Nanjing is still characterized by high-density population distribution. These findings suggest Chinese cities like Nanjing remain highly centralized, with limited level of suburbanization.

One of the distinguished properties of raster population surface is that it can provide an effective visualization of population density from statistical density functions of spatial distributions Bracken, The core area of the CBD, with advanced service functions, had relatively lower population density, immediately surrounded by areas with very high population density.

This contrasts somewhat with cities in the West where CBDs tend to have low population density, with a slow rise and then decline of population density towards suburban areas. The other population center is actually the economic subcenter of Nanjing, i. Since this center functions mainly as residential clusters, the core area of this center did not show patterns of lower-higher population density as found in the CBD.

This also contrasts sharply with Western cities where metropolitan subcenters also tend to have low population density. Figure 7 about here To uncover population distribution with distance to the city center and based on the neighborhood statistic population surface, we select six directions starting from the CBD see Figure 7.

In each direction, we extract population density values in every other m from the population surface. Changing population density with the increasing distance from the CBD is shown in Figure 8.

In general, population density declined with distance from the CBD, with some fluctuations due to neighborhood characteristics and the distribution of satellite towns in the suburban areas. We have found that population density often drops substantially in km from the city center, except for the northwest direction with dense population distribution, indicating the compact nature of Chinese cities. Population density also rose slightly at the far end of this direction where several petrochemical plants are located.

Figures 8b and 8d show similar patterns: population density reached peak values within 2 km from the city center, and then declined quickly. In the direction of the southeast and the west Figures 8c and 8e , population density declined constantly, without significant population subcenters in the suburban areas.

Figure 8f shows high population density and slower decline in the northwestern direction because these are old urban neighborhoods where subcenters of economy and population of Nanjing are located. Both Figures 7 and 8 show the uneven distribution of population in the suburban areas, with several nodes of population distribution, which are masked by the orthodox vector representation in Figure 3.

Population surface also provides a better data source than census areal-units for studying density functions. One of the problems associated with census areal unit-based density function is non-random samples, which means there are more observations near the center than those near the periphery Wang and Zhou, Samples from population surface, however, can avoid such a problem.

For the direction A, we exclude the part with no population over Xuanwu Lake. In the direction F, R2 and b values are relatively low, because of slow decline of population density due to the economic subcenter. We also test the same exponential function with original subdistrict census data Table 2. In contrast to the population surface-based density function, the subdistrict-based density function presents a much lower R2 value 0.

Table 2 about here 5. Spatial Effects of Non-residential Land Use on Population Distribution The above section uses the generated population surface to reveal general spatial patterns of urban population distribution in Nanjing, and in this section, we investigate how population distribution varies with the change of urban land uses.

We use the population surface to analyze spatial associations between land use and population density with exploratory spatial data analysis.

Traditionally, researchers set up density functions based on monocentric or polycentric models to depict the change of population density with the distance from the CBD or economic subcenters.

More recently, Cuthbert and Anderson used bivariate K-function to examine the spatial association of residential and commercial land use patterns and expose the change of urban form in the Halifax-Dartmouth region.

The generated population surface and land-use parcel data allow us to explore spatial associations of non-residential land use and population distribution. For industrial land use the same procedure is performed to assign industrial building floor area to each population cell. For road land use, it is reasonable to only assign the area to each cell.

After all the necessary operations, each cell of the population surface has both population attribute and three land-use attributes indicated by building floor area or area. This suggests that non-residential land uses, from the global view, had significant positive spatial autocorrelations with population density. As urban land use has apparent spatial variations, we are more concerned about how different local land-use activities influence population distribution.

We plot the spatial lag WZ p of population against land-use variables, and use the four quadrants of the scatterplot to identify four types of spatial relationships between land use and population distribution: High-High, Low-Low, High-Low, and Low-High. For example, High-High means a cell with high land-use intensity is surrounded by cells with high population density.

The positive High-High and Low- Low spatial autocorrelation indicate a spatial clustering of similar values high land-use intensity surrounded by high population density and vice versa and the negative High-Low and Low- High indicate a spatial clustering of dissimilar values. Furthermore, LISA statistics are used to assess the significance of such spatial associations. The approach requires that zil of land-use value at location i is held fixed, while z jp of population density values are randomly permutated at all other locations.

These findings confirm that the extent of suburbanization remains limited in Nanjing or the Chinese cities in general , and city centers remain densely populations, with intense commercial and managerial land use patterns. Moreover, new migrants also find home in less expensive suburban areas. The process of suburbanization and urbanization, however, has been poorly managed, and the development of commerce and service activities in those areas has been lagging behind.

Figure 12 shows the general picture of population change at the subdistrict level from to We can see that population of the central city increased slowly in general and several subdistricts lost population.

Population of inner suburban subdistricts increased substantially, especially in some west, southeast and northeast subdistricts. Population of outer suburban subdistricts, however, increased slowly compared with inner suburban subdistricts. Table 3 presents the population change of subdistricts in central city, inner suburb and outer suburb from to Population of central city subdistricts accounted for The share of population in inner suburban subdistricts, however, increased from It is evident that population decentralization has been concentrated in the inner suburb.

Such compact patterns reflect urban planning control in Nanjing, and the widely use of bicycles as the main mode of transportation.

Cells with Low-Low spatial association are mostly found in outer suburban area which remained sparsely developed, and north of the Yangtze River which for years has served as a physical barrier to block the northern, northwestern, and western expansion of Nanjing across the river.

Regarding the significance of the spatial associations, large portions of High-High and Low-Low cells are significant, while much fewer Low-High and High-Low cells are significant. As shown in Figure 10, industrial suburbanization is significant in Nanjing where most of the industrial land uses were located in the outer suburban area, which, however, has not affected population suburbanization substantially.

As to the significance of the spatial associations, most of the High-High cells and Low-High cells in the central city were significant. Most of the positive spatial associations between industrial land use and population distribution were located in suburban areas and fringe of the central city. These areas are where most of the current industries in Nanjing area located, with workers living in the central city or nearby areas.

The negative Low-High spatial association was mainly found in the central city area. With economic reforms and the introduction of land markets, many state-owned industrial enterprises were shut down, privatized, or relocated to suburban areas Wei and Li, Consequently, central cities, including Nanjing, have few industries left, which led to the pattern of low industrial land use and high population density in the central city found in this paper. On the other hand, High-Low spatial association was mainly found in some major outer suburban industrial clusters.

Moreover, with limited development in services such as schools and health care, some workers working in these industrial centers commute to work and prefer living in the central city or inner suburban areas.

Population density in these outer suburban areas tends to be lower than industrial centers in the inner suburban areas. In summary, local analyses suggest that industrial suburbanization has facilitated land-use restructuring in the central city area, but was not the main force underlying population decentralization, which has occurred mainly in the inner suburban area.

The central city area had dense road and population distribution, while cells with Low-High spatial association were found in the eastern and northern fringe of the central city area. The inner suburb in the north, south and east of the central city area still had insufficient roads though we can find dense population clusters in these areas.

This again suggests that the inner suburban areas, despite with increasing population in recent years, lag behind in the development of services, including roads and infrastructure. Not surprisingly, low-density road and population distribution were found in the remote suburb areas, which remain largely rural in nature.