Embed Size (px)
Citation preview
24
GE
O-S
PA
TIA
L D
AT
A A
ND
IN
FO
RM
AT
ION
EN
VIR
ON
ME
NT
AL
MA
NA
GE
ME
NT
ASS
ESS
ME
NT
AN
D M
ON
ITO
RIN
G
G
LOB
AL
EN
VIR
ON
ME
NT
AL
CH
AN
GE
EN
VI
RO
NM
EN
T
AN
D
NA
TU
RA
L
RE
SO
UR
CE
S
WO
RK
IN
G
PA
PE
RMapping globalurban and rural population distributionsEstimates of future global population distribution to 2015
[
]
Mapping globalurban and rural population distributions
AnnexEstimates of future global population distribution to 2015Deborah BalkMelanie BrickmanBridget AndersonFrancesca PozziGreg Yetman
Mirella SalvatoreFrancesca PozziErgin AtamanBarbara HuddlestonMario Bloise
Food and Agriculture Organization of the United Nations Rome, 2005
24
GE
O-S
PA
TIA
L D
AT
A A
ND
IN
FO
RM
AT
ION
EN
VIR
ON
ME
NT
AL
MA
NA
GE
ME
NT
ASS
ESS
ME
NT
AN
D M
ON
ITO
RIN
G
G
LOB
AL
EN
VIR
ON
ME
NT
AL
CH
AN
GE
EN
VI
RO
NM
EN
T
AN
D
NA
TU
RA
L
RE
SO
UR
CE
S
WO
RK
IN
G
PA
PE
R
[
]
Ideas contained in this document are solely those of the authors and do not necessarily represent the view of FAO.
The conclusions given in this report are considered appropriate at the time of its preparation. They may be modified
in the light of further knowledge gained at subsequent stages of the project.
The designations employed and the presentation of material in this information product do not imply the expression of any
opinion whatsoever on the part of the Food and Agriculture Organization of the United Nations concerning the legal or
development status of any country, territory, city or area or of its authorities, or concerning the delimitation of its frontiers or
boundaries.
FAO declines all responsibility for errors or deficiencies in the database or software or in the documentation accompanying it,
for program maintenance and upgrading as well as for any damage that may arise from them. FAO also declines any responsibility
for updating the data and assumes no responsibility for errors and omissions in the data provided. Users are, however, kindly
asked to report any errors or deficiencies in this product to FAO.
All rights reserved. Reproduction and dissemination of material in this information product for educational or other non-
commercial purposes are authorized without any prior written permission from the copyright holders provided the source is fully
acknowledged. Reproduction of material in this information product for resale or other commercial purposes is prohibited
without written permission of the copyright holders.
Applications for such permission should be addressed to:
Chief Publishing Management Service Information Division FAOViale delle Terme di Caracalla00100 RomeItaly
or by e-mail to:[email protected]
© FAO 2005
iii
FOREWORD
Geospatial information is increasingly being used for understanding development issuesand improving decision-making. The Millennium Development Goals (MDGs) and their2015 targets for halving the proportions of people living with hunger and poverty, andimproving living conditions in the sectors of education, gender, health and sanitation haveheightened this awareness. This publication is one of a series under production by FAO toexplore the use of high resolution georeferenced data and GIS-based analysis techniques inorder to pinpoint the precise conditions underlying poverty and hunger in the world.
Rural poverty is often associated with vulnerability to environmental conditions and byrelating population distribution to environmental characteristics the underlying causes ofpoverty can be better understood and addressed. Thus, population distribution is one of thekey variables that, if carefully assessed and analysed, can help to target governmentalinterventions to reduce poverty and improve living conditions.
In this report, existing global georeferenced population datasets and their data sourcesare reviewed, and new methodologies that might provide a consistent way of distinguishingbetween urban and rural population and determining the spatial distribution of rural peopleare described. The terms ‘urban’ and ‘rural’ are defined and the quality of availablepopulation numbers is assessed. A method for estimating global population distribution in2015 – the target date of the MDGs – using subnational data is also presented in the Annex.
We are confident that further exploration and application of GIS-based analysistechniques to deepen our understanding of the links between poverty and the environmentwill demonstrate the usefulness of the method, whilst being immediately applicable to thetask of improving living conditions in vulnerable environments.
FAO is grateful to the Government of Norway for the encouragement and funding ithas provided to the FAO Poverty Mapping project under which this study was carried out.
Jeffrey B. Tschirley
Chief, Environment and Natural Resources Service, FAO
v
ABSTRACT
This monograph is part of a series of reports that explain and illustrate methods forapplying spatial analysis techniques to investigate poverty and environment linksworldwide. Analysing population distribution in relation to poverty and environmentalfactors is increasingly recognized as a valuable element in decision-making processesrelated to development issues. Accurately mapping and assessing vulnerable populationscan provide a solid basis for recommendations on how best to reduce poverty and improveliving conditions in developing countries.
In this report, the various definitions of the terms ‘urban’ and ‘rural’ are reviewed,along with data from the United Nations and other sources, and various georeferencedsources are assessed for their usefulness to the geospatial analysis of populationdistribution. The report examines two widely used global georeferenced populationdatasets, reviews recent methodological developments for distinguishing urban and ruralpopulations spatially and presents a method for creating an urban mask and determiningvariations in the distribution of urban and rural population, by pixel. The report concludeswith a brief discussion of unresolved issues and future challenges. Finally, the Annexdetails a method for estimating global population distribution to the year 2015 using datafrom over 375 000 subnational units.
Mapping global urban and rural population distributionsby Mirella Salvatore, Francesca Pozzi, Ergin Ataman, Barbara Huddleston and Mario Bloise
88 pages, 9 figures, 7 tables, 13 maps
Environment and Natural Resources Series, No. 24 – FAO, Rome, 2005
Keywords:Population, urban, rural, vulnerability, GIS, mapping
This series replaces the following:
Environment and Energy Series
Remote Sensing Centre Series
Agrometeorology Working Paper
A list of documents published in the above series and other information can be found at the Web site:
www.fao.org/sd
vi
]M
AP
PI
NG
G
LO
BA
L
UR
BA
N
AN
D
RU
RA
L
PO
PU
LA
TI
ON
D
IS
TR
IB
UT
IO
NS
[
ACKNOWLEDGEMENTS
The authors have benefited from the advice and input of Jippe Hoogeveen, Emelie Healyand Michela Marinelli.
The authors would like to thank Deborah Dukes, Barbara De Filippis and ReubenSessa for editing; Claudia Tonini and Marta Scapellati for the layout of the document.
This report was prepared as part of the FAO Poverty Mapping Project (GCP/INT/761/NOR), which was funded by the Government of Norway.
vii
CONTENTS
iii
v
vi
xi
1
3
3
4
4
5
6
6
7
8
9
13
13
16
18
22
22
25
25
26
26
27
32
34
37
38
41
43
45
45
46
49
Foreword
Abstract
Acknowledgements
Acronyms
1. INTRODUCTION
2. SOURCES FOR URBAN AND RURAL POPULATION DATASETS
2.1 Definitions
2.2 Statistical sources and databases
2.2.1 Internationally-recognized country-by-country population databases
2.2.2 Other widely known sources of population data
2.2.3 Concluding remarks
2.3 Geospatial datasets
2.3.1 Populated places
2.3.2 The Nighttime Lights of the World
2.3.3 Global Land Cover
3. REVIEW OF EXISTING GEOREFERENCED POPULATION DATASETS
3.1 Gridded Population of the World
3.2 LandScan Global Population Database
3.3 Global Rural Urban Mapping Project
3.4 Population databases for Africa, Asia and Latin America
3.5 Other research efforts to map urban population
4. DETERMINING VARIATION IN THE DISTRIBUTION OF URBAN AND RURAL
POPULATIONS BY PIXEL
4.1 Objectives
4.2 Methodology
4.2.1 Choice of population database
4.2.2 Detection of urban areas
4.2.3 Derivation of population distribution grids
4.2.4 Crosschecking of the urban population results with the UN figures
4.3 Comparison of the PMUe with other similar databases
4.3.1 Comparison of the urban land area results
4.3.2 Evaluation of the geographic coordinates of the human settlements in the PMUe
database
5. UNRESOLVED ISSUES AND FUTURE CHALLENGES
REFERENCES
Sources and notes
Document references
Web site references
viii
]M
AP
PI
NG
G
LO
BA
L
UR
BA
N
AN
D
RU
RA
L
PO
PU
LA
TI
ON
D
IS
TR
IB
UT
IO
NS
[
L IST OF F IGURES, TABLES AND MAPS
29
30
31
32
33
39
40
40
41
28
34
35
38
42
10
12
15
17
21
36
37
FIGURES
Figure 4.1 Geometric correction of Nighttime Lights of the World 2000 to UN
international coastline map: Istanbul area
Figure 4.2 Cumulative distribution of population versus DN value in Italy
Figure 4.3 Average light threshold (LT) value by UN region
Figure 4.4 Cameroon: isolated pixels and very small agglomerations not detectable
by LT
Figure 4.5 Different stages in computing the urban mask for Johannesburg and
vicinity
Figure 4.6 Histogram of the share of urban area in total area, by UN region
Figure 4.7 Comparison of the share of urban area in total area, by continent
Figure 4.8 Comparison of the share of urban area in total area by
developed/developing country
Figure 4.9 Visual comparison of urban extents
TABLES
Table 4.1 List of the 154 countries included in the urban and rural databases
Table 4.2 List of the countries with underestimated urban populations
Table 4.3 List of the countries with overestimated urban populations
Table 4.4 Comparison of urban area by UN regions (km2)
Table 4.5 The human settlements in GRUMP database detected by PMUe with
one kilometre buffer
MAPS
Map 2.1 The Nighttime Lights of the World superimposed on bathymetry
(segment)
Map 2.2 The Global Land Cover, 2000
Map 3.1 Population density in 2000 from GPWv3 adjusted to UN totals
Map 3.2 LandScan Global Population Database, adjusted to UN figure year 2000
Map 3.3 Population density in 2000 from GRUMP adjusted to UN totals
Map 4.1 Spatial distribution of the difference in urban population figure by
country
Map 4.2 UN classification of the world in regions
ix
ANNEXESTIMATES OF FUTURE GLOBAL POPULATION DISTRIBUTION TO 2015
55
59
59
62
65
65
65
67
67
68
69
71
73
1. INTRODUCTION
2. INPUT DATA DESCRIPTION
2.1 Boundary input sources
2.2 Population input sources
3. METHODOLOGY
3.1 The gridding approach
3.2 Extrapolation methodology
4. EXTRAPOLATION PROBLEMS AND SOLUTIONS
4.1 Irreconcilable boundary differences
4.2 Mixed administrative level spatial and population data
4.3 Particularly high population growth rates
5. CONCLUSION
REFERENCES
x
]M
AP
PI
NG
G
LO
BA
L
UR
BA
N
AN
D
RU
RA
L
PO
PU
LA
TI
ON
D
IS
TR
IB
UT
IO
NS
[
L IST OF TABLES AND MAPS
63
64
59
62
64
65
70
74
TABLES
Table 2.1 Countries with the highest and lowest available resolution
Table 2.2 Average level and total number of administrative units, by continent
MAPS
Map 1.1 Global population density in 2015
Map 2.1 Number of administrative units included in GPWv3, by country
Map 2.2 Year of the most recent census data available, by country
Map 2.3 Number of population data years employed, by country
Map 4.1 Number of countries for which sub-national versus national growth rates
were used
Map 5.1 Population density projections for the year 2015 with a focus on selected
urban areas
xi
ACRONYMS
AEZ Agro-Ecological ZonesAVHRR Advanced Very High Resolution RadiometerBUUA Boston University Urban AreaCIAT Centro Internacional de Agricultura TropicalCIESIN Center for International Earth Science Information NetworkDAFIF Digital Aeronautical Flight Information FileDCW Digital Chart of the WorldDMA Defense Mapping AgencyDMSP Defense Meteorological Satellite ProgramDN Digital NumbersDPKO United Nations Department of Peacekeeping OperationsESRI Environmental Systems Research InstituteGIS Geographical Information SystemGLCC Global Land Cover CharacteristicsGNS GEOnet Names SeverGPW Gridded Population of the WorldGRUMP Global Rural Urban Mapping ProjectGRUMPe Global Rural Urban Mapping ProgrammeIDB International Data BaseIIASA International Institute for Applied System AnalysisIPC International Programs CenterIPFRI International Food Policy Research InstituteJNCs Jet Navigation ChartsJRC European Commission’s Joint Research CentreLCCS Land Cover Classification SystemLT Light intensity ThresholdMDGs Millennium Development GoalsMODIS MODerate resolution Imaging Spectroradiometer NCGIA National Center for Geographic Information AnalysisNGA National Geospatial-Intelligence AgencyNGDC National Geophysical Data CenterNIMA National Imagery Mapping AgencyNOAA National Oceanic and Atmospheric AdministrationNTL Nighttime LightsOLS Operational Linescan SystemONC Operational Navigation ChartORNL Oak Ridge National LaboratoriesPMRe5 Poverty Mapping Rural extents at 5 arc-minutesPMRd Poverty Mapping Rural population density
xii
]M
AP
PI
NG
G
LO
BA
L
UR
BA
N
AN
D
RU
RA
L
PO
PU
LA
TI
ON
D
IS
TR
IB
UT
IO
NS
[
PMRp Poverty Mapping Rural populationPMRp5 Poverty Mapping Rural population at 5 arc-minutesPMRSd Poverty Mapping Rural Settlements population densityPMRSe Poverty Mapping Rural Settlements extentsPMRSp Poverty Mapping Rural Settlements populationPMUd Poverty Mapping Urban population density PMUe Poverty Mapping Urban extentsPMUp Poverty Mapping Urban populationPMUR Poverty Mapping Urban Rural databaseSABE Seamless Administrative Boundaries of EuropeTPCs Tactical Pilotage ChartsUN United NationsUNCS United Nations Cartographic SectionUNEP United Nation Environment ProgrammeUNL University of Nebraska-LincolnUNSD United Nations Statistic DivisionUNup United Nations urban population figuresUSGS U.S. Geological SurveyVIIRS Visible Infrared Imaging Radiometer SuiteVMap0 Vector Smart Map level 0VMap1 Vector Smart Map level 1VMap2 Vector Smart Map level 2WRI World Resources Institute
Until recently, demographers and cartographers have had only two sources of informationabout population distribution:
" paper maps and navigational guides showing the location of cities and towns and theadministrative boundaries of countries and subnational administrative units withincountries;
" statistical data obtained from national censuses and other demographic surveys.With this information, it has been possible for some time to present statistical
information about national population size, and the location and population count ofurban centres and other subnational administrative units, in map form. This combinationof statistical data with information about administrative boundaries produces maps thatgive a geographic representation of the concerned parameters, but the usefulness of thesemaps is limited as they are not georeferenced and are not supported by databases thatcapture variations within the mapped units.
Researchers have been developing a number of techniques for mapping globallyvariations of parameters within countries. As these techniques have become moresophisticated, and the capacity of computers to handle very large datasets with great speedhas increased, the interest in developing methods for distributing population data to thegrid cells of GIS maps has also increased. Initially, GIS specialists tended to direct theirefforts towards establishing the coordinates of coastlines and country boundaries, andgenerating georeferenced datasets for physical and environmental variables that could bederived from high-resolution aerial photography and satellite imagery. Less effort wasdirected towards the development of georeferenced socio-economic datasets, mainlybecause such data is collected by censuses and surveys and compiled for political oradministrative units, and direct interpolation techniques to estimate the spatial distributionof socio-economic variables are still lacking (Clark and Rhind, 1992; Deichmann, 1996).
Despite these limitations, improvements in the quality and accessibility ofgeoreferenced environmental data have generated growing demand for more accurate andup-to-date spatial information about the global distribution of population variables. Thisdemand has been driven by two different concerns within the development community.One relates to the interest of demographers, sociologists and urban planners in mappingurbanization processes and defining the location and socio-economic characteristics ofurban populations with more precision. The other relates to the interest of agriculturaleconomists and rural planners in gaining access to current data showing the location andsocio-economic characteristics of rural populations in relation to physical and
1
C H A P T E R 1 INTRODUCTION
2
environmental factors that affect their livelihood options and vulnerability to poverty andfood insecurity.
The first issue to be addressed has been the problem of definition. Somewhat surprisingly,despite the importance of urban growth for development processes worldwide, there is nocommonly accepted definition of what constitutes an urban area, and no commonly acceptedspatial characterization of urban areas. The methods used to enumerate urban and ruralpopulations differ from one country to another, and these national differences are reflected inthe global population statistics maintained by the United Nations.
An exploration of definitions currently in use, along with an assessment of the qualityof available statistical and geospatial data relevant for mapping population distribution, arethe subject of chapter two of this report.
Chapter three describes the evolution of georeferenced population datasets and explainshow information generated from paper maps, high-resolution aerial photography andsatellite imagery has been combined with statistical data to develop GIS datasets of globalpopulation distribution. It presents a method developed by CIESIN, based on its GriddedPopulation of the World (GPW) dataset, which distinguishes urban from rural populationswithin subnational administrative units and maps the location and extents of urban areas. Italso summarizes other work in progress to distinguish and map urban and rural population,in most cases using GPW as the reference GIS dataset for global population distribution.
Chapter four describes a method developed by FAO/SDRN to detect urban areas,using the LandScan Global Population Database, and to generate separate grids for urbanand rural populations at 30 arc-seconds that facilitate spatial analyses of the environmentaland socio-economic vulnerability of rural populations.
While both GPW and LandScan represent significant advances and already haveconsiderable utility for the research community, the methods developed so far for mappingurban and rural populations with these two databases are not fully satisfactory. A briefsummary of unresolved issues and future challenges is given in chapter five.
The Annex describes a method of estimating the distribution of the world’s population tothe year 2015, using GPW version 3 as the database; the particular relevance of 2015 is thatit has been set as the target year for the World Food Summit and Millennium DevelopmentGoals (MDGs).
]M
AP
PI
NG
G
LO
BA
L
UR
BA
N
AN
D
RU
RA
L
PO
PU
LA
TI
ON
D
IS
TR
IB
UT
IO
NS
[
The primary sources for population data are national censuses and other demographicsurveys. The publicly available data include population totals at the country level and byadministrative units, as well as population data for cities – usually above a certain size. Inthis chapter, the ways in which urban and rural areas have been defined are examined, andthe statistical sources of population data and geospatial databases which are used toproduce some of the georeferenced population datasets are described.
2.1 DEFINITIONSThe task of defining urban population has always been particularly challenging. TheUnited Nations itself recognizes the difficulty of defining urban areas globally, stating that,“because of national differences in the characteristics that distinguish urban from ruralareas, the distinction between urban and rural population is not amenable to a singledefinition that would be applicable to all countries” (UN, 1998). Rural areas are usuallydefined as “what is not urban” (UN, 1998 and 2004), and so inconsistencies in thedefinition of what is urban lead to inconsistencies in characterizing what is rural.
Each country defines the term ‘urban’ in its own way, although this is often only interms of other labels; for example, ‘urban centres’, ‘major cities’, ‘administrative centres’ or‘municipalities’. Sometimes the administrative boundaries of human settlements such ascities, towns and villages are available and are used to distinguish urban from rural; thepopulations within these administrative units are classified as urban. When definitions arebased on quantitative thresholds, the minimum population for a place to be consideredurban varies greatly. For instance, in several countries in Latin America and West Africa,the threshold is a population of 2 000, whereas it is 200 in Iceland, and 10 000 in countrieslike Italy and Benin. Alternatively, the definition of an urban population can be verycomplex, involving the socio-economic characteristics of the population or community(UN, 2004).
An urban agglomeration is generally easier to define. The United Nations describes itas a place that “comprises a city or town proper and also the suburban fringe or thicklysettled territory lying outside, but adjacent to, its boundaries. A single large urbanagglomeration may comprise several cities or towns and their suburban fringes” (UN,1998). Nonetheless, the spatial boundaries of the agglomeration or the cities included areusually not provided, yielding great uncertainty about its characterization.
3
C H A P T E R 2 SOURCES FORURBAN AND RURAL POPULATIONDATASETS
4
This lack of commonly accepted definitions makes it extremely difficult to find a globalbasis for defining urban areas.
As will be seen in the description of relevant geospatial datasets, remote sensing can be ahelpful tool in characterizing urban areas consistently on a large scale. With the developmentand refinement of spatial techniques for defining urban boundaries and modelling spatialdistribution of population, the choice of whether to use UN definitions and urban/ruralpopulation counts that conform to national usage but are not consistent across countries, orto use spatial information about human settlements and urban area boundaries derived fromsatellite imagery, will depend on the objectives of specific research applications.
2.2 STATISTICAL SOURCES AND DATABASESRecognized sources for internationally comparable population data are the UNPopulation Division (Web site ref. 1) and the United States Bureau of the CensusInternational Programs Center (IPC) (Web site ref. 2).
Other widely used sources of population data, such as the Web sites of Gazetteer (Website ref. 3) or City Population (Web site ref. 4) supplement data from the previous twosources with information that they obtain from other official in-country sources and localsurvey data for urban agglomerations, cities and towns.
2.2.1 Internationally-recognized country-by-country populationdatabasesThe primary sources of the UN and US Census Bureau are national census and otherdemographic surveys. These are not conducted annually, and the census or survey datesvary from one country to another. Statisticians in both the UN and the US Census Bureaucollect these data and interpolate from one survey to the next to create data series.
The International Data Base (IDB), created in the US Census Bureau’s IPC, is acomputerized source of demographic and socio-economic statistics for 227 countries and areasof the world. The IDB combines data from country sources (especially censuses and surveys)with IPC’s estimates and projections to provide information dating back as far as 1950 and asfar ahead as 2050. Because the IDB is maintained at IPC as a research tool in response to therequirements of its sponsors, the amount of information available for each country may vary.
Through its Demographic Yearbook system, the UN Statistics Division (UNSD) hascollected country-by-country population data from national statistical authorities since1948, through a set of questionnaires dispatched annually to over 230 national statisticaloffices. UNSD’s annual Demographic Yearbooks provide latest available statistics onpopulation size and composition, fertility, adult mortality, infant and foetal mortality,marriage and divorce as well as special topic issues. The 26 tables of the DemographicYearbook 2002 as well as technical notes are available electronically.
UNSD also provides data on the population of capital cities and cities of 100 000 andmore inhabitants for the latest available year. In this database the population data are givenfor the city proper and for the urban agglomeration, including the suburban fringe adjacentto the city boundaries.
]M
AP
PI
NG
G
LO
BA
L
UR
BA
N
AN
D
RU
RA
L
PO
PU
LA
TI
ON
D
IS
TR
IB
UT
IO
NS
[
5
The UN Population Division uses UNSD data as the basis for preparing currentdemographic estimates, standardized time series starting from 1950, and projections to2050 for total population, urban population and rural population for all countries and areasof the world. Standard demographic techniques are used to estimate the population by ageand sex for the current year; these estimates then serve as the base for the projections.International and rural/urban migration, total fertility, life expectancy at birth, infant, childand maternal mortality and increased adult mortality in some regions, as well as thedemographic impact of AIDS, are among the factors taken into account. The results,published annually in World Population Prospects, serve as the standard and consistent setof population figures for use throughout the United Nations system. The entire time seriesis available online.
The UN Population Division also publishes a biennial report, World UrbanizationProspects, which contains summary tables by country and region and also reports thesources of data and the definition of urban and rural when available, for each country.
In the most recent revision of this report (UN, 2004), it is stated that the world’s urbanpopulation reached 2.9 billion in 2000, corresponding to 48 percent of the total population.Much of the population growth that occurred in the past 50 years, and most of what willoccur in the next 30 years, concerns urban areas. The majority of the urban populationgrowth to occur in developing countries, where it is projected to increase by 2.3 percentper year between 2000 and 2030, as opposed to an increase of only 0.5 percent in the moredeveloped countries.
The UN report highlights the differences in urbanization rates and numbers of urbandwellers by region, as well as the size of cities expected to absorb most of the populationgrowth in the next 15 to 30 years. For example, the proportion of people living inmegacities (with population greater than 10 million) across the world is still fairly small,amounting to 4.1 percent in 2000 – a figure expected to rise to 5 percent by 2015. Overall,by 2015 it is expected that only 8.7 percent of the world population will live in cities with5 million inhabitants or more, as opposed to the 27.2 percent expected to be living in urbansettlements with fewer than 500 000 inhabitants.
World Urbanization Prospects also gives estimates and projections of the population ofurban agglomerations with 750 000 inhabitants or more for the period 1950 to 2015.
In both the UNSD and the UN Population Division databases, the geographic informationof latitude and longitude to identify the locations of human settlement is not reported, althoughin most instances these can be derived from other sources (see sections 2.2.2 and 2.3.1).
2.2.2 Other widely known sources of population dataStatistical offices and gazetteers are other widely used sources of population data. There areseveral Web sites listing populated places, usually derived from primary sources, but fewcontain population estimates for those named places. Two Web sites that offer goodinformation about current populations of countries, their administrative divisions, cities andtowns are Gazetteer and City Population. Both report information about population,gathered from official census data, the UN and IDB databases and other official and non-
SOURCES FOR URBAN AND RURAL POPULATION DATASETS
6
official sources; they also produce estimates for the current year, if these are not alreadyavailable from official sources.
In some cases it is in fact difficult to obtain population figures for cities and townsbecause the statistical registration in a country is not very accurate due to civil wars and/orpoverty. Gazetteer admits that the figures presented on its Web site are far from beingofficial, but they are calculated carefully and revised manually if necessary. One issuehighlighted in the documentation of both Web sites concerns the definition of cities andurban agglomerations. Metropolitan areas are important to define, as they are indicators ofa country’s urbanization and economy. However, for many metropolitan areas it is difficultto specify an exact population figure, especially for the fast growing agglomerations indeveloping countries, because they are continuously incorporating cities and urbanizingareas in their environment and their definitions are often not comparable around theworld. Most countries do not specify whether their data is for city or for urbanagglomeration, and some even have data available for both types of place (see Web site ref.3). Both Gazetteer and City Population Web sites take simply what is provided by thecountries in terms of type of settlements. Therefore, the lists provided are to be consideredonly as a rough reference table for the world’s largest agglomerations, and, as with othersources the figures are not necessarily comparable across countries.
2.2.3 Concluding remarksThe main problem with the statistics available concerns the reliability of the populationfigures used. If the city population figures are not very accurate, the calculation of urbanand rural proportions is also incorrect. Furthermore, although some databases that givecity population estimates also include geographic information such as the latitude andlongitude coordinates for points or the centroids of polygons representing the locations ofcities, they rarely contain information about the extent of each urban area.
It is well known that the world population reached 6 billion in 2000 and is projected togrow to 8.9 billion by 2050 (UN, 2003). Even though urban population is about half of thetotal population, the percentage of land occupied by urban areas is only about three percent(Balk et al., 2004a). Current urban population densities are already putting pressure on theenvironment in many parts of the world, and this pressure is likely to increase in fast-growing urban areas. Similarly, high population density, environmental degradation andincreasing poverty are also major issues in traditionally agricultural rural areas. This indicatesthe importance of understanding the spatial distribution of population, in addition to havingaccurate information about the urban/rural proportions.
2.3 GEOSPATIAL DATASETSTwo important georeferenced datasets have been developed during the 1990s to overcomethe limitations of statistical data for spatial studies of population. The initial breakthrough inthe global mapping of urban areas came with the release of the first Digital Chart of theWorld (DCW) in 1992 (Danko, 1992). This was a set of computerised global maps, createdfor the most part by scanning and digitising paper sources. Through DCW, georeferenced
]M
AP
PI
NG
G
LO
BA
L
UR
BA
N
AN
D
RU
RA
L
PO
PU
LA
TI
ON
D
IS
TR
IB
UT
IO
NS
[
7
datasets for settlements, country boundaries and other layers of information were madeavailable by country. The populated places layer of the DCW has evolved into anotherimportant source for spatial information about cities and towns. This is the NationalImagery Mapping Agency (NIMA) points database, which holds the geographiccoordinates of several million human settlements, together with their names and theadministrative units to which they belong, if known.
The other major source of information about urban areas globally is the NighttimeLights dataset from the National Oceanic and Atmospheric Administration (NOAA).Although it has been under development since the early seventies, it is only since 1997 thatthis dataset has been used to derive a global image map showing light sources, includinghuman settlements. All other global georeferenced datasets that include an urban layer relyon either DCW and its subsequent refinements or Nighttime Lights as primary source.
2.3.1 Populated placesThe DCW was developed originally in 1992 by the Environmental Systems ResearchInstitute, Inc. (ESRI) on commission for the US Defense Mapping Agency (DMA) (Web siteref. 5). The DCW is a vector basemap of the world at a scale of 1:1 000 000. The primarysources for this database were the Operational Navigation Chart (ONC) series, co-producedby the military mapping authorities of Australia, Canada, United Kingdom, and the UnitedStates; and the Jet Navigation Charts (JNCs) for the region of Antarctica. Some collateralsources have been used to add extra information about road and railroad connectivitythrough selected urbanized areas, for instance the Digital Aeronautical Flight InformationFile (DAFIF) for the airport data contained in the aeronautical layer, and the Advanced VeryHigh Resolution Radiometer (AVHRR) dataset for the data in the vegetation layer.
The DCW database is organized into 16 thematic layers and one data quality layer.
The DMA subsequently merged with several other agencies to form the NationalImagery and Mapping Agency (NIMA), later renamed as National Geospatial-IntelligenceAgency (NGA). NIMA released an updated and improved version of the DCW database,called Vector Smart Map level 0 (VMap0) in 1997 (Web site ref. 6). VMap0 includes majorroad and rail networks, hydrologic drainage systems, utility networks (cross-country
SOURCES FOR URBAN AND RURAL POPULATION DATASETS
B O X 2 . 1
THEMATIC LAYERS IN DCW
Political/Ocean Hypsography supplemental
Populated places Land cover
Railroads Ocean features
Roads Physiography
Utilities Aeronautical
Drainage Cultural landmarks
Supplemental drainage Transportation structure
Hypsography Vegetation
8
pipelines and communication lines), major airports, elevation contours, coastlines,international boundaries and populated places.
The more recent versions of this database, VMap1 (Web site ref. 7) and VMap2, are notyet completely available in the public domain. The greatest improvement in VMap1 is its1:250 000 map scale resolution, four times higher than VMap0. The structure is quite similar,with the data content held in ten thematic layers. VMap1 data are divided into a rathercomplex global mosaic of 234 geographic zones. However at the present time, NGA is onlyreleasing 55 selected zones from the VMap1 dataset.
The populated places layer in DCW and VMap contains points and polygons thatrepresent human settlements. The points dataset is a collection of latitude/longitudereferences associated with known locations of human settlements. The polygons datasetidentifies urbanized (or built-up) areas of the world that it is possible to represent at 1:1 000000 scale. Their shapes look as viewed from the air and their outlines do not necessarilyconform to political boundaries.
The populated places layer of VMap, i.e. NIMA points database, constitutes perhaps themost comprehensive georeferenced cities database available. In the 1997 version of Vmap0,the populated places remained essentially unchanged from DCW. An updated version ofVMAP0, released in 2000, added the names for most of the unnamed points and polygons,with the result that the database now contains nearly 5,000,000 named settlement points orpolygons. The database is available through the GEOnet Names Server (GNS) of the NGA(Web site ref. 8).
Both points and polygons are good sources of information, but they also have certainlimitations. One drawback is that this points database does not provide populationinformation. The polygons tend to be conservative measures of the urban extents; often theydo not correctly represent the extent of urban agglomeration, and prove to be inconsistentglobally. In many instances where there are multiple settlements with the same name, it is notpossible to ascertain which point and coordinates correspond to the city of interest. Finally,the points do not locate the geographical positions of settlements very precisely, partly dueto imprecision of the source information and partly due to lack of consistent standards forselecting the point within an urban extent that should represent its location.
2.3.2 The Nighttime Lights of the WorldThe Nighttime Lights dataset has been created from data collected by the United States AirForce Defense Meteorological Satellite Program (DMSP) Operational Linescan System(OLS). This instrument has a low-light imaging capability, designed for the observation ofclouds illuminated by moonlight in two spectral bands (visible/near infra-red and thermalinfra-red). In addition to detecting moonlit clouds, the instrument can be used to detect lightsources present at the earth’s surface. Time series data from the DMSP-OLS were used toderive a global dataset and map image showing light sources observed during a six-monthperiod spanning 1994–1995 (Elvidge et al., 2001). The 1994–1995 dataset was released in 1997and has since then been used extensively to map urban areas globally (Elvidge et al., 1997;Sutton, 1997). More recently, the DMSP group at the National Geophysical Data Center
]M
AP
PI
NG
G
LO
BA
L
UR
BA
N
AN
D
RU
RA
L
PO
PU
LA
TI
ON
D
IS
TR
IB
UT
IO
NS
[
9
(NGDC) of the National Oceanic and Atmospheric Administration (NOAA) releasedversion one of a pair of DMSP-OLS ‘Nighttime Lights of the World’ images and relateddatabases, processed specifically to detect change, covering the years 1992-93 and 2000 (Website ref. 9). Map 2.1 shows a focus on Italy superimposed on bathymetry.
The OLS detects lights from human settlements, fires, gas flares, and heavily lit boats(primarily squid-fishing boats). These four types of lights have been distinguished on thebasis of location, brightness/persistence, and visual appearance. Four different datasets areavailable as a result: human settlements (cities, towns, villages and industrial sites), gas flares,fires, and heavily lit fishing boats. These products are usually available as frequency ofdetection (0–100 percent) over cloud-free observations during the time period considered. Inaddition to the percent frequency products, NOAA also provides the number of validcoverages, the number of cloud-free coverages, the number of cloud-free light detections,and the average digital numbers (DN) of the detected lights. The processing for the 1992-1993 and 2000 sets of data included automatic cloud detection and a modified light detectionalgorithm designed to capture dim lighting, but the final products are not radiance- calibratedowing to the lack of on-board calibration and uncertainty in the gain settings. The resolutionof all the lights datasets is 30 arc-seconds (nominally one kilometre at the equator).
There are widely known problems with the data, possibly the most significant of whichconcerns the blooming effect. The blooming effect is an overestimation of the actual extentof urban areas, dependent on to intrinsic characteristics of the sensor (Elvidge et al., 1997,2004). There have been attempts to impose a threshold on the lights in order to reduce thiseffect (Imhoff et al., 1997), but doing so results in the loss of small settlements that are notfrequently lit. The difficulty of finding a unique threshold that could work globally has beenexplored in a recent publication (Small et al., 2005). One other problem concerns fires andgas flares. Using OLS alone, it is difficult to separate gas flares adequately from humansettlements. As a matter of fact, NOAA releases a stable lights dataset (human settlement andfires) as well as its human settlements dataset (from which fires have been removed). Theseparation of city lights and fires is not entirely clear in some parts of the world whereextensive fires are frequent. The other major problem, encountered in the northernhemisphere above almost 40 degrees latitude, is the effect of snow on the extent andbrightness of the lights. Various techniques are being explored to minimize these problemsfor forthcoming releases. A new generation of annual global OLS Nighttime lights iscurrently in production for the 1992-2003 time period using the Visible Infrared ImagingRadiometer Suite (VIIRS) instrument. This series is expected to present substantialimprovements in calibration, spatial resolution and level of quantization (Lee et al., 2004).
2.3.3 Global Land CoverIn the past decade, several efforts have been made to map land cover globally or atcontinent-wide scales using remotely sensed data. One of the most-widely used is theGlobal Land Cover Characteristics (GLCC) dataset (Web site ref. 10), generated by theUnited States Geological Survey (USGS), the University of Nebraska–Lincoln (UNL),and the European Commission’s Joint Research Centre (JRC). The GLCC database was
SOURCES FOR URBAN AND RURAL POPULATION DATASETS
10 ]M A P P I N G G L O B A L U R B A N A N D R U R A L P O P U L A T I O N D I S T R I B U T I O N S[
M A P 2 . 1
The Nighttime Lights of the World superimposed on bathymetry (segment)
Source: National Oceanic and Atmospheric Administration (NOAA)
11
developed on a continent-by-continent basis, based on Advanced Very High ResolutionRadiometer (AVHRR) data for the year from April 1992 to March 1993. The resolution ofthe product is a nominal one kilometre. The dataset was originally released to the public in1997 and has subsequently been updated based on feedback from users.
More recently, the Global Vegetation Monitoring Unit of JRC, in collaboration withmore than 30 research teams, has developed a global land cover product (GLC2000) for theyear 2000 (Web site ref. 11). The GLC2000 is based on SPOT-VEGETATION data, at onekilometre resolution, and on a Land Cover Classification System (LCCS) developed byFAO and the United Nation Environment Programme (UNEP). The hierarchicalclassification system allowed the different partners to choose land cover classes which bestdescribed their region, whilst also providing the possibility to translate regional classes to amore generalized global legend (see Map 2.2).
Both of these global land cover include a land cover class for built-up areas: in the case ofGLCC this comes from DCW; GLC2000 bases its built-up area class on Nighttime Lights1994/95. As these are derived data layer, there is no advantage in using them rather then theoriginal sources. However, the classifications of agricultural and forested areas are unique tothe land cover datasets, and can be used as a basis for distributing rural population acrosspixels, if the average population density for different types of land cover is known or can beestimated.
FAO definitions of and statistics on agricultural areas have been used to estimate totalagricultural area in the world and to verify the accuracy of the area estimates given in theglobal land cover datasets (FAO, 2003).
Another global land cover dataset has been developed by the International Institute forApplied System Analysis (IIASA), in a collaborative effort with FAO. To create this datasetagro-ecological conditions in each pixel have been evaluated for their suitability for differenttypes of crop production and for pasture, and the results matched with research data andagricultural statistics (Fischer et al., 2002). In this dataset, the land cover class for artificialsurfaces and built-up areas is based on GLC2000, with some adjustments to account for thepresence of buildings and infrastructure in rural areas.
Since December 1999 a new generation of satellite images has been produced by MODIS(Moderate Resolution Imaging Spectroradiometer). This instrument detects 17 land covertypes, including 11 categories of vegetation and various non-vegetated surfaces, includingbare soil, water and urban areas (Web site ref. 12). This instrument is operates from NASAsatellites (Terra and Aqua). The frequency and sophistication of the MODIS images offer theprospect of significant future advances in land cover analysis.
SOURCES FOR URBAN AND RURAL POPULATION DATASETS
12 ]M A P P I N G G L O B A L U R B A N A N D R U R A L P O P U L A T I O N D I S T R I B U T I O N S[
M A P 2 . 2
The Global Land Cover, 2000
Source: Global Vegetation Monitoring Unit of the European Commission’s Joint Research Centre (JRC)
The previous chapter reviewed definitions of urban and rural areas, and analysed whatstatistical data are available and what georeferenced datasets could be used as potentialinputs into models of population distribution. In this chapter, the two most widely knownand used georeferenced global population distribution databases that have been developedbased on these sources are reviewed and several recent efforts to model populationdistribution, taking urban and rural areas explicitly into account are described.
The Gridded Population of the World (GPW), originally developed at the NationalCenter for Geographic Information Analysis (NCGIA) and subsequently updated by theCenter for International Earth Science Network (CIESIN) at Columbia University,attributes population to the lowest subnational administrative units for which populationcounts are available. In GPW the population count for each administrative unit isdistributed uniformly across all the gridcells of the unit, without considering whether thegridcell belongs to urban or rural area.
The LandScan Global Population Database, produced by the Oak Ridge NationalLaboratories (ORNL), distributes national populations by land cover category, accordingto a model with assumed coefficients for population occurrence in each type of land cover.
General information about how each database was produced is given below, along withthe main advantages and disadvantages of each. In both cases, the primary sources ofpopulation are data from censuses and surveys compiled for political or administrativeunits. The term global is used to indicate that there is no explicit reference to urban or ruralareas, and only overall total population counts and densities are given. As there is morethan one global database available, each being produced by different methods, the mostsuitable database should be chosen largely on the basis of the type of application for whichit is to be used.
3.1 GRIDDED POPULATION OF THE WORLDThe GPW project was the first major attempt to generate a consistent global georeferencedpopulation dataset. It was originally produced at the National Center for GeographicInformation Analysis (NCGIA) in 1995 (Tobler et al., 1995), and subsequently updated byCIESIN in 2000 (Deichmann et al., 2001) and in 2004 (Balk and Yetman, 2004).
13
C H A P T E R 3 REVIEW OFEXISTINGGEOREFERENCEDPOPULATIONDATASETS
14
GPW was the first global rasterized dataset of population totals based solely onadministrative boundary data and population estimates associated with thoseadministrative units. In the original version, two datasets at 2.5 arc-minutes were producedwith the data for the year 1990:
i) unsmoothed, where the gridding algorithm assigned population in grid cells withmultiple input polygons by a straight majority rule, and
ii) smoothed, where population was distributed based on a smoothing method calledpycnophylactic interpolation (Tobler et al., 1995), which assumes that grid cellsclose to administrative units with higher population density tend to contain morepeople than those close to low density units.
Since that first release, higher resolution population data sets have been compiled forvarious regions of the world. In 2000 CIESIN released an updated second version of GPW.GPWv2 is based on more detailed administrative units, resulting in an improved medianresolution. The median resolution is defined as the ratio of total area of the country tonumber of administrative units; a lower number indicates a larger number ofadministrative units, and therefore a more spatially refined dataset. Nonetheless, no effortwas made to model population distribution, and no ancillary data were used to predictpopulation distribution or revise the population estimates. The only assumption made wasthat population is uniformly distributed within each administrative unit. The latestversion, GPWv3 (Web site ref. 13), is based on the same assumptions as the previousversion but relies on more recent data at higher resolutions (see Map 3.1). In particular, thenumber of administrative units has increased from approximately 128 000 in GPWv2 tomore than 375 000 in GPWv3, and consequently the average median resolution hasdropped from 33 in GPW2 to 18 in GPWv3. This new version contains unadjustedpopulation data for the years 1990, 1995 and 2000, as well as data for those years adjustedto match United Nations population estimates. Data about land area and populationdensity are also included. In order to avoid mismatches at the border between countries,most country boundaries have been matched to standard sources, namely SeamlessAdministrative Boundaries of Europe (SABE, Web site ref. 14) and DCW.
The main advantages in using GPW are that it relies on a very simple area-weightingscheme for reallocation, and on the best possible census and administrative data available.GPW also provides updates every five years, allowing for a (short) time series analysis. Itsmain drawbacks are its coarse resolution of 2.5 arc-minutes, which corresponds toapproximately 5 kilometres at the equator, and the lack of any modelling of populationdistribution within administrative units, causing population to be evenly distributed acrossany given administrative unit. This is unlikely to represent a realistic populationdistribution, especially within large units with significant variation in land covercharacteristics.
]M
AP
PI
NG
G
LO
BA
L
UR
BA
N
AN
D
RU
RA
L
PO
PU
LA
TI
ON
D
IS
TR
IB
UT
IO
NS
[
15
RE
VIE
W O
F EX
ISTIN
G G
EO
RE
FER
EN
CE
D P
OP
ULA
TIO
N D
AT
ASE
TS
M A P 3 . 1
Population density in 2000 from GPWv3 adjusted to UN totals
Source: Center for International Earth Science Information Network (CIESIN), Columbia University and Centro Internacional de Agricultura Tropical (CIAT)
16
3.2 LANDSCAN GLOBAL POPULATION DATABASEThe Oak Ridge National Laboratories developed LandScan (Web site ref. 15) in 1998(Dobson et al., 2000) in order to overcome the limitations of GPW, and originally inresponse to a demand for distributed population data that would show emergency workerswhere populations were likely to be concentrated in the event of a disaster. It wassubsequently updated in 2000, 2001 and 2002. LandScan was conceived as an effort tocapture ambient population, more than decennial population counts. The differencebetween ambient and resident population is not significant as the results are quite coarse inall available population density maps.
LandScan 2003 was released shortly before this report went to press. In this FAO study,a modified version of LandScan 2002 (LandScan–a) was used, as explained in section 4.2.1(see Map 3.2).
The sources used for the LandScan released in 1998, included DCW, Nighttime Lights,GLCC, high–resolution aerial photography and satellite imagery. The methodology wassubsequently updated and the input layers improved.
In the 2000 version of LandScan, the major improvement was the use of VMap1 (seesection 2.3.1) with its superior identification of the road networks, populated places andwater bodies. In the 2001 version, the major improvement was better information aboutsecond order administrative boundaries for population distribution outside the UnitedStates; and, within the United States, newly-available high-resolution (30 metre) land coverdata products. In 2002, refinements were made to the algorithm for its population modelsand MODIS land cover database was used as an input data sources.
The LandScan methodology consists in an automated procedure to allocate populationdata to 30 arc-second cells, which correspond to approximately 1 square kilometre at theequator. The population estimates used as inputs are based primarily on aggregate data forsecond order administrative units compiled by the International Programs Center of theUS Bureau of Census and represent the most recent census information for each country.These population counts are allocated to the individual 30 arc-second cells through a‘smart’ interpolation method that assesses the relative likelihood of population occurrencein cells on the basis of road proximity, slope, land cover, and Nighttime Lights. Probabilitycoefficients are assigned to every value of each input variable, and a composite probabilitycoefficient is calculated for each LandScan cell. The coefficients for all regions are based onthe following factors:
" Roads, weighted by distance from major roads.
" Elevation, weighted by favourability of slope categories.
" Land cover, weighted by type with exclusions for certain types.
" Nighttime Lights of the World, weighted by frequency. The resulting coefficients are weighted values, independent of census data, which can
then be used to apportion shares of actual population counts within any particular area ofinterest. Coefficients vary considerably from country to country even within differentregions of the same country.
]M
AP
PI
NG
G
LO
BA
L
UR
BA
N
AN
D
RU
RA
L
PO
PU
LA
TI
ON
D
IS
TR
IB
UT
IO
NS
[
17
RE
VIE
W O
F EX
ISTIN
G G
EO
RE
FER
EN
CE
D P
OP
ULA
TIO
N D
AT
ASE
TS
M A P 3 . 2
LandScan Global Population Database, adjusted to UN figure year 2000
Source: Oak Ridge National Laboratories (ORNL), Tennessee, USA
18
Control totals can be based on any administrative unit (whether nation, province, districtor minor civil division) or on any arbitrary polygon for which census data are available. Theresulting population distribution is normalized and compared with appropriate controltotals to ensure that aggregate distributions are consistent with census control totals.
The advantages of LandScan, as compared with GPW, include its better outputresolution of 30 arc-seconds, as opposed to 2.5 arc-minutes, and the use of an extensivemodel to predict population distribution within administrative units. Although LandScantakes urban areas into account, it does not distinguish urban and rural populations in thedatabase. However, the input layers are such that urban areas can be inferred by analysingthe population density.
One problem with LandScan concerns the roads database. The model processes theinput layers by country without taking into consideration the spatial continuity of the roadnetworks between them, resulting in uneven changes of population density at countryboundaries. Another problem is that, owing to the way in which the LandScan processingmethods evolved, population comparisons between available revisions of the database arenot possible. Although each revision date of LandScan represents the adjustedmidyear–July population estimates for that year, comparatively, the available 1998, 2000,2001, 2002 and 2003 releases of these data do not represent a time series that can be usedfor pixel-by-pixel analyses or comparisons (see also Dooley, 2005). Also the underlyingmodels have not been published, so the assumptions employed by LandScan to distributepopulation counts to pixels are not known.
3.3 GLOBAL RURAL URBAN MAPPING PROJECTIn a recent project, CIESIN and partners such as the International Food Policy ResearchInstitute (IPFRI), the World Bank and the Centro Internacional de Agricultura Tropical(CIAT), developed a model for redistributing population within administrative units bycombining data from several sources. The description of the method and the datasets, inthe box, draws on the working paper available at the GPW Web site (Balk et al., 2004a).
]M
AP
PI
NG
G
LO
BA
L
UR
BA
N
AN
D
RU
RA
L
PO
PU
LA
TI
ON
D
IS
TR
IB
UT
IO
NS
[
19
REVIEW OF EXISTING GEOREFERENCED POPULATION DATASETS
What does the GRUMP dataset contain?
Human
settlements
database of about
55 000 settlements
points that have a
population of
1 000 or more
Urban extent
database of over
21 000 areas
A global database of cities and towns (points). Each
point, represented as a latitude/longitude pair, has
associated tabular information on its population and data
sources. Population data were gathered primarily from
official statistical offices (census data) and secondarily
from other sources, such as Gazetteer and City
Population. Based on the data available and applying UN
growth rates, population was estimated for the year
1990, 1995, and 2000. When the records for cities and
town did not include latitude and longitude coordinates,
those were taken from the NIMA database, based on a
city name and administrative units match. As mentioned
earlier, due to uncertainties in the positional accuracy of
the NIMA coordinates, some of the cities and towns
might not be accurately geolocated.
The GRUMP urban mask represents an attempt to delineate
extents associated with human settlements globally. The
physical extents of settlements are derived from both raster
and vector datasets. In particular, the team used the
Nighttime Lights dataset for the period 1994–1995 (Elvidge
et al., 1997, 2001), DCW Populated Places, and cities from
the Tactical Pilotage Charts (standard charts produced by
the Australian Defense Imagery and Geospatial
Organization, at a scale of 1:500 000) for selected countries
in Africa. All the sources of urban extent (night-lights, DCW
polygons and TPCs) were combined in order to obtain the
maximum possible coverage for each country. The
population values are assigned to the physical extents from
points within a three kilometre buffer. For points that are
not within the three kilometres buffer of an extent, circles
were created based on the relationship between
population size and areal extents for the points with known
parameters. These newly created circles were added to the
existing ones to create a complete coverage of urban
extents with population information for each country.
B O X 3 . 1
GLOBAL RURAL URBAN MAPPING PROJECT (GRUMP) DATASET
see next page ➥
20
]M
AP
PI
NG
G
LO
BA
L
UR
BA
N
AN
D
RU
RA
L
PO
PU
LA
TI
ON
D
IS
TR
IB
UT
IO
NS
[]
[
Urban–rural
population grid,
with an output
resolution of 30
arc-seconds
The urban-rural population grid was created by using a
mass-conserving algorithm called GRUMPe (Global Rural
Urban Mapping Programme), developed by CIESIN, that
reallocates people into urban areas, within each
administrative unit. In particular the data inputs are the
administrative polygons, containing the total population for
each admin unit, and the populated urban extents. The
reallocation process works iteratively so that the output
urban and rural proportions match, when possible, the UN
ones. Although the UN totals are useful as a benchmark, in
some cases the GRUMP output proportions have not been
matched to the UN ones (when for example CIESIN’s data
includes many more small settlements than those
corresponding to the urban threshold given by the country).
The main advantage of GRUMP is that it uses population data from the census,
rather than predicting it based on probability coefficients or lighted areas.
Also, it makes use of other GIS data to identify urban areas, compensating for
the small settlements in poor countries that are not detected by the Nighttime
Lights. The resulting grid is a dataset at moderate resolution that represents a
more accurate distribution of human population than the existing datasets,
and that makes explicit reference to urban and rural areas.
What are GRUMP’s main advantages?
The lights are known to overestimate the actual extents of urban areas (Elvidge et
al., 2004), but, as previously discussed, applying a threshold would reduce the
number of small settlements that are not frequently lit, as in developing countries.
Given the complexity of finding a single threshold that could work globally (Small
et al., 2005), no light threshold was applied, resulting in an overestimation of the
urban extents in some parts of the world. Although population is estimated for
three time periods (1990, 1995, and 2000), users need to remember that the lights
refer to one point in time only (the 1994/1995 time period), so it would not be
advisable to use these extents for any analysis of change in urban areas.
These data provide the first systematic assessment of the world’s urban land area
– nearly three percent (Balk et al., 2004a), and how population distributions by
ecosystems differ dramatically. Coastal zones are the most urban of all systems,
and sustain the highest population densities, not only in the urban areas, but in
the rural ones as well. The GRUMP grid is one of the key input datasets in the
Millennium Ecosystem Assessment (McGranahan et al., 2005).
What are GRUMP’s main limitations?
21
RE
VIE
W O
F EX
ISTIN
G G
EO
RE
FER
EN
CE
D P
OP
ULA
TIO
N D
AT
ASE
TS
M A P 3 . 3
Population density in 2000 from GRUMP adjusted to UN totals
Source: Center for International Earth Science Information Network (CIESIN), Columbia University; International Food Policy Research Inst. (IPFRI), the World Bank andCentro Internacional de Agricultura Tropical (CIAT)
22
3.4 POPULATION DATABASES FOR AFRICA, ASIA AND LATIN AMERICAPopulation databases for Africa, Asia and Latin America, compiled by the United NationEnvironment Programme (UNEP) and partners (CIAT and CIESIN), build on the GPWtradition but take road networks and populated places into account in the redistribution ofpopulation (Web site ref. 16)
As described in the documentation (Deichmann, 1996a; Hyman et al., 2000; Nelson,2004), a model was created in the following stages. First, information about thetransportation network and urban centres was collected. The transportation networkincluded roads, railroads and navigable rivers using data from DCW, the World BoundaryDatabank II, and Michelin paper maps, while information about urban centres consists oflocation and size of towns and cities from the human settlements database of GRUMP.This information was then used to compute a simple measure of accessibility for each nodein the network. This measure is the so-called population potential, which is the sum of thepopulation of towns in the vicinity of a given node weighted by a function of distance,using network distances rather than straight-line distances. The computed accessibilityestimates for each node were subsequently interpolated onto a regular raster surface. Asimple inverse distance interpolation procedure was used, which resulted in a relativelysmooth surface. Raster data for inland water bodies (lakes and glaciers), protected areasand altitude were then used to adjust the accessibility surface heuristically. Finally, thepopulation totals estimated for each administrative unit were distributed in proportion tothe accessibility index measures estimated for each grid cell. The input administrative units,with corresponding population numbers, are the same as those of GPW. The outputresolution, as for GPW, is 2.5 arc-minutes.
This model undoubtedly represents an improvement upon GPW, in that it takes intoaccount road networks and populated places to achieve a better reallocation of populationwithin administrative units. Unlike LandScan, only roads and populated places are used,and there is no explicit effort to capture the ambient quality of the LandScan approach. Theresolution might still be too coarse for detailed studies at the local/national level, but itprovides consistent population distributions across continents, allowing analysis at theregional scale.
3.5 OTHER RESEARCH EFFORTS TO MAP URBAN POPULATIONIn this section, other recent attempts to model population distribution are described. Thefirst two use GPW as base population input and additional georeferenced datasets, whilefor the third the starting point is country-level demographic statistics. The first is work inprogress, and is not available publicly.
The first one was conducted by CIESIN, in a parallel effort to the GRUMP database.CIESIN pursued a method for improving on the GPW by using the Nighttime Lightsdataset to identify urban areas (Pozzi et al., 2003). The project aimed to overcome some ofthe limitations of LandScan (extensive modelling), GPW (lack of modelling) and GRUMP(extensive data collection) by developing a simple model to redistribute population withinadministrative units according to human settlements. Human settlements are identified bythe Nighttime Lights dataset produced for the year 1994/1995 (Elvidge et al., 1997, 2001).
]M
AP
PI
NG
G
LO
BA
L
UR
BA
N
AN
D
RU
RA
L
PO
PU
LA
TI
ON
D
IS
TR
IB
UT
IO
NS
[
23
The reallocation of population within administrative units is based on a function derivedfrom the relationship between the population density and Nighttime Light frequency fora sample of regions of the world with spatially detailed administrative areas. The result isspatial refinement in areas or countries with relatively large populations but poor spatialdetail for administrative boundaries. As the identification of urban areas is based solely onNighttime Lights, in countries with poor lights coverage (for instance in Africa) theaccuracy of the reallocation may not be very precise.
The second effort was conducted at the Department of Geography and Center ofRemote Sensing at Boston University, as part of a larger project to map global land coverfrom MODIS data (Web site ref. 17). The authors present a method for mapping urbanland cover at spatial resolution of one kilometre by fusing multiple sources of coarseresolution data (Schneider et al., 2003). The objective was to determine the boundaries andthe extents of urban areas more accurately. Population density data were used as one of thesources for determining probable location of urban areas, but no effort was made toactually estimate urban population counts. Two major tasks were involved in this study.First, a supervised decision tree classification method was developed by fusing onekilometre MODIS data and two ancillary sources: the Nighttime Lights data (Elvidge et al,1999) and population density data (GPW, see Tobler et al., 1995; Deichmann et al., 2001).The second task was to establish the best means for evaluating the accuracy of urban landcover maps produced over large regions, an issue that is especially problematic when theclass of interest is a small fraction of the total area mapped. For most parts of the world,multiple data sources were fused to achieve the results. The fusion of these three data typesimproves urban classification results by resolving confusion between urban and otherclasses that occurs when any one of the data sets is used by itself.
For Africa, the ancillary data were too problematic, and Africa was successfullymapped with MODIS data alone. Any city around the globe larger than a few squarekilometres should be represented, barring those areas (such as the majority of the Congobasin) that have continuous cloud cover. In addition, the scale of cities in developingcountries is quite different from the rest of the world, so that most small cities in Africa,India and China (which might only be one pixel) are not represented (Schneider, personalcommunications).
The third project is part of the World Water Development Report II Indicators forWorld Water Assessment Programme (Web site ref. 18). The University of NewHampshire Water Systems Analysis Group has developed a compendium of Earth Systemand socio-economic databases describing the current state of global water resources,including associated human interactions and pressures. Global population fields wereconstructed for the year 2000 using country-level demographic statistics contained in theWorld Resources Institute (WRI) Earth Trends database. The urban and rural populationdata sets were developed by spatially distributing the WRI 2000 country-level urbanpopulation data among DMSP-OLS nighttime stable-lights imagery (Elvidge et al., 1997a)and ESRI Digital Chart of the World populated places points. Country-level urbanpopulation was evenly distributed among the DMSP-OLS city lights data set at one-
REVIEW OF EXISTING GEOREFERENCED POPULATION DATASETS
24
kilometre grid cell resolution with detectable lights in at least ten percent of the cloud freeobservations (Elvidge et al, 1997b). Where available, the spatial extents of major citylocations with known demographic data (Tobler et al, 1995) were superimposed in theDMSP-OLS city lights data set to enhance the accuracy of the urban populationdistribution. Rural population was spatially distributed equally among the DCWpopulated places points falling outside of the DMSP-OLS city lights extent. Totalpopulation is simply the sum of urban and rural population data sets gridded to the 30minute simulated topological river network (STN-30) (Fekete et al., 2001).
]M
AP
PI
NG
G
LO
BA
L
UR
BA
N
AN
D
RU
RA
L
PO
PU
LA
TI
ON
D
IS
TR
IB
UT
IO
NS
[
4.1 OBJECTIVESThis chapter describes the method used by FAO/SDRN to develop gridded urban andrural population databases for inclusion in the FIVIMS Global GIS Database (FGGD)(Huddleston et al., 2005). The main difference between this method and the ones reviewedin Chapter 3 is that it allows making of rural population maps in which pixel values reflectvariations not only between subnational units, but also within the units. The method isbased on detecting and masking out urban areas on the LandScan Global PopulationDatabase in order to make a global rural population grid at the same resolution asLandScan, that is at 30 arc-seconds.
This task has been carried out as part of a larger effort within the context of a PovertyMapping Project, implemented jointly by FAO, UNEP and CGIAR to promote the useof poverty maps in policy - making and in targeting assistance, particularly in the areas offood security and environmental management (Web site ref. 19).
Poverty mapping, defined as the spatial representation and analysis of indicators ofhuman well-being and poverty, provides a means for integrating biophysical and geophysicalinformation with socio-economic indicators to provide a more systematic and analyticpicture of human well-being and equity (Henninger and Snel, 2002).
GIS-based analysis of links between environment and poverty would not be possiblewithout gridded databases and maps showing the spatial distribution of the world’s ruraland urban populations at a very high resolution. The gridded rural population databasedeveloped by FAO/SDRN is particularly useful for comparing the distribution of ruralpopulations with available natural resources and other environmental and geophysicalindicators of the degree of vulnerability of rural livelihood systems in developingcountries. In this context, the aim is to identify the spatial distribution of rural populationglobally, so that reasonable estimates of the numbers of people living in different ruralenvironments around the world and within regions and countries can be generated, suchas in different agro-ecological zones, farming systems or crop zones.
Besides describing the method developed to detect the urban population grid cells inthe LandScan global population database and create the urban area mask, this chapter alsopresents results in map and table formats and compares them with other similar databases.
25
C H A P T E R 4 DETERMININGVARIATION IN THEDISTRIBUTION OFURBAN AND RURALPOPULATIONS BYPIXEL
26
4.2 METHODOLOGYThe task of detecting the urban areas was not straightforward because, as discussed insection 2.1, there is no commonly accepted definition of what constitutes an urban area.Indeed, since most humans tend to congregate in settlements, by some definitions almostall people could be said to live in urban areas. But generally, human settlements occurringin areas that are largely agricultural are considered rural, even though the size of theirpopulation may sometimes be quite large. In this report these are referred to as ‘ruralsettlements’ and the population living in these settlements is excluded from analysesreporting rural population. On the other hand, in some countries, particularly those wheretotal population density is not very large, even some small settlements are consideredurban.
To create the gridded urban, rural and rural settlements population databases, fourprimary sources were used. LandScan 2002 was used as the reference database forpopulation distribution. Nighttime Lights of the World 2000 was used to identify theextent of urban areas. UN population data for each country for the year 2000 were takenas the reference point for urban/rural population and for overall totals. Detailedinformation about these three sources was given in previous chapters. In addition, the UN(DPKO/UNCS) International Boundaries/Coastlines map for 2004 was used to delineatethe country boundaries and coastlines (Web site ref. 20).
The reasons behind the choice of LandScan as the reference database for globalpopulation distribution, and the technical steps for generating the urban mask, are given insections 4.2.1 and 4.2.2 respectively.
4.2.1 Choice of population databaseThree global population datasets – GPW, GRUMP and LandScan – were evaluated inorder to choose the most suitable one for this study. GPW, as described in section 3.1, hasa fairly coarse resolution and the population is uniformly distributed within any givenadministrative unit. The GRUMP database has better spatial resolution (30 arc-seconds)and superior differentiation of urban and rural populations, but both are still uniformlydistributed within any given administrative unit, and in any case the database was notavailable at the time of this study. Therefore, LandScan was chosen as the source databasefor global population distribution, because of its high resolution and its depiction ofvariation in population counts also within each administrative unit, rather than showingonly their averages. An additional advantage of the LandScan database is that, although itdoes not provide direct information about urban and rural areas, its population modeldistinguishes urban and rural populations and their distribution.
The ORNL has released five versions of LandScan (see section 3.2). Each version hasincluded new refinements, reflecting improvements in the quality of the data sources aswell as improved data manipulation.
The 2002 version of LandScan has been used as the source for the spatial distributionof the world’s population because the most recent version, for 2003, was not released untilnear the end of this study. Since the LandScan database does not contain administrative
]M
AP
PI
NG
G
LO
BA
L
UR
BA
N
AN
D
RU
RA
L
PO
PU
LA
TI
ON
D
IS
TR
IB
UT
IO
NS
[]
[
27
boundaries, it was overlaid with the standard UN International Boundaries map in orderto delineate national boundaries and populations.
A comparison of UN population database figures for 2002 with LandScan 2002 showedthat, at the global level and for most of the countries, the differences were insignificant. Asthe year 2000 was selected as the reference year for other time-sensitive variables analysed inthis report series, the LandScan 2002 had to be adjusted to year 2000 population estimates.This was deemed to be a more accurate representation than using the LandScan 2000 databaseitself, which used a less refined population distribution model. To reconcile the two sets ofdata, FAO used the UN 2000 population data for the country totals and LandScan 2002 forthe distribution of the population within each country. In other words, for every countryincluded in the database, the total population numbers derived from LandScan 2002 wereadjusted to the UN figure for 2000. The new totals were then distributed across the pixels inthe same proportion as in the original LandScan 2002 database. The adjustment coefficient iscalculated for each country and is the ratio between the total population from the UNdatabase and the total population from the LandScan database using the UN InternationalBoundaries. The result is a 30 arc-second grid of population distribution that is matched tothe UN figures in terms of total population for each country. From here on, we will refer tothis modified LandScan global population database for year 2000 as LandScan-a.
The FAO/SDRN rural and urban population distribution grids have been generated at 30arc-seconds on LandScan-a. Because these databases were developed to analyse the distributionof rural population in relation to environmental and geophysical factors which were availableonly at 5 arc-minute resolution, it was necessary to convert the rural population grid from 30arc-seconds to 5 arc-minutes. However, an analysis of the country area calculations at 5 arc-minute resolution indicated that at that resolution, GIS analysis in countries with areas lessthan 3 000 square kilometres would not be sufficiently accurate. Therefore such countrieswere not included in the analysis, nor were the countries with a UN total population figureless than 500 000. Table 4.1 lists the 154 countries included in the analysis.
4.2.2 Detection of urban areasSeveral methods were explored for determining urban area boundaries and extents thatreturn urban population counts consistent with UN population data for each country. Thesimplest method is to classify all the pixels in LandScan-a with population density above acertain threshold as urban, for instance all pixels with greater than 1 000 persons per squarekilometre. A variant of this method is to establish a unique threshold for each region orcountry. However, this concept was found to be too simplistic for discriminating urbanand rural populations as it produces a very fragmented urban mask.
Another method that was considered was to use a threshold for the gradient of thepopulation density, rather than the population density itself. This method seemed verypromising as differences in population density between many urban and rural areas werequite easily detected. However this method and even its combination with the populationdensity threshold method described above was also not sufficiently accurate in somecountries, and was not pursued.
DETERMINING VARIATION IN THE DISTRIBUTION OF URBAN AND RURAL POPULATIONS BY PIXEL
28
The third method – and the one selected for producing the urban mask – is based ondelineating urban boundaries depending on the intensity of the lights from populatedareas. In satellite images of the globe taken at night, urban areas appear highly lighted. Thecorrelation between these lighted zones and urban areas had already been explored byother researchers (Imhoff et al., 1997; Sutton, 1997; Elvidge et al., 1997). More recently aresearch on the metrics for quantifying the relationships within geospatial datasets hasbeen developed. A spatial cross correlation between population counts in the LandScandatabase and the Nighttime Lights was computed and the two were found to be highlycorrelated (Ganguly A., personal communication).
The main idea is to determine light intensity threshold (LT) for delineating urban areasusing the human settlements dataset of the Nighttime Lights (NTL) of the World for theyear 2000. In this dataset the values indicate the Digital Number (DN) genereted by the
]M
AP
PI
NG
G
LO
BA
L
UR
BA
N
AN
D
RU
RA
L
PO
PU
LA
TI
ON
D
IS
TR
IB
UT
IO
NS
[]
[
Afghanistan Costa Rica Indonesia Morocco SomaliaAlbania Côte d’Ivoire Iran, Islamic Rep of Mozambique South AfricaAlgeria Croatia Iraq Myanmar SpainAngola Cuba Ireland Namibia Sri LankaArgentina Cyprus Israel Nepal SudanArmenia Czech Republic Italy Netherlands SwazilandAustralia Denmark Jamaica New Zealand SwedenAustria Djibouti Japan Nicaragua SwitzerlandAzerbaijan, Dominican Jordan Niger Syrian ArabRepublic of Republic RepublicBangladesh Ecuador Kazakhstan Nigeria TajikistanBelarus Egypt Kenya Norway Tanzania, United
Rep ofBelgium El Salvador Korea, Dem Oman Thailand
People’s RepBenin Eritrea Korea, Republic of Pakistan Timor-LesteBhutan Estonia Kuwait Panama TogoBolivia Ethiopia Kyrgyzstan Papua New Guinea Trinidad and TobagoBosnia and Finland Laos Paraguay TunisiaHerzegovinaBotswana France Latvia Peru TurkeyBrazil Gabon Lebanon Philippines TurkmenistanBulgaria Gambia Lesotho Poland UgandaBurkina Faso Georgia Liberia Portugal UkraineBurundi Germany Libyan Arab Puerto Rico United Arab
Jamahiriya EmiratesCambodia Ghana Lithuania Qatar United KingdomCameroon Greece Macedonia, Romania United States of
The Fmr Yug Rp AmericaCanada Guatemala Madagascar Russian Federation UruguayCentral African Guinea Malawi Rwanda UzbekistanRepublicChad Guinea-Bissau Malaysia Saudi Arabia Venezuela, Bolivar
Rep ofChile Guyana Mali Senegal Viet NamChina Haiti Mauritania Serbia and Yemen
MontenegroColombia Honduras Mexico Sierra Leone ZambiaCongo, Dem Hungary Moldova, Slovakia ZimbabweRepublic of Republic ofCongo, Republic of India Mongolia Slovenia
T A B L E 4 . 1
List of the 154 countries included in the urban and rural databases
29
Operational Linescan System (OLS) satellite monitoring, where the value of DN correlateswith the light intensity on the ground (see section 2.3.2). The range for the DN values isfrom 0–63. These numbers are the average DN values for the year. The minimum valueidentifies a situation of no lights; the maximum of saturated lights..
Initially, it was found that the images generated from the NTL database did not havesufficiently high positional accuracy for a global analysis. There were considerable non-systematic positional shifts, sometime as high as 15 Km when compared to accuratereference maps.
Once the NTL images were geometrically corrected, there was sufficiently goodregistration between the NTL images and the coastlines map used. Figure 4.1 depicts theuncorrected and the corrected images of a very highly populated metropolitan area – thecity of Istanbul, Turkey. The waterway (Bosphorus) in the centre of the image has a widthof approximately 1 000 metres at the narrowest point and was covered with light in theuncorrected image but not in the corrected one. This is indicative of the positionalaccuracy achieved by the geometric correction applied.
It should be noted that in order to use the NTL images for the delineation of the urbanmask two problems needed to be resolved. First, since the lights are more linearlycorrelated with GDP and electrification than with population density (Doll et al., 2000), agiven urban population density will produce lower light intensity where GDP andelectrification are low than where they are high. Second, as mentioned in section 2.3.2, thelights tend to overestimate the actual extents of the urban areas because of the bloomingeffect (Elvidge et al., 2004).
The solution to the first problem required the determination of a specific LT value foreach country. In order to determine the LT value for a given country, first a histogram and
DETERMINING VARIATION IN THE DISTRIBUTION OF URBAN AND RURAL POPULATIONS BY PIXEL
F I G U R E 4 . 1
Geometric correction of Nighttime Lights of the World 2000 to UN international coastlinemap: Istanbul area
30
then a cumulative distribution of population in that country for each DN value weregenerated. This is done by locating rural population figure on the y-axis of the cumulativedistribution (see Figure 4.2) and finding the DN (i.e. the x-axis) value corresponding mostclosely to it. This value of DN corresponds to the LT value which covers all the urbanpopulation given by the UN figures for each country.
Figure 4.2 depicts the procedure above using the UN population figures for Italy as anexample. In the UN figures the total population for Italy is 57 536 000 and the total ruraland urban populations are 19 020 000 and 38 516 000 respectively.
The DN value of 44 is the value on the x-axis that comes closest to the UN figure forrural population on the y-axis. It identifies 18 772 729 as rural population which is adifference of 1.3 percent compared to the UN figure. In the same way it is possible to talkin terms of urban population, that is 38 763 271 with a difference of one percent.
The above differences are due to rather coarse representation of the DN values that is byonly 64 integer values. In most countries the difference was less than ten percent. Asexplained in section 2.1, the definition of urban and rural areas is controversial and thereforethe UN urban and rural population figures could also be considered controversial.Nevertheless, it was considered essential to use some urban and rural population figures as abenchmark and the UN figures were chosen because of their international acceptance.
There is a large variation in the LT values between countries. Figure 4.3 depicts theaverage LT for all the UN regions. All Africa, except the North, required the lowest valueof DN to detect the urban population. Western Asia, Northern America and Japanrequired greater values of DN on average.
]M
AP
PI
NG
G
LO
BA
L
UR
BA
N
AN
D
RU
RA
L
PO
PU
LA
TI
ON
D
IS
TR
IB
UT
IO
NS
[]
[
F I G U R E 4 . 2
Cumulative distribution of population versus DN value in Italy
31
During the analysis of the NTL images, it was noted that in 50 countries there were notsufficient lights to account for UN urban population figures.
Even taking the lowest DN value as the LT, in these 50 countries the estimates of urbanpopulation were far lower than the UN figures and had an error greater than ten percent.Not surprisingly 49 of them are developing countries and only one is developed, Australia(-11 percent). Of the 49 countries, 69 percent are in Africa, 18 percent in Latin America,ten percent in Asia and two percent in Oceania.
Also, in these countries, there appears to be a number of high population pixels inLandScan-a, which could not be detected by the lights. The reason for this could not beexplained as LandScan population distribution model is not available in the public domain.In most cases, they are isolated pixels with high population density outside the lights (seeblue boxes in Figure 4.4), but in some other cases (see circles in Figure 4.4) they appear tobe in the form of agglomerations. In these 50 countries those pixels were classified as urbanif they had the same or greater population values than the urban population density of thecountry.
Regarding the problem of the blooming effect of the lights for the actual extents of theurban areas, even if the urban population of a country were close to the UN figure, therecould be some local overestimation of extents. That is, some scarcely populated pixels
DETERMINING VARIATION IN THE DISTRIBUTION OF URBAN AND RURAL POPULATIONS BY PIXEL
F I G U R E 4 . 3
Average light threshold (LT) value by UN region
*See section 4.3 for the description of the UN regions.
within or near urban areas can have high nighttime lights. In order to eliminate these pixels,for each country where there was not underestimation, the urban population density wascalculated and all the pixels in which the number of people was less than ten percent of thisvalue were reclassified as rural. Finally, a 3x3 majority filter was applied to the urban maskto reduce the fragmentation. The procedure for generating the urban mask is illustrated inFigure 4.5 for the urban agglomeration of Johannesburg. The red pixels (a) indicate theareas detected by LT for South Africa. Pink areas (b) show the agglomeration afterremoving the pixels with population less than ten percent of the urban density of thecountry. The blooming effect is reduced considerably, but some areas were too fragmented.The blue pixels (c) indicate the final result after the application of the majority filter.
Application of the procedures described above generated an urban mask, called PovertyMapping Urban extents (PMUe), which was then used as a tool for deriving the rural andurban population distribution grids.
4.2.3 Derivation of population distribution gridsAll the pixels of LandScan-a corresponding to the urban extents grid generated the PovertyMapping Urban population (PMUp) distribution grid at 30 arc-seconds. The PovertyMapping Rural population (PMRp) distribution grid at 30 arc-seconds was defined bymasking out the detected urban extents from LandScan-a. These are population distributiongrids, where the value for each pixel represents the number of persons found on that pixel. It
32
]M
AP
PI
NG
G
LO
BA
L
UR
BA
N
AN
D
RU
RA
L
PO
PU
LA
TI
ON
D
IS
TR
IB
UT
IO
NS
[
F I G U R E 4 . 4
Cameroon: isolated pixels and very small agglomerations not detectable by LT
33
should be noted that this number does not represent persons per square kilometre, as pixelareas vary by latitude. Urban and rural population density grids, called PMUd and PMRd,were also computed by dividing each pixel value by the pixel area.
Since almost all the maps of the Poverty Mapping project are at 5 arc-minute resolution,PMRe and PMRp were also converted to this lower resolution. These grids are denoted withthe acronyms PMRe5 and PMRp5. PMRe5 values indicate the percent area occupied in each5 arc-minute pixel by the rural pixels from the 30 arc-second grid. PMRp5 values are the sumof the rural population numbers on 30 arc-seconds pixels in each 5 arc-minute pixel.
In some countries, rural pixels exhibit very high density. Such pixels were classifieddifferently based on a study carried out by the International Institute for Applied SystemsAnalysis (IIASA) for the identification of cultivated areas. Based on data from China andBangladesh, two countries with very high population density in certain areas, therelationship between population density and the land area required for buildings andinfrastructure was estimated. It was determined that almost negligible land area would beleft for agriculture at areas with population density greater than 2 000 persons per squarekilometre Therefore besides rural and urban classes, a third class, called ‘rural settlement’was created and rural pixels with population density values greater than 2 000 wereassigned to that class. A new mask called Poverty Mapping Rural Settlements extents(PMRSe) was generated for this class. The acronyms PMRSp and PMRSd denote the gridscorresponding to population distribution and density grids respectively, corresponding to
DETERMINING VARIATION IN THE DISTRIBUTION OF URBAN AND RURAL POPULATIONS BY PIXEL
F I G U R E 4 . 5
Different stages in computing the urban mask for Johannesburg and vicinity
(a)
(c)
(b)
34
PMRSe. Subsequently a new grid PMURRS was generated in which each pixel belongs toone of the following three classes: Urban, Rural or Rural Settlement.
All the grids described in this section are part of the Poverty Mapping Urban Rural(PMUR) database.
4.2.4 Crosschecking of the urban population results with the UNfiguresAs noted above the UN urban population figures (UNup) were used as a benchmark forcomputing the PMUp grid. A comparison of the PMUp results with the UNup shows thePMUp results to be within -/+ ten percent of the UN urban population figures in 125countries, which represent about 81 percent of the countries included in the analysis. Ofthese, 32 percent are developed countries, and their percent differences are less than +/-five, with the exception of Australia. The remaining 68 percent are developing countriesand of these almost half of them are in Asia, and the other half is distributed equallybetween Africa and Central/South Americas.
The 29 countries which are not within the -/+ ten percent above, are all developing countries.In 25 of these countries, the PMUp underestimates the UNup. Most of these countries are inAfrica (76 percent), with a further 12 percent in Latin America, and the remaining are Mongoliaand Timor-Leste in Asia and Papua New Guinea in Oceania (Table 4.2).
]M
AP
PI
NG
G
LO
BA
L
UR
BA
N
AN
D
RU
RA
L
PO
PU
LA
TI
ON
D
IS
TR
IB
UT
IO
NS
[
Country UNup PMUp Percent difference PMUp - UNup
population in thousands percentage
Benin 2,630 2,316 -12.0 Argentina 32,700 28,737 -12.1 Angola 4,236 3,713 -12.3 Uruguay 3,071 2,636 -14.2 Botswana 846 724 -14.4 Perú 18,885 16,116 -14.7 Timor-Leste 53 44 -16.3 Cameroon 7,395 6,175 -16.5 Djibouti 559 466 -16.6 Madagascar 4,710 3,913 -16.9 Papua New Guinea 928 758 -18.3 Tanzania, United Rep of 11,236 9,019 -19.7 Namibia 584 463 -20.7 Mozambique 5,735 4,498 -21.6 Burundi 561 440 -21.6 Mongolia 1,415 1,061 -25.0 Congo, Republic of 2,254 1,648 -26.9 Mali 3,594 2,571 -28.5 Guinea-Bissau 431 297 -31.1 Gabon 1,024 668 -34.8 Burkina Faso 1,967 1,192 -39.4 Central African Republic 1,530 876 -42.8 Mauritania 1,527 811 -46.9 Sierra Leone 1,618 810 -50.0 Liberia 1,321 602 -54.4
T A B L E 4 . 2
List of the countries with underestimated urban populations
35
The countries in which PMUp overestimates compared to the UNup are only four: Niger,Haiti, Kenya and Afghanistan (Table 4.3).
Map 4.1 shows the geographic distribution of the countries for which the PMUp - UNupdifference is more than ten percent.
In general the countries where there is under and over estimation are more rural. In thecountries where the PMUp estimates are within ten percent of the UNup, the average urbanpopulation share is about 75 percent for developed countries, and about 40 percent in thedeveloping ones. In the countries, where the differences are greater than ten percent, theaverage urban population share is not quite 40 percent for the underestimated ones and 33percent for the overestimated.
DETERMINING VARIATION IN THE DISTRIBUTION OF URBAN AND RURAL POPULATIONS BY PIXEL
Country UNup PMUp Percent difference PMUp - UNup
population in thousands percentage
Kenya 10,194,000 11,577,200 13.57Niger 2,209,000 2,447,335 10.79Haiti 2,857,000 3,254,837 13.92Afghanistan 4,680,000 5,891,543 25.89
T A B L E 4 . 3
List of the countries with overestimated urban populations
36 ]M A P P I N G G L O B A L U R B A N A N D R U R A L P O P U L A T I O N D I S T R I B U T I O N S[
M A P 4 . 1
Spatial distribution of the difference in urban population figure by country
37
4.3 COMPARISON OF THE PMUE WITH OTHER SIMILAR DATABASESIn this section, the PMUe results have been compared with similar data from other sourcesfor the land area of urban extents in square kilometre and for the geographic coordinatesof urban centers. This cannot be considered a measure of the accuracy of the PMUe resultsbecause of the conceptual problems for defining urban areas noted before. First there is nota clear and unique definition of an urban area that is applicable to all countries (see section2.1). Furthermore there is a lack of non-controversial global statistical data aggregated atsufficiently high spatial resolution with accurate geographic coordinates and populationfigures (see section 2.2). However, the comparisons do generally confirm the validity of thePMUe.
The reminder of this section describes the results of the comparisons, aggregated byUN region and/or continents.
UN classification of regions (UN, 2002) is depicted in Map 4.2.
DETERMINING VARIATION IN THE DISTRIBUTION OF URBAN AND RURAL POPULATIONS BY PIXEL
M A P 4 . 2
UN classification of the world in regions
38
4.3.1 Comparison of the urban land area resultsComparison of the urban land area results of PMUe, GRUMP and Boston UniversityUrban Area (BUUA) databases are listed in Table 4.4 by UN region. Generally theGRUMP data produced by CIESIN yielded the largest extents. However, in 49 of thecountries analysed the PMUe results were greater than the GRUMP urban extents and inseven of twenty regions (three in Africa, one in America, one in Europe and two in Oceania)the PMUe identified a larger urban area than GRUMP. The land cover class called ‘built-uparea’ detected by BUUA was the smallest of the three databases, except in Japan.
]M
AP
PI
NG
G
LO
BA
L
UR
BA
N
AN
D
RU
RA
L
PO
PU
LA
TI
ON
D
IS
TR
IB
UT
IO
NS
[
Continent Region PMUe GRUMP BUUA Database withextents built-up area greater urban area
Africa Eastern Africa 44,839 30,228 9,870 PMUeMiddle Africa 17,059 16,402 3,486 PMUeNorthern Africa 55,462 81,379 15,360 GRUMPSouthern Africa 21,285 49,873 10,458 GRUMPWestern Africa 60,416 39,482 13,566 PMUe
Americas Caribbean 20,327 26,629 989 GRUMPCentral America 44,787 121,251 8,683 GRUMPNorthern America 292,374 885,444 124,510 GRUMPSouth America 425,846 372,434 42,221 PMUe
Asia Eastern Asia 266,398 283,229 101,554 GRUMPJapan 26,438 104,210 52,067 GRUMPSouth-central Asia 153,717 349,989 85,313 GRUMPSouth-eastern Asia 83,927 108,044 17,603 GRUMPWestern Asia 74,240 141,586 27,405 GRUMP
Europe Eastern Europe 217,942 299,381 68,212 GRUMPNorthern Europe 89,668 156,289 21,263 GRUMPSouthern Europe 49,435 194,572 48,933 GRUMPWestern Europe 225,256 179,379 52,797 PMUe
Oceania Australia and New Zealand 42,123 44,601 8,990 GRUMPMelanesia 1,957 1,194 50 PMUeDeveloping countries 1,270,260 1,621,720 336,558 GRUMPDeveloped countries 943,236 1,863,876 376,772 GRUMPWorld Total 2,213,496 3,485,596 713,330 GRUMP
T A B L E 4 . 4
Comparison of urban area by UN regions (km2)
39
DE
TE
RM
ININ
G V
AR
IAT
ION
IN T
HE
DIST
RIB
UT
ION
OF U
RB
AN
AN
D R
UR
AL P
OP
ULA
TIO
NS B
Y P
IXE
L
0%
Africa
East
ern
Afr
ica
Mid
dle
Afr
ica
Nor
ther
n A
fric
a
Sout
hern
Afr
ica
Wes
tern
Afr
ica
Carib
bean
Cent
ral A
mer
ica
East
ern
Asi
a
Jap
an
Sout
h-ce
ntra
l Asi
a
Sout
h-ea
ster
n A
sia
Wes
tern
Asi
a
East
ern
Euro
pe
Nor
ther
n Eu
rope
Sout
hern
Eur
ope
Wes
tern
Eur
ope
Aus
tral
ia a
ndN
ew Z
eala
nd
Mel
anes
ia
Nor
ther
n A
mer
ica
Sout
h A
mer
ica
America Asia Europe Oceania
5%
10%
15%
20%
25%
30%
% of urban area for PMUe
% of urban area for GRUMP
% of urban area for BUUA
F I G U R E 4 . 6
Histogram of the share of urban area in total area, by UN region
Figures 4.6, 4.7 and 4.8 compare the share of the urban area in total area by UN region,by continent and by developed/developing country. As expected, developed countries aremore urbanized than developing countries, with the largest difference in GRUMP.In conclusion, the urban land area of the world estimated by PMUe and GRUMP are 1.7and 2.7 percent respectively. The BUUA figures are much smaller, only 0.5 percent.
40
]M
AP
PI
NG
G
LO
BA
L
UR
BA
N
AN
D
RU
RA
L
PO
PU
LA
TI
ON
D
IS
TR
IB
UT
IO
NS
[
F I G U R E 4 . 7
Comparison of the share of urban area in total area, by continent
0,0%
0,5%
1,0%
1,5%
2,0%
2,5%
3,0%
3,5%
4,0%% of urban area for PMUe % of urban area for GRUMP % of urban area for BUUA
F I G U R E 4 . 8
Comparison of the share of urban area in total area by developed/developing country
Developed Countries
% of urban area for PMUe
Developing Countries
0,0%
0,5%
1,0%
1,5%
2,0%
2,5%
3,0%
3,5%
4,0%% of urban area for GRUMP % of urban area for BUUA
41
Figure 4.9 compares boundaries of four urban extents defined by PMUe, GRUMP andBUUA. Three of these examples show the generally larger extents generated by GRUMPcompared to PMUe and BUUA for most urban areas, both the agglomerations and thesmaller settlements. One example from South America shows a larger extent estimated byPMUe, as was typical for that region.
4.3.2 Evaluation of the geographic coordinates of the humansettlements in the PMUe databaseAs noted in section 3.3, the GRUMP human settlements database contains the geographiccoordinates and the estimated population for the years 1990, 1995 and 2000 for eachhuman settlement. Therefore it could be used for comparing the geographic coordinates ofthe human settlements detected in PMUe.
DETERMINING VARIATION IN THE DISTRIBUTION OF URBAN AND RURAL POPULATIONS BY PIXEL
F I G U R E 4 . 9
Visual comparison of urban extents
(a) Atlanta, Georgia, USA (b) Mexico City, Mexico
(c) Santiago, Chile
Global Rural Urban Mapping Project (GRUMP)
Poverty Mapping Urban extents (PMUe)
Boston University Urban Area (BUUA)
Rural Area
(d) New Delhi, India
42
Using the original version of the PMUe (i.e. without a buffer around the humansettlements), it was possible to detect about 72 percent of the GRUMP settlements globally;these held 88 percent of the population living in human settlements in 2000.
The lowest detection percentage was again in Africa, where the results were about 60percent, containing 86.5 percent of the population.
In order to ascertain whether this relatively poor detection performance was merely theresult of a slight difference in the positioning of the urban area and the points, or was causedby a more serious problem, a buffer of one kilometre around the PMUe human settlements wasgenerated. With this buffer the detection results improved considerably. Globally, 92 percent ofthe GRUMP human settlements were captured, corresponding to 97 percent of the population(Table 4.5).
With the buffer, the improvement in terms of the detected points was almost 30 percent, butin terms of population, the increment was only nine percent, as the buffer pixels were nothighly populated. However the size of the global urban population not captured was very small– around three percent.
]M
AP
PI
NG
G
LO
BA
L
UR
BA
N
AN
D
RU
RA
L
PO
PU
LA
TI
ON
D
IS
TR
IB
UT
IO
NS
[
Continent Region N. of human Estimated Percentage Percentage settlements population of human of population
in grump for the human settlements detected withsettlements detected PMUe
in 2000 with PMUe
Africa Eastern Africa 588 44,455,854 79.3 94.9Middle Africa 232 29,202,553 83.2 96.2Northern Africa 504 72,405,126 90.1 96.3Southern Africa 344 30,502,302 95.6 99.0Western Africa 711 68,768,363 90.0 95.9
Americas Caribbean 518 20,326,012 70.7 91.6Central America 994 93,845,559 97.9 99.2Northern America 13,014 226,648,808 96.3 99.2South America 6,810 283,398,079 90.2 97.7
Asia Eastern Asia 2,205 246,255,414 82.9 93.2Japan 863 103,854,253 93.4 95.4South-central Asia 3,095 366,132,233 90.2 96.9South-eastern Asia 851 145,445,943 93.8 95.6Western Asia 888 105,620,869 91.0 98.0
Europe Eastern Europe 2,602 173,285,465 91.0 99.3Northern Europe 2,057 59,720,165 93.2 98.4Southern Europe 3,040 96,234,237 88.3 96.6Western Europe 2,912 129,331,298 95.5 99.5
Oceania Australia and New Zealand 324 21,387,165 97.5 99.7Melanesia 24 635,565 100.0 100.0Developing countries 17,764 1,506,993,872 89.0 96.4Developed countries 24,812 810,461,391 94.3 98.4World Total 42,576 2,317,455,263 92.1 97.1
T A B L E 4 . 5
The human settlements in GRUMP database detected by PMUe with one kilometre buffer
During the past decade a number of important advances have been made in GIStechnologies that have allowed demographers and GIS experts, working together, to beginto map the spatial distribution of urban and rural populations globally. This report hasreviewed the results of these efforts to date, and presented a new method for mappingvariations in the distribution of rural populations by pixel. The unique contributions ofthis new method are:
" creation of urban and rural population distribution grids (both pixel counts anddensities), that assign values to each pixel approximating the actual number ofpeople living in that location;
" creation of a population grid for rural settlements, i.e., human settlements that arenot classified as urban but where the population density is too high for them to beconsidered mostly agricultural.
Both results are important and were required for achieving FAO’s objectives for thePoverty Mapping project. Nevertheless, a number of issues and challenges remain, that, ifresolved, would permit an even more refined analysis of the spatial distribution of thehuman population around the globe.
The first issue, commented upon several times already, is the lack of standarddefinitions for what constitutes an urban area, and the criteria used for distinguishingurban from rural population. A second issue relates to the imprecision of the datasetscurrently available for determining the location of human settlements and their extents(DCW, NIMA points database, Nighttime Lights, GRUMP). Another relates to the lackof comparability of statistical data from different countries and sources, and the need torely on statistical estimation procedures to create time series.
These issues have limited the ability of researchers to validate their results, as noindependent source exists that could serve this purpose. Until now, validation efforts forpopulation distribution grids have been limited to crosschecking results with populationtotals reported by the UN (in the case of GPW, GRUMP, PMUR) or by official sources(in the case of LandScan). For urban extents, the Demographic Health Survey (DHS)points have been used to validate the location and extents of urban areas generated by theNighttime Lights at country level for some countries, but the DHS coverage is not global.
Some of the more promising approaches for resolving these issues include:
" georeferencing of census and survey data at the time of collection and introducingdata collection procedures that allow recording of results for lower leveladministrative units before aggregating to the national level;
43
C H A P T E R 5 UNRESOLVED ISSUES AND FUTURECHALLENGES
44
" developing and publicly releasing models such as that used by LandScan todistribute population counts by pixel, and validating the results with a samplingframe and on-the-ground field surveys;
" further improvement in the quality and accuracy of the Nighttime Lights databasesand images;
" reliance on medium and high resolution images now available from MODIS andother imaging satellites that can detect urban areas more reliably.
One of the most pressing challenges of our time – the reduction and eventual eliminationof poverty and hunger from the globe – cannot be effectively addressed without accurateknowledge about who the poor and hungry are, where they live, and what factors present intheir immediate surroundings are contributing to their distress. Mapping the spatialdistribution of the global population is an essential tool for generating this knowledge;continued effort to resolve remaining challenges will be required to obtain full benefit fromthis potentially powerful tool.
]M
AP
PI
NG
G
LO
BA
L
UR
BA
N
AN
D
RU
RA
L
PO
PU
LA
TI
ON
D
IS
TR
IB
UT
IO
NS
[
45
REFERENCES
Sources and notes
Map 2.1 The Nighttime Lights of the World superimposed on bathymetry (segment)Source: National Oceanic and Atmospheric Administration (NOAA).
Web site http://dmsp.ngdc.noaa.gov/html/download_world_change_pair.htmlNotes: The database has a resolution of 30 arc-seconds.
Map 2.2 The Global Land Cover, 2000Source: Global Vegetation Monitoring Unit of the Joint Research Center (JRC), 2000.
Global Land Cover of the Year 2000, Ispra (VA), Italy.Web site www-gvm.jrc.it/glc2000/
Notes: Copyright European Commission, 2004.
Map 3.1 Population density in 2000 from GPWv3 adjusted to UN totalsSource: Center for International Earth Science Information Network (CIESIN), Columbia
University; and Centro Internacional de Agricultura Tropical (CIAT), 2004. Gridded Population of the World (GPW), Version 3, Palisades, NY, USA. Web site http://beta.sedac.ciesin.columbia.edu/gpw/global.jsp
Notes: GPW grid is a population database at 2.5 arc-minutes resolution.
Map 3.2 The LandScan Global Population Database, 2002Source: Oak Ridge National Laboratories (ORNL), 2002.
LandScan 2002 global population database, Oak Ridge, TN, USA.Web site www.ornl.gov/sci/gist/landscan/
Notes: LandScan 2002 dataset is a worldwide ambient population database compiled on a 30arc-second grid.
Map 3.3 Population density in 2000 from GRUMP adjusted to UN totalsSource: Center for International Earth Science Information Network (CIESIN), Columbia
University; International Food Policy Research Institute (IPFRI), the World Bank; andCentro Internacional de Agricultura Tropical (CIAT), 2004. Global Rural-Urban Mapping Project (GRUMP): Gridded Population of the World,version 3, with Urban Reallocation (GPW-UR), Palisades, NY, USA.Web site http://beta.sedac.ciesin.columbia.edu/gpw/ global.jsp
Notes: GRUMP dataset is a population database at 30 arc-seconds resolution.
Map 4.1 Spatial distribution of the difference in urban population by countrySource: results of the analysisNotes: the map was made by FAO-SDRN GIS Unit following the results of the comparison
the UN figure with the PMUp estimates.
Map 4.2 UN classification of the world in regionsSource: United Nations. 2002. World Urbanization Prospects, the 2001 Revision. United
Nations Publication sales No. E.02.XIII.16Notes: The map was made by FAO-SDRN GIS Unit following the definitions of regions and
major areas contained in the report.
46
Document references
Balk, D. & Yetman, G. 2004. The Global Distribution of Population: Evaluating the Gains inResolution Refinement. Documentation for GPW Version 3 available only athttp://beta.sedac.ciesin.columbia.edu/gpw/docs/gpw3_documentation_final.pdf.
Balk, D., Pozzi F., Yetman, G., Deichmann, U. & Nelson, A. 2004a. The distribution of people andthe dimension of place: methodologies to improve the global estimation of urban extents. Working Paper, CIESIN, Columbia University. Palisades, NY. Documentation for GRUMP also available at http://beta.sedac.ciesin.columbia.edu/gpw/docs/UR_paper_webdraft1.pdf.
Clark, J.I. & Rhind, D.W. 1992. Population Data and Global Environmental Change. Paris,IISC/UNESCO.
Danko, D.M. 1992. The Digital Chart of the World project. Photogrammetric Engineering andRemote Sensing, 58:1125–1128.
Deichmann, U. 1996. A review of spatial population database design and modeling. Technical ReportTR-96-3, National Center for Geographic Information and Analysis, Santa Barbara.
Deichmann, U. 1996a. Asia medium resolution population database documentation. Databasedocumentation and digital database prepared in collaboration with UNEP/GRID Geneva for theUNEP/CGIAR Initiative on Use of GIS in Agricultural Research, National Center forGeographic Information and Analysis, University of California, Santa Barbara.(available also at http://www.na.unep.net/globalpop/asia/index.php3).
Deichmann, U., Balk, D. & Yetman, G. 2001. Transforming population data for interdisciplinaryusages: from census to grid. Documentation for GPW Version 2 available only athttp://sedac.ciesin.columbia.edu/plue/gpw/GPWdocumentation.pdf.
Dobson, J.E., Bright, E.A., Coleman, P.R., Durfee, R.C. & Worley, B. A. 2000. LandScan: a globalpopulation database for estimating populations at risk. Photogrammetric Engineering andRemote Sensing. 66(7): 849–857.
Doll, C.N.H., Muller, J.P. & Elvidge, C.D. 2000. Nighttime imagery as a tool for global mapping ofsocio-economic parameters and greenhouse gas emissions. Ambio, 29(3): 157–162.
Dooley, J.F. 2005. An inventory and comparison of globally consistent GIS databases and libraries.Environment and Natural Resources Series No 19, FAO - Rome.
Elvidge, C.D., Baugh, K.E., Kihn, E.A., Kroehl, H.W. & Davis, E.R. 1997a. Mapping city lightswith nighttime data from the DMSP Operational Linescan System. PhotogrammetricEngineering and Remote Sensing, 63(6): 727–734.
Elvidge, C.D., Baugh, K.E., Kihn, E.A., Kroehl, H.W. & Davis, E.R. 1997b. Relation betweensatellite observed visible-near infrared emission, population, economic activity and electric powerconsumption. Int. Journal of Remote Sensing, 18(6): 1373–1379.
Elvidge, C.D., Baugh, K.E., Dietz, J.B., Sutton, P.C. & Kroehl, H.W. 1999. Radiance calibration ofDMSP-OLS low light imaging data of human settlements. Remote Sensing of Environment, 68:77–88.
Elvidge, C.D., Imhoff, M.L., Baugh, K.E., Hobson, V.R., Nelson, I., Safran, J., Dietz, J.B. &Tuttle, B.T. 2001. Nighttime Lights of the World: 1994–1995. ISPRS Journal of Photogrammetryand Remote Sensing, 56: 81–99.
]M
AP
PI
NG
G
LO
BA
L
UR
BA
N
AN
D
RU
RA
L
PO
PU
LA
TI
ON
D
IS
TR
IB
UT
IO
NS
[
47
Elvidge, C.D., Safran, J., Nelson, I.L., Tuttle, B.T., Hobson, V.R., Baugh, K.E., Dietz, J.B. & Erwin,E.H. 2004. Area and position accuracy of DMSP night-time lights data. Chapter 20 in RemoteSensing and GIS Accuracy Assessment. Eds. R.S. Lunetta and J.G. Lyon, CRC Press, pp. 281–292.
Fekete, B. M., C. J. Vorosmarty, and R. B. Lammers. 2001. Scaling gridded river networks for macroscalehydrology: Development, analysis and control of error. Water Resources Research, 3(77): 1955–1967.
Fischer, G., van Velthuizen, H., Shah, M. & Nachtergaele, F.O. 2002. Global agro-ecologicalassessment for agriculture in the 21st century: methodology and results. IIASA and FAO,Publication RR-02-02.
FAO. 2003. Compendium of agricultural environmental indicators 1989–1991 to 2000. StatisticsAnalysis Service, Statistics Division, Rome.
Henninger, N. & Snel, M. 2002. Where are the poor? Experiences with the development and use ofpoverty maps. WRI and UNEP/GRID-Arendal.
Huddleston, B., Ataman, E., Salvatore, M. & Bloise, M. 2005. A geospatial information frameworkfor analysis of poverty and environment links. Environment and Natural Resources Series, FAO- Rome (forthcoming).
Hyman, G., Nelson, A., Lema, G., Fosnight, G., Singh, A. & Deichmann, U. 2000. Latin Americaand Caribbean Population Database Documentation (available only at http://www.na.unep.net/globalpop/lac/intro.html).
Imhoff, M.L., Lawrence, W.T., Stutzer, D.C. & Elvidge, C.D. 1997. A technique for using compositeDMSP/OLS “city lights” satellite data to map urban area. Remote Sensing of Environment,61(3): 361-370.
Lee, T. F., Miller, S. D., Turk, F. J., Schueler, C., Julian, R., Elvidge, C., Deyo, S., Dills, P. & Wang, S.2004. The Day/Night Visible Sensor aboard NPOESS VIIRS, Proc. of the 13th Conference onSatellite Meteorology and Oceanography, 1.8, American Meteorological Society, Norfolk, VA.
McGrahanan, G., Marcotullio, P., Bai, X., Balk, D., Braga, T., Douglas, I., Elmqvist, T., Rees, W.,Satterthwaite, D., Songsore, J. & Zlotnik, H. 2005. Urban Systems. Chapter 22 in Conditionsand Trends Assessment of the Millennium Ecosystem Assessment. Forthcoming in October 2005.
Nelson, A. 2004. African population database documentation (available only at http://www.na.unep.net/globalpop/africa/Africa_index.html)
Pozzi, F., Small, C. & Yetman, G. 2003. Modeling the distribution of human population with nighttimesatellite imagery and gridded population of the world. Earth Observation Magazine, 12 (4): 24–30.
Oak Ridge National Laboratory (ORNL). 2004. Documentation of LandScan Global Population 1998Database and further releases 2000, 2001, 2002 and 2003 (available only at http://www.ornl.gov/sci/gist/landscan/landscanCommon/landscan_doc.html).
Schneider, A., Friedl, M.A., McIver, D.K. & Woodcock, C.E. 2003. Mapping urban areas by fusingmultiple sources of coarse resolution remotely sensed data. Photogrammetric Engineering andRemote Sensing, 69 (12):1377–1386.
Small, C., Pozzi, F. & Elvidge, C.D. 2005. Spatial analysis of global urban extents from the DMSP-OLS Night Lights. Remote Sensing of the Environment, 96 (3-4): 277-291.
Sutton, P. 1997. Modeling population density with night-time satellite imagery and GIS. Computers,Environment and Urban Systems, 21(3/4): 227–244.
48
Tobler, W.R., Deichmann, U., Gottsegen, J. & Maloy, K. 1995. The global demography project.National Center for Geographic Information and Analysis 95–6, University of California, SantaBarbara, California, 75 pp.
United Nations. 1998. Principles and Recommendations for Population and Housing Censuses.Revision 1. Series M, No. 67, Rev. 1 (United Nations publication, Sales No. E.98.XVII.8).
United Nations. 2002. World Urbanization Prospects, the 2001 Revision. United NationsPublication sales No. E.02.XIII.16
United Nations. 2003. World Population Prospects, the 2002 Revision. United Nations Publicationsales No. E.03.XIII.10
United Nations. 2004. World Urbanization Prospects, the 2003 Revision. United Nations Publicationsales No. E.04.XIII.6
]M
AP
PI
NG
G
LO
BA
L
UR
BA
N
AN
D
RU
RA
L
PO
PU
LA
TI
ON
D
IS
TR
IB
UT
IO
NS
[
49
Web s i te references
1. United Nations Population DivisionWorld Population Prospect: The 2004 Revision Population Database (available at http://esa.un.org/unpp/)
2. U.S. Census Bureau’s International Program CenterStatistical demographic and socio-economic data for 227 countries and areas of the world. (available at www.census.gov/ipc/www/idbnew.html)
3. World Gazetteer (available at www.world-gazetteer.com/)
4. City Population (available at www.citypopulation.de)
5. Environmental Systems Research Institute, Inc. (ESRI)Digital Chart of the World (DCW) (available at www.maproom.psu.edu/dcw/)
6. National Geospatial-Intelligence AgencyVector Smart Map level 0 (available at http://earth-info.nga.mil/publications/vmap0.html)
7. National Geospatial-Intelligence AgencyVector Smart Map level 1 (available at www.mapability.com/info/vmap1_index.html)
8. National Geospatial-Intelligence AgencyGEONet Names Server (GNS) (formerly NIMA points database) (available at http://earth-info.nga.mil/gns/html/index.html)
9. National Oceanic and Atmospheric Administration (NOAA)Database of Nighttime Lights of the World (available at http://dmsp.ngdc.noaa.gov/html/download_world_change_pair.html)
10. Global Land Cover Characteristics (GLCC) dataset(available at http://edcdaac.usgs.gov/glcc/glcc.asp)
11. Global Vegetation Monitoring Unit (GVM). 2004Global Land Cover 2000 database (GLC2000) (available at www-gvm.jrc.it/glc2000/)
12. National Aeronautics and Space AdministrationMOderate Resolution Imaging Spectroradiometer (MODIS) (available at http://modis.gsfc.nasa.gov/)
13. Center for International Earth Science Information Network (CIESIN)Gridded Population of the World version 3 (GPW v3) dataset;Future Estimates 2015 dataset; Data products developed under the Global Urban-Rural Mapping Project (GRUMP).(Downloaded in April 2005)(available at http://beta.sedac.ciesin.columbia.edu/gpw/global.jps)
50
14. EuroGeographicsSeamless Administrative Boundaries of Europe(available at www.eurogeographics.org/eng/04_sabe.asp)
15. Oak Ridge National Laboratory (ORNL)LandScan Global Population Database, 2002(Downloaded in March 2003) (available at www.ornl.gov/sci/gist/landscan/landscan2002/index.html)
16. United Nations Environment Programme (UNEP)Global Resource Information Database(available at http://grid2.cr.usgs.gov/)
17. Boston University’s Department of GeographyUrbanization as a component of global change: a global map of urban areas(available at http://duckwater.bu.edu/urban/modis_map.html)
18. United Nations Educational, Social and Cultural Organization (UNESCO)World Water Development Report II Indicators for World Water Assessment Programme(available at http://wwdrii.sr.unh.edu/index.html)
19. Poverty Mapping Project(available at http://povertymap.net)
20. UN Geographic Information Working Group (DPKO/UNCS)International Boundaries dataset(Downloaded March 2004)(available at http://boundaries.ungiwg.org/ for members of the UN system)
]M
AP
PI
NG
G
LO
BA
L
UR
BA
N
AN
D
RU
RA
L
PO
PU
LA
TI
ON
D
IS
TR
IB
UT
IO
NS
[
CIESIN Columbia University PO Box 1000 Palisades, NY 10964
Please correspond with: [email protected]
Acknowledgments for this annex
We thank Lisa Lukang, Mirella Salvatore, Barbara Huddleston and Ergin Ataman, in particular, for theircontributions. This database and map was prepared as part of the FAO Poverty Mapping Project(GCP/INT/761/NOR) funded by the Government of Norway and also with the support from NationalAeronautics and Space Administration (Contract NAS5-03117) to the Socioeconomic Data andApplications Center (SEDAC) at CIESIN. The data are freely available at the following sites:http://sedac.ciesin.columbia.edu/gpw http://www.fao.org/geonetwork/srv/en/main.searchhttp://povertymap.net/
AnnexEstimates of futureglobal population distribution to 2015Deborah BalkMelanie BrickmanBridget AndersonFrancesca PozziGreg Yetman
DISCLAIMER FOR THIS ANNEX
The designations employed and the presentation of material in this information product do not imply the expression of any opinion whatsoever on
the part of CIESIN or the Food and Agriculture Organization of the United Nations concerning the legal status of any country, territory, city or
area or of its authorities, or concerning the delimitation of its frontiers or boundaries. Ideas contained in this document are solely those of the author
and do not necessarily represent the views of CIESIN and FAO.
COPYRIGHT FOR THIS ANNEX
All rights reserved. Reproduction and dissemination of material in this information product for educational or other non-commercial purposes are
authorized without any prior written permission from the copyright holders provided the source is fully acknowledged. Reproduction of material
in this information product for resale or other commercial purposes is prohibited without written permission of the copyright holders. Applications
for such permission should be addressed to CIESIN or the Chief, Publishing and Multimedia Service, Information Division, FAO, Viale delle Terme
di Caracalla, 00100, Rome, Italy or by e-mail to [email protected].
© FAO and The Trustees of Columbia University in the City of New York 2005
There is considerable interest in the future distribution of human population. The UnitedNations Population Division produces biannual updates to its medium-term projections ofpopulation (UN, 2003) to insure that researchers and policy makers have the most recentinformation upon which to base their analysis and policies. The UN (and otherprojection–making organizations, see review in O’Neill et al., 2001) project population atthe national level only, despite the fact that there is evidence to believe that futurepopulation growth, on average, is more likely to occur in urban areas than rural ones (e.g.UN, 2002). A recent National Research Council study has called for much greaterattention to be paid toward understanding spatial issues in understanding futureurbanization (NRC, 2003). In the near term, however, there are no formal demographicforecasts of population that are spatially explicit. This exercise is a stop–gap measure toaddress a short-term scenario: If the current rates of population growth, as observed in thedecade prior to 2000, continue for 15 years, what would the distribution of populationlook like in the year 2015?
The Gridded Population of the World: Future Estimates, 2015 (GPW2015) providesestimates of the world’s population, by country and continent, for the year 2015 andconverts the distribution of human population from sub-national units to a series of 2.5arc-minute quadrilateral grids. This 2015 data product is entirely derived from the spatialand population input data used to construct the Gridded Population of the World version3 (GPWv3) (CIESIN and CIAT, 2005). This is comprised of administrative boundary andassociated population data.
The 2015 gridded population data was derived from almost 400,000 administrativeunits. For most countries of the world, roughly 75 percent of them, subnational estimatesof population from the two most recent censuses (c. 1990 and 2000) were used as the basisof the extrapolation. Sub-national rates of growth for the 1990–2000 interval were thenapplied, in five–year increments, as described in more detail below. Population estimatesare projected to the year 2015 using the same simple extrapolation methods as the GPWv3and prior GPW databases (Deichmann, Balk and Yetman, 2001; Tobler et al., 1997).
The purpose of our 2015 projection is to show a scenario of future spatial distributionfor the population at a subnational resolution. However, it assumes a continuation ofrecent demographic patterns and is not suitable for generating national population totalsin and of itself. The UN method for projecting population (UN, 2001) follows a cohort-component methodology and incorporates more information about the baselinepopulation (e.g. age structure) and future population trends (e.g. expected fertility and
55
A N N E X
1. INTRODUCTION
56
mortality). Therefore, our 2015 total population estimates are adjusted at the national levelto the United Nations 2015 population projections.
An adjustment factor (A) is applied to our administrative unit population totals(PopNSO2015) via the following calculation, where PopUN2015 represents the UN mediumvariant projected population for 2015:
A=1+
The results of this method are shown in Map 1.1. As in the present, the most denselypopulated places are south and southeast Asia. Similarly, there are expected to be verydensely populated regions of Africa (notably in Nigeria, and east Africa), in Brazil, partsof Central America (including an already dense Mexico City region) and North America(particularly the coastal portions of the urban north east and Los Angeles areas). Europealso continues to be densely populated.
]M
AP
PI
NG
G
LO
BA
L
UR
BA
N
AN
D
RU
RA
L
PO
PU
LA
TI
ON
D
IS
TR
IB
UT
IO
NS
[]]
(PopUN 2015 – PopNSO 2015)PopNSO 2015
57
AN
NE
X- 1
. INT
RO
DU
CT
ION
M A P 1 . 1
Global population density in 2015
Source: Center for International Earth Science Information Network (CIESIN), Columbia University; Food and Agricultural Organization (FAO) and Centro Internacionalde Agricultura Tropical (CIAT)
As previously stated, the input data for the 2015 database mirrors that in GPWv3. Thespecificity of these inputs varies greatly by country due to factors such as: date of mostrecent census, administrative level at which population and spatial data are released, degreeto which the boundaries and population inputs match spatially, the relationship between thenumber of administrative units and the country land area, among other influences. Allinputs are divided between two categories: boundary data and population data, as describedin more detail below.
2.1 BOUNDARY INPUT SOURCESGeographic Information System (GIS) data sets of either administrative or statistical(census) reporting units are produced by national statistical and mapping agencies, researchprojects, and commercial data vendors. GPWv3 relied on a combination of publiclyavailable boundary data sets and additional boundaries from commercial data vendors orstatistical agencies that sell spatial data on license. The level of the spatial inputs utilized inGPWv3 was constrained to the level for which matching population data was available,which varies substantially by country. Levels are commonly ranked from low to high,where the lowest level (level one), refers to the first subnational administrative level belowthe national one, with higher levels representing subsequently finer administrative levelswithin each country.
In general, while there is no consistent pattern between countries with regard to thenumber of administrative units, there tend to be higher levels available for more developedcountries. Differences in administrative levels that can be used to generate our estimates aredue in part, to data availability – i.e., population and spatial inputs for the highest-levelunits are not always available or usable. In addition, the designation of administrative unitsis sometimes ambiguous. Often, administrative units are based on historic boundaries thatare based on geographic and political features that were once historically important butwhich no longer translate to necessarily meaningful divisions. It also should be noted thatfor statistical data-reporting, some countries utilize geographic regions that serve noadministrative purpose and therefore do not match the administrative boundaries. Asdemonstrated by Map 2.1, the number of administrative units included in GPWv3 variesgreatly between countries and is not necessarily proportional to the land area of a nation.
59
A N N E X
2. INPUT DATADESCRIPTION
60
The degree of resolution of administrative units provides a good representation of thisvariability in the number of administrative units. Resolution is calculated as:
Resolution is to some extent determined by the geographic size and average populationdensity of a country. Smaller countries have a relatively higher resolution even beforeadjusting for the number of administrative units. In other words, the national extent of asmall country may already be smaller than an administrative unit of another country.Slovenia is an example of a small country with one of the highest resolutions both becauseof geographic size and number of units. Conversely, many countries with vast, mostlyuninhabited areas tend to have large administrative units resulting in very low resolution(e.g. Mongolia, Libya).
Additionally, the presence of relatively densely distributed populations generallynecessitates a larger number of administrative units than a more sparsely populatedcountry of equivalent size. This results in higher relative resolution. For example, India ismuch more densely populated and has higher resolution than similarly sized, but sparselypopulated Algeria.
Low resolution can be a result of inadequate data, in which higher resolutionadministrative units boundaries exist, but were simply not available for this project. It can
]M
AP
PI
NG
G
LO
BA
L
UR
BA
N
AN
D
RU
RA
L
PO
PU
LA
TI
ON
D
IS
TR
IB
UT
IO
NS
[
M A P 2 . 1
Number of administrative units included in GPWv3, by country
√ (country area) / (number of units)
61
also stem from a combination of data quality, geographic and population density issues. Asa comparison, Mongolia and Saudi Arabia have similarly low resolution, but for differentreasons.
These two countries are similar in geographic size, but Mongolia has approximately onetenth the population size of Saudi Arabia. The measure of average persons per administrativeunit for Mongolia was 108 in 2000, but 672 in Saudi Arabia. Since we would expect higherresolution in more highly populated areas, the data quality for Saudi Arabia is considered tobe inferior to that of Mongolia. For Saudi Arabia, more detailed administrative units wouldhelp considerably in the precise representation of population distribution.
Table 2.1 demonstrates the countries with the highest and the lowest availableresolution (excluding countries and areas smaller than 10,000 square kilometres in size,many of which consist of only one administrative unit).
Within a given country, the mean resolution (across administrative units) dependsconsiderably on a combination of geographic and demographic characteristics, some ofwhich have been described above. Thus, mean resolutions are not always comparablebetween countries. For example, level-three administrative units in Canada can vary froma small, densely populated city-district to large tracts of uninhabited land whereas the sameadministrative level in the continental United States varies much less in area.
By continent, the average level and total number of administrative units used are shownin Table 2.2. There are clear differences, with Europe, Oceania, and North America havinghigher average resolutions. All continents, however, have some countries with high-resolution data, leading to a large number of units for each continent. As compared to thefirst version of GPW, undertaken a decade ago, there is nearly a 20 times improvement inGPWv3.
ANNEX - 2. INPUT DATA DESCRIPTION
10 lowest resolutions Km 10 highest resolutions Km
Saudi Arabia 386 Slovenia 0.01Chad 298 Malawi 1.84Mongolia 265 Switzerland 3.21Angola 264 South Africa 3.54Libyan Arab Jamahiriya 254 France 3.66Svalbard 246 Slovakia 3.87Algeria 219 Ireland 4.09Sudan 171 Portugal 4.49Yugoslavia 159 Indonesia 4.65Botswana 156 Hungary 5.23
T A B L E 2 . 1
Countries with the highest and lowest available resolution
62
2.2 POPULATION INPUT SOURCESPopulation data were collected for each country via national statistical agencies and censusbureaus. The most recent year and most detailed administrative level were acquiredwhenever possible. A large portion of the data was publicly available, however it was alsonecessary to purchase population information for many areas. Population data constraintssuch as censuses occurring in different years and inconsistent data availability result indisparities related to the most recent population data year employed for each country. Thisis illustrated by Map 2.2 below.
]M
AP
PI
NG
G
LO
BA
L
UR
BA
N
AN
D
RU
RA
L
PO
PU
LA
TI
ON
D
IS
TR
IB
UT
IO
NS
[
10 lowest resolutions Mode of administrative Sum of the Average AverageLevels Number of units Resolution Persons per unit
Africa 2 109,172 16 7Asia 2 99,781 18 36Europe 2 98,926 15 7North america 2 74,527 17 6Oceania 1 2,191 62 14South america 2 15,150 34 22World total 2 399,747 18 15
T A B L E 2 . 2
Average level and total number of administrative units, by continent
M A P 2 . 2
Year of the most recent census data available, by country
63
Where possible, two data points were collected as close to the target years of 1990 and2000 as possible. Obviously, the closer the data points were to 1990 and 2000, the lessinterpolation was required.
The greatest source of uncertainty in the dataset occurred in cases where the availablepopulation data was far from the target years, and where only one population data yearwas available.
Countries with only one data point occurred most often in areas where new dataobtained for GPWv3 was at a higher spatial resolution than in past GPW iterations, thusaffecting our extrapolation method (see section 3.2). Map 2.3 displays the number ofpopulation data years employed globally.
ANNEX - 2. INPUT DATA DESCRIPTION
M A P 2 . 3
Number of population data years employed, by country
In the following paragraphs we describe the methodology used to create the GPW2015,both in terms of the gridding approach used to produce the final raster grids, and in termsof the extrapolation methodology used to calculate the population distribution in 2015.
3.1 THE GRIDDING APPROACHThe GPWv3 administrative and population input data were used to produce raster gridsdemonstrating the estimated number of people residing in each grid cell. When theadministrative units are converted to grids it is possible for more than one unit to fall intothe same grid cell and for some units to be smaller than a single grid cell. To ensure that noadministrative information is lost in the gridding process, we implemented a proportionalallocation of population from administrative units to grid cells. Proportional allocationworks on the assumption that the variable being modelled – in this case population – isdistributed evenly over the administrative unit. Grid cells are assigned a portion of the totalpopulation for the administrative unit they fall within, dependent on the proportion of thearea of administrative unit that the grid cell takes up. A simple example of proportionalallocation (also known as areal weighting) would be an administrative unit with apopulation of 5 000 that is filled exactly with 100 grid cells. For this case, each grid cellwould be assigned a population of 50. In the creation of the population grids, the actualimplementation of areal weighting uses the administrative unit’s population density andthe area of overlap between administrative unit and grid cell to calculate each unit’scontribution to the cell population total (further description is given in Deichmann et al.,2001, and a comparison between this and other methods is given in Deichmann, 1996).
3.2 EXTRAPOLATION METHODOLOGYThe methodology for the extrapolation of population data to 2015 is similar to that usedfor extrapolating population data in GPWv3 to 1990, 1995, and 2000. In both instances, thepopulation inputs were collected for the most recent years and smallest sub-national units.The majority of these data were obtained via national censuses or official estimates. For theGPW2015, the official population estimates were then extrapolated forward by computingan average annual geometric growth rate that was then applied to the most recentpopulation data. Because population numbers do not typically rise or fall in a linearfashion, a geometric growth rate was calculated for these estimates.
65
A N N E X
3.METHODOLOGY
66
The formula employed for calculating the growth rate is:
r =
where, LN = the natural log, P1 and P2= population counts for the first and secondreference years, t1 and t2 = time periods 1 and 2.
The forward extrapolations are thereby computed with the following formula:
ert *P1
where, r= the geometric growth rate (as defined above), t= the number of years the initialestimate will be projected forward/backward, P1= population counts for the first referenceyear.
These extrapolations are not meant to be formal projections. As indicated initially, this isan extrapolation method that is commonly used for short-term projections and is nottypically employed for longer-term projections because it lacks information useful for thelonger-term adjustments to population composition and dynamics. The growth rates are heldconstant and the populations are accordingly estimated for 2015 without the aid of additionalinformation. In the next section, we address when and under what circumstances adjustments– beyond that of adjusting the national population totals to the UN medium-run project –were made.
]M
AP
PI
NG
G
LO
BA
L
UR
BA
N
AN
D
RU
RA
L
PO
PU
LA
TI
ON
D
IS
TR
IB
UT
IO
NS
[
LN [(P2/P1)](t2-t1)
4.EXTRAPOLATIONPROBLEMS ANDSOLUTIONS
In a number of instances, outstanding obstacles impeded our use of the abovemethodology for growth rate and projection calculations. Problems were dealt with on acase-by-case basis. Descriptions of the setbacks we encountered and explanations of oursolution procedures are described below.
4.1 IRRECONCILABLE BOUNDARY DIFFERENCESIn general, geographic boundaries are not static. Unfortunately, however, if an administrativeunit changes size or shape between two data years it is impossible to use the above methodto calculate a population growth rate for that particular unit. Thus, when faced withirreconcilable boundary differences between two data years, we implemented a three-tieredapproach for determining a growth rate to be used in the population projections:
(a) Whenever viable, we created hybrids of the administrative unit polygons (and theirassociated population figures) in order to form matching subnational datasets fortwo time periods. In instances where hybrids were created, our administrative unitsdo not match those politically defined by the country of origin, but are still spatiallyand demographically accurate.
(b) If a polygon-based hybrid was impractical, the next step was to consider using acoarser administrative level to calculate the growth rate. For example, if there weresubstantial boundary changes at the second administrative level, but the firstadministrative level remained unchanged, then a growth rate was computed at thefirst administrative level and applied to the higher resolution data.
(c) When neither option (a) nor (b) were feasible, national level growth rates werecalculated using United Nations population estimates and projections (UN, 2001).These rates were then uniformly applied to the most detailed and recent subnationaldata at our disposal. In cases were we suspected the data to be largely erroneous,United Nations derived growth rates were implemented as well.
Map 4.1 illustrates countries for which we used subnational growth rates or, wherenecessary, national growth rates.
67
A N N E X
68
4.2 MIXED ADMINISTRATIVE LEVEL SPATIAL AND POPULATION DATAAdministrative and population data are often collected and released by separategovernmental offices as well as in unconnected years. Because the two types of data are notpublished together, a matched dataset at the lowest available administrative level may beunattainable. In this situation, two potential data scenarios occur:
(a) The population data are at a smaller administrative level than the spatial data. Whenthis situation transpired, it was necessary to aggregate the population data to thecoarser level of the spatial data. As a result, we were unable to use the more detailedlevel of population data and will continue to be incapable of doing so until spatialdata are made available at the same level. For example, if we had populationestimates for the Delaware counties of Kent, New Castle, and Sussex but only hadspatial boundaries for the state of Delaware, it would be necessary to combine thepopulation figures up to the state level.
or(b) The spatial data are at a smaller administrative level than the population data.
Under these circumstances, the population growth rate for the larger unit could beapplied to the smaller spatial units it encompassed. In this scenario, the moredetailed geographic level was maintained. For example, if we only had populationestimates for the state of Delaware but had spatial boundaries for counties of Kent,New Castle, and Sussex, we could calculate a Delaware growth rate and apply thissame rate to each of the three counties.
]M
AP
PI
NG
G
LO
BA
L
UR
BA
N
AN
D
RU
RA
L
PO
PU
LA
TI
ON
D
IS
TR
IB
UT
IO
NS
[
M A P 4 . 1
Number of countries for which sub-national versus national growth rates were used
69
4.3 PARTICULARLY HIGH POPULATION GROWTH RATESLocal area estimates of population are bound to have higher levels of error than largerunits: rapid growth appearing in a small region would be absorbed by estimates of a largerregion; and small area rates of growth may be unlikely to persist in the long-run. Rather,they may be localized in space and time. Even over a ten-year period highly localizedgrowth may not be sustained. Thus, there is an optimal level of the administrative data atwhich to apply rates of growth, neither too coarse nor too fine. In general, if a country hasvery high-resolution data (such as level four or five), we do not use that information as thebasis of the growth rate, rather we use a coarser unit (e.g., counties rather than tracts in theUS) and apply those growth rates for the units that nest beneath it.
We used a benchmark growth rate of five percent, because such a high level of growthis unusual for large administrative units (e.g. countries). Similar benchmarks have beenimplemented by the World Bank in a comparable exercise, in the World DevelopmentReport (WDR, 2002).
(a) If population growth rates were higher than five percent for less than ten percent ofall administrative units in a given country, growth rates were manually set to the fivepercent benchmark for the administrative units concerned.
(b) If population growth rates were higher than five percent for more than ten percentof all administrative units in a given country, we suspected that the data were tooflawed or unreliable to use; and United Nations–derived growth rates wereimplemented as explained in section 4.1c of this annex.
ANNEX - 4. EXTRAPOLATION PROBLEMS AND SOLUTIONS
The Gridded Population of the World: Future Estimates, 2015 is a useful tool in conjunctionwith the UN 2015 projections as it shows a future scenario of the spatial distribution ofpopulations. As already stated outright, this method has limitations for even short-runforecasting. Future investments should include further data development such that morerigorous estimates of future population, along with estimates of associated uncertainty, canbe made at a subnational level.
When shown with urban area extents for 2000, it is possible to see how the urban areasmight grow over the next decade both in spatial extent and in population density comparedto the year 2000. Map 5.1 shows scenarios for select urban areas from CIESIN’s GlobalRural-Urban Mapping Project database. They clearly emerge as much more denselypopulated than surrounding rural areas. Further improvements in resolution to theunderlying population and boundary data will make it possible to gain greater insight inthe expected future population of urban areas, and current and future peri-urban areas.
Data constraints result in varying degrees of accuracy in the projected estimatesbetween countries, making comparisons difficult in some circumstances, particularly forparts of Africa and Asia – two regions of high concern for future urban and ruraldevelopment. Recent investments in more timely, high-resolution, reliable population andboundary data have been made in many countries, such as Malawi, South Africa,Cambodia, Indonesia and Kenya. Using these countries as models for other nations in thesame regions to follow, would go a long way to contributing to regional and global effortsto understand current and future population dynamics in urban and rural areas.
71
A N N E X
5.CONCLUSION
72 ]M A P P I N G G L O B A L U R B A N A N D R U R A L P O P U L A T I O N D I S T R I B U T I O N S[
M A P 5 . 1
Population density projections for the year 2015 with a focus on selected urban areas
73
REFERENCES
Balk, D. & Yetman, G. 2004. The global distribution of population: Evaluating the gains in resolutionrefinement. (available at http://beta.sedac.ciesin.columbia.edu/gpw/docs/gpw3_ documentation_final.pdf)
Deichmann, U., Balk, D. & Yetman, G. 2001. Transforming population data for interdisciplinaryusages: from census to grid (available at: http://sedac.ciesin.columbia.edu/plue/gpw/GPWdocumentation.pdf)
Deichmann, U. 1996. A review of spatial population database design and modeling. Technical ReportTR-96-3, National Center for Geographic Information and Analysis, Santa Barbara.
National Research Council. 2003. Cities transformed: demographic change and its implications inthe developing world. Panel on Urban Population Dynamics, Eds. M.R. Montgomery, R. Stren,B. Cohen, & H.E. Reed, Committee on Population, Division of Behavioral and Social Sciencesand Education. Washington, DC, The National Academies Press.
O’Neill, B., Balk, D., Brickman, M. & Ezra, M. 2001. A guide to global population projections.Demographic Research 4(8): 199–288.
Tobler, W., Deichmann, U., Gottsegen, J. & Maloy, K. 1997. World population in a grid of sphericalquadrilaterals. International Journal of Population Geography, Volume 3, Issue 3, pp. 203–225.
United Nations Department of International Economic and Social Affairs. 2001. World populationprospects: 2000 Revision. Volume 1 [Comprehensive Tables]. New York.
United Nations. 2002. World urbanization prospects, the 2001 revision. United Nations Publicationsales No. E.02.XIII.16
United Nations. 2003. World population prospects, the 2002 revision. United Nations Publicationsales No. E.03.XIII.10
World Bank. 2002. World development report: Sustainable development in a dynamic world.Washington DC.
75
FA
O-
EN
VI R
O
NM
E N TA N D N A T U R A L
R E S OU
RC
ES
SE
RV
IC
E
GEO-SPATIAL DATA AND INFORMATION
ASS
ES
SM
EN
TA
ND
MO
NIT
OR
ING
ENVIRONMENTAL MANAGEMENT
GLO
BA
LE
NV
IRO
NM
EN
TA
LC
HA
NG
E
F A O E N V I R O N M E N T A N D N A T U R A L R E S O U R C E S S E R I E S
1. Africover: Specifications for geometry and cartography, 2000 (E)
2. Terrestrial Carbon Observation: The Ottawa assessment of requirements, status and next
steps, 2002 (E)
3. Terrestrial Carbon Observation: The Rio de Janeiro recommendations for terrestrial and
atmospheric measurements, 2002 (E)
4. Organic agriculture: Environment and food security, 2003 (E and S)
5. Terrestrial Carbon Observation: The Frascati report on in situ carbon data and
information, 2002 (E)
6. The Clean Development Mechanism: Implications for energy and sustainable agriculture
and rural development projects, 2003 (E)*
7. The application of a spatial regression model to the analysis and mapping of poverty,
2003 (E)
8. Land Cover Classification System (LCCS), version 2, 2005 (E)
9. Coastal GTOS. Strategic design and phase 1 implementation plan, 2005 (E)
10. Frost Protection: fundamentals, practice and economics- Volume I and II + CD, 2005 (E)
The FAO Technical Papers
are available through the authorized
FAO Sales Agents or directly from:
Sales and Marketing Group - FAO
Viale delle Terme di Caracalla
00100 Rome - Italy
Availability: March 2005
Ar
C
E
Arabic
Chinese
English
F
P
S
French
Portuguese
Spanish
Multil
*
**
Multilingual
Out of print
In preparation
76
]M
AP
PI
NG
G
LO
BA
L
UR
BA
N
AN
D
RU
RA
L
PO
PU
LA
TI
ON
D
IS
TR
IB
UT
IO
NS
[]
[]
[]
[
ENVIRONMENT AND NATURAL RESOURCES WORKING PAPERS
1. Inventory and monitoring of shrimp farms in Sri Lanka by ERS SAR data, 1999 (E)2. Solar photovoltaics for sustainable agriculture and rural development, 2000 (E)3. Energia solar fotovoltaica para la agricultura y el desarrollo rural sostenibles, 2000 (S)4. The energy and agriculture nexus, 2000 (E)5. World wide agroclimatic database, FAOCLIM CD-ROM v. 2.01, 2001 (E)6. Preparation of a land cover database of Bulgaria through remote sensing and GIS, 2001 (E)7. GIS and spatial analysis for poverty and food insecurity, 2002 (E)8. Enviromental monitoring and natural resources management for food security and sustainable
development, CD-ROM, 2002 (E)9. Local climate estimator, LocClim 1.0 CD-ROM, 2002 (E)10. Toward a GIS-based analysis of mountain environments and populations, 2003 (E)11. TERRASTAT: Global land resources GIS models and databases for poverty and food insecurity mapping,
CD-ROM, 2003 (E)12. FAO & climate change, CD-ROM, 2003 (E)13. Groundwater search by remote sensing, a methodological approach, 2003 (E)14. Geo-information for agriculture development. A selection of applications. (E) **15. Guidelines for establishing audits of agricultural-environmental hotspots, 2003 (E)16. Integrated natural resources management to enhance food security. The case for community-based
approaches in Ethiopia, 2003 (E)17. Towards sustainable agriculture and rural development in the Ethiopian highlands. Proceedings of the
technical workshop on improving the natural resources base of rural well-being, 2004 (E)18. The scope of organic agriculture, sustainable forest management and ecoforestry in protected area
management, 2004 (E)19. An inventory and comparison of globally consistent geospatial databases and libraries, 2005 (E)20. New LocClim, Local Climate Estimator CD-ROM, 2005 (E)21. AgroMet Shell: a toolbox for agrometeorological crop monitoring and forecasting CD-ROM, 2005 (E) **22. Agriculture atlas of the Union of Myanmar (agriculture year 2001-2002), 2005 (E)23. Better understanding livelihood strategies and poverty through the mapping of livelihood assets: a pilot
study in Kenya, 2005 (E)
The FAO Technical Papersare available through the authorized FAO Sales Agents or directly from:
Sales and Marketing Group - FAO
Viale delle Terme di Caracalla
00100 Rome - Italy
Availability: October 2005
Ar
C
E
Arabic
Chinese
English
F
P
S
French
Portuguese
Spanish
Multil
*
**
Multilingual
Out of print
In preparation
Printed on ecological paper
Nations and other sources, and various
georeferenced sources are assessed for
their usefulness to the geospatial analysis of
population distribution. The report examines
two widely used global georeferenced population datasets,
reviews recent methodological developments for
distinguishing urban and rural populations spatially and
presents a method for creating an urban mask and
determining variations in the distribution of urban and
rural populations, by pixel. The report concludes with a
brief discussion of unresolved issues and future challenges.
Finally, the Annex details a method for estimating global
population distribution to the year 2015 using data from
over 375 000 subnational units.
This monograph is part of a series of
reports that explain and illustrate
methods for applying spatial analysis
techniques to investigate poverty and
environment links worldwide. Analysing population
distribution in relation to poverty and environmental
factors is increasingly recognized as a valuable element in
decision-making processes related to development issues.
Accurately mapping and assessing vulnerable populations
can provide a solid basis for recommendations on how best
to reduce poverty and improve living conditions in
developing countries.
In this report, the various definitions of the terms ‘urban’
and ‘rural’ are reviewed, along with data from the United
Environment and Natural Resources Service (SDRN) publicationswww.fao.org/sd/enpub1_en.htm
SDRN contact: [email protected]
Food and Agriculture Organization of the United Nations (FAO)www.fao.org
FA
O-
EN
VI R
O
NM
E N TA N D N A T U R A L
R E S OU
RC
ES
SE
RV
IC
E
GEO-SPATIAL DATA AND INFORMATION
ASS
ES
SM
EN
TA
ND
MO
NIT
OR
ING
ENVIRONMENTAL MANAGEMENT
GLO
BA
LE
NV
IRO
NM
EN
TA
LC
HA
NG
E
