Uwe Deichmann
National Center for Geographic Information and Analysis, Department of Geography,
University of California, Santa Barbara, CA 93106, USA, email: uwe@ncgia.ucsb.edu
I. Introduction
The debate about sustainable agriculture focuses on the question of how to keep up food production in the presence of increasing demand from growing populations in developing countries. These issues have been discussed at length - typically in policy debates about population planning, environmental degradation and agricultural policy (e.g., Ehrlich et al, 1993; see also United Nations, 1994, and Bogue et al. 1993). In light of this debate, the task for the CGIAR and similar institutions is to help increase the food production capacity of agricultural lands to support an expanding population while preserving the resource base upon which agriculture relies. Food supply depends upon two factors: natural resources such as land, water and plants; and human resources employed in food production. Human population thus enters on the demand side, as well as on the supply side as a factor of production. Demographic analysis is thus an integral part of agricultural research touching upon a wide range of topics from the development and targeting of agricultural technology (Dvorak 1993) to the prevention and alleviation of severe food shortages as exemplified by the famine early warning systems supported by FAO or USAID. Information about demographic conditions and processes consequently need to be incorporated in agricultural research along with climatic, soils, vegetation and other socioeconomic factors.
II. Data requirements
It is clearly impossible to anticipate all potential uses of population data for agricultural research. The following is therefore meant to stimulate a discussion which will hopefully lead to consensus regarding the most important variables to be collected in a systematic way.
For demographic analysis in a GIS, two components are required: (1) the spatial reference system - either administrative unit boundaries or point features representing the location of settlements; and (2) the demographic variables of interest that are linked to spatial features. Political boundaries that are the basis for census taking are also used as a reference system for other indicators, such as production data from agricultural censuses, or aggregate education and health statistics. For analysis at smaller cartographic scales, administrative units at the 2nd or 3rd subnational level (for example, districts, municipios or sous-prefectures) are usually sufficient. This will be referred to as medium resolution level as it bridges the gap between national level analysis using the databases compiled by various UN agencies and NGOs, and high resolution data at large cartographic scales such as block level urban databases.
Sources of population data are censuses or sample-based surveys, and are compiled regularly by national statistical offices. The results are generally published within a few years after data collection in the form of census reports or statistical yearbooks. Among the demographic variables collected in censuses, the following list contains those that are the most important for agricultural research. Many socioeconomic characteristics - for example, concerning employment or education - are also important but will not be discussed here.
· distribution and size of the population by administrative unit and/or settlements,
· population by gender and age groups, ideally by 5-year intervals, but at least including the number of children, working age population and elderly. From this information, along with the number of births in the year prior to the census, a number of indirect measures of population dynamics can be estimated, such as birth and fertility rates. Sex ratios by age groups provide a good indirect measure of population mobility (see for example, National Research Council 1993). Low sex ratios - indicating a surplus of women - typically indicate out-migration by the male labor force who work in urban or mining areas. This has consequences for agricultural productivity since labor shortages pose one of the major barriers to increased agricultural productivity;
· rural versus urban population; this indicator, although very important for agricultural research, is often frowned upon by demographers. The reason for this is that definitions of what constitutes an urban settlement vary widely between countries; e.g., from towns with at least 500 people to those of 20,000 or more. Sometimes urban population only refers to the capital, or a functional definition is followed such as "major administrative centers" (see United Nations 1993). It is clearly necessary to design methods of indirectly estimating consistent rural/urban figures for cross-national analysis;
· population movements; migration is a relatively neglected field in demography most likely due to the problem of obtaining direct measurements. Such data are collected using a recall element in the census, but are often considered unreliable. Migration occurs within and among countries. This poses problems in cross-national applications where absent nationals are counted as part of the de jure population in the country of origin, but also as part of the de facto population in the receiving country. A specific aspect of migration are refugee flows due to conflict or environmental crises. These can result in very sudden movements of large numbers of people, and often result in long term displacement. Such sudden population dynamics put a large strain on the resource base of the receiving area and consequently are very important from an agricultural management perspective;
· temporal dynamics are important in explaining current agricultural production patterns and to make informed planning decisions affecting future developments. Ideally one would like to obtain spatially referenced time series of at least the most important demographic parameters such as total population, broken down by male/female and urban/rural residence.
III. Available Data Sets
After discussing the "ideal", minimum demographic data set, the following review of existing data sets may be somewhat disillusioning. Although we have come a long way since Arendal I, there are still very few comprehensive data sets available for developing countries. However, with the virtual explosion of GIS applications in a diverse range of sectors in recent past, many spatially referenced databases have been developed. The challenge is to make these databases compatible, and to expand their resolution, quality and scope in order to obtain the comprehensive "family of population databases" envisaged by Clarke and Rhind (1992). The following survey is by no means complete. A full, up-to-date review would be desirable but is beyond the scope of this paper.
Global data sets
Population data sets at the global level are of primary interest to the global change community, for example, to assess the impacts of CO2 emissions on climatic patterns. Typically, population data need to be converted from the irregular pattern of administrative (political) units to regular raster surfaces used in bio-physical models. There have been several initiatives to produce such data sets that so far resulted mostly in coarse resolution grids. A recently completed project at NCGIA (supported by CIESIN and ESRI) has produced a GIS database of about 19,000 administrative units with associated population totals, and several interpolated raster surfaces of total population and population density by 5' grid cells (Tobler et al 1995). This work will hopefully be continued to improve the resolution, scope and accuracy in areas not well covered in the current database. The raster surfaces are distributed by CIESIN, while the public domain administrative unit boundaries will be extracted and made available through CIESIN, UNEP/GRID, WRI and similar agencies. The quality of the data sets included varies however. What is quite suitable for global level applications may not necessarily be appropriate for high resolution agricultural research.
Continental and regional data sets
Closer to the needs of the CGIAR centers are databases at the continental or regional scale. Several such databases have been developed (some of which were used in the global database):
· CIAT has produced a set of administrative boundaries for Central and South America. Work on South America is still on-going, but most of Central America (excluding Mexico) is well covered at the 2nd to 3rd subnational level. Currently, these databases do not contain attribute data beyond the total population figures that were added in the global demography project. Additional sources of boundary data are listed in the UNEP/GRID Database of Digital Data Archives in Latin America.
· A medium resolution population database for Africa has been available for about a year and has been used by various agencies (Deichmann 1994). This data set, which is distributed by WRI, is based on boundary data from FAO, USGS, and various CGIAR centers augmented by additionally digitized boundaries. It includes total population from the latest enumeration and a consistent estimate for 1994.
· The OECD/Club du Sahel compiled a large amount of population data for their West Africa Long Term Perspective Study or WALTPS (Snrech 1995) covering 19 countries. Coordinated by WRI, these data sets have been converted to Arc/Info vector format and additional boundaries from the USAID/FEWS project and other sources have been added. The resulting database contains rural and urban population figures for 1960, 70, 80 and 90 for about 2000 administrative units as well as agricultural production estimates for major food crops for about 425 agricultural census units. A proposal is currently under preparation at WRI, UNEP/GRID and NCGIA to merge the African continental coverage, the WALTPS data and several other higher resolution databases and to compile time series of total, urban and rural estimates for all of Africa.
· A Spatial Information Infrastructure for Asian Studies in Australia (SIIASA) has been set up at Griffith University in cooperation with other Australian universities. One component of this project is the development of administrative unit boundary data sets. The objective is to compile data sets for all of Asia, using DCW as the base map.
· The US Bureau of the Census, Center for International Research, has been compiling gridded demographic databases at the national level for many countries and regions of the World. These data sets are gradually being released for public access through CIESIN. The resolution of these data vary between 5' by 7.5' and 20' by 30'.
National level data sets
National level data sets are often compiled for specific projects, or by commercial companies for marketing applications. Some examples of data sets available or under construction are listed below.
· CIESIN is working on a detailed database of Chinese counties in cooperation with Chinese institutions;
· the UN Statistical Division's Software Development Project distributes a desktop mapping package (POPMAP) to national statistical offices, provides training, and supports pilot projects. Two flagship applications have been produced for Uganda and Indonesia. While POPMAP currently does not support a complete set of GIS functions, this effort will lead to increasing availability of spatially referenced census data right from the source;
· many project-specific national data sets have been developed. For example, the Rwanda Society-Environment project at Michigan State University, or the Nepal database developed as a case study for the GIS project at the United Nations Statistical Division;
· Demosphere International, Inc., is one of the few commercial providers of socioeconomic data for developing countries. They sell GIS databases for Mexico (municipios) and India (districts). Both data sets contain a comprehensive set of demographic indicators.
Towns
Data sets of the location and size of towns and cities are even more scattered than those referenced by administrative units. For Africa, a useful effort has been made by the USAID/FEWS project to spatially reference village locations in Burkina Faso, Mali and parts of Niger. A major problem in these efforts is to match village names in census lists to those listed in often out-dated maps. There are various efforts to compile global data sets of city locations (e.g., the Birbeck College cities database distributed by GRID, the non-spatial Urban Database maintained by the United Nations Statistical Division, and a project at Environment Canada). However, most of these are limited to larger cities of at least 50,000 people.
Special purpose data collection
Many relevant socioeconomic indicators are collected using sample surveys, several of which are conducted at regular intervals. Examples are the USAID sponsored Demographic and Health Surveys (DHS), the UN National Household Survey Capability Program, the World Fertility Survey, the World Bank's Living Standard Measurement Surveys, and the Social Dimensions of Adjustment Program. Few of these have so far been systematically referenced to spatial coordinates. A project undertaken jointly by the US Census Bureau and USAID/REDSO Abidjan is currently testing the feasibility of spatially referencing the DHS surveys and linking these data to other surveys. Clearly, much could be gained through better utilization of existing survey data.
IV. Next Steps
A set of recommendations regarding the development and maintenance of population databases for agricultural research should obviously come out of the Arendal II meeting. Below are a few suggestions aimed at stimulating discussion.
· pool existing national level data sets to create consistent regional or continental databases of population data by administrative units and by settlements;
· assign regional responsibilities for coordinating this work, for distribution, and for updating and upgrading existing regional databases;
· develop a standard database template as a guideline for developing population databases for agricultural research, including recommended variables, definitions, conventions for spatial and temporal data interpolation, relational database structures, and update procedures;
· investigate options for converting administrative unit level population data to regular raster surfaces to increase compatibility with physiographic databases.
V. References
Bogue, D., E.A. Arriaga, and D.L. Anderton, eds. (1993), Basic readings in population research methodology, 8 vols., United Nations Population Fund, New York.
Clarke, J.I. and D.W. Rhind (1992), Population data and global environmental change, Paris, IISC/UNESCO.
Deichmann, U. (1994), A medium resolution population database for Africa, Technical paper and digital database, National Center for Geographic Information and Analysis, Santa Barbara.
Dvorak, K.A., ed. (1993), Social science research for agricultural technology development. Spatial and temporal dimensions, CAB International, Wallingford.
Ehrlich, P.R., A.H. Ehrlich, and G.C. Daily (1993), Food security, population and environment, Population and Development Review, 19, 1:1-32.
National Research Council (1993), Demographic change in sub-Saharan Africa, National Academy Press, Washington, D.C. Additional titles on demographic issues in Africa appeared in the same series.
United Nations (1993), World Urbanization Prospects, Department for Social and Economic Information and Policy Analysis, Population Division, New York
United Nations (1994), Population and the environment in developing countries: Literature survey and research bibliography, Department for Social and Economic Information and Policy Analysis, Population Division, New York
Plane and Rogerson (1994), The geographical analysis of population, Wiley, New York.
Snrech, S. (1995), West Africa Long Term Perspective Study - synthesis report, Club du Sahel, Organization for Economic Cooperation and Development, Paris.
Tobler, W., U. Deichmann, J. Gottsegen and K. Maloy (1995), The global demography project, Technical Report TR-6-95, National Center for Geographic Information and Analysis, Santa Barbara.
VI. Contacts
Following is a list of email or fax numbers for contact persons at agencies mentioned in the text. UNEP/GRID and CGIAR contexts have not been included.
CIESIN Bob Chen bob.chen@ciesin.org
ciesin.info@ciesin.org
Demosphere International, Fairfax, VA Kent Hargesheimer fax: +1-703-802-0102
Environment Canada Arthur Lee fax: +1-416-739-4288
Michigan State University, Rwanda Project David Campbell fax: +1-517-336-1671
OECD/Club du Sahel, Paris Serge Snrech Serge.SNRECH@oecd.org
SIIASA, Griffith University, Australia Lawrence Chrissman Crissman@ASIAN.gu.edu.au
UNHCR / GIS coordinator, Geneva Jean-Yves Bouchard bouchard@unhcr.ch
UN Statistical Division, POPMAP Project Vu Du Man vu@un.org
Patrick Gerland pgerland@un.org
USAID/FEWS EROS DC, Sioux Falls, SD Andrew Nadeau nadeau@edcserver1.cr.usgs.gov
or contact GRID/Sioux Falls
USAID/REDSO, Abidjan Glenn Rogers grogers@usaid.gov
US Census Bureau, Center for Int'l Research Richard Turnage richard_turnage@smtp-gw.census.gov
WHO / GIS coordinator, Geneva Isabelle Nuttall nuttalli@who.ch
World Resources Institute, Washington DC Jake Brunner jakeb@wri.org
Norbert Henninger norbert@wri.org
Last updated May 10, 1996 by Lorant Czaran /