Land Use/Cover Data ¹-- Background Information

Introduction

This unit is derived from an initial assessment of global data for the study of land use/cover change, emphasizing only a limited set of LULC changes, the sources that generate them, the manner in which they are generated, and their accuracy (Young et al. 1990). Such a critical data assessments must precedes a meaningful examination of the relationships among various causes of change and the actual land use/cover changes.

The types of land use/cover examined here are cultivation, forest conversion (i.e., deforestation and use of once-forested land for other purposes), livestock, settlement, wetlands, and surface water. These six priority land use/covers were identified by the working group on Human Interactions, Committee on Global Change, National Research Council (NRC), and the working group on Land-Use Change, Committee for Research on Global Environmental Change, Social Science Research Council (SSRC). The latter committee identified the first three as the highest priority with the next three being of second priority. These committees also identified the problems of data as a major hindrance to documenting the spatial extent and rates of change in these covers and uses of the land and assessing the causes of this change. In the following sections, we offer a trial assessment of the data sets for each land use/cover that portend to be "global" in scale. (The discussion of all the land use/cover types and their associated data problems is rather lengthy. We therefore recommend that you split into six groups with each group focussing on just one land use type and then reporting back to the class with a summary of the information contained in each section.)

In Unit 3 later on, we offer a cursory quantitative assessment of the relationships among certain human "macro-forces" of land use change and the first three priority land changes. These primary or global-scale forces of global environmental change include: population change, technological change, economic development, institutions, and attitudes/beliefs (Turner et al. 1990). We examine only the first three forces (or surrogates of them) to provide examples of the kinds of relationships found and the problems with them.

---

Cultivation

• how land use/cover changes contribute to global environmental change,
• how global change will affect land use/cover, and
• what socio-economic and other factors co-determine land use/cover change.

With regard to cultivation, we need to recognize that every form of cultivation is essentially linked with global nutrient cycles, i.e., the take-up, processing, temporary storage, and release of nutrients. The most important nutritional elements with regard to climate change are carbon (C), nitrogen (N), and less so, sulphur (S) -- each one occurring in a variety of molecular forms. For example, paddy rice cultivation releases significant amounts of methane (CH4) -- a greenhouse gas contributing to global climate change -- into the atmosphere. Whether or not paddy rice cultivation is expanding, and by how much, is therefore important to know. Similarly, we want to know whether the production of food crops like wheat, rye, rice, soy, sorghum, and corn can be maintained or increased under changing global environmental conditions. Given the uneven food security situation across the globe and the fast-growing populations in most developing countries, questions about the potential for continued and even intensified cultivation are crucial. To begin even to discuss these questions we need good cultivation data.

Definition: Cultivation refers to the land use for the production of domesticated plants. Cropland is a collective term for land in cultivation. Croplands are commonly divided into two categories: arable land and land under permanent crops. The meanings of these terms vary from one data source to another. The U.N. Food and Agriculture Organization's (FAO) definition of arable land includes lands under temporary (annual) crops (double-cropped areas are counted only once), temporary meadows for mowing or pasture, lands under market and kitchen gardens (including the cultivation of grass), and land temporarily fallow or lying idle (FAO 1989a). The U.S. Central Intelligence Agency (CIA) also designates as arable land those area which are cultivated for crops that are replanted after each harvest (e.g., wheat, rice) (CIA 1990).

FAO and CIA definitions for land under permanent crops coincide: land under crops that do not need to be replanted after each harvest such as cocoa, coffee, and rubber; it includes land under shrub, fruit trees, nut trees, and vines, but excludes land under trees grown for wood-fuel, timber, and wood by-products.

A problem with most typologies of this sort is that a land use of cultivation may also be counted as a land use of livestock production. For example, temporary meadows and croplands during fallow period, used for rearing of livestock, could be reported by some countries as pasture land, as well (see section on livestock). Also there is no definition of land under slash-and-burn agriculture within the arable land category.

Sources: Data sources for currently cultivated lands are diverse; although not numerous, they do provide uniform coverage of the world. Undoubtedly, the FAOProduction Yearbook is the primary source for many analyses and other data outlets, such as the UNEP Environmental Data Report and World Resources. The FAO data are supplied annually by national governments and therefore lack control for uniformity and accuracy. For incomplete data, which is common for developing countries, the FAO provides unofficial or estimated numbers for land use.

At least the last ten editions of the FAO Production Yearbook contain information about land use dynamics for every country by providing current data and 20 years back-data (in hectares). Additionally, the UNEP Environmental Data Report (1987-1989) gives the same information in percentages of land area. Some regional studies (e.g., Hart 1984; Honrad 1987; Whitby and Ollernshaw 1988) contain data for ten years (for Europe), 20 years (for Central America) and 50 years (for the United States).

Other sources include the CIA World Factbook 1999 or World Factbook 1995 which provides data on the percentage of arable land and land under permanent crops by countries. Although FAO and CIA data may have been collected in different ways, the results are similar. An untapped potential source of information is various world maps depicting land use and cover. For example, the Map of the Main Land Use Units of the World (1: 15,000,000), compiled by Soviet geographers (from Moscow State University and the Institute of Geography, Moscow) in 1987 is based on systematic analysis of numerous statistical and cartographical sources and space images -- but not on FAO data. Digitizing these maps might well provide interesting quantitative comparisons for FAO figures, although other attempts to do so have led to questions about the temporal congruence of the data (David Skole, personal communication). Such maps may well prove useful in outlining general changes decade by decade, but they are too general for more specific needs.

Estimates: Total area of land under cultivation is about 11% of world land area (1,447,509,000 ha), with 10% in arable land and 0.78% under permanent crops. The distribution of croplands varies greatly from one region or country to another. The majority of the world's cropland is concentrated in Asia (about 30%), North and Central America (about 18%), and the area of the former USSR (about 16%). Nevertheless, if we consider cropland as a percentage of total land area, Europe is the most "cultivated" region of the world, followed by Asia and North America.

The constant growth of cropland area, though already slowed from the beginning of the 1970's (only 2.7% between 1975-77 and 1985-87), has -- at least for now -- been interrupted. The world cropland area in 1993 was 0.4% less than that in 1988 (FAO 1994). The highest rate of expansion was observed in South America and Oceania (12-13%), the lowest ones in North America (1.8%) and Africa (4%). Meanwhile, Europe has experienced a decline of total area cultivated on average by 0.4% per year because of afforesta-tion, expansion of urban areas, and abandonment of marginal lands in favor of intensifying cultivation on more productive farmland.

Cropland availability is another important measure because the level of current use influences trends in land use for the future. The total area of potentially cultivable land is 24% of the ice-free surface of the earth, and its distribution among regions is uneven: 23% of it is in Africa, 21% in South America, 20% in Asia, 15% in North America, 11% in the area of the former USSR, 5% in Australia, and 5% in Europe (Revelle 1984).

About 46% of the potentially cultivable area of the world (i.e., the land mass of the earth on which climate conditions permit cultivation) is cultivated, but there are contrasts among regions in the degree of use of their agricultural land potential. As of 1987, Africa cultivated only 23.5% of its potential cropland, although including fallow land increases this proportion considerably. According to the statistics, Southwest Asia cultivates an area larger than its potentially cultivable area, raising serious questions about the measure of arable land! Southeast Asia is also approaching the limits of its suitable land. South America uses the smallest proportion of its potential of any region, only 17.3% in 1987 (FAO 1984a).

Data Quality: Assessing the data quality for cropland is difficult, because the primary source, the FAO, does not provide controls for uniformity or measure. In addition, contradictions exist in FAO reports. For example, data from FAO Production Yearbook Data, Computer Tape (cited in Urban and Volltran, 1984) for the years 1961-65, 1970, 1975, 1980 differs from FAO Production Yearbook, vol.40 (1986) data for the same years. Data reported by governments and/or FAO estimations are corrected over time. It seems reasonable, therefore, to rely more on the most recent publications of FAO. Problems also arise from the broad FAO definition of arable land and from the multiple use of some types of land cover during the year or over a period of years. As mentioned above, temporary meadows and land temporarily fallow can be used for livestock raising and therefore might be reported as pasture land. Also, it is difficult to figure out the area of "net arable land" or the land that was tilled in a certain year because the areas of the land lying fallow are counted within arable land and are not reported separately.

Forests

All of these questions indicate the many linkages between this land use/cover type with the global environment, including the socio-economic environment. Whether you want to include this important land cover type in global models, or whether you want to predict forest changes in just one region, good forest data are crucial.

Definition: Definitions of forest vary with data sources and publications. The FAO defines forest and woodland in the Production Yearbook as "land under natural or planted stands of trees, whether productive or not, and includes land from which forests have been cleared but that will be reforested in the foreseeable future" (FAO 1988a: 3). The FAO cautions users of the Production Yearbook that "it should be borne in mind that definitions used by reporting countries vary considerably and items classified under the same category often relate to greatly differing kinds of land" (FAO 1988a [emphasis added]). The 1980 Tropical Forest Assessment, a joint publication of the FAO and the United Nations Environmental Program (UNEP), does not include planted stands of trees in calculations of open and closed forest. While definitions of open and closed forest vary by source, they generally refer to the extent of tree cover over the ground or percentage of area covered by tree crowns.

Sources: Most of the professional users (e.g., Allen and Barnes 1985; Brown 1974, 1989; Williams 1989) cite the FAO and the World Resources Institute (WRI) as the major sources for world forest data. The FAO has included forest and woodland data in its annual Production Yearbook since 1950. The Yearbook reports data as provided by individual countries through annual questionnaires and national agricultural censuses. Unofficial data and estimates are used when necessary. Data are not available for all countries.

In association with UNEP, the FAO also published the 1980 Tropical Forest Assessment in 1982. This assessment, covering 76 countries, was the first for global tropical forests. This information has been expanded to cover 129 developing countries and is revised annually (WRI 1990). Sources of data include: national forestry institutes; land use and survey institutions; photographic surveys (all or parts of five countries); side-looking airborne radar (all or parts of four additional countries); and satellite imagery (all or parts of 19 countries). The FAO adjusts the data to fit its definitions.

The U.N. Economic Commission for Europe (UNECE), in conjunction with the FAO, published The Forest Resources of the ECE Region (which includes Europe, the former USSR, and North America) in 1985 (FAO/UNECE 1985). Data were obtained through questionnaires completed by individual countries, official estimates, and FAO reports. The report presents general forest resource inventory data along with volume and mass of trees and other woody biomass.

The World Resources Institute compiles data from the FAO, the UNECE, and more recent country reports into its annual report. Tables in that report show the extent of forest and woodland, average annual deforestation rates, and average annual reforestation for the 1980s (WRI 1990). Eight countries have produced individual studies of domestic deforestation independent of FAO and UNECE (WRI 1990). These studies provide recent and comprehensive data illustrating the situation in the respective countries.

Estimates: In 1970, about 32 % of the land surface of the world was covered with forest and woodland (FAO 1987a). By 1985, forest and woodland occupied 31.24 %; a net loss of over 104 million hectares (approximately 1 million km2). Only two regions reported increases in forest and woodland cover over the same time period: Asia gained almost eight million hectares and Europe reported an increase of just over five million hectares. All other regions reported net losses. South America and Africa reported the greatest losses, 62.2 million hectares and 46.2 million hectares, respectively.Preliminary figures released by the FAO in an interim report presented to the Commission on Forestry in Rome in September, 1990, indicate an average annual deforestation rate of 17 million hectares over the last 10 years (Henninger 1990).

Data Quality: While it is universally recognized that major forest conversion has taken place, specific estimates on the amount and on current rates of deforestation and afforestation are few and controversial. Insufficient historical data, lack of consensus on definitions and data-gathering techniques, and the subjectivity of data interpretation are impediments. Existing estimates are rife with likelihoods, estimations, suppositions, and guesses. Williams (Williams 1990a: 179) summarizes the problematic nature of calculating both historical and current deforestation:

Crucial to a discussion of deforestation is the calculation of how much forest has been cleared. But the task is not an easy one. Even today, with all the modern aids of land use censuses and satellite imagery, there is no unequivocal inventory of contemporary woodland (Allen and Barnes 1985). If this calculation is a problem now, how much more difficult for past ages. ... No overall synthesis of reliable data is at hand. ... The best that can be done is to indicate the magnitude of likely change that could be expected to occur.

Current deforestation rates could be derived by comparing data on land use under forest and woodland from the annual FAO reports. A comparison of the 1961-1965 data to that of 1970 reveals a large shift inconsistent with that of later years. This raises questions as to the accuracy of the data, particularly before 1970. FAO's inclusion in the Production Yearbook of land to be reforested in the 'foreseeable future' raises questions: when is the foreseeable future?, how does the FAO know there will be replanting?, and what proportion of the figures relate to this future planting?

FAO-UNEP (1982) forest data are better than those found in the FAO production publications because the data are qualitatively assessed. The forest data are distinguished by country in terms of the quality of its data (Lanly 1983; Rudel 1989).

WRI has the most comprehensive data set on forest land, reforestation rates, and deforestation rates, but it is not problem-free. WRI uses data adjusted by the FAO for 129 countries and compares them to the situation in 1980 (baseline); the ECE data are from the early 1980s but were not adjusted to a baseline year. The UNECE compilation does not detail what information-gathering techniques were used to complete the questionnaires and estimates. The FAO reforestation rates as reported in WRI have been criticized because the "trees are not seen for the forest" (trees planted in configurations that do not correspond to the definition of forest) (UNESCO 1989). Reforestation data may or may not include regeneration (either natural or through forest management) or trees planted for non-industrial use.

Perhaps most importantly, local experts are highly suspicious of many of the FAO and WRI data for specific locales. Harold Brookfield (personal communication) argues that the FAO figures for Malaysia are so poor that he will not use them in his assessment of land use change there. David Kummer, after extensive study of the issue on his part, notes the same for the WRI figures for the Philippines (Kummer 1990b, personal communication). Also, the WRI does not make it clear that "the 1980's" has two different meanings depending on the category of forest resources. For extent of forest and woodland, it means "the end of 1980 unless otherwise noted"; for deforestation and reforestation rates it means "1981-1985 unless otherwise noted" (Henninger 1990).

Other Issues: In 1992 FAO and UNEP published the 1990 Forest Assessment. They did not overcome the recurrent problems of definitions and data interpretation, but the report can be considered the most comprehensive source of data on forest cover, deforestation, and reforestation at the time. UNEP and FAO also made a comprehensive attempt to remedy the definition and data problems by compiling renowned experts= advice on the environmental parameters that are to guide future global forest assessment (UNEP/FAO 1993).

Any statement concerning global or regional forest cover, reforestation, or deforestation must take into account the variety of data sources, incompatible time frames, and varying definitions. The reliability of forest data is contentious, and drawing conclusions regarding any aspect of forest trends (other than simplistic ones) at the global or regional level is risky. Global statements will be of questionable value until exhaustive work is done at the country level, using common definitions and agreed-upon standards of coverage, quality, and accuracy.

Livestock

Farmers have to weigh many factors (market value for different products, cultural preferences, physical capabilities of the land, availability of technical know-how and other inputs) to arrive at a compromise between how much land they should devote to crop versus livestock production (although, as shown below, there is some overlap which makes data assessment very difficult). Global changes ranging from climate change to population pressures and changing economic forces are linked in an intricate manner to affect livestock production.

Definition: Livestock refers to domesticated animals (non-pets) and their relationship to land used for their production and maintenance. It provides two areas of concern for global land use change:2 the amount of land used for livestock rearing, and the total number of head of livestock (or land intensity of livestock). Some of the land covers used for livestock include: open pasture land (both improved and degraded), open land (meadows, marshes, tundra, steppe, savanna, desert), forest (open, closed, plantations), pasture (improved grasslands), and cropland (often during fallow periods, but also during crop growth). Typically, however, land cover/use associated with livestock is reported as either rangeland or pasture.

Rangeland designates the land use of livestock production (WRI 1987, 1995), and constitutes areas that provide forage for free-ranging livestock and wild animals.3 Rangeland may have physical limitations that make it unsuitable or uneconomical (at the time) for agriculture or intensive forestry, although many examples exist where livestock rearing and other land uses are compatible (agropastoral systems).

Permanent meadows and pastures refer to land used permanently (five years or more) for herbaceous forage crops, either cultivated or growing wild (wild prairie or grazing land). This use normally implies attempts to improve fodder conditions for animals, from burning to planting grasses.

Data on the extent and change in rangeland are sparse. Because rangeland includes a number of land covers, the calculations of rangeland can vary considerably from region to region. Systems of livestock herding (nomadism/ranching) also vary from region to region, making regional comparisons difficult. There are also conflicts about which land covers constitute rangeland. The total number of livestock is somewhat more comparable because there is not a great variation between the specific species of animals around the world. The data, however, are disaggregated by age or sex of animals. Also, when considering total numbers of head of livestock for methane production, one must consider what the changes have been in wild ruminant populations as well.

Sources: Knowledge of the extent, condition, and use of rangeland is incomplete: no comprehensive global assessment or data sets exist to our knowledge. The most complete analysis of rangeland has been done by the World Resources Institute (WRI) and the International Institute for Environment and Development in their World Resources Series (WRI 1990). They use data that have been gathered by the FAO and are published in their annual Production Yearbooks (FAO 1989a). In addition, they use data from the FAO/UNECE published in their Forest Resources 1980 book (FAO/UNECE 1985).

The WRI's global, regional, and country estimation of rangeland is based on adding the total amount of FAO's permanent pasture plus all of FAO/UNECE's open forest land plus one half of FAO's "other land" category (WRI 1987). The permanent pasture (noted above) and other land data have been collected and published yearly by FAO since 1946. Other lands includes unused but potentially productive land, built-on areas, wasteland, parks, ornamental gardens, roads, lanes, barren land, and any other land not specifically listed under arable land, land under permanent crops, permanent meadows and pasture, and forests and woodland (FAO 1986a, 1994). The open forest land -- woodlands that have a relatively continuous grass cover on the forest floor and the canopy covers more than 5% of the area, but no more than 20% -- is based on 1980 estimates. Also this land must not be used primarily for agriculture or forestry (FAO/UNECE 1985).

Data for number of livestock are gathered by the FAO and published in their annual Production Yearbook. The data for the FAO Production Yearbooks come from official figures supplied by governments through questionnaires or from government publications and reports to the UN or FAO. When official figures are not available, data are taken from "reliable," unofficial sources, or are estimated by FAO and are indicated as such. As with the forest data, the definitions used by reporting countries vary considerably, and items classified under the same category often relate to greatly differing kinds of land (FAO 1989a). As a result, the land use data collected by the FAO are not completely compatible, and the aggregate estimates are rough. These data, however, are perhaps the best on a world-wide basis.

For historical changes, John Richards (Richards 1986; Richards 1990) uses the WRI's data and FAO data (as noted above) as well as reconstructing vegetation from soil studies (Houghton et al. 1983; Richards, Olson, and Rotty 1983) and estimating agricultural land from world population (McEvedy and Jones 1978). The WRI has created historical land use data by extracting from vegetation maps and determining rangeland in combination with population and growth estimates (Richards 1990; WRI 1987, Table 18.3).

Estimates: According to the WRI, the amount of rangeland world-wide in 1983 was 67 million km2 or 51% of the total ice-free land area of the earth. Of this total, permanent pasture is 31 million km2 or 24% of the total ice-free land area (WRI 1987). According to FAO Production Yearbooks, permanent pasture land has changed from 30.46 million km2 to 33.62 million km2 between 1965 and 1994. Regionally, according to FAO data, the world has not experienced much change in the quantity of permanent pasture, except perhaps Asia.

The total number of domesticated ruminants4 world-wide as of 1994 was 3,152 (in millions of individuals) (FAO 1994). The total number of all domesticated animals in 1994 world-wide was 16,957 (in millions of individuals). The total number of ruminants has changed from 2,583 to 3,152 (in millions of individuals) between 1965-1994. Regionally there have been few changes, except perhaps in Asia. An important aspect of analyzing both the pasture land and livestock data is that the two do not vary together. Livestock numbers have been increasing at a greater rate than pasture land.

Data Quality: Quantifying the extent of rangeland is difficult because of overlapping definitions of lands that provide forage; different studies include different land covers for forage. Problems also arise owing to multiple land use and changes in these uses per unit of land and time. For example, some forests are closed to livestock, some pasture land is left fallow for years, and some areas are only used during part of the year. Since each country may count these units different, the potential error in comparisons is large.

The aggregate numbers for rangeland probably underestimate forage area because they do not necessarily consider the land area from which extensive biomass used for forage is obtained, including seasonal browsing areas (i.e., forage land used only during certain times of the year) or peasant fodder systems, such as "backyard" livestock fed on crop residues. In addition, the aggregate numbers hide the regional differences in rangeland conditions, productivity, and intensity. This requires us to consider: land area under different covers used for forage, total forage production, and total land used.

The FAO Production Yearbook data compiled in the 1940's and 1950's are not as accurate as the data from the 1970's and 1980's. This is because the earlier data were based more on general estimates that vary widely. In addition, the earlier data were not recorded for the same year for each country, and therefore it is difficult to ascertain change for aggregate numbers. The aggregate numbers in the WRI's rangeland data are rough estimates; they may be appropriate at the global level, but regionally there are difficulties. WRI claims that one-half of FAO's Aother land@ is rangeland. Yet one-half of Algeria's "other land" is desert, while this half for Sweden is tundra where reindeer forage. For the historical reporting that the WRI and John Richards provide, there are problems of arriving at a new estimate from sources that also estimate data (WRI 1987; Richards 1986).5 This leads to the problem of compound error.

Other Issues: The most complete data on regional range conditions and changes are for North America. This is because the rangeland has been intensively used only for the past 100 to 150 years and because the U.S. government has kept extensive statistics (US Congress 1982; Stoddart et al. 1975; US Forest Service 1980). When the western part of the United States was opened to livestock, overgrazing lead to degradation of much of the land. Laws in the post-Dust Bowl era have helped improve the rangeland so that now it is in better overall condition than it has been in any time this century (BLM 1985). The U.S. has 312 million hectares of rangeland (WRI 1995). Earlier indications of its quality are as follows: 32% in good, 28% in fair, and 12% in very poor condition (US Forest Service 1980). Canada has followed similar paths as the United States.

Three types of rangelands are recognized in Latin America: (1) natural grasslands, woodlands and savannas; (2) high-elevation natural grasslands and shrub lands; and (3) cultivated pastures, established in areas once occupied by forests. About 70% of the rangeland is in the first category (WRI 1990). Much of the rangeland of Latin America has been overgrazed and degraded like that of the North. The greater Amazonian region has amassed much concern because of short-term pasture created from tropical forests. The principal countries with rangeland includes: Argentina, Paraguay, Uruguay, and Brazil. Rangeland covers about one-third, or 700 million hectares, of Latin America's land surface. Permanent pasture land occupies about 569 million hectares (WRI 1990).

Australia has a similar history to North America in terms of settlement, introduction of European species, and overgrazing of rangelands. Much of Australia's rangeland is arid and is easily disturbed. A unique problem to Australia was the introduction and subsequent overpopulation and overgrazing of the European rabbit. According to the FAO Production Yearbook, in 1993 Australia had 413.800 million hectares of pasture land (FAO 1994).

About 65% or 1,945 million hectares of Africa is rangeland (WRI 1990). Much of Africa's rangeland has complementary and competing land uses such as, respectively, cropland used for seasonal grazing, and wildlife reserves and fuelwood supplies in which grazing takes place. Like parts of Asia, Africa's long history of pastoralism, nomadic or other forms, has claimed almost all dry forest, savanna, and semi-xeric and xeric lands for livestock production, if only for brief periods of time throughout the year and for relatively few livestock (per unit area of land). Much of this area is also the grazing domain of large herding animals.

Thirty-three percent of Europe is rangeland, including open land and open forest. Of this, highly productive permanent pasture land makes up 18% (WRI 1987). Most of Europe's pasture land is intensively managed and is, on the average, the most productive in the world.

The Middle East and Central Asia have an ancient tradition of livestock grazing, primarily through various forms of nomadism. Rangelands, therefore, account for a large percentage of the land area in this region. Interestingly, this area has some of the driest continually grazed areas in the world (WRI 1987, Table 5.8). According to the FAO Production Yearbook, in 1986 the Middle East had 267.652 million hectares of pasture land (FAO 1987a).

Asia, including all lands from Siberia into the Indian subcontinent, has the most varied rangelands in the world (including: tundra, steppe, desert grasslands, opened monsoon forests, tropical forests). The People's Republic of China and Mongolia have over half of Asia's permanent pasture lands (FAO 1994). China has about 400 million hectares of permanent pasture; Mongolia has 125 million hectares of permanent pasture land (World Bank 1984; FAO 1987a, 1994). On the Indian subcontinent permanent pasture land makes up 11.4 million hectares or about 4% of land surface and permanent pasture. About 49 million hectares of other land can be considered rangeland (FAO 1987a). Much of the rangeland in the subcontinent has been overgrazed and degraded. There are, however, some successful programs in managing the common rangeland and rehabilitating lands (WRI 1988).

Settlements

Definition: Settlement refers to the land used for human habitation. In land use/land cover classifications such as that of the US Geological Survey, this area is referred to as urban or built-up land. In the USGS system, this category includes cities, towns, villages, strip developments along highways, transportation, power, and communication facilities, and industrial, commercial and institutional sites (Anderson et al. 1976, cited in Lillesand and Kiefer 1979).

Sources: Reflecting the dominant interests of researchers and policy makers, data are readily available on urban populations, but not on the area occupied. Occasionally, works on global environmental change include estimates of the planet's urbanized area, but the reliability of these estimates is questionable. Some researchers have estimated settlement area by extrapolation from urban population numbers.

International agencies such as the United Nations and the World Bank publish voluminous data on global population, including estimates of the urban portion of the population. The main data sources include the UN's Demographic Yearbooks and World Population Prospects Reports.6 The Demographic Yearbook for 1988, gives urban population by country for each year between 1979 and 1988 (UN 1990). A special edition published in 1979 summarizes population data by year from 1948 to 1978 (UN 1979). All of this information is based on census data supplied to the UN by national governments. The Population Reference Bureau (PRB) is considered by some to be the authoritative source on population, but it also draws heavily upon UN data. In addition, a variety of general urban studies and case studies on urbanization have been published. For example, the World Commission on Environment and Development commissioned four background papers (Burton 1985; Hardoy and Satterthwaite 1986a; Hardoy and Satterthwaite 1986b; Sachs 1985). Some of these studies include data on urban areal extent, but these data do not appear to have been compiled.7 Chandler (1987: 6-7) mentions that in the case of compiling his historical database on urban populations he sometimes used city area to calculate population;8 he does not, however, publish areal data. Another useful, but incomplete, data source is research estimating the conversion of agricultural land to urban and other non-agricultural uses.9

Estimates: Estimates of the extent of land occupied by human settlements have been made by researchers of global environmental change, by extrapolating from urban population numbers, and locally in case studies. Within this context, L'vovich, White, and collaborators (1990: 246) estimated that the mid-1986 urban population of approximately 2.2 billion, Aincluding industrial enterprises and roads, occupied an area of about 1.2 - 1.4 million km2." Earlier, in an article on anthropogenic albedo changes, Sagan and colleagues (1979) assumed an urbanized population of one billion, and estimated global urbanized area at one million km2, or about 0.2% of the earth's surface; they estimated the annual rate of change as 20,000 km2, or about 0.004% of the earth's surface.

The magnitude of urban populations and an example of the type of data available are illustrated by the following figures. The UN report, World Population Prospects as Assessed in 1982 (UN 1985), gives estimates and projections of urbanization, urban and rural populations (in absolute numbers, percentages) and population density -- from 1950 to 2025 (see Table 5). These data are available by country and region, and as world totals. The Population Reference Bureau's mid-1990 estimate of global urban population was 2.18 billion (PRB 1990), up from 600 million in 1950 (Brown and Jacobsen 1987).

Data Quality: The reliability of global estimates is questionable. L'vovich and White (1990) do not explain how their estimates were derived. Sagan, Toon, and Pollack (1979) freely admitted that it is impossible to estimate global land use changes to an accuracy greater than a factor of two. Their estimate was based on work by Wong (1978), who extrapolated urban area from urban population. Wong, citing Pire (1976) for a portion of his methodology, claimed that the average California city dweller requires 1,000 m2 of urban space, each urban Briton 600 m2, and therefore, the average urban dweller uses 800 m2 (Wong 1978). Such a methodology is obviously inadequate. A single conversion factor is not suitable, given spatial and temporal differences in urban patterns, to say nothing of cultural variability.10

Further problems also arise from the use of urban population figures to extrapolate area. Some increases in urban population numbers reflect changing urban boundary delineations as well as actual increases within a particular space (cf. WCED 1987: 258, note 4). Researchers who use data sources other than aerial photography or satellite imagery are likely to run into this problem of municipal boundaries. The jurisdictional limits of a 'city' or other data-reporting unit will not necessarily reflect actual land cover or land use.11

More detailed concerns are raised by UN documents. According to the 1988 Demographic Yearbook, the major constraints on data reliability for urban population estimates are under-enumeration, distinguishing de jure and de facto populations and varying definitions of urban; the latter is the most significant factor, seriously limiting comparability (UN 1990). What is considered 'urban' varies according to each national census; the various definitions are listed at the end of each UN table. An impressionistic review of the definitions in the 1979 and 1988 editions indicates that in the 1980's most nations considered settlements with 2,000-5,000 residents as the minimum threshold for being urban; however, some nations simply list the population of the specific towns declared to be the nation's urban areas. In the 1950's, the minimum threshold ranged from towns of several 100 for some countries to towns of 1,000-2,000 for most nations, and to towns of 5,000 for a few developed countries.

Another implication of relying on national census data is that actual counts are only available for the years in which a particular country happened to conduct a census. The yearbooks do give references for earlier censuses, even those predating the founding of the UN; the 1988 yearbook, for example, gives references going back to 1920. The 1979 special edition lists, by country, years for which urban census data are available and gives urban definitions by country by census year (UN 1979).

Wetlands

Definition: Wetlands may be defined as "lands transitional between terrestrial and aquatic systems where the water table is usually at or near the surface of the land or is covered by shallow water" (Cowardin et al. 1979; also Orme 1990). Comprising an ecotone between dry land and aquatic ecosystems, albeit one with unique ecological characteristics, wetlands form a continuous gradient between the terrestrial and the aquatic, and the upper and lower boundary limits in definitions are, therefore, arbitrary. For example, flood frequency has been a source of controversy in US definitions; other US controversies occur over the type, sizes, location, and conditions that may be defined as wetlands. Thus, there is no universally accepted definition, and those that are used, tend to be colored by the purpose of the agency using them (Mitsch and Gosselink 1986).12

Sources: No authoritative global database on wetlands is indicated by the literature. Very few global estimates of wetland extent or loss appear to have been published at all. In general, it seems that wetlands became a major issue only in the 1970's, when some countries instituted national wetland inventories (Williams 1990b). There is no indication in the literature that any global agency is monitoring these national compilations; it is not even clear that very many countries are doing inventories. The United States has the most complete wetlands data (Williams 1990b); but "little progress has been made outside North America and especially in the Third World" (Maltby 1988: 6).

Losses are especially difficult to estimate. Even in the US, accurate baseline data really only date to the mid-1970's, and there has been "no comparable comprehensive national survey elsewhere" (Maltby 1988: 6). Large European losses appear in the historical record for specific regions or locales. For the rest of the world, data come from case studies (see, for example, Williams 1990b), although region-wide compilations have been made for some regions (e.g., Carp 1980; Karpowicz 1985; Scott and Carbonell 1986; Scott 1989).

The World Resources Institute's 1990-91 world data "Wetlands/Marsh" category can be found in its "Habitat Loss, 1980's" table (WRI 1990, Table 20.4). This table pieces together data from a number of local and regional studies. The table lists data by country; these entries are not summed into continental or global figures. Elsewhere in the literature, the most widely cited figure (see below) for the areal extent of global wetlands is that of Maltby and Turner (1983). This estimate was published without citations in a popular magazine. It was apparently based on biogeographical information compiled by Soviet geographers (Bazilevich, Rodin, and Rozov 1971); this, in turn, is based on a 1964 Soviet atlas. These figures were later reworked by Mitsch and Gosselink (1986). Another Soviet figure (see below) appears as an estimate of "marshland" in the context of a study on global water storage/water balance (UNESCO 1978).

Estimates: Maltby and Turner (1983: 13) estimated total wetlands for eleven "thermal belts and bioclimatic regions" of the globe, and concluded that wetlands comprise 6.2% of the earth's land area, or 8,228,000 km2. Mitsch and Gosselink (1986) reworked these figures, making more consistent use of the Russian data from which they were derived; their estimate placed total world wetlands at 8,558,000 km2 or 6.4% of total land area. These figures are still cited (Williams 1990b). Apparently using a more restrictive definition, the Soviet water balance study for UNESCO estimated Amarshland@ at only 2,682,000 km2, or 2% of the earth's land surface (UNESCO 1978).

Much concern is expressed about the extent and rate of wetland loss, but estimating this is very difficult given the state of current inventories and lack of baseline data. The World Resources Institute's compilation of data led them to estimate a 50% global loss of wetlands --presumably in the 1980's, although this is not made clear. Nor do they explain how the 50% loss figure was derived; however, totaling the WRI estimates gives 4,106,541 km2, or 49.9% of Maltby and Turner's 8,228,000 km2 figure calculated from the 1971 Russian figures.

According to Maltby (1988), the greatest potential for future wetland losses to development lies in the Third Word, where pressures to increase agricultural land and reduce water-borne diseases combine with irrigation and hydroelectric projects to threaten wetlands.

Some of the problems inherent in estimating current wetlands and wetland loss are illustrated by the case of the United States, the acknowledged leader in wetland inventory and study (Maltby 1988). Between 1907 and 1987, fourteen different estimates have been made, with little agreement (Williams 1990b). Two definitions of wetlands are used by the US government. The Fish and Wildlife Service definition is used for scientific work, inventory, mapping and classification, while the Army Corps of Engineers/Environmental Protection Agency definition is accepted by managers and regulatory agencies (Mitsch and Gosselink 1986). In the mid-1970's, the first definition resulted in a 99 million acre (or 40,095,000 hectares) US wetland inventory; the second definition resulted in one of 64 million acres (or 25,920,000 hectares) (U.S. OTA 1984). Loss estimates are similarly variable, depending on the agency doing the estimating (Horwitz 1978); the Council on Environmental Quality estimates a long-term loss of 53% (CEQ 1990), while other loss estimates were in the 30-40% range. Loss estimates are further complicated by the creation of artificial wetlands, although it is not clear that these can replicate the functioning of natural systems (Gosselink and Maltby 1990).

Data Quality: Some of the World Resources Institute's figures are based on regional research specifically concerned with wetlands (e.g., Carp 1980; Karpowicz 1985; Scott and Carbonell 1986; Scott 1989; Canada 1988). Some data are taken from the FAO's agro-ecological zones project (FAO 1978); however, there are no data for some nations, and definitions of 'wetland' are inconsistent across countries. The WRI cautions that their figures probably underestimate actual wetland extent. The Maltby and Turner (1983) estimate has several major problems. First, it is based on research done by Bazilevich, Rodin, and Rozov (1971) for the purpose of quantifying plant productivity (biomass production) in different geographical regions. These regions are taken from the soil and vegetation maps in the Soviet Physical-Geographic Atlas of the World (1964) (American Geographical Society 1965). None of the assumptions made in the atlas or in the estimates by Bazilevich and colleagues -- or their implications for wetlands -- are made explicit.

Second, Maltby and Turner (1983) were inconsistent in deciding which of the Soviet bioclimatic vegetation categories to include as wetlands and which to leave out; some forests with small bogs were included, others were not. They also classified floodplains and humid tropical meadows as wetlands. The revised estimates by Mitsch and Gosselink (1986) corrected this second problem but not the first.

Surface Water

Definition: For the purpose of this assessment, surface water is defined as inland water as represented by lakes, rivers, reservoirs, and ponds, but not wetlands (see above). It also does not include glaciers or the coastal bays and inlets classed as 'inland water' by official territorial boundaries.

Sources: Given that every encyclopedia and atlas lists the area of each nation, and given that water is so easily visible in the infrared bands of remote sensing platforms, one might assume that data on the areal extent of surface water would be readily accessible. According to hydrologist Harry Schwarz (personal communication) it is not, however, probably because there is not a demand for data on surface area; water resources experts are interested in volume not area, and national surface areas are calculated on the basis of territory controlled, violating the limitation on 'inland water' (see Definition above).

It is surprisingly difficult to derive even a one-time tabulation of total surface water for the globe. There does not appear to be any ongoing monitoring of changes in the areal extent of surface water, nor is there a single agreed-upon baseline figure to compare such changes against. UNESCO's 1978 World Water Balance comes the closest to being an authoritative source. The surface areas of large lakes and inland seas are published in atlases and similar references such as the CIA's World Factbook; accuracy and measurement methodologies change, however, so these data cannot be used as time series. The World Register of Dams lists surface areas of large reservoirs. Fairly complete hydrological data, including surface areas, are available for North America, Europe, parts of Asia, and Australia, but not the rest of the world (L'vovich 1979). Although they comprise a significant area, data on small water bodies -- natural and artificial -- is not generally available (Nace 1970; L'vovich 1979).

Estimates: According to Nace, a foremost US expert, Ainland water areas of the world probably exceed 1 million km2"; however, "only very crude estimates are available" (Nace 1970). These estimates follow by categories.

Lakes: Natural lakes comprise the largest inland surface water area. UNESCO (1978: 43) estimated the area of the world's lakes at 2,058,700 km2, or about 1.4% of the earth's total land area; of this, 1,236, 400 km2 is fresh water and 822,300 km2 is salt water. Globally, UNESCO identified 145 large (over 100 km2) lakes and estimated that they cover 1,300,000 km2; they are thought to contain over 95% of total water volume (UNESCO 1978). Citing "USGS" as their source, Botkin and Keller (1987) give the global surface area of freshwater lakes at 855,000 km2.

The significance of small water bodies is illustrated by a Soviet example. Bochkov, Chebotarev, and Voskresensky (1972) estimated that the former USSR has 2,850,000 lakes with a total surface area of about 500,000 km2 -- about 2% of the country. More than 98% of these are small lakes less than 1 km2; the total area of the 17 large lakes with a surface area over 1,000 km2 is 173,000 km2 (Bochkov, Chebotarev, and Voskresensky 1972), but those lakes contain over 98% of water volume (UNESCO 1978).

Reservoirs: The most significant land use change affecting surface water is the creation of reservoirs. Petts (1984: xiii) noted the magnitude of this only recently appreciated human impact: "without doubt the damming of rivers has been one of the most dramatic and widespread, deliberate impacts of Man [sic] on the natural environment." As with lakes, global data are available for large dams and reservoirs -- which hold most of the water -- but the many small structures are not well-documented (L'vovich 1979).

Several estimates of the total water surface of global reservoirs have been made. A 1972 estimate put the total at 600,000 km2; not counting the lakes included in backwater lake areas, the total water surface for reservoirs proper was estimated to be 400,000 km2 (UNESCO 1978). L'vovich, White, and colleagues (1990) gave two estimates of the maximum water surface of global reservoirs. Apparently referring to reservoirs of more than 100 million m3 capacity built since 1951, they estimated global reservoir surface area at 590,000 km2; they also cited Voropaev and Avakian's (1986) estimate of the surface area of all large reservoirs when full as 390,000 km2.

The 1972 UNESCO approximation of reservoir surface area was based on an estimated 10,000 reservoirs, mostly in Europe, the former USSR, and North America (UNESCO 1978). UNESCO also reported a total of 143 reservoirs with a capacity greater than 5 km3 volume (UNESCO 1978). L'vovich (1979) cites the estimate of Avakian and Ovchinnikova (1971) that there were 1,350 large reservoirs (with a storage capacity greater than 100 million m3) in 1971, as well as thousands of smaller ones, perhaps numbering 10-20,000. A World Register of Large Dams -- describing dams higher than 15 m -- has been issued by the International Commission on Large Dams periodically since the early 1970's (SCOPE 1972; van der Leeden 1990). The vast majority of large dams are in North America (Beaumont 1978). Although the land areas flooded by dams through history is not known, several authors have traced the history of dam building. SCOPE (1972) noted centuries of small lake construction such as the tanks of Sri Lanka and the fish and mill ponds of Europe. Lakes larger than 100 km2 surface area were not built until 1915 when new concrete and earth moving technologies became available. By 1970, at least 40 reservoirs had been built which covered more than 1,000 km2 and 260 between 100-1,000 km2 in area were in operation all over the world -- as well as countless small dams (SCOPE 1972).

Beaumont (1978) identified three distinct periods of worldwide dam building between 1840-1971. Before 1900, there was increasing building activity, but the overall impacts were still relatively small. From 1900 to 1945 there was moderate activity, concentrated mainly in North America, W. Europe, SE Asia, and Japan, with gaps during wars and the depression. Between 1945 and 1971, 8180 major dams were built; this "phenomenal burst of building activity" peaked in 1968 with the commissioning of 548 structures. During 1962-68, more than 200 large projects were completed each year (Beaumont 1978: 40).

Rivers: No estimates of flowing water surface areas were noted in the literature; interest centers on measures of volume and flow. In theory, surface area could be calculated from existing data (in well-inventoried regions) on stream, river, and canal miles by using approximate width values based on stream order and, for canals, engineering or navigational data.

Data Quality: A number of data problems have already been mentioned. Surface water data are limited by a lack of interest in areal data, lack of monitoring at the global level, and a lack of baseline measures. Changing accuracy standards and measurement methodologies preclude the use of published figures for estimating changes through time. In addition, data are more complete for industrialized nations and for large water bodies. Information theoretically available from hydrographic offices may be very time consuming to compile.

Even when data are available, a number of definition and measurement problems remain. For example, our definition of inland water does not address the boundary problem posed by estuaries or coastal wetlands. Inland, other problems are posed by fluctuating water levels and ephemeral water bodies. Most natural water bodies fluctuate so little in size (UNESCO 1978) that changes in measurement accuracy would probably overshadow actual variations (H. Schwarz, personal communication). However, closed-basin lakes -- which include some of the world's largest -- may vary in area by a factor of 4 to 10 (UNESCO 1978). L'vovich (1979) gives a number of examples. This variability could serve to skew aggregate data. Many reservoirs fluctuate seasonally -- by as much as 40-60% -- depending on whether they are full or drawn-down (H. Schwarz, personal communication). Some data sources record maximum area, others average area.

A historical note on areal measurement is also instructive. Until very recently, calculating areas from maps was an arduous process; lake areas were often estimated by treating the water body as if it were rectangular. Comparing Russell's US Lake Survey (1895) with Greswell and Huxley's lakes and rivers encyclopedia (1965) shows similar concerns about measurement and rectangularity. This has obvious implications for the use of historical data in time series.

Finally, while remote sensing data on water body area is readily available in principle, the cost of image processing and gaps in coverage mean that the information is not necessarily accessible.

Key Terms / Phrases

Data problems

Cultivation

Importance of cultivation

Linkage to global nutrient cycles

Cropland definition

Arable land

Land under permanent crops

Cropland data sources

Cropland estimates

Cropland data quality

Forests

Importance of forests

Forest definition

Forest data sources

Forest estimates

Forest data quality

Data uncertainty

Deforestation rates

Forest assessments

Livestock

Importance of livestock

Livestock definition

Rangeland

Permanent meadows Pastures

Livestock/rangeland data sources

Livestock/rangeland estimates

Ruminants

Domesticated animals

Livestock data quality

Regional issues

North America

Latin America

Australia

Africa

Europe

Middle East and Central Asia

Asia (incl. Siberia, India, China and Mongolia)

Settlement

Importance of settlement

Settlement definition

Settlement data sources

Settlement estimates

Urban and rural population estimates and projections

Settlement data quality

Wetlands

Importance of wetlands

Wetlands definition

Wetland data sources

Wetland estimates

Wetlands in the U.S.

Wetland data quality

Surface water

Importance of surface water

Surface water definition

Surface water data sources

Surface water estimates

Lakes

Reservoirs

Rivers

Surface water data quality

Endnotes

2. An obvious issue for this category is the usefulness of examining livestock or pasture/grassland. The latter is the typical land cover associated with major livestock rearing -- the land use. To be consistent with the land use emphasis of this module, livestock has been favored conceptually over grassland/pasture, but either emphases create problems. As noted in the section on cultivation, some lands used for that purpose are also grazed during the fallow season; hence a count of land cover would typically miss this land use. Likewise, a count of livestock does not necessarily inform us of the associated land cover.

3. In this case, wild animals refers to non-domesticated animals that are used by humans. This potentially expands the area under rangeland by quite a bit in certain regions of the world.

4. Ruminants include cattle, buffalo, camels, sheep and goats. Domesticated animals are all ruminants plus pigs, chickens, ducks, and turkeys (after FAO Production Yearbook 1986).

5. John Richards creates new estimates for the extent of agricultural land based on estimates of population by McEvedy and Jones (1978).

6. For example, the World Bank's 1990 World Development Report gives tables of urban population (as percentages) and changes between 1965 and 1988; the main sources are the UN's Prospects of World Urbanization (UN 1988) and its report, Patterns of Urban and Rural Population Growth (UN 1980).

7. An example of data available in case studies is provided by Brown and Jacobsen (1987), who give the population (P) and area (A) of Sao Paulo for the years 1930 (P= 1 million; A=150km²), 1962 (P= 4million; A= 750 km²), and 1980 (P= 12 million; A= 1400 km²). They also note that, in the mid-1980s, Hong Kong's population was 5 million and its area 1000 km² including extensive urban agriculture.

8. As delineated by city wall. Except in Britain, walls encompassed virtually all cities until 1890.

9. See, for example, Crosson (1982) for the United States.

10. There have been some attempts to use remote sensing data to track land use changes associated with urbanization: for example, an NTIS newsletter reports on a Utah Study which tested the use of Landsat MSS data as a means for detecting conversion of agricultural land to urban land use (NTIS 1985). It is not known how much coverage of urban areas is available or how many such analyses have been done. Aerial photography provides an excellent means of tracking urbanization trends; the USDA, for example, commissioned a 1976 air photo study of 53 US counties (Zeimetz 1976). Coverage is expensive and therefore probably very spotty; it is also difficult to track down, especially on a global scale.

11. For example, the overview of Third World cities by Hardoy and Satterthwaite (1986) highlighted several interesting issues. They note that many cities--including Sao Paolo, Bombay, Delhi, Bangkok, Manila--contain hundreds or thousands of hectares of undeveloped land which is being held by speculators; on the other hand, in Colombia, speculators are causing the rapid urbanization of the best agricultural land. In Egypt, or 10% of the prime agricultural land has been urbanized, mostly by squatters and by subdivision. Since 1990 the Delhi urban area has increased thirteen-fold, eating up over 100 agricultural villages, and including the phenomenon of using topsoil to make bricks. Official enumeration boundaries may not reflect these changes as they occur.

12. The US Army Corps of Engineer's definition is: "areas that are inundated or saturated by surface or ground water at a frequency and duration sufficient to support, and that under normal circumstances do support, a prevalence of vegetation typically adapted for life in saturated soil conditions" (cf. e.g., US Corps of Engineers 1987).