Abstract. Nighttime satellite images of the Earth showing city lights were merged with census data and a digital soils map in an effort to estimate the extent of developed land in the United States and the impact of development on soil resources. The urban areas defined by "city lights" had mean population densities of 1,033 persons/km² and 427 housing units/km² (4.13 persons and 1.7 households/acre). Urban areas accounted for 2.7% of the surface area in the United States, an area approximately equal to the state of Minnesota or one-half the size of California. A United Nations Food and Agriculture Organization soils map of the United States was overlaid on the nighttime "city lights" image to determine which soil types are most impacted by development. The more limiting factors a soil has, the more difficult or expensive it is to farm; consequently a soil fertility classification system based on physical factors that limit agricultural production was used to rank soils. Results for the United States show that the residential, commercial, and industrial development, known as "urban sprawl," appears to be following soil resources, with the better agricultural soils being the most affected. Some unique soil types appear to be on the verge of being entirely covered by urban sprawl. The conversion of good agricultural soils to nonagricultural use may have long-term ramifications for sustainable development at the local, regional, and global levels.

The postagricultural growth of human populations, combined with technological advancement, has led to the widespread transformation of natural ecosystems into those dominated and heavily managed by human beings. The potential impact of this process on Earth's biological and geochemical systems is a current subject of debate, and concerns range from those dealing with biosphere-atmosphere interactions and global climate change (Kates et al. 1990) to the preservation of biodiversity, sustainable development, economics, and agricultural productivity (Vitousek et al. 1986; Ehrlich and Wilson 1991; Raven 1991; Ehrlich et al. 1995).

The conversion of natural systems to agricultural production has been the primary basis for the successful growth of human populations for the last 9,000 years (Kates et al. 1990). The conflict between urban and agricultural land use, however, is only now becoming a subject of controversy. The transformation of productive agricultural land to urban use under burgeoning populations has become a contentious element in debates over sustainable development and food security (Ehrlich 1989; Daily and Ehrlich 1992; Ehrlich and Ehrlich 1992). As more land is converted to urban uses, the question arises as to whether this trend represents a systematic reduction in our ability to produce food by placing our infrastructure on the most productive soil resources. A disturbing consequence of this urbanization process is a growing dependence on ever greater yields per unit area (on soils that remain) or a reliance on more distant soil resources and agricultural production.

Given present demographic trends, it is important that issues of agricultural versus urban land use be resolved. An increasing number of regional populations may be at risk of food shortages in the future as a result of sociopolitical and economic instability (e.g., war, economic depression, social upheaval, etc.) with their consequent effects on global food supplies. While the reality of some agricultural land loss is accepted, both the magnitude and the potential effect are hotly debated. Central to much of the debate is the difficulty in acquiring accurate measurements of the area of urban land use, monitoring changes in urban land use, and assessing the impact of these changes on agricultural land area or production in a way that can be used in rational, cost-beneficial analyses (Parsons 1977; Meyers and Simon 1994).

Mapping Urban Sprawl Using Remote Sensing

Surprisingly, measuring the extent of urbanization using conventional methods has been difficult, even in the United States where modern census procedures are used. In 1977 for example, the U.S. Department of Agriculture (USDA) announced figures concerning the loss of agricultural land in the United States due to urban sprawl. The figures, which were derived by using census data, sparked a controversy which resulted in two significant revisions between 1977 and 1987 before finally being set at a total area of 116.4 million hectares of land converted to urban use (Meyers and Simon 1994). Thus, census data are a valuable resource but their interpretation is controversial, i.e., they are not a substitute for direct observations of the land surface.

The use of satellite remote sensing is an obvious alternative for detecting and monitoring global change and for classifying land-use transformation. Satellite sensors such as the Advanced Very High Resolution Radiometer (AVHRR), Landsat Thematic Mapper (TM), and the French SPOT system have been used successfully for measuring deforestation, biomass burning, and other landcover changes including the expansion and contraction of deserts (Skole and Tucker 1993; Tucker et al. 1991; Sellers et al. 1995). However, remote sensing techniques are just beginning to be used to monitor the conversion of agricultural land to infrastructure (i.e., the process of urbanization). Land conversion to urban, suburban, or commercial and industrial use is one of the land-use processes most disruptive to vegetation. Yet urbanization is a difficult process to measure on a global scale because the spectrally diverse landcover types found within urbanized areas are easily confused with nonurban areas when interpreting satellite data. In addition, coarse-resolution data sets (pixels > 1 km²) have inadequate spatial and spectral resolution for reliably determining urban infrastructure, while the higher resolution data sets (100 - 1,000 m²) present problems of analysis due to the vast data volumes required for processing, high costs of acquisition, and difficulties in automating interpretation. Most remote sensing literature dealing with urban areas emphasizes detection and mapping of traditional land-use/landcover types using high-resolution image data sources such as SPOT and/or Landsat TM data (Ridd 1995). Although some mapping of urban areas has been done with lower resolution sensors such as AVHRR (Ridd 1995), the development of techniques to accurately identify urban lands using this system is still in its infancy.

New satellite imaging approaches are needed to provide accurate information about the location and extent of human habitation and urbanization at regional, continental, and global scales. One approach uses nighttime satellite data sets as a means of detecting and mapping urban land use and merging those data sets with soils maps to estimate the potential impact of urbanization on basic soil resources. Nighttime images of the Earth acquired at visible wavelengths by the Defense Meteorological Satellite Program's Operational Linescan System (DMSP/OLS) provide a dramatic picture of urbanization through the detection of city lights as they are seen from space (Kramer 1994; Elvidge et al. 1997; Fig. 3-1). The stark contrast between lighted and unlighted lands provided by this type of image, combined with the sensor's moderate spatial resolution (2.6 km) and large area covered per scene, makes it an obvious choice for classifying and mapping land transformation to urban and suburban uses over large areas.

Satellite data acquired at night from the DMSP/OLS were merged in a computer with digitized soils maps and census data to make a first approximation of how urban land use in the United States may be affecting agricultural potential. Merging the data sets allowed us to look at how each data set related to the other, much like overlaying map transparencies on top of each other. In this way, we could overlay the "city lights" images on the soils maps to examine whether or not the most productive soils were being lost to urban sprawl. Results of a continental overview of the conterminous United States (lower 48) and regional analyses for California, Illinois, Wisconsin, and Florida--the four states with the largest agricultural economies ranked by total market value of products according to the 1987 U.S. Census of Agriculture--are described in this chapter (see Imhoff et al. 1997a, 1997b).

Nighttime Views of the Earth from the DMSP/OLS Satellite (City Lights)

The DMSP/OLS satellite sensor is sensitive to very low intensity light sources, including city lights, lightning, moonlit clouds, fires, and other bright surfaces. The sensor was originally designed to allow the observation of moonlit cloud cover for meteorological forecasting and nighttime flight operations by the U.S. Air Force. Pointed down at Earth's surface, this satellite sensor takes dramatic pictures of Earth's cities on new moon nights when the city lights are the brightest objects on the surface. A "city lights" data set based on this sensor is particularly attractive because it allows easy identification of lit and nonlit places, which correspond to heavily populated and less populated areas (see Welch 1980; Welch and Zupko 1980).

The satellite image data set used here is a composite generated by the National Oceanic and Atmospheric Administration (NOAA) from over 230 cloud-free nighttime images of the United States gathered between 1994 and 1995 (Elvidge et al. 1997). After some initial processing, a final map product (Fig. 3-1) was generated. The final map product compared very favorably with census estimates of urban area in the United States (Imhoff 1997b).

Fig. 3-1. Nighttime "city lights" image for the continental United States derived from a 231-orbit composite of the U.S. Air Force Defense Meteorological Satellite Program's Optical Linescan System. The nation's highway system is overlaid on the image. Urban areas are shown as yellow and represent areas heavily lit at night as observed from the satellite.

In order to determine what housing and population densities are found in the lit urban areas on the "city lights" satellite image, we merged United States census data on housing density and population for a number of large metropolitan areas with the satellite data. The housing and population density statistical information was extracted from the 1990 Census of Population and Housing and associated maps (U.S. Bureau of the Census 1991, 1992). Merging census and satellite data allowed us to "tag" the lit areas on the nighttime satellite image with numbers on population and housing density. The mean housing density for the urban (lit) areas was 427 housing units/km² and the mean population density was 1,033 persons/km². Since the population and housing density of the lit areas almost certainly precludes major agricultural production beyond small home gardens and subsistence agriculture, we classified these areas as being lost to large-scale agricultural production.

Assessing Soil Resources -What Makes One Soil Better Than Another?

In order to assess the soil resources of the United States, we chose to make our first estimate using the United Nations Food and Agriculture Organization's (UNFAO) Digital Soils Map of the World (1975, 1992). At present, the UNFAO map is the only global digital soils database available. While it is coarse in resolution (1:5,000,000), it has been used extensively in global- and continental-scale analyses of terrestrial resources, biological processes, and global change (Potter et al. 1993, 1994). While other, finer scale digital soils maps are available for the United States, they are not available for all parts of the globe. We selected the UNFAO map so we could compare the U.S. results with those from the rest of the world for an analysis of global food security.

The UNFAO soils map identifies approximately 315 soil units for the United States based on their physical and chemical composition, their topographic situation, and the local climate. The soil units were ranked using a soil fertility index known as the Fertility Capability Classification (FCC) system (Buol et al. 1974; Sanchez et al. 1982). This system provides a basis for comparing and rating soils world-wide for their suitability for agricultural use. The ranking system assesses the textural, structural, chemical, and climatic characteristics of the soil that have a direct relationship to fertilizer, plowing, and irrigation requirements for successful farming. The FCC system assigns condition modifiers, or limiting factors, to each soil type, if any are present. Condition modifiers include such factors as dry or seasonally dry conditions, low cation exchange capacity (CEC), aluminum toxicity, acidity, iron and phosphorus fixation, gleying, basic reaction, potassium deficiency, salinity, etc. As the number of limiting factors present in a soil unit increases, so does the cost of agricultural production, because the limiting factors need to be overcome through liming, irrigation, fertilization, or other processes. In a sense, ranking the soils by the number of limiting factors is like an economic rating of the costs required to farm a plot of land. The best soils are those with the fewest limiting factors or with those factors most easily and cheaply overcome. A soil with no limiting factors is good for farming as-it-already-is--there is no need to add to it to make crops grow. Soils with many limiting factors may be prohibitively expensive to farm. While not perfect, the FCC system can be quickly applied to the UNFAO soils maps and can be interpreted as a surrogate "native soil productivity" rating for agriculture.

For our analysis, we ranked soil units in the United States as ranging from those having zero or no limiting factors to those with eight limiting factors. Soils with no limiting factors are the best soils for agriculture (in terms of least cost of production), and soils with seven or eight limiting factors were considered nearly impossible to cultivate because of the presence of a cumulative number of impediments (dry, saline conditions, shallow soils, high slope, etc.). Looking at the soils data for North America, it must be said that the United States contains a lot of good soils within its borders (Fig. 3-2). About 53% of the total soil area of the United States is made up of soils with two or fewer limiting factors, and over 90% of the country is composed of soils with four or fewer limiting factors. While some limiting factors are more economically significant than others, the United States is still capable of impressive agricultural production relative to the rest of the world. Many countries in the world have relatively few soils with two or fewer limiting factors and are forced to farm soils with many obstacles. The huge investment in terracing hillsides in China is an extreme example of attempts to overcome the limiting factor of severe slope.

 
Fig. 3-2. A soils map of the United States created from the United Nations Food and Agriculture Organization digital soils map of the world. Soils are classified by their number of agronomic limiting factors. Soils with a high number of limiting factors are problematic and require remediation for agricultural production. The best soils for agriculture have no or few limiting factors.

An estimate of urban land use and its impact on agricultural soil resources was made by classifying the "city lights" images into urban and nonurban land use classes and then comparing them to the fertility-ranked soils map. The amount of soil area in urban use (under city lights) was calculated for each soil type mapped for the entire conterminous United States. Any soil units found under the "city lights" were considered to be in urban use and therefore unavailable to mechanized agriculture. The soil units under "city lights" were assessed individually (all 315 units separately) and then grouped by the number of FCC limiting factors they had (nine classes: zero to eight limitations). An additional computer map overlay showing state boundaries was used to select data for individual states for a more localized view of soil losses due to urban sprawl.

Urban Occupation of Soils in the United States

By our satellite-based assessment, about 2.67% of the total land area of the conterminous United States is now in urban use. While this may seem like a small area, if all of this urban/suburban or commercial/industrial land were coalesced into a single entity, it would be a city the size of the entire state of Minnesota or a city that would occupy all of California from San Jose to San Diego. Furthermore, when considering the effects of land conversion, it is important to address the issue of quantity versus quality. For example, in our analysis, only about 3% of the land surface is urbanized, but the best soils are being developed first. If the situation is examined from a continental scale, it is clear that the percentage of the land in development generally increases as the number of limiting factors decreases (Fig. 3-3; Table 3-1). These findings support the argument that the best soils are the first victims of urban sprawl. There is evidence, however, that some preservation of the very best soils may be taking place. For example, the graph shows that soils with no limiting factors (i.e., the very best agricultural soils) are a little less developed (2% urban) than the soils with one limiting factor (4.22% urban). If it is true that this is evidence of preservation of the very best soils, however, it is also true that it is happening at the expense of the second best agricultural soils.

Table 3-1. U.S. soils (United Nations Food and Agriculture Organization) grouped by number of physical factors limiting agricultural production.

---

Number of limiting factors	Total area of soil class (km 2)	Soil area covered by lights (km 2)	% soil area covered by lights	% urban area from 1990 census
0	1,282,805	26,558	2.07
1	1,663,218	70, 248	4.22
2	1,172,880	32,723	2.79
3	1,608,131	37,060	2.30
4	1,353,633	30,872	2.28
5	216,542	5,267	2.43
6	309,090	1,846	0.60
7	15,681	233	1.49
8	93,234	979	1.05
Totals	7,715,214a	205,786b	2.67	2.9c (222,977 km2)

Since these soils often exist closely to one another, it is understandable that if the best soils are prorected, development would be pushed onto the soils immediately around them. It then becomes important to address the question of how soils are distributed locally and what constitutes the "best" agricultural soils in a specific region. For example, in Fig. 3-3, which shows all the soils in the United States grouped together by limiting factors, urban development appears to be more or less evenly distributed across soils with two, three, or four limiting factors. This observation would lead one to believe that development is mostly taking place on soils with less agricultural capacity. What is not evident in this graph (or at the national level), however, is that in many parts of the country, soils with two, three, or four limiting factors are the "best" agricultural soils in those regions. Any loss of those soils could lead to significant reduction in agricultural potential there. California offers a prime example of this situation, where the soils of the central valley all have one limiting factor, yet they are the most important agricultural soils in the state and perhaps the entire United States.

Fig. 3-3. Percent of soil area covered by lights in the United States. The soils are grouped into nine classes depending on the number of limiting factors they have. The greater the number of limiting factors, the more effort required to use a given soil unit for agriculture. The soils with fewer limiting factors are bearing the brunt of urbanization, but soils with zero limits are less urbanized than others, possibly giving evidence of marginal preservation.

It is not unusual that the best soils should fall to urban development. Most major cities of the world are located on or near river flood plains or deltas, which are also the most productive farming areas. Unfortunately, poorly rated soils are often unsuitable not only for agriculture but also for construction because of steep slopes or other limiting physical factors. The economics of urbanization, however, make it easier to overcome some soil-related problems.

Fig. 3-4. Percent of soil area covered by lights for all 315 soil units for the United States. Soil units are from the United Nations Food and Agriculture Organization digital soil map of the world. The soil units are rated by using the Fertility Capability Classification system. Note that some soils are nearly completely covered by "city lights," indicating a high degree of urbanization.

Another result of this analysis is the apparent loss of unique soil types. The pressure of urban sprawl is not equally distributed over all of the soil types. For example, between 50% and 70% of some FAO soil units (types) are now in urban use. Two mesic humid soils suitable for field crops in eastern New York (Bd24-2a) and southern New York and northern New Jersey (Bd20-2b) are 53% and 71% covered by lights, respectively (Fig. 3-4). These soil types can be considered endangered since the physical alteration of the soil profile is substantial. The potential loss of entire soil mapping units, with their unique history of formation and biology, gives rise to the issue of the loss of biological diversity inherent in those soils.

Urban Occupation of Soils in the Four Top Ranked Agricultural States

To further examine trends in urbanization, we applied our analysis to California, Illinois, Florida, and Wisconsin, the states with the highest agricultural production in the United States (market value according to the U.S. Bureau of the Census 1987). Results are considerably less ambiguous than the continental view and show a disturbing, positive correlation between urban sprawl and the loss of productive agricultural soils. In these states, urban development appears to be taking place on the soil groups best suited to agriculture (Table 3-2).

California and Illinois perhaps best demonstrate this trend, with urban development occupying soils with the fewest FCC limitations to agriculture. In California, nearly 16% of the surface area of the best soil for agriculture is in urban use. In Illinois, 8.27% of the soils that have no limitations is lost to urban use. Almost 5% of the total land surface of California is in urban use, and 5.5% of the total land area in Illinois is urbanized. In both cases the percent of soil surface area in urban use drops substantially as the number of limiting factors increases.

In Wisconsin, however, the soil group having the most surface area in urban use is soils with one limiting factor, leaving the soils with zero limiting factors less developed. In the case of Wisconsin, this pattern may reflect preservation of the best of farmland and deserves further examination. If the "best" two soils in Wisconsin (those with zero and one limiting factors) are tallied together, though, then most of the development is clearly still on the best soils, leaving those with two or more limiting factors relatively undeveloped. According to our inventory, about 2.6% of the total surface area of Wisconsin is urbanized.

The pattern for Florida resembles that of Wisconsin, although the economic and social pressures at work there are extremely different. Even though Florida generates a considerable income from agricultural products (nearly $4 billion annually), the highly sought after living environment of its coastlines causes most of the urban development to occur there, where the soils are less suited to agriculture. This situation may give a false appearance of farmland preservation. Of the soil types with two or fewer limiting factors, Florida has lost 12.1% to urban use, while over 10% of the entire surface area of Florida is urbanized.

Long Term Impacts - Food Security and Conservation

At the continental scale the trend in land use conversion indicates that development is favoring the soils with fewer agricultural limiting factors. While preservation of the very best soils does seem to take place in some cases, it is occurring at the expense of the next best soils, which are often located nearby. This pattern is evident in the four most economically important farm states, except that in some states the very best soils do not appear to be preserved at all (e.g., California). Our data tend to lend credence to the theory that development is associated with soil resources with higher production potential. Since many of the very same physical properties that make soils good for agriculture are also good for construction, the relationship is logical. Economic incentives, then, are negatively synergistic: "good" soils bring economic wealth, thereby encouraging development, but are also themselves attractive soils on which to build. Short-term economic forces may tend to undervalue agricultural use relative to urban development, especially when agricultural production can be economically shifted to distant areas due to inexpensive transportation. That the top four agricultural states are experiencing this trend, however, bodes ill for sustained productivity in the United States.

Table 3-2. Results for soils (United Nations Food and Agriculture Organization) of four top producing agricultural states; soil units grouped and ranked by number of physical factors limiting agricultural production (light data from DMSP/OLS composite 89+% lit).

---

Number of limiting factors	1987 agricultural production ($ billions)a	Total area of soil class in state (km 2)b	Soil area covered by lights (km2)	% soil area in state covered by lights	% urban area from 1990 censusc
California	$12.71
1		53,149	8,499	15.99
2		39,619	3,510	8.86
3		112,186	1,812	1.61
4		57,451	1,630	2.84
5		61,538	3,881	6.31
8		80,627	788	0.98
Totals		404,571	20,120	4.97	5.2 (21,176 km2)
Illinois	$4.97
0		68,536	5,670	8.27
1		66,479	1,960	2.95
3		10,445	364	3.49
Totals		145,460	7,994	5.50	5.5 (7,854 km2)
Wisconsin	$4.18
0		23,648	481	2.03
1		54,519	2,823	5.18
2		18,390	125	0.68
3		23,706	192	0.81
4		23,064	87	0.38
Totals		143,328	3,707	2.59	2.9 (4,086 km2)
Florida	$3.93
1		8,780	617	7.03
2		37,458	4,996	13.34
3		80,789	7,368	9.12
4		11,331	927	8.18
Totals		138,358	13,908	10.05	9.5 (13,282 km2)

While the overall agricultural potential of the United States may not be seriously diminished at present, if the trend is allowed to continue, the country may soon experience a decline in agricultural production. Currently, there is less and less reliance on local agricultural products in the United States. Many grocery stores are stocked mainly with produce generated in a few primary agricultural zones in the United States and abroad. As the local soils are converted to nonagricultural uses, those localities will be even more reliant on their access to national or international markets. As such they will be vulnerable to changes in those markets and will be in direct competition with a very broad and, in some cases, wealthy customer base for the products. If the need arises to revitalize local agriculture to support growing populations nearby, only the poorer soils will be available for use. These soils will require more fertilizer and other inputs since more limiting factors will have to be overcome to make the soil produce a crop. The need to farm poorer soils will tend to increase the cost of production and the price of food. An example of this potential can be found in the state of Pennsylvania. Traditionally rich in farmlands, university experts estimate that Pennsylvania is losing 1% of its prime agricultural land to development each year, according to recent estimates (G.W. Petersen, Pennsylvania State University, personal communication). If the trend continues, in 100 years there will be no more prime agricultural land in the entire state. At that time, the human population will be much larger, suggesting that Pennsylvania will become increasingly dependent on outside agricultural resources.

In a future world of large human populations, where will those critically needed soil resources be found? Many countries are all depending upon the surplus production of the United States and other productive regions of the world to help carry their growing populations through the next 50 years. However, the United States, too, is depending on its current surplus capacity to feed its growing population. In fact, the U.S. agricultural production capability as it is now may be overcommitted by a factor of three by 2050 since a large percentage of the world's population expects that the surplus production will be available to them. Given this possibility, it would be prudent to protect the best agricultural soils from development. Not only should the best soils be protected, but it is vitally important that the farmland conservation effort take place at the local level and not simply at the national level. Consideration should be made for sustainable development at the local level, so that there is not the forced reliance of local populations on the interstate transportation systems that consume huge amounts of fossil fuels and are deteriorating under heavy use. While more detailed local analyses are needed to shed light on how each region of the country will be affected by the loss of soil resources, one certain outcome is that depletion of productive soil will bring with it a dependence on more distant resources and require ever higher yields per remaining acre on poorer soils.

Another concern brought out by this study is the potential loss of certain soil types or unique soil units to urbanization. Our results indicate that four soil types, as classified in the UNFAO system in the United States, may be in danger of disappearing under urban/suburban structures. Is the loss of "soil diversity" meaningful in a biological or economic sense? The study of soil biodiversity is a relatively new field, yet recent studies indicate that great diversity may exist in soils, with their unique physical structure, environment, and history of formation (Huston 1993). The loss of soil types may therefore represent loss of whole biological communities unique to that soil type. The conservation of soil diversity may bring into question the wisdom of converting to agriculture soils that have not previously been cultivated. Agriculture seriously disrupts the soil by changing its chemistry, structure, and ecological dynamics. Many of the soils that have already undergone agricultural transformation are in locations that (for the most part) limit soil loss to erosion and other adverse impacts (even so, soil erosion is an increasingly severe problem). As stable soils become unavailable to agriculture through conversion to urban/suburban infrastructure, soils less suited for cultivation may be used for farming. Farming such marginal soils could increase erosion resulting in the destruction of many soils, right down to the bedrock or parent material. This process has already occurred in many parts of the world such as the Caribbean Islands (Haiti) and areas in South and Central America, Asia, and Africa (Ehrlich 1997).

Our research demonstrates the use of nighttime images of the Earth, from the DMSP/OLS sensor, to measure the extent of urbanization and assess its impact on soil resources. Results show that urban development follows soil resources and that, in general, the best soils are being converted to nonagricultural uses by urban sprawl. In addition, some distinct soil types, with their unique physical structure and history of formation, may be in danger of elimination, likely resulting in a substantial loss of below ground and above ground biodiversity. The conservation of soil diversity is only now being discussed in scientific circles. With the increasing global demands on agricultural production, protection of the best agricultural soils emerges as an important priority, especially when considered from the viewpoint of future generations.

The paramount importance of soil resources suggests a need for a global assessment of urban sprawl and its effect on soil productivity. This assessment should include not only impacts on the soil resource but a more robust analysis of potential loss of global agricultural productivity. Such an analysis, using soil- and climate-based production models, will help develop forecasts for management, planning, and climate change research. Given the very real possibility of temporal and spatial displacement of climatic regimes under global change scenarios, it is critical that any future work include not only basic soil fertility characteristics but a realistic assessment of the impacts of climate change on agricultural systems.

This work was supported by the National Aeronautics and Space Administration's EOS Interdisciplinary Science Program. We would like to thank Dr. Herbert Kroehl of the National Oceanic and Atmospheric Administration's National Geoscience Data Center for the DMSP/OLS data sets and Dr. Norman Bliss of the U.S. Geological Survey's EROS Data Center for his help with the UNFAO-FCC system.

American Farmland Trust and Department of Geography, North Illinois University. 1993.

Farming on the edge. Map. American Farmland Trust, Washington, D.C.

Buol, S. W., P. A. Sanchez, R. B. Cate, Jr., and M. A. Granger. 1974.

Soil fertility capability classification, Centro International de Agricultura Tropical, Colombia.

Daily, G. C., and P. R. Ehrlich. 1992.

Population, sustainability, and Earth's carrying capacity. BioScience 42:761-771.

Ehrlich, P. R. 1997.

A world of wounds: ecologists and the human dilemma. Ecology Institute, Ohlendorf/Luhe, Germany. 210 pp.

Ehrlich, P. R. 1989.

The limits to substitution: meta-resource depletion and a new economic-ecological paradigm. Ecological Economics 1:9-16.

Ehrlich, P. R., and A. H. Ehrlich. 1992.

The value of biodiversity. Ambio 21(3):219-226.

Ehrlich, P., A. Ehrlich, and G. Daily. 1995.

The stork and the plow. Grosset/Putnam, New York, N.Y.

Ehrlich, P., and E. O. Wilson. 1991.

Biodiversity studies: science and policy. Science 253:758-762.

Elvidge, C., K. E. Baugh, E. A. Kihn, and E. R. Davis. 1997.

Mapping city lights with nighttime data from the DMSP Operational Linescan System. Photogrammetric Engineering and Remote Sensing 63:727-734.

Huston, M. 1993.

Biological diversity, soils, and economics. Science 262:1676-1680.

Imhoff, M. L., W. T. Lawrence, D. C. Stutzer, and C. D. Elvidge. 1997a.

A technique for using composite DMSP/OLS "city lights" satellite data to accurately map urban areas. Remote Sensing of Environment 61:361-370.

Imhoff, M. L., W. T. Lawrence, C. D. Elvidge, T. Paul, E. Levine, M. Prevalsky, and V. Brown. 1997b.

Using nighttime DMSP/OLS images of city lights to estimate the impact of urban land use on soil resources in the U.S. Remote Sensing of Environment 59:105-117.

Kates, R. W., B. L. Turner, II, and W. C. Clark. 1990.

Pages 1-17 in B. L. Turner, editor. The earth transformed by human action.Cambridge University Press, Cambridge, U.K.

Kramer, H. J. 1994.

A.19 DMSP (Defense Meteorological Satellite Program) and A.19.1 DMSP data availability - visible and infrared imagery. Pages 61-67 in Observation of the Earth and its environment. Survey of missions and sensors. Springer-Verlag, Berlin.

Meyers, N., and J. Simon. 1994.

Scarcity or abundance: a debate on the environment. Norton, New York, N.Y. 254 pp.

Parsons, J. 1977.

Population fallacies. Elek/Pemberton, London, U.K. 286 pp.

Potter, C., P. Matson, and P. Vitousek. 1994.

In W. K. Michener, editor. Environmental information systems Analysis. Taylor and Francis, London, U.K.

Potter, C., J. Randerson, C. Field, P. Matson, P. Vitousek, H. Mooney, and S. Klooster. 1993.

Terrestrial ecosystem production: a process model based on global satellite surface data. Global Biogeochemical Cycles 7(4): 811-841.

Raven, P. H. 1991.

Pages 121-137 in Global change and the human prospect: issues in population, science, technology, and equity. Sigma Xi, Research Triangle Park, N.C.

Ridd, M.K. 1995.

Exploring a V-I-S (vegetation-impervious surface-soil) model for urban ecosystem analysis through remote sensing: comparative anatomy for cities. International Journal of Remote Sensing 12:2165-2185.

Sanchez, P. A., W. Couto, and S. W. Buol. 1982.

The fertility capability soil classification system: interpretation, applicability, and modification. Geoderma 27:283-309.

Sellers, P. J., B. W. Meeson, F. G. Hall, G. Asrar, R. E. Murphy, R. A. Schiffer, F. P. Bretherton,

R. E. Dickinson, R. G. Ellingson, C. B. Field, K. F. Huemmrich, C. O. Justice, J M. Melack, N. T. Roulet, D. S.

Schimel, and P. D. Try. 1995.

Remote sensing of the land surface for studies of gobal change: models--algorithms--experiments. Remote Sensing of Environment 51:3-26.

Skole, D., and C. J. Tucker. 1993.

Tropical deforestation and habitat fragmentation in the Amazon: satellite data from 1978 to 1988. Science 1905-1909.

Tucker, C. J., H. Dregne, and W. Newcomb. 1991.

Expansion and contraction of the Sahara Desert from 1980 to 1990. Science 253(5017):299-301.

United Nations Food and Agriculture Organization. 1992.

The digitized soil map of the world notes, Release 1.1. United Nations Food and Agriculture Organization, Rome, Italy.

United Nations Food and Agriculture Organization. 1975.

Soil map of the world, Vol. II: North America. United Nations Food and Agriculture Organization, Rome, Italy.

U.S. Bureau of the Census. 1987.

1987 Census of Agriculture. U.S. Department of Commerce, Washington D.C.

U.S. Bureau of the Census. 1991.

Summary Tape File 1B. U.S. Department of Commerce, Washington, D.C.

U.S. Bureau of the Census. 1992.

Topologically Integrated Geographic Encoding and Referencing System (TIGER) Line Files. U.S. Department of Commerce, Washington, D.C.

Vitousek, P.M., P. Ehrlich, A. Ehrlich, and P. M. Matson. 1986.

Human appropriation of the products of photosynthesis. BioScience 36:368-373.

Welch, R. 1980.

Monitoring urban population and energy utilization patterns from satellite data. Remote Sensing of Environment 9:1-9.

Welch, R., and S. Zupko. 1980.

Urbanized area energy patterns from DMSP data. Photogrammetric Engineering and Remote Sensing 46(2):201-207.