Home


Monitoring Changes in Landscapes from Satellite Imagery


by
Thomas R. Loveland
U.S. Geological Survey
H.L. Hutcheson
South Dakota State University
It has been said that "a model without data has no predictive power" (Rasool 1992). The need to model the extent, condition, and trends in biological resources is a central element for most environmental assessments. Whether the issues involve biological diversity or the effects of changing biogeochemical cycles, accurate baseline data are essential to the environmental monitoring and modeling of future environmental conditions.
Methods and tools for monitoring natural vegetation at the level of plots to small sitesfrom a single square meter to millions of square meters are well developed and widely used (Küchler and Zonneveld 1988), but at the national level there is a lack of comprehensive environmental data from which we can assess national patterns of environmental diversity. The early western explorers conducted extensive surveys of regional geological, topographic, and ethnographical resources but did not collect enough detailed biological data that could provide us with a starting point for understanding the environmental transformations that have taken place since the nation was founded. More recently, Klopatek et al. (1979) tried to assess the modification of natural vegetation in the United States but concluded that the exercise was difficult because recent land-use changes were typically undocumented. As a result, assessments of current environmental conditions are too frequently based on decades-old data.

Current Estimates of Vegetation Patterns
Perhaps the best estimate of vegetation patterns of the conterminous United States before European settlement is from Küchler's potential natural vegetation (Küchler 1964). His map of the potential natural vegetation divides the country into 116 potential vegetation types. He defines potential natural vegetation as the vegetation that would exist today if humans were removed from the scene and if the resulting plant succession were telescoped into a single moment.
There are, however, limitations in using potential natural vegetation as an indicator of pre-European settlement vegetation patterns, including problems related to the coarse scale of the Küchler map (1:3,168,000), the processes of succession, and the determination of climax vegetation types (Klopatek et al. 1979). Küchler, for example, attempted to show the potential climax stage of vegetation, although some ecosystems never reached climax because of natural controls such as fire. Küchler also pointed out the difficulties and the assumptions in using the terms "natural" and "original" vegetation. His map, however, probably represents the best approximation available today of the continent's vegetation before European settlement.
The most current picture of national land-cover vegetation patterns is from a 1990 data set produced by the U.S. Geological Survey (USGS; Loveland et al. 1991). The USGS land-cover data were interpreted from 1990 satellite imagery from the Advanced Very High Resolution Radiometer (AVHRR) sensor aboard the National Oceanic and Atmospheric Administration's polar-orbiting meteorological satellites. The USGS map of land cover is limited in its use for local applications because of the coarse ground resolution of AVHRR data and its subsequent inability to distinguish vegetation structure, seral stages, and exotic versus natural vegetation. It does, though, provide a picture of vegetation and land-cover patterns at the national level. For example, in the lower 48 states, about 38% of the land is forested, 29% is rangeland or grassland, and 23% is agricultural land. While the USGS land-cover study did not identify urban lands, information from the Defense Mapping Agency's Digital Chart of the World shows that at least 14,500 km2 (5,655 mi2) or 1.0% of the conterminous United States is urbanized (Danko 1992).
It must be noted that a comprehensive assessment of accuracy of the 1990 land-cover map has not been completed, although an independent study shows that the classification of forest lands is within 4% of the estimate of the U.S. Forest Service (Turner et al. 1993). Comparisons with selected state land-cover maps and U.S. Department of Agriculture crop area statistics have also shown general correspondence between land-cover estimates at the national level (Merchant et al. 1995).

Change in Natural Vegetation
The estimated extent of change in the natural vegetation since European settlement is derived by comparing Küchler's potential natural vegetation (Küchler 1964) with the 1990 land-cover data set produced by the USGS (Loveland et al. 1991). Both potential natural vegetation (Fig. 1) and 1990 land cover (Fig. 2) have been generalized to show six vegetation groups: needleleaf forest, broadleaf forest, mixed forest, grassland, shrubland, and grassland-shrubland. Note that the 1990 land-cover classification does not distinguish between natural and altered vegetation (e.g., an even-age tree plantation is mapped as forest even though it does not have the ecological value or function of a natural forest). The 1990 land-cover map (Fig. 2) also includes four additional categories: urban areas, cropland, cropland-woodland mosaics, and cropland-grassland mosaics.

Fig. 1. Grouped categories of potential natural vegetation aggregated from Küchler (1964).
A representation of the percentage of land modified from its natural state by either cultivation or urban development was produced by calculating the percentage of 1990 agriculture and urban lands found within each Küchler vegetation type (Fig. 3). Although more than 61% of the conterminous United States is covered with the same dominant vegetation as Küchler suggests, the percentage varies considerably by region. Almost 92% of the western forests region remains covered with tree species, while only 29% of the central and eastern grasslands region remains as grasslands.

Fig. 2. Grouped categories of 1990 land cover depicting 1990 conterminous U.S. land cover that was developed from 1990 AVHRR imagery.
It must be understood that a low percentage of agricultural or urban lands in a region does not imply that the landscape exists in a pristine, natural state. In some cases, the "natural" vegetation may be altered substantially by local land-use practices such as grazing and logging or changed by the introduction or invasion of non-native vegetation. Küchler (1964) recognized overgrazing as having long altered the central grassland. He also mentioned Kentucky bluegrass (Poa pratensis) as an exotic that has become the dominant grassland in regions including the Black Hills of South Dakota. As a result, many areas that are not affected by agriculture or urbanization are far from their natural state and do not perform the same ecological role as did the original ecosystem. The coarse nature of the AVHRR data and the lack of detailed baseline data on original vegetation conditions do not allow for the detection of these important landscape qualities. While these assessments have limitations, the comparisons represent the type of analysis and monitoring that can be done with a properly designed operational vegetation monitoring system.

Fig. 3. Percentage of Küchler's potential natural vegetation types (Küchler 1964) that have been converted to agricultural and urban land cover. The lighter tones represent the higher levels of human modification. Percentages of modification are displayed as deca-percentiles.
The areas with the highest percentage of land modified from its natural condition are in the central United States. With one exception, the most intensively cultivated areas coincide with Küchler's grassland or mixed grassland-forest types. The exception is the elm-ash forest south and west of Lake Erie (91.03% cropland). This vegetation type covers a relatively small area (23,103 km2; 9,010 mi2). The principal vegetation type that is now more than 90% cropland or mixed cropland is Küchler's bluestem prairie, which covers 271,990 km2 (106, 076 mi2), 3.5% of the conterminous United States. The 1990 land-cover data indicate that 90.28% is predominately cropland.
The least cultivated of Küchler's types are grama-tobosa prairie (0.18%), trans-pecos shrub savanna (0.28%), creosote bush (0.60%), and blackbrush (0.66%). These four types are all part of the western shrub and woodland group. In the eastern United States, the most "uncultivated" of Küchler's types is the mixed-mesophytic forest (5.07%), which covers an area of 496,790 km2 (193,748 mi2) and has been noted as having the highest species diversity of all the eastern broadleaf forests (Braun 1950).
There are several other ways to evaluate the differences in the 1990 landscape versus the potential natural state. For example, certain Küchler types retain the highest percentages of areas not covered by agriculture or urban land cover (Table 1), although these areas may be presently highly disturbed by logging, road building, strip mining, grazing, or other activities. With the exception of the mixed-mesophytic forest and the relatively small southeastern spruce-fir forest, these are all in the western part of the country. Vegetation types from the Küchler map that have the highest percentage of urbanization on the USGS map (Table 2) are relatively small and are all coastal. Some, like coastal sagebrush, are types that are considered threatened (see Stoms and Davis, this section). For selected grassland types of the Great Plains and central lowlands, there is a decrease in percentage of cultivation from east to west (Table 3), reflecting the role of annual precipitation in conversion of grassland areas to cultivation. Table 1. Küchler vegetation types least modified by urbanization and agricultural developments.

Type and location Unaltered
%
Grama-tobosa prairie (Arizona, New Mexico) 99.80
Trans-pecos shrub savanna (Texas, New Mexico) 98.82
Oak-juniper woodland (Arizona, New Mexico, Texas) 98.67
Southeastern spruce-fir forest (southern Appalachia) 98.18
Silver fir-Douglas fir (Oregon, Washington) 97.08
Cedar-hemlock pine forest (northern Rocky Mountains) 97.30
Grama-tobosa shrub-steppe (Arizona, New Mexico) 97.14
Creosote bush-tarbush (Arizona, New Mexico) 96.04
Chaparral (California) 96.03
Blackbrush (Utah, Arizona) 95.67
Montane chaparral (California) 95.36
Redwood forest (California, Oregon) 94.70
Mixed mesophytic forest (Pennsylvania, West Virginia, Ohio, Kentucky, Tennessee) 94.57

Comparing forested areas from the USGS map with the Küchler map would indicate that about 57% of the potential forested area is currently covered by tree species (Turner et al. 1993). The potential impacts of these changes are significant. For example, the loss of forest cover since before European settlement (43%) has increased both albedo and carbon dioxide levels. A rise in albedo has been shown to cause a decrease in mesoscale rainfall (Charney et al. 1975). Increases in irrigated agriculture can result in a decrease in albedo, which can cause an increase in mesoscale rainfall (Barnston and Shickedanz 1984). Also, a shift from forest to grasses results in a decrease in primary productivity by a factor of two, thus reducing the rate of atmospheric carbon fixation. Table 2. Küchler vegetation types most affected by urbanization, their locales, and associated urban areas.

Type and location Urbanized
%
Fescue oatgrass (western slopes of northern coast ranges, California, San Francisco) 24.00
Subtropical pine forest (southern Florida, Miami) 21.07
Coastal sagebrush (coastal regions of southern California, Los Angeles) 15.87
Pine-cypress forest (coastal California) 6.10
Northeastern oak-pine forest (coastal New England to New Jersey, New York, Newark, Philadelphia) 5.86

Continuing Transformations
The comparison of 1990 land cover with potential natural vegetation illustrates the magnitude of change that has possibly occurred in the past 250 years. Changes in the landscape are not exclusive to that period, however; in fact, the 1990 view of United States land cover is already becoming outdated in some regions as natural and human forces continue to transform the landscape. For example, a comparison of 1970's and 1980's satellite images from the Landsat Multispectral Scanner (MSS; see box) shows that significant changes in some areas selected for examination are taking place. Landsat MSS images have been acquired over most of the United States since July 1972. With approximately 80 m x 80 m (260 ft x 260 ft) resolution, they provide a means to map in more detail the changes that have occurred in the past 22 years. Table 3. Selected grassland types arranged by percentage cultivation.

Grassland type and location Cultivation
%
Bluestem prairie (North Dakota and Minnesota southward to Oklahoma) 90.28
Wheatgrass-bluestem-needlegrass (North Dakota, South Dakota, Nebraska) 82.43
Bluestem-grama prairie (Kansas, Nebraska, Colorado, Oklahoma) 76.24
Nebraska sandhills prairie (Nebraska, South Dakota) 73.32
Wheatgrass-needlegrass (North Dakota, South Dakota, Montana, Wyoming, Colorado) 32.71
Grama-buffalograss (New Mexico, Colorado, Wyoming, Nebraska, Kansas, Oklahoma, Texas) 25.75
Wheatgrass-grama-buffalograss (South Dakota) 10.21
Grama-needlegrass-wheatgrass (Wyoming, Montana) 7.39

Future Possibilities
The vignettes presented here illustrate both the potential and the limitations associated with modeling and monitoring of environmental conditions and processes with satellite images. Clearly, baseline data are an essential starting point for these applications. Also needed is a sound framework from which baseline data can be collected, calibrated, and used in a monitoring system to target and assess environmental changes.
Remote-sensing images from orbiting satellites can play an important role in the collection of baseline vegetation data and in monitoring their status. Coarse-resolution data such as 1-km (0.62-mi) AVHRR imagery offer a means to view landscapes with daily frequency, thereby allowing the monitoring of vegetation condition both within a growing period and between years. Over a long period, AVHRR may provide a means for monitoring the subtle changes in the vegetation that may relate to such events as long-term drought. AVHRR data are not adequate for assessing the effects of more local changes. Landscape changes at the local level will be better understood with higher resolution imagery such as that provided by Landsat systems. Improved data from the sensors planned as part of the National Aeronautics and Space Administration's (NASA) Mission to Planet Earth's Earth Observing System will likely provide even better remote sensing systems for environmental monitoring.
Many components needed for a national environmental monitoring system already exist. A robust system that provides mechanisms for targeting and quantifying changes in the landscape will need to include both the synoptic overview capabilities from Earth-orbiting satellites and detailed site-specific observations of biological processes. The National Biological Service's Gap Analysis Program (GAP) provides an essential high-resolution inventory of habitat and natural vegetation for the United States by using Landsat Thematic Mapper imagery with 30 m x 30 m (98 ft x 98 ft) resolution along with substantial amounts of ancillary information such as field reconnaissance and air photos (Scott et al. 1993). Regional monitoring of the stressors to the natural systems is needed to improve the predictive capabilities of an operational monitoring system. Those systems, tied together with an integrated sampling and assessment framework, could provide a synergistic means for long-term environmental monitoring.
For further information:
Thomas R. Loveland
EROS Data Center
U.S. Geological Survey
Sioux Falls, SD 57198
References
Barnston, A.G., and P.T. Schickedanz. 1984. The effect of irrigation on warm season precipitation in the southern Great Plains. Journal of Climatology and Applied Meteorology 23:865-888.

Braun, E.L. 1950. Deciduous forests of eastern North America. Blaskiston Co., Philadelphia, PA. 596 pp.

Charney, J., P.H. Stone, and W.J. Quick. 1975. Drought in the Sahara: a biogeophysical feedback mechanism. Science 187:434-435.

Danko, D.M. 1992. The digital chart of the world. GeoInfo Systems (2)1:29-36.

Klopatek, J.M., R.J. Olson, C.J. Emerson, and J.L. Joness. 1979. Land-use conflicts with natural vegetation in the United States. Environmental Conservation (6)3:191-199.

Küchler, A.W. 1964. Potential natural vegetation of the conterminous United States: a map and manual. American Geographical Society Special Publ. 36. Princeton Polychrome Press, Princeton, NJ. 116 pp.

 Küchler, A.W., and I.S. Zonneveld, eds. 1988. Vegetation mapping. Kluwer Academic, Norwell, MA. 635 pp.

Loveland, T.R., J.W. Merchant, D.O. Ohlen, and J.F. Brown. 1991. Development of a land-cover characteristics database for the conterminous U.S. Photogrammetric Engineering and Remote Sensing (57)11:1453-1463.

Merchant, J.W., L. Yang, and W. Yang. 1995. Validation of continent-scale land cover data bases developed from AVHRR data. Pages 63-72 in Proceedings: Pecora 12 Symposium on Land Information from Space-based Systems. American Society of Photogrammetry and Remote Sensing, Bethesda, MD.

Rasool, S.I., ed. 1992. Requirements for terrestrial biospheric data for IGBP core projects. International Geosphere-Biosphere Programme-Data and Information System Working Paper 2. Université de Paris VI, Paris, France. 48 pp.

Scott, M.J., F. Davis, B. Csuti, R. Noss, B. Butterfield, C. Groves, H. Anderson, S. Caicco, F. D'Erchia, T. Edwards, J. Ulliman, and R.G. Wright. 1993. Gap analysis: a geographic approach to protection of biological diversity. Wildlife Monographs 123. 41 pp.

Turner, D.P., G. Koerper, H. Gucinski, C. Peterson, and R. Dixon. 1993. Monitoring global change: comparison of forest cover estimates using remote sensing and inventory approaches. Environmental Monitoring and Assessment 26:295-305.


Home