Many of the hypothesized ecosystem responses to climate change will cause changes in ecosystem structure (Woodmansee, 1988). For example, increases in sea level will dramatically restructure coastal dune systems (Carter, 1991). Global climatic change may lead to increased biological invasions (D'Antonio and Vitousek, 1992), changes in nutrient cycling regimes (Field et al, 1992), and increased levels of disturbance (Graham et al, 1990). Reconstructions of the biotic response to climate change in the past have shown wholesale changes in species composition and distribution of forest and grassland communities (Webb and Bartlein, 1992). Although changes in structure may have occurred over intervals of decades to centuries in the past, human induced climatic changes and landscape modification may accelerate these processes such that changes in structure may be observable over much shorter intervals (Spies et al, 1994; Simmons et al, 1992).
Landscape ecology is the science of relating pattern to process at large scales (Turner, 1989). The relationship between pattern and scale is considered a central theme of landscape ecology (Turner, 1989). As result in recent years landscape ecologists have striven to create statistical measures of landscape pattern, both in terms of structure and scale (Turner et al, 1991; O'Neill et al, 1988a). Evaluations over large portions of the United States have been made using USDA landuse maps (Hunsaker et al, 1994; O'Neill et al, 1988a), which have shown that landuse patterns can be summarized using indicies like contagion, dominance and fractal dimension. Fractal dimension appears to be correlated with human manipulation of the landscape (O'Neill et al, 1988a). Theoretical studies relating landscape pattern to process have shown that the rates of some ecological processes like propagation of disturbance (Turner et al, 1989) and the spread of disease (Burdon et al, 1989) could increase non-linearly if a threshold of connectivity is reached (O'Neill et al, 1992). However more empirical studies are required for development and confirmation of these models.
Many of these structural measures are directly applicable to remote sensing since it is a direct and complete sampling of landscape pattern (Turner et al, 1991). Remotely sensed images represent the true heterogeneity of the landscape which is sometime lost in maps of landuse or vegetation. Moreover because remote sensing data is not categorical like land use maps, so techniques like textural analysis (Musick and Grover, 1991) and spectral analysis (Turner et al, 1991) can be used to evaluate gradients and scale relationships in landscape structure.
Given recent developments in remote sensing technology and landscape analysis, it is possible to synthesize these lines of research to quantitatively evaluate ecosystem structure and scale across a large variety of landscapes. Although remote sensing data is almost universally assumed in the context of global change research, it is not clear what the consequences of remote sensing scale (defined by pixel size and swath width) are on the interpretation of landscape scale and structure (defined by ecosystem process). Indeed it is generally not known what the generic "scale" of a given ecosystem is (S. Ustin, personal communication) nor how well a given sensor will be able to capture it. There are a variety of opinions in the literature regarding the sensor scale and system scale (Dunn et al, 1991; Lillesand and Kiefer, 1994; Woodcock and Straler, 1988), but in general no consensus has been reached, as can be seen in the range of spatial and spectral resolutions on current and planned satellite sensors. What is generally agreed, however, is that remote sensing is the only practical tool for consistently monitoring regional to global scale processes over time (Wessman, 1992; Ustin et al, 1991).
Thus, there is a need for a research program to consider explicitly the relationships between remotely sensed scale of observation, the observed scale of the landscape, and the consequences of these two scales for quantifying landscape structure. Such research contributes to NASA's responsibilities for global change research in several ways. First explicit evaluations of scale and structure across a large number of landscapes will help us to design field
experiments over larger scales than are traditionally used by ecologists. By establishing the structural scale of a system, we determine at what lag distances landscape units are uncorrelated in space. Placement of sample points by the investigator can then be made with respect to units of statistical homogeneity, sampling either within or between identified units (Simmons et al, 1992; Milne, 1991).
Second landscape pattern stands in a feedback relationship to a large number of earth processes which make it a useful indicator ecosystem response to climate change. For example if an annual grassland/oak woodland environment becomes more xeric, patches of grass may become smaller and further apart, resulting in lower contagion between grassy patches and increased dominance of the oak cover, which is more resistant to drought. Moreover different landscape types (e.g. grassland, boreal forest, semiarid scrub) may have distinctive landscape structures which can be attributed to functional attributes of those systems. Determining these structural relationships is a fundamental step in developing our understanding of how pattern relates to process. If one does not know what the "metric scale" of a landscape is, for example, in terms of the number of patches, the dispersion of patches, and the shape of patches, it is difficult to evaluate how organisms are interacting with the environment on their own "biotic scales." (Milne, 1991)
Third a thorough understanding of landscape structure is prerequisite for scaling methodologies from fine scale measurements of process to regional and global scales, even with remote sensing (Moloney et al, 1992). Little is known about the translation of traditional ecological measurements made at meter scales to the 500 km scales of global circulation model (GCM) grid cells. Yet even if we had a technology which allowed us with surety to evaluate (for example) photosynthetic rate from space, we would still need to understand how to aggregate photosynthetic measurements from the 35 Thematic Mapper images which cover California into the single GCM cell which does. Knowing the scale of the landscape gives us a first guess as to what can be averaged safely and what level of detail must be preserved, which is the goal of scaling studies (Wessman, 1992) .
Finally we can use a categorization of landscapes by structural relationships to illuminate the causal links between pattern and process. The dataset described below is extensive enough to for statistically significant comparisons to be made across continental differences in climate, vegetation type, and physiographic factors. By comparing the structure of western coniferous forest in British Columbia, Washington, Oregon and California, salient similarities and differences can be identified. Western coniferous forest can be compared to boreal forest from Saskatchewan and Maine, deciduous forest from Massachusetts and Virginia, and tropical, seasonally dry forest from Mexico. The result will be a series of hypotheses of how landscape structure relates to climate, which can then be interpreted based on known functional relationships in ecosystems (Field, 1992) and models of landscape dynamics (O'Neill et al, 1992).
The goal of this phase of the project will be to determine how observed landscape scale and structure vary spectrally and spatially, and particularly the relationship between sensor characteristics, like pixel size and band width, and landscape parameters. These investigations will be carried out by aggregating the finer scale data available from AVIRIS to simulate other sensors, both those which are currently available (Landsat-MSS, Landsat-TM, SPOT) and those which are planned (MODIS, ASTER). Once scale has been identified at a finer pixel size (using the quantitative techniques discussed below), will it be preserved with a larger pixel size? What is the relationship between pixel size and landscape grain? Related questions involve spectral resolution. To what extent does band placement and width influence the measurement of landscape scale and pattern? How spatially independent are measurements of scale in different bands? How much can band passes be widened and still preserve the same view of landscape structure? Finally there are many techniques in the literature for creating derived images based on remote sensing data. Some of these are very commonly used (e.g. NDVI) (Tucker, 1979) while others have been developed more recently (e.g. Spectral Mixture Analysis) (Ustin et al, 1986). How do these analyses reveal new aspects of landscape scale/structure, or do they? Sites for this portion of the study will be located in California in regions for which some background information is available, yet represent distinctive landscapes. Tentative sites include Jasper Ridge (oak woodlands/grassland), San Pablo Bay (tidal wetlands), Davis-Winters (Central Valley agriculture), West Tahoe (conifer forest/alpine wilderness) and Death Valley (desert).
The second phase of the study will examine how ecosystem scale and structure vary over a large (>100) variety of landscape sites using remotely sensed data acquired in the AVIRIS program. Much reduced versions of AVIRIS data cubes are available called "quick looks." Quick looks are one-band images, averaged every two lines and two samples, available for every site AVIRIS imaged between 1992-1994, which constitutes over 4000 scenes from over 100 different sites throughout North America. Quick looks have several advantages for the survey of ecosystem structure. They are small (0.5 Mb) and easy to manipulate, yet retain a pixel size (40 m) which approximates the spatial resolution of current land observation satellite systems like Landsat -TM, -MSS and SPOT. The single band is centered at 700 nm (nominally 10 nm wide), which avoids major atmospheric absorbtions (Lillesand and Kiefer, 1994), yet is in a region where basic terrestrial materials, like water, soil/rock, vegetation, and snow/ice can generally be distinguished (Lillesand and Kiefer, 1994). Simple image processing techniques (contrast stretching, level slicing) will allow the scenes to be classified into spectrally distinct landscape components for analysis. Though only a crude approximation of the landscape, these methods will allow each scene to be analyzed in terms of the same component classes while retaining the spatial heterogeneity of the landscape.
A broad range of ecosystems representing a variety of climatic, geological and ecological conditions are represented in the quick look dataset. The dataset spans locations from 18º N to 56º N latitude and from 68º W to 126º W longitude, from boreal forest to tropical uplands, from coastal British Columbia to Key West, Florida. Eighteen of the twenty-seven North American vegetation types identified by Barbour and Billings (1988) and eleven of the fourteen North American physiographic regions identified by Vankat (1979) are represented in the dataset, including several examples each of montane conifer forest, boreal forest, temperate deciduous forest, desert scrub, Mediterranean scrub, prairie grasslands and tidal marshes. A short list of sites includes Mt. Rainier, WA, Biscayne Bay, FL, Los Angeles, CA, Rocky Mountain National Park, CO, Dismal Swamp, VA, Organ Pipe, AZ, Boreas, Canada, and Teloapan, Mexico. Additional sites may also become available during this project: proposed AVIRIS schedules include flightlines in Alaska, Brazil and Australia. For most archive locations, several scenes are available, and for a few locations, multiple dates as well. Thus for a few scenes, I will be able to evaluate constancy of landscape structure seasonally and interannually. Each scene in a particular area will be considered an independent sampling of that landscape type, such that the variability within landscapes as well as between landscapes will be evaluated.
Textural measures (Haralick, 1979) will be useful for evaluating gradients in the data. Angular second moment (ASM) and inverse difference moment (IDM) are indicies of data homogeneity (Musick and Grover, 1991) which have been applied to Landsat-TM data. ASM was useful for determining qualitative differences of type in the data and may be useful in the classification process, while IDM was useful for showing shallow gradients in the data. Textural measures require intensity rather than categorical data.
Fractal analyses are useful for identifying self-affinity in a system (Milne, 1991). For example, Lathrop and Peterson (1992) identified structural self-affinity in mountainous landscapes using Digital Elevation Models. Milne (1991) shows that measurements of self-similarity are useful for identifying ecotones, characterizing area-perimeter relationships across scales, and describing diffusion across a heterogeneous landscape. For our purposes, fractal analysis will be used to compare self-affinity in different ecological communities and to link differences to functional causes.
After evaluation for scale and texture, each quick look scene will be classified using a supervised classification into exposed soil/rock, vegetation, water, and snow/ice. Although this is a crude reduction of the variation of ecosystem structure, it also represents the most basic components of any ecosystem. Patches of like materials will be grouped and delineated by polygon boundaries for the pattern analysis. Clouds and shadows of clouds will masked out of the images. Landscape statistics including contagion, dominance and edge density will be calculated for each patch type. Patch size distribution, dispersion and number will be determined and compared both within and between scenes. Patterns will be summarized and compared across landscape types based on vegetation and physiographic factors and across different climatic conditions. Observed patterns will also be compared with published results from analyses of landuse maps (O'Neill et al, 1988a), digital elevation models (Lathrop and Peterson, 1992), and theoretical studies (O'Neill et al, 1992) to generate hypotheses against which other scenes can be compared.
Our first idea is to stratify the analyzed sites in terms of known ecological function, physiographic factors, and climatic regime to find instances of landscape convergence and divergence. For example convergence under similar climatic conditions in disparate parts of the world is well-known, such as in schlerophyllous shrub communities in Mediterranean climates. Despite a variety of phyletic origins, these communities have converged physiognomically and functionally in response to similar climates. Does convergence extend to the landscape level, in terms of ecosystem structure? If current hypotheses are correct about resource allocation in response to environmental forcing factors (Field et al, 1992), then we expect to find predictable patterns in landscape structure across broadly similar climatic regimes. The extent to which such patterns are stable under conditions of increased temperature and atmospheric CO2 is unknown, yet changes seem likely given the responses of animal and plant communities in the past and present to changes in climatic conditions.
Burdon, J.J., Jarosz, A.M., and Kirby, G.C. 1989. Pattern and patchiness in plant-pathogen interactions--causes and consequences. Annual Review of Ecology and Systematics 20:119-136.
Carter, R.W.G. 1991. Near-future sea level impacts on coastal dune landscapes. Landscape Ecology 6:29-39.
Curran, P.J. 1988. The semivariogram in remote sensing: an introduction. Remote Sensing of Environment 24:493-507.
D'Antonio, C.M. and Vitousek, P.M. 1992. Biological invasions by exotic grasses, the grass/fire cycle, and global change. Annual Review of Ecology and Systematics 23:63-87.
Dunn, C.P., Sharpe, D.M., Guntenspergen, G.R., Stearns, F. and Yang, Z. 1991. Methods for analyzing temporal changes in landscape pattern, in eds. Turner, M.G. and Gardner, R.H. Quantitative Methods in Landscape Ecology: The Analysis and Interpretation of Landscape Heterogeneity. Springer-Verlag: New York.
Field, C.B., Chapin, F.S., Matson, P.A., and Mooney, H.A. 1992. Responses of terrestrial ecosystems to the changing atmosphere: a resource-based approach. Annual Review of Ecology and Systematics 23:201-235.
Forman, R.T.T. and Godron, M. 1986. Landscape Ecology. Wiley: New York.
Gardner, R.H. and Turner, M.G. 1991. Future directions in quantitative landscape ecology, in eds. Turner, M.G. and Gardner, R.H. Quantitative Methods in Landscape Ecology: The Analysis and Interpretation of Landscape Heterogeneity. Springer-Verlag. New York.
Graham, R.L., Turner, M.G., and Dale, V.H. 1990. How increasing CO2 and climate change affect forests. Bioscience 40: 575-587.
Hall, F.G., Strebel, D.E., and Sellers, P.J. 1988. Linking knowledge amoung spatial and temporal scales: vegetation, atmosphere, climate and remote sensing. Landscape Ecology 2:3-22.
Haralick, R.M. 1979. Statistical and structural approaches to texture. Proceedings of the IEEE 7:786-804.
Hunsaker, C.T., O'Neill, R.V., Jackson, B.L., Timmins, S.P., Levine, D.A., and Norton, D.A. 1994. Sampling to characterize landscape pattern. Landscape Ecology 9:207-226.
Lathrop, R.G. and Peterson, D.L. 1992. Identifying structural self-similarity in mountainous landscapes. Landscape Ecology 6: 233-238.
Lillesand, T.M. and Kiefer, R.W. 1994. Remote Sensing and Image Interpretation, Third Edition. Wiley: New York.
Milne, B.T. 1991. Lessons from applying fractal models to landscape patterns, in eds. Turner, M.G. and Gardner, R.H. Quantitative Methods in Landscape Ecology: The Analysis and Interpretation of Landscape Heterogeneity. Springer-Verlag: New York.
Moloney, K.A., Levin, S.A., Chiariello, N.R., and Buttel, L. 1992. Pattern and scale in a serpentine grassland. Theoretical Population Biology 41: 257-276.
Musick, H.B. and Grover, H.D. 1991. Image textural measures as indicies of landscape pattern, in eds. Turner, M.G. and Gardner, R.H. Quantitative Methods in Landscape Ecology: The Analysis and Interpretation of Landscape Heterogeneity. Springer-Verlag: New York.
Norton, D.J. and Slonecker, E.T. 1990. The ecological geography of EMAP. Geo Info Systems 1:33-43.
O'Neill, R.V., Gardner, R.H., Turner, M.G., and Romme, W.H. 1992. Epidemiology theory and disturbance spread on landscapes. Landscape Ecology 7:19-26.
O'Neill, R.V., Krummel, J.R., Gardner, R.H., Sugihara, G., Jackson, B., DeAngelis, D.I., Milne, B.T., Turner, M.G., Zygmunt, B., Christensen, S.W., Dale, V.H., and Graham, R.I. 1988b. Indicies of landscape pattern. Landscape Ecology 1:153-162
O'Neill, R.V., Milne, B.T., Turner, M.G., and Gardner, R.H. 1988a. Resource utilization scales and landscape pattern. Landscape Ecology 2:63-69
Sanderson, E.W., Zhang, M.H., Rejmankova, E., and Haxo, R.S. 1995. Validating spatial structure in canopy water content using geostatistics. JPL Proceedings AVIRIS Workshop, January 23-24, 1995.
Simmons, M.A., Cullinan, V.I., and Thomas, J.M. 1992. Satellite imagery as a tool to evaluate ecological scale. Landscape Ecology 7:77-85.
Spies, T.A., Ripple, W.J., and Bradshaw, G.A. 1994. Dynamics and pattern of a managed coniferous forest landscape in Oregon. Ecological Applications 4:555-568.
Tucker, C.J. 1979. Red and photographic infrared linear combinations for monitoring vegetation. Remote Sensing of Environment 8: 127-150.
Turner, M.G. 1989. Landscape ecology: the effect of pattern on process. Annual Review of Ecology and Systematics 20: 171-197.
Turner, M.G. and Gardner, R.H. 1991. Quantitative methods in landscape ecology: an introduction. in eds. Turner, M.G. and Gardner, R.H. Quantitative Methods in Landscape Ecology: The Analysis and Interpretation of Landscape Heterogeneity. Springer-Verlag: New York.
Turner, M.G., Gardner, R.H., Dale, V.H., and O'Neill, R.V. 1989. Predicting the spread of disturbance in heterogeneous landscapes. Oikos 55: 121-129.
Turner, S.J., O'Neill, R.V., Conley, W., Conley, M.R., and Humphries, H.C. 1991. Pattern and scale: statistics for landscape ecology, in eds. Turner, M.G. and Gardner, R.H. Quantitative Methods in Landscape Ecology: The Analysis and Interpretation of Landscape Heterogeneity. Springer-Verlag: New York.
Ustin, S.L., Adams, J.B., Elvidge, C.D., Rejmanek, M., Rock, B.N., Smith, M.O., Thomas, R.W., and Woodward, R.A. 1986. Thematic Mapper studies of semiarid shrub communities. Bioscience 36: 446-452.
Ustin, S.L., Wessman, C.A., Curtis, B., Kasischke, E., Way, J., and Vanderbilt, V.C. 1991. Opportunities for using the EOS imaging spectrometers and synthetic aperture radar in ecological models. Ecology 72: 1934-1945.
Vane, G., Green, R.O., Chrien, T.G., Enmark, H.T., Hansen, E.G., and Porter, W.M. 1993. The Airborne Visible/Infrared Imaging Spectrometer (AVIRIS). Remote Sensing of the Environment 44:127-143.
Vankat, J.L. 1979. The Natural Vegetation of North America. Wiley: New York.
Webb, T. and Bartlein, P.J. 1992. Global changes during the last 3 million years: climatic controls and biotic responses. Annual Review of Ecology and Systematics 23:141-173.
Wessman, C.A. 1992. Spatial scales and global change: bridging the gap from plots to GCM grid cells. Annual Review of Ecology and Systematics 23:175-200.
Woodcock, C.E. and Strahler, A.H. 1987. The factor of scale in remote sensing. Remote Sensing of Environment 21:311-332.
Woodmanse, R.G., 1988. Ecosystem processes and global change. in eds. Rosswall, T., Woodmansee, R.G., and Risser, P.G. Scales and Global Change. Wiley: New York.
Return to Eric's Current Research