Crawford, Thomas W. (forthcoming) Associations Between Biophysical, Social, and Spatial Factors and Indicators of Economic Status: An Analysis of Human-Environmental Linkages Among Villages in Nang Rong District, Thailand (Under the direction of Dr. Stephen J. Walsh).

Abstract

This research will analyze linkages between the physical environment and economic outcomes at the village level in a rural region of northeast Thailand. The main questions that will be addressed are: (1) How do villages differ from each other across biophysical, social, and spatial domains?, (2) What are the relationships between village differences and economic development outcomes, and, more specifically, what are the strengths of associations between environmental resource variation, landuse/landcover change, and economic outcomes?, and (3) Methodologically, how does one define village regions in the absence of cadastral information using GIS and social survey data? Answers to this third question support the previous two by enabling the construction of village biophysical variables in a GIS environment.

These questions are important because they represent attempts to link spatially social and environmental data in analyses where social variables, in this case economic indicators, are the main outcomes of interest. Much of the previous work linking social and environmental data has focused on human drivers of environmental change including landuse/landcover change (LULCC), land degradation, and climate change. Relatively little research has worked in the opposite direction. This study will analyze relationships between natural resource endowments, LULCC, and selected indicators of economic development along with other relevant population data.

Another contribution of this research is that it operates at a scale that is relatively untreated in the literature. Much of the social sciences literature on economic development seems to work with macro or micro-level data. Macro examples include country level analyses, and micro examples include household or individual level analyses. This study operates at a meso-scale by using the village as a unit of analysis. In doing so it brings a geographic perspective that stresses the importance of differences among places -- in this case, villages. Much research in developing countries has emphasized striking urban-rural differences in a variety of socio-economic indicators. Relatively unexplored is the nature of rural-rural place differences. This study will examine differences among a sample of 51 villages located within 50 km of each other in a single district. Results may indicate to policy makers that, even within a relatively limited geographic area, there exists significant variation among villages that should be considered when designing and implementing programs regarding economic development.

Crews-Meyer, Kelley. (forthcoming). Landuse/Landcover Change in Northeast Thailand: Integration of Infrastructure and Environemntal Policy in Modeling Resource Utilization Patterns and Social Responses (Under the direction of Dr. Stephen J. Walsh).

Abstract

My dissertation focuses on the analysis of landuse/landcover change (LULCC) in Northeast Thailand, with particular emphasis on agricultural extensification and deforestation. I am examining spatial and temporal patterns of LULCC in light of increased geographic accessibility through road and hydrographic infrastructure enhancements over a decade, culminating in the 1994 landscape. This work examines landscape-level population-environment interactions via spatial, statistical, and policy analyses. The methodology includes digital image processing of a time series of multispectral imagery for classification and change detection, which is then used to further classify the 1994 landscape based on the trajectories of LULCC. This classification is then statistically and spatially related to GIS-derived accessibility surfaces as well as ancillary biophysical and social survey data. Accessibility surfaces are created from transportation and hydrographic infrastructure that were captured via digital analysis ofa regional air photo mosaic. The work is performed in a UNIX environment utilizing primarily ERDAS Imagine and ESRI's ArcInfo in conjunction with pattern metrics and several spatial and statistical software packages. Accuracy assessment will be conducted in the next few months in Thailand under support of a Fellowship and using GPS technology.

Valdivia, Gabriela. (forthcoming). Cultural Landscapes and Remotely Sensed Imagery of Northeastern Ecuador. (under the direction of Dr. Stephen J. Walsh).

Abstract

The Ecuadorian Amazon is characterized by a rather fragmented landscape of multiple land use patterns. The various settlement patterns within the larger Oriente region are a function of the interacting cultural groups inhabiting the region and related human processes, such as migration, deforestation, and cultural assimilation. In the case of comunas, territories owned by indigenous groups in close contact to colonization areas, increased contact with agricultural colonization is changing indigenous landuse methods and perceptions of the surrounding natural environment. The goal of this paper is to evaluate the change in patterns of land use at the comuna level by comparing individual comunas across the Oriente region, based on historical differences in development, indigenous group origin, and proximity to colonization settlement sectors. The landscape composition and structure will be assessed through the analysis of patterns of land use in the remotely sensed imagery, together with supporting in situ data. The results will be used to support the hypothesis that cultural assimilation influences the conversion of forested landscapes into agricultural landscapes.

Evans, Tom P. 1998. Integration of Community-level Social and Environmental Data: Spatial Modeling Boundaries in Northeast Thailand (Under the direction of Dr. Stephen J. Walsh).

Abstract

Researchers in the fields of geography and population studies have struggled with the integration of social and environmental data. Geographic Information Systems (GIS) offer spatial tools and data structures to facilitate this integration, but considerable data transformations are often necessary to enable this spatial linkage. These transformations have important implications for the nature of the resulting integration and for analyses performed using generated products. This dissertation explores methods of integrating social, spatial, and biophysical data and introduces new methods and concepts to the existing research. In addition, this research examines fundamental space partitioning considerations for linking people to the landscape through the definition of village boundaries set through land use considerations and not by arbitrary political units. The human-environment linkage may be used to explore topics such as the relationship between deforestation and population migration or the relationship between landcover change and labor demand. One methodological hurdle contributing to the difficulty in linking population and environmental data is the question of how to define a meaningful boundary representation for communities. This research discusses how to represent land ownership boundaries of rural communities in an agrarian environment where the predominant settlement pattern is characterized by nucleated household aggregations. Broad research objectives include: (1) how to spatially represent the household area of village communities, (2) how to spatially represent the area surrounding a community that is linked to that community through land use and land ownership patterns, and (3) how to use these representations to link community level data to spatially continuous biophysical data. The concept of spatial data transformations serve as the theoretical framework for this research and are used to address each of the above objectives within a GIS framework. A series of vector-based point-to-polygon transformations are presented as one method of generating community boundaries. A region growing method is developed as an alternative method of generating boundaries. The region growing algorithm is a powerful tool allowing various social, spatial, and biophysical features which affect the spatial pattern of community- level land ownership to be integrated into the boundary- generation model.

Townsend, Philip A. 1997. Environmental Gradients and Vegetation Patterns on the Roanoke River Floodplain, North Carolina (Under the direction of Dr. Stephen J. Walsh).

Abstract

This research examined landscape-scale processes that influence vegetation patterns on the Roanoke River floodplain of North Carolina. The results include the development of spatially-explicit, quantitative representations of the major environmental factors that influence forest patterns and regeneration in the study area. The research has involved the following: development of a detailed vegetation classification of ecologically significant terrestrial and aquatic communities from multi-temporal Landsat TM imagery; creation of an extensive digital database with important environmental variables for research and resource management; development of a digital elevation model, from which models of potential flood inundation were derived; production of a model of potential hydroperiod regime, incorporating both spatial and temporal properties, using multi-temporal synthetic aperture radar (SAR) imagery; and integration of field data, GIS variables, and remote sensing analyses to examine species and community distributions on the lower Roanoke River floodplain. SAR was used to successfully identify patterns of flood inundation beneath the forest canopy. These results were applied directly to analyses of the vegetation of the region, and helped explain community and species distributions on the floodplain. The analyses of vegetation data indicated a distinct series of gradients related to topographic position in the floodplain, hydroperiod, and alluvial processes (sedimentation, organic matter accumulation). Modeling of species distributions on the floodplain suggested that vegetation communities in many locations are mismatched with the current environmental processes. Specifically, the relatively flood-intolerant species Fagus grandifolia is modeled to be much more prominent on the floodplain than it presently is. The modeling results are corroborated by data on species regeneration, which indicate the widespread abundance of beech seedlings in forests that are currently dominated by other species. This indicates that the conditions under which the current forests established differ from those present today. There are several potential causes for vegetation change in the region, including, most prominently, changes in hydrology due to the construction of dams on the river upstream of the study area. This research represents the foundation for a watershed-scale analysis of the response of an ecosystem to regional and global change.

Allen, Thomas Richard. 1995. Relationships between Spatial Pattern and Environment at the Alpine Treeline Ecotone, Glacier National Park, Montana (Under the direction of Stephen J. Walsh).

Abstract

The alpine treeline ecotone is a major landscape boundary separating subalpine and alpine ecosystems in Glacier National Park, Montana. This project mapped and analyzed spatial patterns of alpine treeline using an integrated geographic information system. Multitemporal Landsat Thematic Mapper satellite data was classified into a detailed ecological characterization of treeline vegetation using a hierarchical digital image classification approach, differentially corrected Global Positioning System training sites, and aerial photointerpretation. Treeline patterns were assessed using spatial pattern metrics in topographically defined hillslopes identified by digital terrain analysis. Cluster analyses showed that spatial patterns of treeline comprised six general types related to the diversity of vegetation types present, the juxtaposition of subalpine and alpine types, the degree of edge complexity, and texture of patches. Canonical correlation analysis related spatial pattern attributes to environmental variation in the treeline ecotone. The hypothesis that environmental gradients explain the most spatial variation of treeline was affirmed. Subordinate factors affecting treeline patterns were snow avalanche path disturbance and the presence of sills, faults, and perennial snowfields. Alpine treeline pattern- environment linkages included: contagion and elevation, edge density and slope angle, forest-tundra juxtaposition and insolation variability, and an association between vegetation class diversity and topographic wetness variation. The research identified areas where treeline patterns are potentially climatically controlled and not restricted by disturbances, topographic controls, or geologic structures.

McGregor, Stephen J. 1995. Topographic Modeling of Habitat Suitability in the Alpine Tundra Ecosystem: An Integrated Geographic Information Systems Approach (Conifer Growth) (Under the diretion of Dr. Stephen J. Walsh).

Abstract

The basic intent of this research was to examine the spatial pattern and composition of alpine tundra using an integrated geographic information system (IGIS) approach. This study examined the alpine tundra ecosystem in terms of the biophysical factors that affect its spatial pattern and composition, to determine if there are areas within the ecosystem and the forest-alpine tundra ecotone (FATE) that are suitable for conifer growth, and hence, the migration of treelinecomponents. Biophysical factors used in the analysis included topoclimatic, biogeographic, and edaphic controlling factors. Biophysical data were collected for four alpine tundra sites in Glacier National Park, Montana (Logan Pass, Siyeh Pass, East Flattop Mountain, and Scenic Point) using remote sensors, in-situ techniques, and ancillary sources, and integrated in a raster-based GIS framework. The IGIS database was analyzed using digital image processing techniques and GIS cartographic modeling techniques, to create landscape feature sets. Feature sets contained individual biogeographic, topoclimatic, and edaphic data layers at two geographic scales, regional/local-scale (spatial resolution 30 x 30m) and micro-scale (spatial resolution 10 x 10m and 5 x 5m). These data layers represented key biophysical variables observed to control the pattern and composition of alpine tundra and the FATE. The landscape feature sets were recombined, using a function-oriented additive approach, to create suitability models containing an index of habitat suitability potential for conifer growth, for each regional/local-scale and micro-scale site. Areas identified as suitable habitat were surveyed in the field to verify the reliability of the modeling approach. Other biophysical factors affecting the alpine ecosystem and the FATE, not included in the habitat suitability model, and recommendations for future research are discussed. Significant contributions of this research include: (1) development of a multi-scale IGIS database which can be used as a baseline for future studies of alpine tundra and environmental change; (2) development of a model for identifying the habitat suitability of sites within the alpine tundra ecosystem for future study and monitoring; and (3) a methodology for integrating remote sensor, in-situ, GPS survey, and GIS techniques for modeling biophysical processes affecting the alpine tundra, the FATE, their spatial pattern a nd distributions, and their potential for change.

Brown, Daniel George. 1992. Topographical and Biophysical Modeling of Vegetation Patterns at Alpine Treeline (Under the Direction of Dr. Stephen J. Walsh).

 Abstract

Alpine treeline represents the transition from environments which sustain dense stands of subalpine forest to alpine sites which are unable to support trees in any form. Availability and abundance of several important topo-climatic elements explains much of the variation in the patterns of vegetation along the treeline ecotone. Active geomorphic and biophysical disturbance regimes in alpine areas introduce additional variability on the treeline transition. The purpose of this research was to model the spatial patterns of vegetation communities along the treetine ecotone relative to topo-climatic and disturbance processes. Surrogates of several topographically controlled climatic elements (solar radiation potential, soil moisture potential, and wind/snow potential) were constructed from digital elevation models (DEMs) for a study area in east-central Glacier National Park, Montana. Vegetation communities in the study area were characterized through statistical classification of Landsat Thematic Mapper digital data, field calibration, and validation. Topographical Empirical Models of Treeline (TEMTREEs) were constructed to examine the relative importance of factors which affected the treeline transition. TEMTREEs were also evaluated as predictive tools for extending the analysis through additional variables representing multi-scale processes. Empirical models were constructed for selected elevation zones (from 1600 to 2350 meters a.m.s.l.) at 150 meter intervals. The Kappa statistic was used to assess the predictive ability of the models.The results suggested that the elevation gradient accounted for much of the variation in the vegetation patterns at alpine treeline. Geomorphic disturbance patterns, characterized by mapping talus slopes, snow avalanche paths, and slope angles greater than 34 degrees, consistently explained significant variations in the vegetation patterns. This finding suggests that predictions of treeline responses to climatic change must account for potential changes in the frequency and magnitude of geomorphic process disturbances. Spatial autocorrelation in the model residuals suggested that additional variables should be used to account for the patterns of treeline. Analysis of the residuals indicated that basin-scale variables accounted for some of the unexplained variation between the observed and expected pattern of alpine treeline characterized through closed-canopy forest, open-canopy forest, tundra, and bare surfaces.

Bian, Ling. 1991. Effects of Spatial Scale on Estimating the Relationship between Vegetation and Topography in a Mountainous Environment (Montana) (Under the direction of Dr. Stephen J. Walsh).

Abstract

The effects of spatial scale are inherently involved in landscape studies. Little research, however, has focused on the effects of scale dependencies on the relationships between landscape and environment from a systematic perspective. Such a study is important since it relates both to ecological and spatial analyses in a manner that focuses on landscape pattern, process, and scale. This research systematically studied the effects of spatial scale on estimating the relationship between vegetation and topography in Glacier National Park, Montana. The vegetation index, Reflectance/Absorptance, derived from the Landsat TM data were chosen to represent vegetative landscape to relate three topographic variables, elevation, slope angle, and slope aspect, derived from the Digital Elevation Models. The effective ranges of spatial scales within which the two sets of variables were spatially dependent, and the degree of the spatial dependences were characterized through semivariogram and fractal analyses. The correlations between the vegetation index and the topographic variables were obtained through regression analysis at a series of aggregated spatial scales. Most topographic variables and the vegetation index were spatially dependent within a scale range representing the ridge-valley distance in the study area. Elevation was the most spatially dependent topographic variable. The vegetation index was less spatially dependent than all the topographic variables because it represented the collective results of many processes and factors functioning at various scales. The correlations between vegetation index and topographic variables were low at fine scales but increased with data aggregation levels; high at the characteristic scales; and slightly depressed at coarser scales for all correlations except for slope angle. The maximum correlations occurred at a scale of the ridge- valley distance. Variables aggregated at this level represented the most basic spatial structures in the study area. At finer or coarser scales, factors other than topography were more effective in affecting the variation in vegetation. The relative importance of the topographic variables varied with scale in a similar manner as that of correlations. The elevation was the most important at all scales.

Stewart, James D. 1995. Topography and Fire: Factors Affecting Landscape Productivity in Glacier National Park, Montana (Under the direction of Stephen J. Walsh).

Abstract

The basic intent of this research was to evaluate the impacts of elevation, slope angle and aspect, and time since fire on the pattern of landscape productivity in a remote, topographically complex area of Glacier National Park, Montana. The measure of landscape productivity was provided via the calculation of a Normalized DifferenceVegetation Index (NDVI) based on Landsat Thematic Mapper (TM) bands 3 and 4. The topographic variables for analysis were derived from a U.S. Geoglogical Survey (USGS) 1:24,000 base-scale Digital Elevation Model (DEM). Time since fire information was based on data maintained by Park officials. The statistical relationships betweeen the variables were examined using a combination of correlation analysis, simple linear and multiple regression, and analysis of variance (ANOVA). Elevation was by far the most important topographic variable influencing the level of productivity in the sutdy area, whereas aspect was the least important. Pixels that corresponded with water bodies significantly affected the statistical analysis of topography in relation to productivity. The statistical explanation of procuctivity in relation to elevation and slope angle water increased dramatically with the exclusion of water pixels from the analysis; it improved only slightly in relation to aspect. Time since fire and topography (elevation, slope angle and aspect) were negatively correlated. Time since fire explained only about 16 percent of the variance in the NDVI-based productivity measure. Based on forward multiple regression analysis, elevation explained the greatest amount of productivity (approximately 24 percent including water pixels) and time since fire provided approximately 7 percent additional explanation. Slope angle and aspect explained lesser amounts of the observed landscape productivity (approximately 1 and .2 percent, respectively). The research was conducted using digital image processing techniques within a geographic information system (GIS) environment.

Daniel G. Brown. 1992. Modeling Alpine Lake Turbidity Levels through Morphometric Basin Variables: Integration of Remote Sensing and Statistical Approaches (Under the direction of Stephen J. Walsh).

Abstract

The statistical relationship between spectral radiance values from lake surfaces (evaluated using Landsat Thematic Mapper) and lake turbidity levels (measured at ground sample points) was examined. Estimations of turbidity levels, in lakes for which no field data were available, were made using a multiple regression approach. Estimated lake turbidity values were compared with basin biogeogmorphic variables through a clustering procedure and analysis of variance (ANOVA) tests. The

objective was to assess the degree to which the selected variables (representing basin form, drainage network characteristics, and lake form) affected lake turbidity levels. A significant statistical relationship existed between basin geogmorphology and lake turbidity levels. Areal extent of glacial ice and total stream length within the basins were the most significant variables for explaining lake turbidity levels.

Thaddeus J. Bara. 1993. A probabilistic Approach to Landscape Regionalization in Glacier National Park, Montana (under the supervision of Dr. Stephen J.Walsh).

Abstract

A complex mountainous landscape in Glacier National Park, Montana, wasregionalized into landscape units defined by homogeneous probabilitiesof vegetated landcover occurrence. Probability surfaces were generated for each of ten different landcover types using non-parametric statistical

curve-fitting. Natural groupings of probabilities were identified with density linkage clustering, and then mapped as landscape units in the study area. A total of five landscape unit types were identified, including homogeneous, dense canopy, subalpine forest primarily along valley floors, a more diverse subalpine forest on gentle slopes and low ridge tops, a highly diverse upper subalpine zone, characterized by open canopy forests, subalpine meadows, and snow avalanche paths, on steep

valley slopes and in high-mountain valleys, fire-disturbed areas, and the high elevation alpine zone. The probabilistic approach used in this thesis represents a potentially useful way to identify and map

ecologically meaningful landscape units in a diverse and complex natural landscape.

Barry F. Doll. 1993. Use of Topographic Information in Estimating Areal Precipitation in ththe Mountain Environment of Western North Carolina (Under the direction of Dr. Stephen J. Walsh).

Abstract

Estimates of areal precipitation are based on observations at points.Conventional spatial interpolation methods fail to account for orographiceffects on rainfall between measured points in mountainous regions, oftenresulting in unrealistically low areal estimates. This research usedmultiple regression analysis to relate long-term annual and seasonalprecipitation to topographic and locational attributes at and around guagelocations, for the purpose of predicting precipitation at ungauged points. Primary precipitation data were monthly records from 49 stations during1951-1980; topographic variables were derived from USGS Digital Elevation Models, extracted through GIS techniques. One selected regression model explained 76 percent of the variance in annual precipitation. Predictionsform the model were combined with corrections from maps of regression residuals to make compound estimates at 17 additiaonal stations plus 10 unguaged points. Incorporating the additional point data resulted in higher estimated areal precipitation vs. estimates made at the 49 original stations. The mapped residuals weere also used to identify potential new variables and sub-regions for further analysis.

Bonnie Morris Henderson. 1991. Plowed, Paved, or in Succession: A Study of Landscape Change in Orange County, North Carolina (Under the direction of Dr. Stephen J. Walsh).

Abstract

The landscape of the North Carolina Piedmont is a patchy mosaic of forests, fields, farmland, and urban areas; this particular configuration of land cover types relects past and present land use patterns. The research conducted for this thesis employed historical aerial photographs and spatial analysis to examine landscape change in Orange County, North Carolina during the years 1938, 1955, 1972, and 1982. Spatial, historical, and ecological perspectives of landscape change were investigated and reported. Results indicated that changes occurred in the extent and pattern of land cover types on the landscape: agricultural land cover decreased, forested land cover increased, urban land cover in the southern part of the county increased, and the landscape in general became more diversified. These changes have numerous ecological implications and can be traced to historical changes in land use patterns.

Kelley, Nina M. 1991. Role of Topgraphy in the Establishment and Maintenance and Treeline Ecosystem Components in Glacier National Park, Montana: An Integration of Remote Sensing Methods and Digital Elevation Models (Under the direction of Dr. Stephen J. Walsh).

Abstract

This research assessed the sensitivity of Landsat TM satellite data for the evaluation of treeline components, and investigated the relationship between topography and the treeline components. The treeline information was provided by a Landsat TM spectral reflectivity score, the topographic variables were provided by USGS DEMs. The image data were first spectrally enhanced, and then classified. The classification of the spectral enhancements was successsful in discriminating the treeline components (forest, distrubance, mosaic, krumholz, tundra, and bare surfaces); peviously, ecosystem mapping was the most detailed level of vegetation indentification. Univariate and multivariate statistical techniques were performed on the individual treeline components and the underlying topography to understand the control of topography on the components of the treeline trasition zone, and the zone as a whole. The regression resulted in an expanation oby the topograhic variables of 31% of the variance found within the spectral reflectance score data. The research introduced issues of scale dependencies, and are discussed in the conclusion.

Hinton, Curtis A. 1990. Modeling Nonpoint Pollution oand Site/Factors for the Locaiton of Water Retention Ponds (Under the direction of Stephen J. Walsh).

Abstract

This study demonstrated how a Geographic Information System (GIS) can be used in solving complex spatial problems. Disparate spatial data were input into a GIS to aid in locating areas at high risk due to nonpoint pollution (NPP). A variety of variables, such as land use, slope, and soils, were entered into the GIS as independent spatial layers. Each variable was assessed and weighted as to its impact on NPP generation. The weighted variables were combined into an empirical model that indicated areas at risk for NPP in two selected test watersheds in northern Durham county, North Carolina. Derived NPP levels were evaluated to reduce their movement throughout the hydrologic system by the strategic placement of retention ponds, that serve as sediment, chemical, and high water traps. The results of the NPP location model were utilized as input, along with other spatial data layers, for a subsequent model which dete (Under the rmined optimal site for retention pond location.