Vegetation distributions in semi-arid environments: Spatial analysis for climate and landscape characterization
AdvisorMarsh, Stuart E.
MetadataShow full item record
PublisherThe University of Arizona.
RightsCopyright © is held by the author. Digital access to this material is made possible by the University Libraries, University of Arizona. Further transmission, reproduction or presentation (such as public display or performance) of protected items is prohibited except with permission of the author.
AbstractSpatially explicit knowledge of land cover is increasingly important for environmental modeling and decision support for land managers. Such knowledge is often provided over large regions by thematic maps produced from remotely sensed satellite data. Remote sensing of vegetation in semi-arid areas is complicated, however, by high levels of landscape spatial heterogeneity, resulting in large part from spatially varying soils, topography, and microclimates. Increased understanding of spatial distributions of vegetation and the factors affecting them will enhance our ability to inventory and monitor natural resources, and to model potential consequences of land management alternatives and larger issues such as global climate change. In addition, the uncertainty in spatial knowledge must be made spatially explicit in order to determine where more information is needed and where predictions maybe less reliable. Geostatistical kriging and multiple linear regression interpolation were used to map climate spatial distributions over the San Pedro River watershed, southeastern Arizona. Both methods used climate station location and elevation and climate data. Although mean interpolation errors were similar, kriging climate with elevation as external drift was preferred due to the patterns of spatial bias in regression errors. Interpolation results provided a step toward understanding climate influence on vegetation in this area. Accuracies of four land cover maps covering the upper San Pedro watershed, mapped from remotely sensed data, were determined using aerial photography, digital orthophoto quadrangles, and airborne video data reference data sets as alternatives to contemporaneous ground-collected data. Overall map accuracies were 67--75%; class accuracies varied more for smaller classes than for larger ones. Finally, the uncertainty of occurrence of the low-accuracy Mesquite Woodland class was mapped using simple indicator kriging with locally varying means and data derived from accuracy assessment information. Enhanced class discrimination in an independent validation data set confirmed the utility of this procedure. The results of these analyses can provide direct input for use in environmental modeling and can inform land management decision making, and the methods can be employed in other settings where spatial variability and uncertainty play large roles in the landscape.
Degree ProgramGraduate College
Arid Lands Resource Sciences