UBC Theses and Dissertations
Chemical terrain variability : a geomorphological approach using numerical and remote sensing techniques Schreier, Hanspeter
The variability of chemical parameters over the landscape was examined in this research. A terrain hierarchy based on genetic geo-morphological unit concepts was developed in two Quaternary landscapes in the Fraser Valley and in the Peace River area in British Columbia. The relative variability within and betueen different hierarchical units ranging from "sites" to "landform units" to "landform unit types" was compared. The variability was large in all units but was smaller at the site scale than at the landform unit scale, within single landform units chemical parameters were shown to be closely related to type of drainage. Available Ca, Mg, Na, H, and Si were found to be the most important differentiating parameters for all units. Site categories which reflected units of similar parent material, form, and inferred genesis were determined by application of a cluster analysis procedure. Sites grouped by this method were not coextensive with individual land-form units, thus suggesting that the environmental imprint on the.chemical conditions was not as strong as that of the genesis. The best grouping was obtained with the Peace River data where more natural conditions prevailed. A data screening through factor analysis prior to the grouping improved the landform unit type classification in the Fraser Valley where the chemical conditions were complicated by a more complex and intensive land use pattern. Multispectral remote sensing techniques were used to assess the potential of predicting chemical ground conditions from spectral measurements. Multispectral photography combined with density slicing and additive color viewing techniques were used to quantify chemical ground conditions on the photographic image. Areas'of different soil moisture and percent Carbon content could readily be identified and quantified by this means. Exchangeable Ca, Mg, and Na could partially be differentiated probably as a result of direct correlation with Carbon. The wider band photography (color k00-700 nm and color IR 500-900 nm wavelength range) produced better results than the narrow band black and white images (50D-6QD nm and 6DD-7DD nm wavelength range). The trends in detecting chemicals were consistent for both vegetated and non-vegetated surfaces; the sliced color film image was slightly more useful for analyzing exposed soil surfaces, while the sliced color IR image proved to be more useful for the interpretation of vegetated surfaces. Direct digital reflection measurements were made with a multichannel spectrometer from the air, and Dn soil samples on the ground and in the laboratory. In the field the A-DO-1000 nm spectral wavelength range was used and the laboratory analysis was extended to the 35D-25D0 nm wavelength range. Only bare soil surfaces were investigated. Correlation and regression analysis revealed that % Carbon, % Fe, exchangeable Mg, and exchangeable K could be predicted from spectral reflection values. There is evidence thattthe spectral-chemical relationship follows a curvilinear function, but adequate predictions were obtained with linear relationships at low chemical concentration levels. Despite differences in measuring techniques similar regression trends were obtained for all three methods and the 50D-11D0 nm wavelength range was found to be most useful in this analysis. The total spectral reflectance curve was found to be of importance since soils from similar parent materials produced characteristic curves which could readily be differentiated by all three types of measurements.
Item Citations and Data