Global climate change is a significant research focus area in contemporary Earth science. Changes in climatic patterns have already resulted in shifting energy flows with associated changes in hydrologic and ecologic systems. More specifically, changes in vegetation distribution and abundance are one of the most visible and potentially significant effects of a changing climatic regime. However, to monitor and predict future changes in vegetation, the initial conditions must be well characterized. This thesis examines the distribution of vegetation in a semiarid mountain watershed in three important ways: (1) quantifying the factors affecting the distribution of broad classes of vegetation at the hillslope scale (e.g., 30-100 m), (2) quantifying factors affecting the organization of vegetation at sub-hillslope scales, and (3) quantifying the factors influencing the distribution of vegetation water content. The first of these themes is aimed at producing a hillslope-scale classification map of 4 broad classes (sagebrush steppe, Douglas fir, ponderosa pine, and deciduous/riparian) of vegetation within Dry Creek Experimental Watershed (DCEW) at the 30 m spatial resolution using remote sensing and geospatial data, field data, and 2 supervised learning classifiers known as artificial neural network (ANN) and classification and regression tree (CART). We investigated possible drivers of vegetation distribution by partitioning 11 topographic and remote sensing inputs into the ANN and CART models. Results show that the ANN had better overall accuracy (82.3%) than the CART (77.22%); however it is less informative of the input variables used to classify due to the complex nature of the ANN architecture. Therefore, the CART model was used to determine that 5 of the 11 predictors were significant drivers of vegetation distribution.
At the sub-hillslope scale we quantified the percent cover of specific biotic and abiotic cover types (grass, forb, shrub, bare ground, and etc.) contained within the sagebrush steppe class; the diameter at breast height (DBH), frequency, and density for the conifer classes; and percent composition and width of the “green line” of riparian reaches. We also investigated influences of topographic effects on abundance within the sagebrush steppe, Douglas fir and ponderosa pine ecosystems. Preliminary results of sagebrush steppe indicate percent cover of bare ground was 47.89% ± 3.29%, grass was 17.716% ± 4.6%, forb was 13.03% ± 2.3%, shrub was 16.0% ± 4.09%. Moreover, there was a significant (p =0.05) association between grass cover and topographic aspect, with North slopes having a grass cover of 14.58% ± 7.33% and South slopes with 11.67% ± 6.0% at the 95% confidence level (P = 0.05). The average DBH for mature Douglas fir and ponderosa pine was 84.6 cm ± 19.4 and 82.4 cm ± 24.2, respectively. Frequency and density of Douglas fir (> 1.4m in height) was 100% and 17.57 per 30 m x 30 m pixel while ponderosa pine was 57.14% and 8.77 per pixel. The most prevalent trends in riparian vegetation suggest that willow, water birch, rose, and Lewis’ mockorange all decrease in percent composition with elevation gain, while mountain maple and mountain alder increase in percent composition. Also, the breadth of the “green line” decreases with gaining elevation.
Finally, the National Aeronautics and Space Administration (NASA) has been using microwave passive and active remote sensing to develop algorithms to retrieve global soil moisture. However, there are issues with the above ground vegetation over laying the soil in which leaf water content attenuates the signal through the leaves and increases sensitivity to moisture below the top centimeter of the soil surface (Njoku and Li 1999). Therefore, we investigated spatial and temporal changes in the vegetation water content (VWC) of grass/forb life form and the relative vegetation water content (RVWC) of shrubs within the sagebrush steppe ecosystem for NASA’s model development in the Soil Moisture Active Passive (SMAP). Our data suggest that to define this value, calculations are spatially and temporally dependent and to determine the VWC across DCEW, further data acquisition and processing must be completed.