Plants on the Peak: Field and Remote Sensing Variables in Alpine Biodiversity Models

Abstract

This thesis investigated the ability of field-based and remote sensing variables from an unmanned aerial vehicle (UAV) to predict plant species diversity in an alpine ecosystem on Snodgrass Hillslope, located adjacent to Mount Crested Butte, Colorado. Using multiple linear regression models across 88 field plots, soil moisture and slope aspect were the most consistent predictors of species diversity. Other variables in models included slope angle, soil texture, soil temperature, and infrared temperature. Field-based models explained between 37% and 77% of the variance in plant species diversity, with the strongest performance in the upper forested grid. The incorporation of remote sensing variables, including Normalized Difference Vegetation Index (NDVI) and estimated percent vegetation coverage, improved model performance in most cases and most significantly improved performance when modeling the combined lower and upper plots. Remote-only models using NDVI minimum and slope aspect provided a visualization for species diversity across the entire grids but explained less site-level variance with R2 values of 0.6663 (Adj. R2 = 0.555) for the upper grid and an R2 of 0.4123 (Adj. R2 = 0.2227) for the lower grid. These results highlight both the advantages and limitations of solely UAV-based modelling in mountainous terrain. Future projections using SnowClim v1.0 climate projections demonstrated that areas such as Snodgrass Hillslope will become warmer and more stressed for late-season moisture as snowpack decreases, snowmelt timing advances, and summer temperatures rise. These climate shifts may reduce species diversity in water-limited habitats and alter plant community composition in alpine landscapes. Keywords: alpine plant species diversity, multiple linear models, volumetric water content (VWC), slope aspect, soil texture, remote sensing, unmanned aerial vehicles (UAVs), SnowClim v1.0