Remotely sensing and modeling semi-arid rangeland grass biodiversity in the greater Yellowstone ecosystem

Abstract

Western rangelands and their ecosystem services are integral to Montana's economy and culture. Researchers and land managers are increasingly relying on remote sensing techniques for monitoring and modeling ecosystems at scales far greater than possible when primarily relying on conventional field surveys. Despite growing academic interest in the value of rangelands and their ecosystem services, gaps remain regarding the degree to which remotely sensed products can assist in monitoring and modeling rangeland biotic communities. In this study, I examine two applications of remotely sensed data to study rangeland biodiversity, introduced in Chapter 1. In Chapter 2, I utilize EMIT imaging spectroscopy to capture signals of beta-diversity in GYE rangelands outside of the growing season, detecting around half of the differences in plant community between sites based on differences in spectral variance from off- season images. While the degree to which this signal is driven by biotic and abiotic landscape features is unclear and warrants future interrogation, our findings suggest that imaging spectroscopy data can be valuable for understanding and monitoring rangeland biodiversity. In Chapter 3, I utilize remotely sensed environmental variables and rangeland grass species occupancy data from three separate studies/field seasons to develop species distribution models. These models are used to predict species occupancy across our study site and assess the degree to which internal model fit can be used to select models with reliable predictive accuracy. I also explore how this relationship between fit and predictive performance is impacted by the environmental variables used. I found significant correlations between model fit and assessments of model transferability, but these relationships are too weak to be useful for decision making. Additionally, predictive performance is highly dependent on the species being modeled and the environmental space within which predictions are being made. These two studies explore how remotely sensed products can be used to study biodiversity and species distribution at scales that are unrealistic when relying on traditional species surveys. In Chapter 4, I provide a synthesis and the conclusions that can be drawn from this research.