If you have any problems related to the accessibility of any content (or if you want to request that a specific publication be accessible), please contact us at email@example.com.
Quantifying and understanding vegetation change in the conterminous United States using geospatial data
AuthorParra, Adriana Sofia
AdvisorGreenberg, Jonathan A
Ecology, Evolution and Conservation Biology
AltmetricsView Usage Statistics
Human activities are a pervasive driver of vegetation change either through land-use change or through human-derived climate change. Alterations in climatic conditions that affect plant establishment, growth, and mortality will result in impacts on the distribution and composition of vegetation. Considering that terrestrial vegetation is an essential component of land-atmosphere interactions, alterations in vegetation cover and composition can affect the fluxes of water and energy in the earth system, and thus affect important environmental services for human livelihoods. Therefore, quantifying and understanding vegetation change at relevant spatial and temporal scales is a key step to improve predictions of vegetation response to climate change and to evaluate potential derived impacts on ecosystem function, and vegetation feedbacks to the earth system.In this dissertation, I assessed the changes in vegetation lifeform cover across the conterminous United States (CONUS) from 1986 to 2018 with the ultimate goal of understanding how climate impacts these rates of change. To accomplish this, in my first chapter I demonstrated the benefits of fusing airborne multisensor data with Landsat imagery in machine learning models to construct tree, shrub, and herbaceous fractional cover maps and their spatially-explicit uncertainty. I constructed annual maps of fractional percent cover, with an improved temporal and thematic resolution than other available products of percent cover. The map construction workflow used a simple set of threshold rules for creating the training and validation dataset, and I only constructed one model per lifeform type which avoids seamlines between mapping zones that appear when constructing separate models for different locations. For my second chapter, I further investigated errors in my approach to constructing these maps, as well as related approaches that included ancillary data, by explicitly exploring the spatial accuracy of the products. I found that, regardless of the set of response variables used for map construction, the resulting vegetation maps do not accurately describe the heterogeneity of the true landscape, and that overall accuracy is not strongly connected to spatial accuracy. Finally, I used the maps developed in the first chapter to quantify the rates of change in lifeform cover and to construct maps of climate potential rates of change. Observed vegetation rates of change showed a “greening” CONUS, with net increases in vegetation cover in most of the country over the last three decades, but localized areas with vegetation loss, particularly across the southwest region of the country. I found a large climate potential for vegetation gain for all lifeforms across most of the country, but a comparison between this climate potential and observed change shows that vegetation change across 90% of the land area of CONUS is controlled by non-climatic factors, most likely related to human activities.