This thesis presents a case study on the application of deep learning methods for the dense prediction of land cover types in oblique ground-based photography. While deep learning approaches are widely used in land cover classification of remote-sensing data (i.e., aerial and satellite orthoimagery) for change detection analysis, dense classification of oblique landscape imagery used in repeat photography remains undeveloped. A performance evaluation was carried out to test two state-of the-art architectures, U-net and Deeplabv3+, as well as a fully-connected conditional random fields model used to boost segmentation accuracy. The evaluation focuses on the use of a novel threshold-based data augmentation technique, and three multi-loss functions selected to mitigate class imbalance and input noise. The dataset used for this study was sampled from the Mountain Legacy Project (MLP) collection, comprised of high-resolution historic (grayscale) survey photographs of Canada’s Western mountains captured from the 1880s through the 1950s and their corresponding modern (colour) repeat images. Land cover segmentations manually created by MLP researchers were used as ground truth labels. Experimental results showed top overall F1 scores of 0.841 for historic models, and 0.909 for repeat models. Data augmentation showed modest improvements to overall accuracy (+3.0% historic / +1.0% repeat), but much larger gains for under-represented classes. / Graduate

Master of Landscape Architecture / Department of Landscape Architecture/Regional and Community Planning / Blake Belanger / In recent decades, traditional methods of oil and gas extraction in the United States have been fortified by hydraulic fracturing, or fracking. The process of fracking involves injecting water, aggregates, and chemicals into the earth to rupture rock that is trapping oil and gas. This process has unlocked access to once unobtainable reserves, and as a result, U.S. oil and gas production has continued to increase despite recurring forecasts that supplies would peak. While increased production has strengthened some sectors in the U.S. economy, it has also renewed a reliance on non-renewable energy, compromised the well-being of communities, and poses serious environmental threat. While research into the process of hydraulic fracturing and its effects are common, little discussion has been generated regarding the broader impacts of the systems required to construct, supply, and maintain fracking operations. The processes of hydraulic fracturing contain a dense array of components that effect both the present and future state of communities, environments, and economies. As energy demands grow and resources deplete, these millions of facilities will demarcate the wicked problems of a post-oil and gas future and reveal a dense system of derelict infrastructure and underutilized lands. This report presents the Fracklands. Fracklands are a comprehensive telling of the landscapes of hydraulic fracturing. They offer insight into what a dynamic and complex system of modern oil and gas extraction infrastructure looks like. After first defining fracking and discussing current practices and policies as grounds, I present a classification framework for defining the Fracklands. Organized by four approaches – Systems, Typologies, Trends and Futures – this Framework utilizes a set of descriptive methods conducted in three U.S. regions to present and discuss the Fracklands. Results reveal a more complete picture of fracking’s effects on the American landscape today, while giving hints of what the Fracklands will present in the future. The Fracklands are a little understood system of components and processes that profoundly affect land, people, place, and society. By presenting the Fracklands framework in this report, I aspire that planners, designers, and decision-makers will have a clear outline for better understanding the nature of this wicked problem. As a point of departure, I propose three unique design-based alternatives to address the future of the Fracklands and dilemmas yet materialized. With the Fracklands revealed, footholds are set for a methodology to be adapted and used in future study for understanding the ever-changing landscape of hydraulic fracturing.

In landscape ecology, it is widely acknowledged that landscape is as much a social and cultural entity as it is biophysical, and that people and place must be jointly considered to fully understand the evolution of spatial pattern. This thesis explores the overlapping biophysical and human dimensions of landscape in the context of an (i) ecological and (ii) ethnoecological classification on the local landscape of the Tayal Mrqwang indigenous people in north central Taiwan. The goal of the ecological classification was to identify ecosystem types for a ~3000 acre landscape by relating vegetation patterns to gradients of physiography, soil, humidity, light, pixel brightness, and human modification across 76 transect sample plots. Using multivariate analyses, seven ecosystem types were identified, ranging from xeric through submesic pine, bamboo, alder, and laurel forests to mesic evergreen broadleaved and mixed coniferous forests. At the broad scale, ecosystems were distributed along gradients of elevation, soil, humidity and human modification, while factors related to local variability in physiography and soil development were more important at the fine scale (i.e., within elevational ecoregions). Within lower elevation sites in particular, patterns of forest variation and soil development were resonant of ancestral practices, including shifting cultivation, terrace farming, arboriculture, and selective extraction. The objective of the ethnoecological classification was to explore whether the Mrqwang people categorize landscape variation according to systematic or multidimensional knowledge structures. Results of interviews and free-listing exercises revealed an unsystematized classification that recognizes a continuum of forest variation through the intersection of three overlapping categories: history of disturbance, topography and substrate, vegetation. These categories are modified through land tenure and toponyms. The classification appears accommodating of personal experience, and it is theorized that this flexibility results in dynamic knowledge that evolves with time, generating variable characterizations of forest variation rather than formalized categories. The thesis concludes that despite the lack of formalization, the Tayal are nonetheless highly cognizant of how current forest variation coincides with the environment and the activities of their ancestors. This knowledge represents immense local expertise and must not be excluded from conservation and co-management projects in the local area. / NSERC CGS-D

Current approaches of anthropogenic impact on the ecosystems and ecosystem services are within the scope of this master's thesis. The aim of the thesis is, on the basis of current landscape classifications, landscape typologies, and their combination with approaches of ecosystem services classification, to appoint socio-ecological units as functional socio-ecological types of the territory. In other words, on the basis of classification listed, this thesis attempts to prove, that socio-ecological units are functional socio-ecological types of the territory, which should be given proper attention. Socio-ecological landscapes are defined according to degree of anthropogenic impact, selected socioeconomic characteristics, and selected ecosystem services. The thesis focuses on three of four categories of ecosystem services defined in Millennium Ecosystem Assessment - provisioning services, regulating services, and cultural services. As a methodology, the quantitative research was chosen. The research was carried out by data collection and the analysis of statistical data of Czech Statistical Office, Czech Office for Surveying, Mapping and Cadastre, and Statistical Yearbook of Soil Services. The thesis main finding is that social and ecological components of socio-ecological systems are in mutual...

No description available.