Ground-level and overhead images provide complementary viewpoints of the world. This thesis proposes methods which leverage dense overhead imagery, in addition to sparsely distributed ground-level imagery, to advance traditional computer vision problems, such as ground-level image localization and fine-grained urban mapping. Our work focuses on three primary research areas: learning a joint feature representation between ground-level and overhead imagery to enable direct comparison for the task of image geolocalization, incorporating unlabeled overhead images by inferring labels from nearby ground-level images to improve image-driven mapping, and fusing ground-level imagery with overhead imagery to enhance understanding. The ultimate contribution of this thesis is a general framework for estimating geospatial functions, such as land cover or land use, which integrates visual evidence from both ground-level and overhead image viewpoints.

ABSTRACT ROSS CARTER Traffic-related Pollution Exposure Assessment of Fulton County and Atlanta Public Schools (K-12) in Proximity to Major Highways and Expressways Background: A number of studies have linked traffic-related pollutant exposures to asthma in children. Health conditions such as asthma can contribute to school absenteeism and missed learning opportunities as well as place a major burden on health resources. Although children spend a significant amount of time in school, few states have adopted school siting policies that prevent the placement of schools near major highways. Furthermore, schools often fail to take appropriate steps (e.g. adequate HVAC and air filters) to address indoor air quality concerns for students. The study was designed to identify the number of schools and the number of children in Fulton County and Atlanta Public Schools that attend schools in proximity to major highways and expressways where they are likely to have greater exposure to traffic pollution. The results of this study can be used to build an evidence base for stricter school siting guidelines, for planning safe routes to school, and for mitigation strategies to limit pollutant exposures for children who attend high-risk schools. Methods: Highway, expressway, county, and school shapefiles were overlaid using ArcMap in ArcGIS version 10.1 (ERSI, USA). A circular buffer with radius 0.5 mile (~ 800 m) was created for each school. ArcMap geospatial tools were used to identify major highways and expressways with these buffers. Results: A total of 119 of the 225 schools in Fulton County and Atlanta Public school districts were identified as being located within 0.5 miles of a major highway or expressway. Of the 119 schools meeting the intersection criteria, 72.2% (86 of 119 schools) were designated Title I. Conclusion: Approximately half of schools were located within 0.5 miles of a major highway or expressway. This may result in elevated levels of traffic-related air pollution on the school campuses and potential increased exposure for students. Proper air filter selection, HVAC maintenance, and air quality programs as well as land use, planning, and assessment measures are recommended for these schools to help mitigate exposures.

Crime associated with alcohol consumption poses a serious problem. There is a variety of approaches that try to conceptualize this relationship. One of the most progressive approaches is that of emphasizing the effects of alcohol outlets on crime associated with alcohol consumption. Even though it is a well-established field in other regions, in Czech Republic, there are no available studies dealing with the relationship of alcohol outlets and crime associated with alcohol consumption. This diploma thesis deals with the association of on-premise alcohol outlets and crime in Czech Republic. The unit of analysis used in this diploma thesis was police districts. In the final data set, we had 517 police districts. For all of the police districts, the incidence rates of crime, on-premise alcohol outlet densities and other characteristics were obtained. Analysis by the means of linear regression and geographically weighted regression was performed on data. The results of analysis indicate that on-premise alcohol densities are associated with all examined crime incidence rates (except of road accidents), even after adjusting for other variables. Also, there is a non-trivial spatial variation in data. The regression models had high explanatory power. The results of this diploma thesis imply that it is...

Abstract Background Measles is a serious infectious disease that contributes significantly to the burden of disease in many developing countries. In most developed nations, such as Canada, endemic transmission of measles has been declared eliminated thanks to rigorous vaccination programs, but isolated outbreaks of the disease continue to happen. Therefore, a thorough understanding of the factors contributing to these outbreaks is necessary. Objectives There were two main objectives of this thesis. The first objective was to assess the geospatial distribution of reported measles cases in Ontario with a goal of identifying clusters of reported measles. For this objective, the main hypothesis was that measles cases would not be randomly distributed across Ontario and instead would cluster in certain regions. The second objective was to explore some of the factors that may be associated with measles clusters. For this objective, the main hypothesis was that the proportion of immigrants, population density, low-income prevalence and education level would be associated with measles clusters. Methods The first objective was achieved through a thorough geospatial analysis using SaTScan and R. Individual forward sortation areas were used as the spatial unit of analysis. The analysis leveraged data from multiple sources: 2016 Census data, Ontario measles cases data from iPHIS from 2008 to 2019, a shapefile of all forward sortation areas in Canada from Statistics Canada and centroid coordinates of forward sortation areas that were obtained using web scrapping techniques on the geolocation service of Natural Resources Canada. The maximal window size of the geospatial analysis was chosen using the maximum clustering heterogeneous set-proportion technique. The geospatial analysis was run with 99,999 Monte Carlo repetitions under a Poisson distribution using the purely spatial analysis. The Ontario population from the 2016 Census was used as the population at risk. Any cluster with a p ≤ 0.05 was deemed statistically significant. The second objective was achieved through a case-control study: Forward sortation areas that were within statistically significant measles clusters were considered as cases and the rest of the forward sortation areas were considered as controls. Demographic data necessary to assess the factors of interest were extracted from the 2016 Census. A univariable logistic regression model was run to compute the odds ratio and test the association between the factors of interest and measles clusters. 95% confidence intervals were computed for each odds ratio. Data-curation techniques and data analysis were performed in R 4.0.4. Results From 2008 through 2019, 178 measles cases were identified. 82% of cases lacked necessary vaccination or vaccination records against measles, 35% of cases were linked to traveling outside of Ontario, 20% of cases reported being in contact with a known case, and 72% of cases were less than 5 years old or older than 21. Ten measles clusters were identified of which six were deemed statistically significant. These six significant clusters represented 7% of the population at risk but contained nearly 40% of all reported measles cases between 2008 and 2019. Measles clusters had a strong association with the proportion of immigrants living within them, population density and prevalence of low-income. No association was found between education level and measles clusters. Conclusion The results indicate that most measles cases in Ontario are unvaccinated or lack proof of vaccination; arise through secondary transmission within the province; arise from undetected transmission; and are adults or infants. Additionally, it is possible to see that the risk of reported measles cases is not randomly distributed across the province, but instead measles cases tend to cluster in certain regions. Such clusters tend to be characterized by specific population-level factors that may be contributing to the risk of reported measles. Targeted and equitable interventions are needed as we continue on the path to eradication.

Abortion, the termination of pregnancy, is safe when provided as a surgical procedure by a trained provider or when the correct dosage of the drugs mifepristone and/or misoprostol are used. Despite this, many barriers to abortion care exist. In the United States (US), targeted state-level abortion restrictions create barriers to care, which make it so that people who wish to utilize abortion care face difficulty or are unable to do so. Such barriers to care have important public health implications, as studies have shown that individuals who cannot access wanted abortion care have poorer psychological, physical, social, and economic outcomes than those who obtained care. This dissertation aims to examine one component of abortion access, accessibility, operationalized as the drive time from a woman’s home to the nearest abortion-providing facility. We employ a novel measure of abortion accessibility constructed from three data sources: (1) the Advancing New Standards in Reproductive Health facility database; (2) US Census estimates and shapefiles; and (3) OpenStreetMap data. In the first study, we used geographic information systems (GIS) to explore the effect of programmatic and policy changes related to telemedicine for medication abortion services (TMAB) on population-level measures of abortion accessibility, or drive time to the nearest abortion-providing facility. We found that either expansions in TMAB services or removal of TMAB bans could improve abortion accessibility in the US. For these two exposure scenarios, compared to the current abortion provision scenario, increases in the proportion of women within a 30-, 60-, and 90-minute drive time of an abortion-providing facility ranged from 1.25 percentage points, or an additional 781,556 US women aged 15-44 years with accessibility, to 5.66 percentage points, or an additional 3,530,423 US women aged 15-44 years with accessibility. In the second study, we used GIS to assess the potential effect of the geographic unit of analysis (i.e., block group, ZIP code tabulation area [ZCTA], or county) on misclassification of the proportion of US women of reproductive age within a 30-minute drive time of an abortion-providing facility relative to a measure calculated using Census blocks. We found that block group- or ZCTA-based estimates of abortion accessibility were an underestimate, but resulted in little misclassification relative to measures constructed using Census blocks at the national level; however, county-based measures substantially underestimated abortion accessibility compared with Census block-based measures. Nationwide, the Census block-based abortion accessibility estimate was 0.35 percentage points greater than the block group-based estimate, 2.72 percentage points greater than the ZCTA-based estimate, and 24.21 percentage points greater than the county-based estimate. By state, the Census block-based abortion accessibility estimate ranged from 0 to 8.51 percentage points greater than the block group-based estimate, from 0 to 27.86 percentage points greater than the ZCTA-based estimate, and from 0 to 79.49 percentage points greater than the county-based estimate. Given that state-level ZCTA-based estimates could be substantially different from the Census block-based estimate, ZCTA-based estimates are likely not appropriate for state-level analyses or US analyses stratified by state. Finally, in the third study, we assessed the relationship between level of accessibility in an abortion client’s home ZCTA and the gestational age at which the client obtained abortion care, using fine stratification by propensity score to control confounding. We found that compared with living in a ZCTA with >0% accessibility, living in a ZCTA with 0% accessibility was associated with a decreased risk of being at or beyond 14 weeks’ gestation at abortion visit. These unexpected findings could be due to a selection bias induced by limiting the sample to those who obtained abortion care, uncontrolled or poorly controlled confounding, misclassification of exposure and/or outcome, and/or unidentified effect measure modification by state abortion provision landscape. Through these three dissertation studies, we highlighted the potential impact on abortion accessibility in the US with different changes in programming and policy, quantified misclassification of abortion accessibility, and examined how misclassification varied by geographic measure and location. The third study in this dissertation suggests a need for more research to identify how selection bias may affect studies of abortion access in the US that rely on data only from those who are able to access care.

Water shortages resulting from increased demand or reduced supply may be addressed, in part, by redirecting recycled water for irrigation, industrial reuse, groundwater recharge, and as effluent discharge returned to streams. Recycled water is an essential component of integrated water management and broader adoption of recycled water will increase water conservation in water-stressed coastal communities. This study examined spatial patterns of recycled water use in Florida in 2009 and 2015 to detect gaps in distribution, quantify temporal change, and identify potential areas for expansion. Databases of recycled water products and distribution centers for Florida in 2009 and 2015 were developed by combining the 2008 and 2012 Clean Water Needs Survey databases with Florida’s 2009 and 2015 Reuse Inventory databases, respectively. Florida increased recycled water production from 674.85 mgd in 2009 to 738.15 mgd in 2015, an increase of 63.30 mgd. The increase was primarily allocated to use in public access areas, groundwater recharge, and industrial reuse, all within the South Florida Water Management District (WMD). In particular, Miami was identified in 2009 as an area of opportunity for recycled water development, and by 2015 it had increased production and reduced the production gap. Overall, South Florida WMD had the largest increase in production of 44.38 mgd (69%), while Southwest Florida WMD decreased production of recycled water by 1.68 mgd, or 3%. Overall increase in use of recycled water may be related to higher demand due to increased population coupled with public programs and policy changes that promote recycled water use at both the municipal and individual level.

Karst springs are an essential source of private water supply for about 10% of households in Tennessee. However, these springs, which can be easily polluted, are unregulated. This study, therefore, assesses water quality spatial patterns and water quality rating of roadside springs in northeast Tennessee. Karst spring water samples collected from 50 springs were assessed using EPA Standard methods for pathogens, nutrients, radon, and physicochemical parameters. Springs generally met federal and state standards for physicochemical parameters, 90% of samples contained E. coli, and all samples contained fecal coliform. High E. coli was spatially clustered causing a fecal contamination hot spot on the border of Washington and Sullivan Counties, Tennessee. 60% of springs exceeded radon concentrations of 300 pCi/L. Water quality ratings were very poor or unfit for drinking, with 4% of springs ranked “good”. Therefore, microbial pollution purification procedures are advised before using these springs as a drinking water source.

The Ebola outbreak in West Africa in 2014 to 2016 had more than 28,000 suspected, probable, and confirmed cases. It was the largest Ebola outbreak in history. Of the 28,000 cases in the three Ebola-affected countries, Liberia had 10,000 cases with almost 5,000 deaths. The Ebola Virus Disease (EVD) entered Liberia along the border of Guinea and moved to the capital city of Monrovia where the virus spread. Ebola Treatment Units (ETUs) were constructed throughout the response in locations where there were available facilities versus distance to care challenges. This study examined the association of distance from villages to ETUs and mortality. Using Geographic Information System (GIS) and statistics framed within the Social Ecological Model and the GIS Framework, this study geolocated the Ebola cases by village, mapped the travel routes and calculated the distance to the ETU. A logistic regression was then used to determine if there was an association between distance and mortality, with and without controlling for age and gender, and, to calculate the odds ratio. A logistic regression model showed there is an association between distance and mortality and that Ebola patients living within 12 kilometers of the ETU were 1.8 times less at risk of mortality (OR = 1.778, 95% CI [1.171 - 2.7]) than those living more than 12 kilometers. In addition, males had a 1.4 times lower risk of death due to EVD. This understanding can inform future outbreak responses and placement of treatment units. In addition, this information can lead to social change with respect to individual understanding of access to care, community expectations, and national health care planning.

Digital soil mapping typically involves inputs of digital elevation models, remotely sensed imagery, and other spatially explicit digital data as environmental covariates to predict soil classes and attributes over a landscape using statistical models. Digital imagery from Landsat 5, a digital elevation model, and a digital geology map were used as environmental covariates in a 67,000-ha study area of the Great Basin west of Fillmore, UT. A “pre-map” was created for selecting sampling locations. Several indices were derived from the Landsat imagery, including a normalized difference vegetation index, normalized difference ratios from bands 5/2, bands 5/7, bands 4/7, and bands 5/4. Slope, topographic curvature, inverse wetness index, and area solar radiation were calculated from the digital elevation model. The greatest variation across the study area was found by calculating the Optimum Index Factor of covariates, choosing band 7, normalized difference ratio bands 5/2, normalized difference vegetation index, slope, profile curvature, and area solar radiation. A 20-class ISODATA unsupervised classification of these six data layers was reduced to 12. Comparing the 12-class map to a geologic map, 166 sites were chosen weighted by areal extent; 158 sites were visited. Twelve points were added using case-based reasoning to total 170 points for model training. A validation set of 50 sites was selected using conditioned Latin Hypercube Sampling. Density plots of sample sets compared to raw data produced comparable results. Geology was used to stratify the study area into areas above and below the Lake Bonneville highstand shoreline. Raster data were subset to these areas, and predictions were made on each area. Spatial modeling was performed with three different models: random forests, support vector machines, and bagged classification trees. A set of covariates selected by random forests variable importance and the set of Optimum Index Factor covariates were used in the models. The Optimum Index Factor covariates produced the best classification using random forests. Classification accuracy was 45.7%. The predictive rasters may not be useful for soil map unit delineation, but using a hybrid method to guide further sampling using the pre-map and standard sampling techniques can produce a reasonable soil map.

Mitigating energy poverty requires a multi-criteria decision protocol integrating socio-economic, cultural, environmental, and technical systems, influencing energy access, and consumption. Situations of energy poverty are typical in rural and urban poor households, particularly in sub-Saharan Africa. These situations are commonly prevalent in informal settlements, sprawling across the periphery of South African metros. Majorities of informal households lack access to grid-electricity and consume local energy sources for their energy needs. There are ongoing government efforts directed to mitigating energy poverty among energy-poor households, such as informal households, through policies and subsidies. Socio-economic and cultural environments also redefine the extent to which energy poverty is mitigated in these households. At present, informal households are constantly and rapidly growing, and as a result, compromise policy effectiveness and other functional strategies, targeting to mitigating energy poverty in these households, and achieving universal energy access in South Africa. Accordingly, this research study adopted a multidisciplinary approach to understanding related matters of energy poverty based on energy policies; electricity access, and pricing; geospatial analysis; energy use and access; and management strategies, with emphasis on informal settlements in South Africa. The first part of the study reviewed energy pro-poor policies, relevant to improving energy access and energy-use efficiency in energy-poor households in South Africa. The study also investigated electricity access (access rates), connection costs (access costs), and electricity tariffs to understand historical precedents and forecast scenarios, and the relationships to gaining complete electricity access by 2030 in the City of Cape Town. The third part mapped and monitored informal areas to understand landscape processes and poverty with energy poverty propagations by Land Cover (LC) and Land-Cover Change (LCC) in the City of Cape Town. The fourth part of the research investigated energy-use patterns and other energy-related matters in a selected informal settlement - a typical case study of an energy-poor community in South Africa and sub-Saharan Africa. The last part proposed and designed a novel System Reinforcing Model (SRM), an Energy Access Sustainability (EAS) management scheme, applicable to mitigating energy poverty in any energy-poor community. The study review validated government efforts in improving energy access in energy-poor households through commissioned energy pro-poor policies but not without drawbacks and proposed recommendations to support future policy reforms. The research also revealed iv A novel systems approach to energy poverty in sub-Saharan Africa: A South African informal settlement as case study. increasing patterns in historical trends of access rates, costs, and tariffs, and relationships between parameters within the assessment period (from 2010 to 2018). The forecast analyses (from 2019 to 2030) demonstrated that total electricity access could not be reached by 2030 without a shift in Business-As-Usual (BAU) patterns in the City of Cape Town. The LC conversions of informal areas revealed poverty with energy poverty propagations through landscape degradation processes - Persistence and Intensification - in the City of Cape Town. The research study further revealed poor energy use patterns and behaviour in the target Settlement. Informal households in the settlement mainly adopted local energy fuels and appliances in satisfying household energy needs. The novel part of the research study described the application of a systems approach - Systems engineering (SE) and Systems Thinking (SsT) - into energy poverty and access processes to developing the new SRM. SE and SsT concept analyses were employed in identifying and integrating four operating system interfaces in these processes into the new SRM. The new SRM simulated complex systems and elements within the interfaces and categorized them as design decisions and system designs. These systems and elements were grounded in energy-use patterns and behaviour, energy access, and EAS, as well as socio-economic, cultural, technical, and environmental features. Arrays of feedback loops in reinforcing patterns in the new SRM modelled the interactions between, and within, design decisions and system designs, for future energy access rebranding, based on significant sustainability outcomes of favourably coalesced system interfaces. SRM was applied in the target settlement, where the model’s significance was validated. Based on its multi-criteria decision approach, among its many features, SRM revealed system parts instigating energy poverty situations and limiting EAS in the target settlement. SRM tailored energy access solutions, whilst integrating significant outcomes of the whole research study, to advancing energy poverty mitigation and EAS in the target settlement. / Thesis (PhD (Technology Management))--University of Pretoria, 2020. / UP Postgraduate Bursary / International Council on Systems Engineering (INCOSE) / Graduate School of Technology Management (GSTM) / PhD (Technology Management) / Unrestricted