Doctor of Philosophy / Department of Geography / Douglas G. Goodin / Prescribed burning in tallgrass prairie affects a wide range of human and natural systems. Consequently, managing this biome based on sound science, and with the concerns of all stakeholders taken into account, requires a method for mapping burned areas. In order to devise such a method, many different spectral ranges and spectral indices were tested for their ability to differentiate burned from unburned areas at both the field and satellite scales. Those bands and/or indices that performed well, as well as two different classification techniques and two different satellite-based sensors, were tested in order to come up with the best combination of band/index, classification technique, and sensor for mapping burned areas in tallgrass prairie. The ideal method used both the red and near-infrared spectral regions, used imagery at a spatial resolution of at least 250 m, used satellite imagery with daily temporal resolution, and used pixel-based classification techniques rather than object-based techniques. Using this method, burned area maps were generated for the Flint Hills for every year from 2000-2010, creating a fire history of the region during that time period. These maps were compared to active fire and burned area products, and these products were found to underestimate burned areas in tallgrass prairie.

The evolution of software applications from single desktops to sophisticated cloud-based systems is challenging. In particular, applications that involve massive data sets, such as geospatial applications and data science applications are challenging for domain experts who are suddenly constructing these sophisticated code bases. Relatively new software practices, such as Microservice infrastructure and DevOps, give us an opportunity to improve development, maintenance and efficiency for the entire software lifecycle. Microservices and DevOps have become adopted by software developers in the past few years, as they have relieved many of the burdens associated with software evolution. Microservices is an architectural style that structures an application as a collection of services. DevOps is a set of practices that automates the processes between software development and IT teams, in order to build, test, and release software faster and increase reliability. Combined with lightweight virtualization solutions, such as containers, this technology will not only improve response rates in cloud-based solutions but also drastically improve the efficiency of software development. This thesis studies two applications that apply Microservices and DevOps within a domain-specific application. The advantages and disadvantages of Microservices architecture and DevOps are evaluated through the design and development on two different platforms---a batch-based cloud system, and a general purpose cloud environment. / Graduate

n/a

Optimal scale of study and scaling are fundamental to ecological research, and have been made easier with remotely sensed (RS) data. With access to RS data at multiple scales, it is important to identify how they compare and how effectively information at a specific scale will potentially transfer between scales. Therefore, my research compared the spatial, spectral, and temporal aspects of scale of RS data to study biophysical properties and spatio-temporal dynamics of the northern mixed prairie vegetation. I collected ground cover, dominant species, aboveground biomass, and leaf area index (LAI) from 41 sites and along 3 transects in the West Block of Grasslands National Park of Canada (GNPC; +49°, -107°) between June-July of 2006 and 2007. Narrowband (VIn) and broadband vegetation indices (VIb) were derived from RS data at multiple scales acquired through field spectroradiometry (1 m) and satellite imagery (10, 20, 30 m). VIs were upscaled from their native scales to coarser scales for spatial comparison, and time-series imagery at ~5-year intervals was used for temporal comparison. Results showed VIn, VIb, and LAI captured the spatial variation of plant biophysical properties along topographical gradients and their spatial scales ranged from 35-200 m. Among the scales compared, RS data at finer scales showed stronger ability than coarser scales to estimate ground vegetation. VIn were found to be better predictors than VIb in estimating LAI. Upscaling at all spatial scales showed similar weakening trends for LAI prediction using VIb, however spatial regression methods were necessary to minimize spatial effects in the RS data sets and to improve the prediction results. Multiple endmember spectral mixture analysis (MESMA) successfully captured the spatial heterogeneity of vegetation and effective modeling of sub-pixel spectral variability to produce improved vegetation maps. However, the efficiency of spectral unmixing was found to be highly dependent on the identification of optimal type and number of region-specific endmembers, and comparison of spectral unmixing on imagery at different scales showed spectral resolution to be important over spatial resolution. With the development of a comprehensive endmember library, MESMA may be used as a standard tool for identifying spatio-temporal changes in time-series imagery. Climatic variables were found to affect the success of unmixing, with lower success for years of climatic extremes. Change-detection analysis showed the success of biodiversity conservation practices of GNPC since establishment of the park and suggests that its management strategies are effective in maintaining vegetation heterogeneity in the region. Overall, my research has advanced the understanding of RS of the northern mixed prairie vegetation, especially in the context of effects of scale and scaling. From an eco-management perspective, this research has provided cost- and time-effective methods for vegetation mapping and monitoring. Data and techniques tested in this study will be even more useful with hyperspectral imagery should they become available for the northern mixed prairie.

River ice is an important component of the traditional way of life for the communities along the Slave River both culturally and economically. During the winter, a stable ice cover provides local residents with safe access to their traditional hunting, trapping, and fishing grounds along the river. Periodic spring ice breakup flooding is required to maintain the ecological balance along the Slave River Delta. Recently, however, local observations have indicated changes in ice cover characteristics (e.g. air pocket formation, double layer ice, ice cover flooding) during the winter, which increase the risks of travelling on the ice. Also prolongs dry periods during the spring are leading to rapid growth of invasive vegetation that reduces the lake and channel areas of the Delta. Although some attempts have been made to understand the patterns of spring flood frequency in the Delta, very little is known about the Slave River’s ice cover characteristics and behaviour. Remote sensing techniques and field surveys were used in this study to understand the ice cover progression and to examine ice cover characteristics along the river during the winters of 2013-2014 and 2014-2015. RADARSAT-2 satellite imagery captured the changes in the ice cover and identified different types of ice during the winter seasons at two primary study sites – downstream of Fort Smith and the Slave River Delta. The mechanism of ice cover growth, with the formation of air pockets and layers underneath the ice cover was investigated. Steeper channels and several open water sections appear to be contributing to significant amounts of air entrainment into the water in winter. Changes in the hydraulic characteristics due to flow regulation and ice cover progression can also change the quantity and distribution of air pockets along the river ice cover. Additionally, the impact of flow fluctuations on the ice cover (e.g. ice cover flooding) was also observed. Increases in discharge cause the ice cover to crack or dislodge from the river banks, leading to water seeping onto the ice and flooding it, which has implications for the muskrat and beaver populations. A geospatial model was developed to determine the spatial patterns of ice cover breakup along the river from Fort Fitzgerald to the delta. This model successfully identified the areas of breakup initiation and persistence of ice until the end of the breakup. MODIS satellite imagery was used to describe the temporal patterns and evolution of breakup events between the years 2008 and 2011. In addition to geomorphological influences, air temperature and flow conditions also have strong impacts on the spatial and temporal patterns of the ice cover breakup.

No description available.

Snow occurs in the High Drakensberg of southern Africa approximately eight times per annum. Snow cover is frequently captured by Landsat satellite imagery, which provide data for the monitoring of snow cover in other regions of the world. Together with a digital elevation model, repetitive snow cover data are used to analyse the distribution of snow cover in the High Drakensberg study area. The effect that the regional and local topography, latitude, and climatic conditions have on the spatial distribution of snow and the function that temperature, wind, altitude, aspect and slope gradient play in the preservation of snow cover are examined. The results of the spatial study allow for the identification of sites that support the accumulation of snow. Specific active and relict geomorphological features were surveyed and correlated spatially to the contemporary snow cover. Among such features are linear debris ridges on south-facing valley slopes in the High Drakensberg. These appeared similar to glacial features found elsewhere in the world and are thus significant in a long-standing and highly conjectured debate over the validity of possible plateau, cirque and niche glaciation in the region. Late-lying snow cover favours gently sloping south- and southeast-facing aspects at altitudes from 3000 m ASL to just below the highest peaks in the region near 3450 m ASL, above which higher insolation levels on the flat mountain summits provides unfavourable conditions. Snow cover immediately adjacent to the Drakensberg escarpment ablates quickly whilst snow cover at high altitudes in the Lesotho interior experiences better preservation conditions. Latitude has no obvious impact on the distribution of snow cover due to the dominant role of topography in the High Drakensberg other than a limiting of snowfall to regions south of 29°S in late spring. Various synoptic conditions produce snowfall in the region, with cold fronts associated with midlatitude cyclones producing the majority of snow cover. A strong correlation exists between the spatial distribution of snow cover and specific geomorphological features. Observed linear debris ridges are located on slopes that experience frequent contemporary snow cover, lending credence for a glacial origin of the ridges during a period of colder environmental conditions.

Le contrôle de la qualité de l'eau est fondé naturellement et traditionnellement sur des mesures et des prélèvements in situ. Des images satellites étalonnées à partir des données mesurées in situ fournissent une information quantitative et continue sur le milieu aquatique et peuvent être employées pour estimer, avec une approximation raisonnable, les facteurs affectant la qualité de l’eau L’objectif de ce travail consiste à modéliser les propriétés optiques de l’eau de mer et les paramètres physico-chimiques qui caractérisent les eaux côtières. L'application est basée sur le développement d’un Système d’Information Marin caractérisant un système de gestion de base de données géoréférencié POlGIS dédié à la gestion de l'information marine dans le cas de contrôle, suivi et surveillance de la pollution. Nous présentons des modèles exprimant les variables indicatrices de la qualité des eaux du littoral Algérois et la réflectance calculée de chaque pixel à partir d’un modèle physique de correction radiométrique. Les mesures in- situ sont effectuées pour des zones de différentes qualités d’eaux et leurs réflectances sont calculées à partir des images satellites SPOT, Landsat TM, MSS, IRS1C et Seawifs Finalement, des modèles sont établies avec les réflectances permettent d’obtenir des images satellites indicatrices de la pollution et de la bathymétrie des zones côtières à partir du logiciel de traitement d’image PCSATWIN développé afin d’estimer pour chaque pixel le degré de pollution du milieu. / In order to protect the natural medium and to control the pollution caused by such rejects, it is necessary to achieve a continuous survey of the reject zones. The goal of this study is a developed a methodology for modelling pollution and bathymetry from the digital satellite images.Indeed, our objective consists of the development of a software POLGIS intended for the management of the marine databases for the control and the monitoring of the pollution Satellite imagery can be used to estimate, with a reasonable accuracy, the factors affecting the water quality .It has a great importance to achieve the necessary continuous monitoring of the relevant area with an overall analysis of its pollution. A modelling analysis between the pollution contents and the reflectance calculated by the satellite images allow us to transform rough images into images treated and combined using a software of satellite image processing PCSATWIN developed in this study. This complex phenomena us developed an analytic model of radiatif transfer simulation in water coupled to an atmospheric model in order to simulate measure by satellite. This direct model permits to follow the solar radiance in his trajectory Sun-Atmosphere - Sea - Depth of sea- sensor. The goal of this simulation is to show for every satellite of observation (Spot, Landsat MSS, TM) possibilities that can offer in domain of bathymetry.The reflectance coefficient is calculated from satellite image, the detection and the possible determination of the zones contaminated by pollution using the space techniques constitute an effective means to intervene in order to ensure the monitoring of the Algerian coasts. The analysis shows us that each sensor offers useful information and that the combination between these various informations makes it possible to propose a procedure of maps establishment that can be interpreted as pollution maps.

<div>Convolutional neural networks lie at the heart of nearly every object recognition system today. While their performance continues to improve through new architectures and techniques, some of their deciencies have not been fully addressed to date. Two of these deciencies are their inability to distinguish the spatial relationships between features taken from the data, as well as their need for a vast amount of training data. Capsule networks, a new type of convolutional neural network, were designed specically to address these two issues. In this work, several capsule network architectures are utilized to classify objects taken from overhead satellite imagery. These architectures are trained and tested on small datasets that were constructed from the xView dataset, a comprehensive collection of satellite images originally compiled for the task of object detection. Since the objects in overhead satellite imagery are taken from the same viewpoint, the transformations exhibited within each individual object class consist primarily of rotations and translations. These spatial relationships are exploited by capsule networks. As a result it is shown that capsule networks achieve considerably higher accuracy when classifying images from these constructed datasets than a traditional convolutional neural network of approximately the same complexity.</div>

Accurate and up-to-date road information is essential for both effective urban planning and disaster management. Today, very high resolution (VHR) imagery acquired by airborne and spaceborne imaging sensors is the primary source for the acquisition of spatial information of increasingly growing road networks. Given the increased availability of the aerial and satellite images, it is necessary to develop computer-aided techniques to improve the efficiency and reduce the cost of road extraction tasks. Therefore, automation of image-based road extraction is a very active research topic. This thesis deals with the development and implementation aspects of a semi-automatic road extraction strategy, which includes two key approaches: multidirectional and single-direction road extraction. It requires a human operator to initialize a seed circle on a road and specify a extraction approach before the road is extracted by automatic algorithms using multiple vision cues. The multidirectional approach is used to detect roads with different materials, widths, intersection shapes, and degrees of noise, but sometimes it also interprets parking lots as road areas. Different from the multidirectional approach, the single-direction approach can detect roads with few mistakes, but each seed circle can only be used to detect one road. In accordance with this strategy, a RoadModeler prototype was developed. Both aerial and GeoEye-1 satellite images of seven different types of scenes with various road shapes in rural, downtown, and residential areas were used to evaluate the performance of the RoadModeler. The experimental results demonstrated that the RoadModeler is reliable and easy-to-use by a non-expert operator. Therefore, the RoadModeler is much better than the object-oriented classification. Its average road completeness, correctness, and quality achieved 94%, 97%, and 94%, respectively. These results are higher than those of Hu et al. (2007), which are 91%, 90%, and 85%, respectively. The successful development of the RoadModeler suggests that the integration of multiple vision cues potentially offers a solution to simple and fast acquisition of road information. Recommendations are given for further research to be conducted to ensure that this progress goes beyond the prototype stage and towards everyday use.