Christopher David Storie

The variability of urban growth in the Mexico City conurbation reflects the complexity of changes and uncertainty experienced in many urban areas. The goal of this project is to identify how the urban expansion of the Mexico City conurbation reflects the changes in socioeconomic variables of the region over recent decades. The authors first spatially quantified the rate of urban growth from 1990 to 2020 using Landsat data and then identified the socioeconomic variables associated with this urban expansion. Results showed a progressive loss in vegetated land and an increase in urban land. The population’s access to roads and the marginalization index had the highest positive correlation with the observed urban growth. Although these variables were highly correlated with each other, access to roads was not found to be a variable of importance for projecting urban growth. Finally, two distinct zones of urban growth were determined using cluster statistics; the first showed no growth, which corresponded with more established, older municipalities closer to the city centre, and the second zone had significant growth, which corresponded to municipalities away from the centre, coinciding with urban decentralization and new lifestyle patterns. The identification of urban growth zones and socioeconomic variables associated with that growth will assist with effective planning, infrastructure development, and resource management.

Deep learning semantic segmentation algorithms have provided improved frameworks for the automated production of land use and land cover (LULC) maps, which significantly increases the frequency of map generation as well as consistency of production quality. In this research, a total of 28 different model variations were examined to improve the accuracy of LULC maps. The experiments were carried out using Landsat 5/7 or Landsat 8 satellite images with the North American land change monitoring system (NALCMS) labels. The performance of various convolutional neural networks and extension combinations were assessed, where Visual Geometry Group Network with an output stride of 4, and modified U-Net architecture, provided the best results. Additional expanded analysis of the generated LULC maps was also provided. Using a deep neural network, this work achieved 92.4% accuracy for 13 LULC classes within southern Manitoba representing a 15.8% improvement over published results for the NALCMS. Based on the large regions of interest, higher radiometric resolution of Landsat 8 data resulted in better overall accuracies (88.04%) compare to Landsat 5/7 (80.66%) for 16 LULC classes. This represents an 11.44% and 4.06% increase in overall accuracy compared to previously published NALCMS results, including larger land area and higher number of LULC classes incorporated into the models compared to other published LULC map automation methods.

Deep learning has become one of remote sensing scientists' most efficient computer vision tools in recent years. However, the lack of training labels for the remote sensing datasets means that scientists need to solve the domain adaptation problem to narrow the discrepancy between satellite image datasets. As a result, image segmentation models that are then trained, could better generalize and use an existing set of labels instead of requiring new ones. This work proposes an unsupervised domain adaptation model that preserves semantic consistency and per-pixel quality for the images during the style-transferring phase. This paper's major contribution is proposing the improved architecture of the SemI2I model, which significantly boosts the proposed model's performance and makes it competitive with the state-of-the-art CyCADA model. A second contribution is testing the CyCADA model on the remote sensing multi-band datasets such as WorldView-2 and SPOT-6. The proposed model preserves semantic consistency and per-pixel quality for the images during the style-transferring phase. Thus, the semantic segmentation model, trained on the adapted images, shows substantial performance gain compared to the SemI2I model and reaches similar results as the state-of-the-art CyCADA model. The future development of the proposed method could include ecological domain transfer, a priori evaluation of dataset quality in terms of data distribution, or exploration of the inner architecture of the domain adaptation model.

Deep learning has become one of remote sensing scientists' most efficient computer vision tools in recent years. However, the lack of training labels for the remote sensing datasets means that scientists need to solve the domain adaptation problem to narrow the discrepancy between satellite image datasets. As a result, image segmentation models that are then trained, could better generalize and use an existing set of labels instead of requiring new ones. This work proposes an unsupervised domain adaptation model that preserves semantic consistency and per-pixel quality for the images during the style-transferring phase. This paper's major contribution is proposing the improved architecture of the SemI2I model, which significantly boosts the proposed model's performance and makes it competitive with the state-of-the-art CyCADA model. A second contribution is testing the CyCADA model on the remote sensing multi-band datasets such as WorldView-2 and SPOT-6. The proposed model preserves semantic consistency and per-pixel quality for the images during the style-transferring phase. Thus, the semantic segmentation model, trained on the adapted images, shows substantial performance gain compared to the SemI2I model and reaches similar results as the state-of-the-art CyCADA model. The future development of the proposed method could include ecological domain transfer, {\em a priori} evaluation of dataset quality in terms of data distribution, or exploration of the inner architecture of the domain adaptation model.

Estimating surface water using satellite data can improve estimates and predictions for hydroelectric energy generation. The goal of this research is to identify the best threshold method and image band combination to detect seasonal surface water at the basin scale. This paper explores Otsu, Minimum Entropy, and average of Otsu-Minimum Entropy threshold algorithms using Sentinel-2 narrow near-infrared band 8A (N-NIR), shortwave infrared bands 11 and 12 (SWIR-11 or SWIR-12) images for the Assiniboine River Basin (ARB). The highest precision (0.87) of surface water detection was obtained in autumn using Minimum Entropy (N-NIR & SWIR-11) threshold values for the ARB, however, the Minimum Entropy also had poor detection of surface water (low recall). Use of the naïve Bayesian classifier did not improve detection of surface water compared to using threshold values alone, and the use of threshold values generated for one basin was not transferable to another basin. We concluded that threshold methods should be assessed seasonally, at a per-tile scale before combining surface water products at a watershed scale, and products should be evaluated using non-balanced statistical measures for watersheds with high land-to-water ratios. Accuracy assessment is hampered by validation data which represents static hydrologic conditions, not contemporary, seasonal conditions.

Highly accurate classifications of urban Land Use/Land Cover (LULC) are critical for many remote sensing applications in urban geography. A review of the literature shows that there is no one method that consistently achieves this. Therefore, the goal of this study is to investigate the ability of out-of-box, pixel-based, medium-resolution image fusion classifications to increase the thematic classification accuracy of urban LULC mapping. The objectives identified to achieve this goal were to (1) establish a baseline of nonfusion classification results using supervised and unsupervised algorithms and (2) test 2 fusion methods to explore whether they would improve classification results. The unsupervised optical classification and the image fusion using radar products (H/A/α) with optical and near-infrared (NIR) bands both achieved excellent overall accuracies (89.4% and 89.1%, respectively). Lower accuracies were obtained using a principal components analysis of both optical and radar data. Overall, optical-radar image fusion improved urban LULC classification. The choice of how LULC classes are aggregated and which algorithm is chosen were found to be critical in achieving adequate results for urban mapping. Classifications très précises de l’utilisation des terres urbaines de la couverture terrestre ⟨⟨Land Use/Land Cover⟩⟩ (LULC) sont essentiels pour de nombreuses applications de la télédétection dans la géographie urbaine. Une revue de la littérature montre qu’il n’y a pas une méthode qui permet d’obtenir de façon constante cette. Par conséquent, l’objectif de cette étude est de déterminer la capacité de résolution moyenne classifications de fusion d’images à base de pixels hors-boîte pour augmenter la précision de la classification thématique de la cartographie des LULC urbain. Les objectifs identifiés pour atteindre cet objectif sont (1) d’établir une base des résultats de la classification nonfusion en utilisant des algorithmes supervisés et nonsupervisés et (2) un essai sur deux méthodes de fusion d’explorer si elles améliorer les résultats de classification. La classification optique sans surveillance, et la fusion de l’image en utilisant des produits de radar (H/A/α) avec des bandes optiques et ⟨⟨near-infrared⟩⟩(NIR) fois obtenu d’excellents exactitude globale (89,4% et 89,1%). Précision plus faibles ont été obtenues en utilisant une analyse en composantes principales des deux données optiques et radar. Dans l’ensemble, optique-radar fusion d’image améliorée classification LULC urbain. Le choix de la façon dont les classes LULC sont agrégées et l’algorithme a été choisi a été jugée essentielle à la réalisation des résultats adéquats pour la cartographie urbaine.

The urban landscape is an interspersed mixing of residences, shops, theatres, parks, natural areas, and a multitude of other uses. From the early days of the

The world is continuing to urbanize at an alarming rate. The ability to detect, map and monitor change within the urban system is of critical importance as urban growth happens largely at the expense of agricultural loss. Mapping land use/land cover change within urban systems using fully polarimetric synthetic aperture radar has been limited because of the complexity of scattering mechanisms within the urban system. This research examines the use of 2-component polarimetric decompositions for the identification of the urban-rural boundary. Using multi-temporal, fully polarimetric RADARSAT2 fine beam SAR imagery the Neumann 2-Component decomposition was most effective of the four decompositions tested.