Population living along the coast are at risk of losing land, households, and economic resources due to the hazards of coastal erosion. Scientific research has indicated that 70% of the planet's sandy coastal environment is being impacted by coastal erosion. Due to the different characteristics of the lands in the coastal zone versus other areas, it is important to understand how the hazard of shoreline erosion contributes to subsequent land use change in affected coastal regions. This study analyzes how the level of erosion, land loss, and developed land loss performs when added with the default land use change parameters such as existing developed land proximity, proximity to forested areas, population, transportation, etc. Sample points of 1020 from 10 years and 15 years of shoreline erosion data for the lower Meghna River estuary of Southeast Bangladesh have been obtained and from there, different erosion statistics have been developed. Developed land use data has been collected from ESA's World Settlement Footprint dataset and other datasets are also collected from secondary data sources. Logistic regression modeling shows that there are verified contributions of proximity to erosion and amount of land loss with the probability of developed land use conversion in the study area. Adding the variables of environmental hazards increases the prediction accuracy by 2-3% and overall, the models are at least 85% accurate. / Master of Science / Population living along the coast are at risk of losing land, households, and economic resources due to the hazards of coastal erosion. The coast of the Lower Meghna estuary in Bangladesh is a region experiencing chronic and severe shoreline erosion that causes the land to be lost to estuarine waters. This research quantifies the amount of land lost to erosion with a special focus on the amount of developed land that is lost. Developed land in this study is defined as a built-up area typically composed of buildings and roads. The research also evaluates the effects of lost land on the subsequent conversion of interior land from a non-developed to developed status. The main contribution is to quantitatively identify the association between the erosion-induced land loss to future land use conversion. Using statistical modeling and digital mapping methods, results show that loss of land is associated with the subsequent conversion of non-developed land to developed land use. In particular, conversion has a higher probability at sites that are located more distant from the eroding shoreline that also are proximal to shoreline sites with higher rates of erosion-induced land loss. These results are suggestive of a relocation process where previously lost developed land is reestablished at interior sites within five kilometers of the eroding shoreline.

The transition towards flexible work arrangements, notably work-from-home (WFH) practices, has prompted significant discourse on their potential to reshape urban landscapes. While existing urban growth models (UGM) offer insights into environmental and economic impacts, There is a need to study the urban phenomena from the bottom-up style, considering the essential influence of individuals' behavior and decision-making process at disaggregate and local levels (Brail, 2008, p. 89). Addressing this gap, this study aims to comprehensively understand how evolving work modalities influence the urban form and land use patterns by focusing on socioeconomic and environmental factors. This research employs an Agent-Based Model (ABM) and Artificial Neural Network (ANN), integrated with GIS technologies, to predict the future Land Use and Land Cover (LULC) changes within Roanoke, Virginia. The study uniquely explores the dynamic interplay between macro-level policies and micro-level individual behaviors—categorized by employment types, social activities, and residential choices—shedding light on their collective impact on urban morphology. Contrary to conventional expectations, findings reveal that the current low rate in WFH practices has not significantly redirected urban development trends towards sprawl but rather has emphasized urban densification, largely influenced by on-site work modalities. This observation is corroborated by WFH ratios not exceeding 10% in any analyzed census tract. Regarding model performance, the integration of micro-agents into the model substantially improved its accuracy from 86% to 89.78%, enabling a systematic analysis of residential preferences between WFH and on-site working (WrOS) agents. Furthermore, logistic regression analysis and decision score maps delineate the distinct spatial preferences of these agent groups, highlighting a pronounced suburban and rural preference among WFH agents, in contrast to the urban-centric inclination of WrOS agents. Utilizing ABM and ANN integrated with GIS technologies, this research advances the precision and complexity of urban growth predictions. The findings contribute valuable insights for urban planners and policymakers and underline the intricate relationships between work modalities and urban structure, challenging existing paradigms and setting a precedent for future urban planning methodologies. / Doctor of Philosophy / As more people start working from home, cities might change unexpectedly. This study in Roanoke, Virginia, explores how work-from-home (WFH) practices affect urban development. Traditional city growth models look at big-picture trends, but this study dives into the details of workers' individual behaviors and their residential choices. Using advanced computer models such as machine learning and geographic information systems (GIS), predictions are made on how different work arrangements influence where workers live and how cities expand. Surprisingly, fewer people work from home than expected. This hasn't caused cities to spread out more. Instead, Roanoke is expected to become denser in the next ten years because on-site workers tend to live in urban centers, while those who work from home prefer suburban and rural areas and, sometimes, urban. Different work arrangements lead to distinct residential preferences. By including the workers' individual behaviors in the models, the model's accuracy increased from 86% to 89.78%. Logistic regression analysis highlights the factors influencing land use changes, such as proximity to roads, slopes, home values, and wages. This research helps city planners and policymakers understand working arrangement trends and create better policies to manage urban development. It shows the complex relationship between work practices and city structures, providing valuable insights for future city planning.

Geospatial analysis involves application of statistical methods, algorithms and information retrieval techniques to geospatial data. It incorporates time into spatial databases and facilitates investigation of land cover (LC) dynamics through data, model, and analytics. LC dynamics induced by human and natural processes play a major role in global as well as regional scale patterns, which in turn influence weather and climate. Hence, understanding LC dynamics at the local / regional as well as at global levels is essential to evolve appropriate management strategies to mitigate the impacts of LC changes. This can be captured through the multi-resolution remote sensing (RS) data. However, with the advancements in sensor technologies, suitable algorithms and techniques are required for optimal integration of information from multi-resolution sensors which are cost effective while overcoming the possible data and methodological constraints. In this work, several per-pixel traditional and advanced classification techniques have been evaluated with the multi-resolution data along with the role of ancillary geographical data on the performance of classifiers. Techniques for linear and non-linear un-mixing, endmember variability and determination of spatial distribution of class components within a pixel have been applied and validated on multi-resolution data. Endmember estimation method is proposed and its performance is compared with manual, semi-automatic and fully automatic methods of endmember extraction. A novel technique - Hybrid Bayesian Classifier is developed for per pixel classification where the class prior probabilities are determined by un-mixing a low spatial-high spectral resolution multi-spectral data while posterior probabilities are determined from the training data obtained from ground, that are assigned to every pixel in a high spatial-low spectral resolution multi-spectral data in Bayesian classification. These techniques have been validated with multi-resolution data for various landscapes with varying altitudes. As a case study, spatial metrics and cellular automata based models applied for rapidly urbanising landscape with moderate altitude has been carried out.