Successful Dissertation Defense


Congratulations to Dr. Jin Xing, who successfully defended his dissertation.

Scale handling for land Use/cover change in an era of big data

Abstract Big data promises numerous benefits for Land Use/Cover Change (LUCC) research, in terms of increased volume, velocity, and variety of remotely sensed imagery datasets. However, it challenges traditional approaches to identifying LUCC. The increased volume and velocity of big data mean that existing data handling frameworks may not be able to effectively distribute spatial data and computation across a large number of computers. Previous LUCC workflows are not designed for big data and they cannot be easily deployed on big data computing tools such as cloud computing or the Hadoop framework. High levels of scale heterogeneity mean that images can cover different spatial and temporal granularities and extents. Theoretically, it becomes difficult to handle the data because these multiple and conflicting scales exist contemporaneously. Because we are working with big data, geographic entities may be recorded at different granularities and extents than should be detected as LUCC, but cannot be. Finally, no one has yet combined each of these distinct problems to fully examine all of the big data challenges facing LUCC. I present six advances to address each of the big data challenge in LUCC: (1) a theoretical concept called Scope, (2) a spatially sensitive decomposition/recomposition method, (3) a scale invariant change detection method, (4) a spatial-temporal model for LUCC big data, (5) a change boundary optimization algorithm, and (6) a LUCC-specific Geospatial CyberInfrastructure. In this manuscript, I first propose Scope as a concept to model spatial-temporal scales by explicitly merging granularity, extent, time, and property. Second, I develop a new decomposition/recomposition framework to manage data decomposition, distribution, and recombination in a distributed computing environment. Third, a scale invariant change detection method identifies LUCC by combining regional and point features from datasets at multiple spatial granularities and extents. Fourth, I theorize a spatial-temporal object model to improve the integration of space and time within LUCC research. The spatial-temporal object model and, the fifth advance, a change boundary optimization algorithm handle data noise and better organize the spatial-temporal object changes. Finally, a Geospatial CyberInfrastructure combines these separate approaches with cloud computing and distributed computing frameworks as a holistic approach for the big data challenge in LUCC research. These six advances are tested in a series of case studies using datasets collected from 2005-2012, at the Greater Montreal Area.

Biography: A Ph.D. candidate co-supervised by Prof. Renee Sieber and Prof. Margaret Kalacska in the Department of Geography at McGill University, Jin Xing earned his M.S. degree from the School of Computer Science at McGill University and an honors B.E. degree from Hebei University of Technology in Tianjin, China. As a computational geographer, Jin contributes unique research that combines spatial-temporal analysis, geo-computing, and machine learning algorithms to address geospatial big data challenges in Land Use/Cover Change (LUCC) study. His PhD research built a multi-scale big data analytical methodology for LUCC and applied it to an examination of urban changes in Montreal. This work has been recognized by several awards including Rathlyn GIS Award, has been published in the top GIScience journals including International Journal of GIScience, and has been presented at prominent GIScience/Remote Sensing conferences such as IEEE International Geoscience and Remote Sensing Symposium. His interest in applying GIScience to real-world problems also led Jin to enter the ESRI App Challenge 2015 (where his team won the second place) and the 2017 McGill Dobson Cup (where his team made it to the semi-finals). Jin continues to instruct McGill University students in GIScience and actively pursue new research on the application of big data analytics to social and physical urban change studies.