GEOG 506 Presentations

Here are the eleven 2017 GEOG 506, Advanced GIScience, presentations Dec 7th and 11th.

Hannah Ker, Applications of Temporal (Event) GIS

This project is an investigation into the spatiotemporal characteristics of a dataset of geotagged tweets from the City of London, England during a week in April 2016. I am specifically looking at coincident clustering in space and time through the use of space-time scan statistics. The resulting spatiotemporal clusters are examined to determine if they correspond to events taking place at specific locations in spacetime. This project is one of many that utilize Twitter data to detect event occurrences. Much existing research in this field, however, exclusively examines the semantic content of tweets and ignores their spatiotemporal characteristics. This project also addresses practical challenges including handling large volumes of data, rationalizing differing spatial and temporal granularities, and implementing spatiotemporal data visualization techniques. Fundamentally, this project attempts to move beyond the “snapshot” data model that is commonly used in temporal GIS and instead focus on ways to simultaneously consider both time and space.

Fanny Amyot, UAVs Surveying Permafrost

Unmanned aerial vehicle (UAV) photogrammetry is increasingly being used in physical geography because it provides a highly accurate, relatively affordable and unintrusive alternative to traditional survey methods such as Total Stations or dGPS. UAV photogrammetry involves using the computer vision technique of Structure from Motion (SfM) coupled with Multi-View Stereopsis (MVS) to create 3D point clouds and digital surface models (DSM). Though UAV photogrammetry is increasingly popular, there remains some serious shortcomings, including the high potential for error accumulation due to the specialized GIScience knowledge required to carry out the process, the black box software, and the lack of a comprehensive best-practice framework. In order to develop a best practice framework, 593 photos collected at the Eureka Station retrogressive thaw slump on Ellesmere Island, Nunavut, during the summer of 2016 were corrected and processed in Pix4D, a commercially available software. A framework developed based on this analysis provides insight into issues of coordinate systems, scale and error propagation, in addition to shedding light on the algorithms and processes within the black box software.

Caroline Thompson, Critical GIS

In September 2017, the City of Boston announced the launch of the Economic Mobility Lab, a research office that will identify factors of intergenerational poverty and make recommendations for mitigation strategies. This presentation introduces the stated intentions of the Lab via interviews with involved City officials. Through these interviews, it is made clear that there should be a reinvigoration of critical GIS research, focusing on the changing nature of governments and data collection, in addition to the use of qualitative and quantitative data in research. This underscores the understanding that data is not neutral and, in the same vein, increasingly tied to spatial information.

Deboleena Mazumdar, Scale and Crime

The use of GIS by public entities has created big datasets which contain valuable information about citizen demographics, resource management, pollution, socioeconomic variables, and public safety. “Open data” refers to information collected by a given group or entity which anyone has the right to view and manipulate. Urban big data sets allow for an unparalleled opportunity to study human behaviour in a real world environment. One of these applications is to the phenomena of urban crime behaviour. A prominent view of the phenomenon of urban crime seeks to distinguish crime at three scales; place-based, neighbourhood-based, and regional. The scale at which crime is measured has a profound influence on subsequent statistical analyses and the visualization of crime incidents in maps. Ultimately, these visualizations inform the viewers perception of the relative safety of a particular place. The insights that we gain from big datasets of crime are largely dependent on the scale at which the data is analysed. In this paper, I will use the python pandas library and QGIS to gauge how the scale at which crime is measured in New York City influences the visual representation of crime incidents and trends on a map. The data will be sourced from an open dataset of all incidents of crimes reported to the NYPD in 2014. In doing so, I will address discuss the process of acquiring data from a municipal government database, challenges in dealing with large datasets, and how choosing to examine the phenomena of crime at different scales influences its representation in a traditional mapping format.

Sam Lumley, Geovisualization of Projection

Web maps are a popular choice for spatial data visualisation and scientific communication, providing a platform for interactive data navigation and information sharing. In web mapping applications, the Web-Mercator projection has become the de facto standard, despite long having been deemed as inappropriate for scientific data representation for its considerable area distortion of high latitude regions. The presented research draws from cognitive geovisualisation and experimental psychology literature to assess the implications of these distortions for data interpretation. Specifically, we look to whether people identify distortions, and if they do, whether they are able to account for them in subsequent data interpretation. One hundred and twenty participants were recruited via Amazon’s online recruitment platform Mechanical Turk to complete an online survey assessing the influence of map projections on data interpretation. We employed a between-subjects experimental design with two conditions corresponding to the maps viewed by participants–one Lambert cylindrical equal-area quasi control group and one Web-Mercator treatment group. Participants were asked to make estimations of the area of the earth’s surface covered by five coloured regions represented on a global map. On average, participants failed to discount for the projections and interpreted the two maps differently as a result. These results were further corroborated by participants’ responses to questions about map preference and suitability. The findings provide an empirical basis for claims about the distortion effects of the Mercator projection, implicating its appropriateness for the display of global scientific data in web maps. They also contribute towards aspirations held by the GIScience community for further empirical geovisualisation research assessing heterogeneous, non-expert groups.

Cameron Power, Spatial Data Uncertainty in Linking Points to Lines in Hydrology

To enable global or regional hydrological assessment, point data such as monitoring stations, dams, or sources of effluents must first be referenced to an existing river network. However, locational uncertainty associated with rivers and point data often complicates the georeferencing process. Whereas conventional methods may rely entirely on the locational uncertainty of the point data, recent techniques partition uncertainty in the georeferencing process between a point's location and hydrological attributes. In either case, parameterization of the hydrological georeferencing algorithm is informed by the user's assessment of uncertainty associated with the point data or river network. The objectives of the project are twofold: to propose a framework for reasonable parameterization of the more recent hydrological georeferencing algorithm, and to evaluate how uncertainty in the output might be attributed to uncertainty in the algorithms inputs. The results of the project will provide reference to future users of the hydrological georeferencing algorithm to present robust discussion of uncertainty associated with its application.

Allen Zheng, Spatial Data Mining

This spatial data mining research focuses on periodic pattern mining (PPM) in resident trajectories. The datasets “GeoLife GPS Trajectories” were collected in Microsoft Research Asia by 182 users from 2007 to 2012. Since resident’s activities are limited by time and space, individuals are likely to have periodic behaviors. For the community, group behaviors also can be detected through adopting appropriate approaches, which can infer urban travel patterns and benefit for addressing urban sustainability challenges. In this project, I propose a straightforward but effective approach to investigate the periodic patterns of individuals. Since the data only include information about locations collected in constant time intervals, discovering periodic patterns from these data is typically called one dimensional PPM. Existing techniques for PPM include Apriori and Max-subpattern Hit Set, as well as the popular Fourier transform and autocorrelation methods. The automatic detection (i.e., Apriori and Max-subpattern Hit Set) might lead to redundant periods, and the user-specific detection (i.e., Fourier transform and autocorrelation) could miss periods. Besides, these methods simply working on points and ignore the interrelationships between them that is a significant issue in spatial context. Therefore, I suggest to regarding the “typically” one-dimensional data as location values change over events (i.e., different trajectories) and time. Base on the new perspective, I apply two density-based scanners to cluster trajectories and time slides (i.e., points in different trajectory at the same time) to detect the periodic patterns.

Noe Shultz, VGI and Completeness of OpenStreetMap

Volunteered Geographic Information (VGI), a term referring to user generated geographic content was first termed by Goodchild though is fundamentally a new medium for public participation which is a very old concept. Although beyond this old field, certain new aspects need to be looked at more closely as VGI veers towards citizen science, and begins to be treated as a credible source. Understanding how to assess quality, accuracy, completeness, and who is actually contributing are key to being able to use VGI to its fullest capability. OpenStreetMaps is currently the largest accessible database for VGI, and has already been reported to be “80% complete” for its street network (Barrington-Leigh, 2017). What constitutes complete however is not always clear, as many important features such as street name, speed limit, etc., are omitted from contributions leaving holes in the data, which could render it useless in many cases. Although the idea of a 1:1 scale / complete map brings Borges’ tattered map lying across the countryside to mind, I aim to assess OSM data completeness. With these concepts in mind, I aim to compare OpenStreetMap points of interest against an authoritative dataset for the city of Toronto, to compare completeness of the two datasets, and look for a correlation between OpenStreetMap completeness in points of interest, and household median income per census tract of the city.

Matt Poole, Locational Privacy and Intergenerational Perspectives

The goal of this paper is to consider discrepancies in location privacy concerns between different age groups in an attempt to better understand the use of geospatial technologies by different generations. By highlighting differences between digital natives and digital immigrants, preferences regarding privacy and use should become apparent. Research consisted of a survey of over 130 individuals to examine differences in opinions, concerns, and knowledge regarding location sharing, geosurveillance, and privacy. Among other variables, results show that the youngest demographic are only slightly less concerned than the oldest demographic about their privacy despite a significantly higher use of location based services/location sharing.

Nic Levy, Network Analysis for Accessibility

McGill's downtown campus is inhospitable to those with reduced mobility. Steep slopes and scant entrance ramps are particularly prohibitive to wheelchair access. A comprehensive survey of building entrances has been combined with a hi-res DEM to construct an enhanced network of McGill's roads and footpaths. Travel directions can now be queried for a specific set of mobility restrictions. The continuation of this project seeks to increase the ease of mobility for all students and visitors on McGill's campus.

Rudy Quinn, Analyzing the Taxonomy of Squirrel Movements in Laurier Park

The analysis of movement data is one of the most dynamic ways in which GIS can make sense of the world and biological processes. This project delves into the categorization and taxonomy of different types of squirrel movement patterns in Sir Wilfrid-Laurier Park in the Plateau Montreal. This project has attempted to categorize these movement patterns through the use of video recording and visual analytics. It attempts to use the realm of visual analytics to categorise these movements within a real-world setting. It takes inspiration from Nathan’s Movement Ecology paradigm (2008) as a basis for the movement’s behaviours, as well as Dodge et al’s taxonomy of movement patterns (2008). Through this, it was able to identify various kinds of movements which were then recorded on visual maps that were created through a combination of ArcMap data and pen-and-paper cartography.