GI-Forum Winter 2016/17

Date	Topic	Presenter	Affiliation
18.10.2016	Institute meeting
25.10.2016	Modelling and demonstrating the safety of medical devices and systems within their context	Michael Harrison	UCL/Newcastle University
The talk will first describe my recent work related to the demonstration of use related safety requirements for a range of medical devices (IV infusion pumps, a paediatric dialysis machine, a decision aid related to trauma induced coagulopathy). I will then speculate about use related safety requirements that might be appropriate when considering smart systems of such devices within operating rooms, hospital wards and waiting rooms. I will briefly describe models of systems based on notions from distributed cognition, models of location and the role of stochastic models to describe qualitative measures such as clinician workload and patient experience.
01.11.2016	No GI forum (All Hallows)
Not available yet, sorry. Please check back later.
08.11.2016	Where Citizen Science and Geographic Information Science meet?	Muki Hacklay	UCL
Citizen Science and Geographic Information Science emerged in the early 1990s, and are both evolving in their own paths, drawing on many knowledge areas, with many of them overlapping (for example, Human-Computer Interaction, spatial statistics). The seminar will provide an introduction to the range of activities that fall under citizen science and in particular, we look at volunteered geographic information (VGI) and Citizen Science, exploring the societal and technological trends that led to their (re)emergence in the past decade. With an understanding of the wider changes and the commonalities and differences of VGI and citizen science, we can explore what GIScience can do for Citizen Science, and what Citizen Science can do for GIScience.
15.11.2016	Challenges for national mapping agencies in smart cities and for future geospatial information	Jeremy Morley	Ordnance and Survey, UK
These are changing times in the geospatial world. The technology to create and manage map data at traditional scales is widely available. Meanwhile, technologies in physical infrastructure and the digital world are developing outside of the geospatial domain but which still present significant challenges. We are seeing the rise of a range of new techniques and technologies, from increased intelligence about the built environment from BIM through to new opportunities to sense and interact with that environment in real-time presented by the Internet of Things. Running around our roads in the next few years will be increasingly automated and eventually autonomous vehicles. And, in our backbone infrastructure, increasingly adaptable, smart grids are being attached to homes and businesses with smart metering, home generation and personal battery storage. All of these technologies are linked by several characteristics: location or geography as defining components of service operation; potential disruption in the sectors into which they are being introduced; the need for new supporting analytical methods; and relative independence in their development from traditional geographical information. They are individually challenging and more so in combination – how will autonomous vehicles interact with the digital built environment and IoT frameworks, e.g. to book and find the right spot in a parking garage? At OS our research & innovation is exploring these new opportunities for geo and the future of our national mapping, from autonomous vehicle projects such as Atlas, modelling geoinformation data & system requirements, to the Manchester IoT city demonstrator project, CityVerve. In the talk I will examine in more detail some of the challenges, and how the OS R&D programme aims to find new roles and capabilities for Britain’s national mapping agency.
22.11.2016	A plea for combining scenario projections and Pareto frontiers	Judith Verstegen	IfGI, WWU Münster
Environmental impact assessment is used for a priori evaluations of the consequences (positive and negative) of a plan, policy, project, or foreseen event in an environmental system. Broadly speaking, two approaches exist in such impact assessment studies. The first approach is the projection of a continuation of current trends in the system of interest, using a statistic or geosimulation model. Herein it is common to use scenarios representing different strategies, e.g. different policies, to compare their respective environmental impacts. The second is spatial optimization, in which an ‘optimal’ system is designed, given one or more objectives (often low costs and positive environmental impacts). For most optimization problems, there is an unlimited number of optimal solutions, together forming the Pareto frontier. Both scenario projection and spatial optimization have distinct advantages and disadvantages. In this presentation I aim to show how the approaches can complement each other. This is illustrated by an impact assessment of increased biofuel production in Goiás, Brazil, for 2030.
29.11.2016	Institute meeting
06.12.2016	Bridges between GIS and R	Roger Bivand	NHH Bergen, Norway
Because GIS can be used as databases, and their tools can be better suited to some analyses and operations, it may be sensible to use one in addition to data analysis software. There is an extra effort required when using linked software systems, because they may not integrate easily. Since R is open source, and R spatial packages use open source components, staying with open source GIS means that many of the underlying software components are shared. This certainly applies to R and GRASS, and through GRASS, also to R and QGIS --- QGIS is more file-based than GRASS, which has an underlying data storage specification. R as GIS is becoming more feasible as new contributed packages appear, and/or R bridges to GIS to permit GIS operations to be done in GIS but statistical operations in R. Mention will be made of bridges between R and GRASS GIS, and between R and ArcGIS (32-bit >= 10.3.1; 64-bit Pro) (ESRI now an R Consortium member); file formats and data structures are GIS-based in all cases. Here we'll be using GRASS in an example, repeated in R itself without using an external GIS. GIS interfaces can be as simple as just reading and writing files using loose coupling, once the file formats have been worked out. The GRASS 7 interface rgrass7 on CRAN is the current, stable interface. We'll use the John Snow Soho Cholera example to show how GIS operations can be scripted from R, and compate this with an R-only version.
13.12.2016	From location-based to connection-based visualization	Liqiu Meng	TU Munich
With its ubiquitous accessibility, the digital world is nurturing a steady growth of the mobile population on the one hand and fostering an upcoming ecosystem of self-regulating big data on the other hand. The everyday information overload, however, is increasingly depriving us of our attention and patience. Being exposed to the lasting perceptual stress, our eye-mind has become less sensitive to external stimuli and more prone to ignoring important but visually inconspicuous messages. This talk is dedicated to visually driven geoinformation services with the aim to match the graphic design with users’ limited perceptual capacity. Starting from location-based map services that try to answer questions such as “what” and “how much” is at “where” and “when”, the presenter will review the evolving role of cartography and reason the necessity of identifying the connections embedded in complex and dynamic geodata. With the aim to answer questions such as “what”, “how” and “why” is happening between the known locations, she appeals a connection-based view which is committed to representing events, processes, behaviors, correlations and causal relations. Some case studies on the acquisition and visual analysis of open source events will be demonstrated.
20.12.2016 27.12.2016 03.01.2017	No GI forum (Christmas)
10.01.2017	Cancelled.
In this presentation, I will discuss, based on a range of examples, how spatial data analysis depends on implicit assumptions about the meaning of data, and how these assumptions influence not only our choice of analysis methods, tools, and data sources, but also the usefulness of analysis results according to the questions we ask. Turning such implicit semantic structures into an explicit form is therefore a central task for eScience in general, and for spatial information science in particular. I propose to handle implicit semantic structures for spatial analysis based on ontologies, linked data and latent semantics. I will discuss the cases of choropleth maps and statistical summaries of social statistics in geoportals, the Modifiable Area Unit (MAUP) problem, and show how implicit semantic structures can be added to geoprocessing workflows in ordinary GIS. The goal is to enrich, modularize and share spatial analysis methods together with data on the Web.
17.01.2017	Model-based approaches to community ecology	Christoph Scherber	WWU Münster
To analyze patterns in community structure, ecologists often employ distance-based statistical approaches such as principal components analysis (PCA) or related methods (NMDS, DCA, PERMANOVA etc.). However, recent developments in machine learning and software development have allowed for so-called model based approaches. Essentially, these work without dimensional reduction of a dataset. That is, a whole matrix of response variables is entered into statistical models. The simplest example is a binomial generalized linear model, where the abundance of two species is treated as two response vectors. This situation can easily be extended to multinomial models, where a matrix of >2 species is used. In this talk, I will provide an overview of frequentist and Bayesian model-based approaches to analyze community data, with extensions on how to account for spatiotemporal non-independence, overdispersion and other aspects of real-world datasets. I give recommendations on software packages (with a focus on R) and useful functions, with their strengths and limitations. Overall, the talk will show how model-based approaches can help unravel the
24.01.2017	Analysis and Enrichment of Trajectory Data	Monika Sester	Hanover University
In the talk current work at the Insitute of Cartography and Geoinformatics at Leibniz Universität Hannover will be presented. There will be a focus on trajectory data, e.g. their analysis, interpretation, and enrichment with other data.
31.01.2017	Real-time Geo-information Fusion	Florian Hillen	IfGI, WWU Münster
In recent years the amount of sensors that provide geo-information experienced a major growth. The resulting flood of geo-information builds the basis for new, time-critical geo-applications that would have been inconceivable a decade ago. The real-time characteristics of geo-information, which are also getting more important for traditional sensors (e.g. remote sensors), require new methodologies and scientific investigations regarding aggregation and analysis that can be summarised under the term geo-information fusion. In my talk I will introduce the basic idea of geo-information fusion and will present an agent-based modelling use case in the context of crowd monitoring. The scenario is designed to emphasise the benefits of fusing geo-information from different sources as well as to demonstrate the need for up-to-date information and real-time processing.
07.02.2017	Institute meeting