The Research Groups Geoinformatics and Social Geography are inviting:
GIB LECTURE SERIES – Summer term 2022
Geospatial Big Data and Societal Transformations
Modern data science and the large volume and diversity of data stimulate a huge number of novel and ethical research questions and greatly influence our life. But what opportunities and challenges are presented for us to understand societal and environmental phenomena and processes in space and time? Compared to non-spatiotemporal data analysis, spatiotemporal data analysis additionally takes into concern spatio-temporal dependency, spatiotemporal heterogeneity, and the modifiable areal unit problem. What is state-of-the-art and does modern data science bring an evolution in spatiotemporal data analysis and contribute to our understanding in societal and environmental geography? Machine learning methods seem to suit well for predicting and analyzing commonly complex and non-gaussian environmental challenges, but how is the application of machine learning methods affected by the characteristics of spatiotemporal phenomena. How do we integrate data from multiple sources and how do we efficiently represent (possibly abstract) spatiotemporal phenomena (e.g. migration) or objects (e.g. slum mapping) in a computational system?
The lecture series approaches these issues by bringing together speakers with a research background in geospatial data analysis and researchers who are looking into the societal questions of human-technology-interactions and the subsequent effects on how future societies may deal with the transformational issues. This bringing together of Geoinformatics with Social Geography makes room for critical questions such as: What are the societal consequences and implications of digitalization? How does big geospatial data influence – and how will it influence – our everyday lives?
The Lecture Series takes place every Tuesday in May 2022 from 4:15pm to 5:45pm at the University of Bayreuth (Hörsaal H33, Building INF/AI).
A livestream will be available! Please visit the website of the event for further information (see link at the end of the document).
Dr. Simon Schneider (Associate Professor – Department of Earth Sciences – Utrecht University)
Valid statistics with amounts in geographic information
The question under which circumstances some statistical operation is valid, i.e., rightfully applicable to geographic information for a given purpose, is not only crucial for understanding and correctly applying statistics to spatial data, it is also a major bottleneck in automating data science in Geography and the Geosciences. Conceptual models of geographic information are required for both handling geodata resources as well as geographic question answering (geoQA). In this talk, I argue that we need better models of quantities in spatial information. I present a novel theory of geographic amounts that can be used to classify data sources and to constrain possible data transformations to those valid for a certain purpose.
Dr. Xiang Ye (Postdoctoral Researcher – Research Institute for Smart Cities – Shenzhen University) (via ZOOM)
Linear regression, model specification errors, and the modifiable areal unit problem
Spatially aggregated data sets are very commonly adopted by geographers and experts from other nearby disciplines. This way of organizing the observations provides many amenities including facilitating statistical works, protecting privacy, and cooperating smoothly with the administration, but comes with a price: the modifiable areal unit problem (MAUP). The MAUP refers to the sensitivities or inconsistencies of analysis results brought by the different spatial configurations of the same study area. Many empirical studies have documented the extensive impacts brought by the MAUP; however, the mechanism it reshapes the analysis results is not as straightforward as one may anticipate. Sometimes, it does not cause the trouble directly, but leverage the aftermath of other imperfections. In this talk, I will share some findings of how the MAUP will affect the regression results through the channel of model specification errors, and the opportunities that how we can possibly benefit from this ‘problem’ in alternative scenarios.
Dr. Jiong Wang (Assistant Professor – Digital Society Institute – Faculty of Geo-Information Science & Earth Observation (ITC) – University of Twente)
Deriving information from space for urban environmental risk management
The ongoing practice of applying Earth Observation (EO) technology shows great potential of mapping every aspect of urbanization, hence understanding the interactions, resilience and risks at the human-environment interface. However, deriving information regarding the two basic components of risk assessment, exposure and vulnerability, is subject to uncertainty and limited scientific validity. In this presentation, I will show, first of all, how environmental variables are mapped in general and how information can be uncertain due to problems of sampling and scale of the target geographic patterns, which are not discussed sufficiently these days. I will then continue with presenting the evolution of techniques used to map the other component, vulnerability, by using EO based datasets. I will especially focus on the experience in mapping urban poverty to the Global South to showcase how the remote sensing community has been trying to derive knowledge about socioeconomic status, where I argue that the ongoing practice of mapping socioeconomic status may lead to problems of reduced interpretability and limited scientific validity. This argument is then followed by suggested alternatives to avoid these problems as well as inform practical actions for risk management.
Dr. Britta Ricker (Assistant Professor – Copernicus Institute of Sustainable Development Environmental Sciences – Utrecht University)
Cartography, GeoAI, and the United Nations Sustainable Development Goals
To ameliorate wicked problems facing our world, from poverty, hunger, biodiversity preservation, access to water, education, equal rights, and more, the United Nations has identified a set of seventeen global Sustainable Development Goals (SDG). To achieve each of these goals, a set of targets and over 200 indicators have been recognized and agreed upon among UN member states. Well-designed maps and graphics portraying SDG indicator data that inform decision-makers to decide where to place interventions to reach the targets. Design choices influence what will then be understood from the map and decisions that are subsequently made. Currently, indicator data are collected in an uneven heterogeneous manner, aggregated at an imprecise national level, and data are missing in some of the most impoverished parts of the world. This is where Geospatial Artificial Intelligence (GeoAI) could help. Through the use of unmanned aerial vehicles to collect remotely sensed imagery, combined with computer vision software systems to process it, data collection, and visualization strategies could be improved worldwide. In this talk, I will present specific considerations related to data sovereignty and GeoAI practices, cartography and the UN SDGs.
Dr. Fran Meissner (Assistant Professor – Critical Geodata Studies and Geodata Ethics – University of Twente)
Geo-data ethics beyond the data: towards sustainable geodata ecosystems
If some say that data is the new oil, then spatio-temporal data is the new palladium. It is extremely valuable and increasingly crucial for how our social life is structured and made possible. The analogy obviously does not work that well. While palladium is a rare metal – spatio-temporal data is produced and mined in abundance. For spatial analysts, including geographers, this new abundance of spatio-temporal data has led to a rush in finding different applications for this data. Those concerned with the ethics of new data sources and geodata technologies have mainly focused on how the data and its characteristics may harm individuals and groups. This talk will focus on the ethics of spatio-temporal data beyond the data. I will discuss how broader data and technology ethics concerns link in specific ways to the use of spatio-temporal data. With the talk, I make a case for a more proactive and continuous re-visibilising of geo-spatial data infrastructures – the infrastructures needed to make big-data geo-analysis possible. I contend that doing so is crucial if we want to work within ethically sustainable geodata ecosystems.
Further information on date and venue, see: https://www.geographie.uni-bayreuth.de/de/Veranstaltungen/GIB-Lecture-Series/index.html
Jun. Prof. Dr. Meng Lu
Chair of Spatial Big Data
Dr. Matthias Gebauer & Prof. Dr. Eberhard Rothfuß
Chair of Social Geography