GI-Forum

The GI-Forum takes place every Tuesday from 12 to 2 pm during semester in room 242 at the Institute for Geoinformatics (Heisenbergstraße 2, 48151 Münster). The Forum is open to the public and all interested are welcome to attend.

These open sessions present an opportunity to discuss a variety topics across GI Science, generally in the form of presentations of around 45 minutes followed by 15-30 minutes for questions and general discussion, before closing with light refreshments.

Speakers are drawn from a range of academic disciplines, as well as from industry and policy areas, and present results of ongoing research and/or practical outcomes.

GI-Forum Winter 18/19

Date	Topic	Presenter	Affiliation
09.10.2018	Institute Meeting
no GI Forum will take place
16.10.2018	Working with Images in R	Jeroen Ooms	rOpenSci / UC Berkeley
Images in various forms are used for numerous applications across scientific disciplines. Whether you are observing through satellite or microscope, looking at MRI scans or petri dishes, trying to find patterns or abnormalities, the data is in the image. Unfortunately the tools for working with images are traditionally highly fragmented by field, and often narrow in scope. At rOpenSci we are working on a suite of general purpose packages based on powerful c/c++ libraries. These provide an extensible and interoperable foundation for working with images in R, which can be used to implement domain specific-methods. This talk gives a taste of things we can currently do with images in R, and highlights some of the ongoing developments and challenges.
23.10.2018	Use of Unmanned Aircraft Systems for high-resolution remote sensing in landscape ecology applications	Jan Lehmann	Institute for Landscape Ecology
Unmanned Aerial Systems (UAS) presents itself as one of the most recent remote sensing technologies in high-resolution spatial and temporal data acquisition. This talk will address the question of how well, this remote sensing technology can be successfully integrated into landscape ecology, in regards to both basic and applied research. Furthermore, perspectives and frontiers of UAS-based remote sensing within landscape ecology are also brought into light. The proficiencies of low-cost camera sensors for vegetation mapping, monitoring and management are evaluated in particular. Newer, more refined image analysis strategies are therefore presented and appraised in regards to accuracy and ecological reliability. Consequently, different case studies take precedence, each serving as models of innovative remote sensing applications highlighting significant topics in current landscape ecological research.
30.10.2018	An Immersive Science of Place - Paradigm Shifts for Geosciences?	Alexander Klippel	Pennsylvania State University
Sorry, no abstract available yet.
06.11.2018	New insights from monitoring deformations of the earth from space	Ramon Hanssen	TU Delft
Using satellite radar interferometry, we can now observe the dynamics of the earth surface with unprecedented resolution and precision. However, it is a genuine big data problem, where many different physical phenomena are juxtaposed, and the main challenge is to reliably disentangle the different components. In this presentation, we will review the basics of the techniques as well as the new insights we derived from the plethora of data.
13.11.2018	Imputing missing values in satellite data: From parametric to non-parametric approaches	Reinhard Furrer	University of Zurich
Sorry, no abstract available yet.
20.11.2018	Geomarketing: providing location insights to enhance economic efficiency	Tim Appelhans	GFK
Sorry, no abstract available yet.
27.11.2018	Institute meeting
no GI Forum will take place
11.12.2018	Mapping inequality in access to resources in R.	Andy Nelson	University of Twente
Sorry, no abstract available yet.
18.12.2018	From daily to decadal movement of snow and ice seen from space	Bas Altena	University of Oslo
Sorry, no abstract available yet.
08.01.2019	Internationally scalable and reproducible methods and software for active travel planning	Robin Lovelace	University of Leeds
Sorry, no abstract available yet.
15.01.2019	Does Index Insurance Help Households Recover from Disaster? Evidence from IBLI Mongolia	Kati Krähnert	DIW
Sorry, no abstract available yet.
22.01.2019	Improving the usability of 3D city models	Hugo Ledoux	TU Delft
Sorry, no abstract available yet.
29.01.2019	Institute Meeting
no GI Forum will take place

GI-Forum Summer 18

Date	Topic	Presenter	Affiliation
10.04.2018	Institute Meeting
Sorry, no abstract available yet.
17.04.2018	Visual Tracking of Small Animals in Cluttered Natural Environments Using a Freely Moving Camera	Benjamin Risse	WWU Münster
Image-based tracking of animals in their natural habi-tats can provide rich behavioural data, but is very challeng-ing due to complex and dynamic background and target ap-pearances. We present an effective method to recover the positions of terrestrial animals in cluttered environments from video sequences filmed using a freely moving monocu-lar camera. The method uses residual motion cues to detect the targets and is thus robust to different lighting conditions and requires no a-priori appearance model of the animal or environment. The detection is globally optimised based on an inference problem formulation using factor graphs. This handles ambiguities such as occlusions and intersec-tions and provides automatic initialisation. Furthermore, this formulation allows a seamless integration of occasional user input for the most difficult situations, so that the effect of a few manual position estimates are smoothly distributed over long sequences. Testing our system against a bench-mark dataset featuring small targets in natural scenes, we obtain 96% accuracy for fully automated tracking. We also demonstrate reliable tracking in a new data set that includes different targets (insects, vertebrates or artificial objects) in a variety of environments (desert, jungle, meadows, urban) using different imaging devices (day / night vision cameras, smart phones) and modalities (stationary, hand-held, drone operated).
24.04.2018	GIScience for Dynamic Transportation Systems	Anita Graser	Austrian Institute of Technology
Transport research is an interdisciplinary field with a multitude of research topics that overlap with GIScience since movement, by its very nature, happens in space and time across multiple scales and within different context. Applications range from monitoring, understanding, and guiding local movement, to city-wide and global movement analysis and optimization. GIScience provides a strong foundation for understanding spatial aspects as well as the context of movement but the dynamic nature of movement data is a challenge for many established GI approaches. This talk highlights recent GIScience topics, open research questions and challenges relating to pedestrian navigation, as well as analysis of massive movement data at the AIT Dynamic Transportation Systems group. Paper: Graser, A. (2017). Towards landmark-based instructions for pedestrian navigation systems using OpenStreetMap, AGILE2017, Wageningen, Netherlands. https://agile-online.org/images/conference_2017/Proceedings2017/shortpapers/46_ShortPaper_in_PDF.pdf
01.05.2018	May 1
Sorry, no abstract available yet.
08.05.2018	Acquisition and consolidation of hierarchical representations of space	Wiebke Schick	WWU Münster
Navigation – the ability to reach targets which are no visible from the current position depends on the correct recall of the desired target and the environment between one's current position and this target. The content of these representations are subject to influences from different modalities, e.g. vision, and language. A place can be recognized through different cues, e.g. due to a salient object, but also because of the angles of the routes at an intersection, or a name. The location of places as well as the routes connecting them can be integrated and memorized in an allocentric, survey-like representation. Depending on the amount of detail, the granularity level of a representation can be coarser or finer; the different levels are organized hierarchically. Characteristics of a superordinate category, like a region, can affect the perception of its constituting elements, the places; an inheritance of qualities from region to place levels is possible. The formation of superordinate categories depends both on environmental factors as well as individual ones: what is recognized, what is remembered, and which predictions are drawn from this representation? I examine the acquisition of representations of space, in order to identify features that are well suited for being remembered and auxiliary for navigation. I have two research foci: First, I examine the impact of language by using different hierarchically structured naming schemes as place names. Wiener & Mallot (2003) found that characterizing places only with landmarks belonging to different semantic categories influenced route choice as well as the representations of space. I compare these findings to the impact of different naming schemes. I show that there are naming schemes that may influence behavior in a similar way as a landmark does, but that seeing something and reading its name is by far the same thing. The second part focuses on the content of representations established during navigation. With three different navigation experiments, I examine the content of the concepts of space that are acquired during navigation. What is remembered - the location of places, the routes, or the hierarchical structure of the experimental environment? Are there features that are more likely to be consolidated during sleep, e.g., the transfer of concrete knowledge about places and routes into an abstract, survey-like representation? I show that there are improvements in one wayfinding task correlated to sleep. In the other experiments, learning effects were found for both groups. I also address the question of suitable parameters for measuring survey knowledge.
15.05.2018	Visualization of Uncertainty: Empirical Studies	Sarah Fabrikant	University of Zurich
Sorry, no abstract available yet.
22.05.2018	Pentecost Holidays
Sorry, no abstract available yet.
29.05.2018	Institute Meeting
No GI Forum will take place
05.06.2018	Assessing students' individual progress: Concept, trajectories, and effects	Elmar Souvignier	Institut für Psychologie in Bildung und Erziehung, WWU
Assessing students’ individual progress: Concept, trajectories, and effects Elmar Souvignier, University of Münster Does every single student take benefit from my instruction? In contrast to (typical) diagnostic approaches that focus on students’ level of achievement, the idea behind learning progress assessment is to measure change. The concept of ‘progress monitoring’ was designed with the aim to support teachers’ instructional decision making (data-based decision making). In other words: Measuring progress by use of multiple points of measurement (e.g. weekly assessments) should help teachers to differentiate instruction based on objective diagnostic information. This, as a consequence, should lead to positive learning effects. In our research program we designed a web-based system for learning progress assessment with parallel tests for reading and mathematics from grades 1 to 6. First, we analyzed the technical adequacy of these tests (reliability, validity, sensitivity to change, parallelism). As a second research question we were curious about childrens’ learning trajectories over the period of one school year. Analyzing latent classes of learning progress we found that children not only differ with respect to their level of achievement but also according to individual progress of learning. The third focus of our research program was to evaluate effects of providing teachers with diagnostic information about their students’ progress. Throughout three replications we found that learning progress assessment leads to small but reliable positive effects on reading achievement. Future studies deal with the question of optimizing the concept of learning progress assessment by adding feedback-routines and providing teachers with material for differentiated instruction.
12.06.2018	The Earth System Lab: A light-weight data cube approach to study land-atmosphere interactions at the global scale	Miguel Mahecha	Max Planck Institute for Biogeochemistry, Jena
Scientists today are confronted with a plethora of Earth observations (EO) and derived downstream data products. These data streams are extremely valuable for studying processes at the global scale, for instance on how climate extremes influence primary production among others. To jointly exploit these datasets, scientists have to deal with a multitude of data formats, data usage policies, inconsistent resolutions and formatting, as well as getting access to suitable computation and data storage facilities. This time-consuming work slows down scientific progress. The “Earth System Data Lab” is a novel infrastructure for scientists to overcome limitations of this kind. It not only provides a set of harmonized data cubes, but also developed a strong data analytics toolkit in tandem with access to computational facilities to jointly analyze the datasets. The data cube consists on a selection of diverse Earth observation datasets including multiple atmospheric variables, Land-Atmosphere fluxes, biophysical parameters and is currently extended to ocean variables. These datasets are analysis-ready, harmonized on different temporal and spatial resolutions to allow rapid exploitation. The core data analytic toolkit allows users, if proficient in a language of scientific computing like Python or Julia, to apply their own analysis methods on the multivariate data cube. Examples implemented include nonlinear time series analytic tools, dimensionality reduction approaches, or anomaly detectors, among others. To showcase the potential of the approach we will present some scientific case studies demonstrating the full potential of the infrastructure and seek for participants in our early-adopter user engagement call. Our latest paper on using the cube is this one - still in discussion: https://www.biogeosciences-discuss.net/bg-2018-130/bg-2018-130.pdf
19.06.2018	Gaze behavior in natural environments – Why some objects and areas may be more salient than others	Rul von Stülpnagel	Uni Freiburg
When navigating natural environments, some of the elements surrounding us or even entire areas appear more salient than others. They may thus be used, for example, as landmarks that enable us to retrace a route or provide a route description for others. Based on the differentiation of semantic, visual, and structural salience of landmarks, we investigated the gaze behavior of participants during incidental and intentional route learning. We found that visually salient landmarks attracted more gazes in general, independent of learning intention. However, structurally salient landmarks (e.g., landmarks into the turning direction) were particularly relevant during intentional learning. Further evidence for the benefits of structurally salient landmarks on memory for turning directions comes from a controlled lab-study. In a related research vein, we aimed at linking the perception of subjective risk during urban cycling to the gaze distribution in the local environment. Less surprising, people tend to focus on those areas they assume to be dangerous. More interestingly, preliminary findings imply that although cyclists identify most potentially dangerous areas correctly, some spots are literally overlooked, thus contributing to the level of objective risk for an accident.
25.06.2018	Schematic Mapping from an Algorithmic Perspective	Prof. Martin Nöllenburg	TU Wien
Schematic maps deliberately distort topographic reality and use a simplified and rather abstract depiction of spatial objects and relations in order to facilitate selected types of map use. For example, in a schematic public transport map, the most important information is the connectivity of stations and the course of each transit line so as to allow for fast visual route planning in the transport system. In a schematic road map, the important information for a driver who wants to get from A to B is, how to turn at road junctions to get to the destination. Creating schematic network maps automatically is a challenging problem that has also been addressed from a formal, mathematical perspective in the fields of graph drawing and computational geometry. Combining rigorous mathematical modeling and methods from algorithm engineering, the aim is to design and implement efficient algorithms with proven quality guarantees and evaluate them experimentally. This talk gives an overview of different algorithmic approaches for network schematization and their respective properties.
03.07.2018	Interacting with Space and Spatial Data	Ioannis Giannopoulos	TU Wien
Understanding how humans interact with their surroundings during spatial decision-making is crucial for the understanding of several processes, such as navigation. Furthermore, during spatial decision-making humans also with spatial data often presented to them through a display device. Through eye tracking and machine learning we are able to come closer to an understanding, optimize the relevant interaction dialogues, classify relevant interaction spaces and assist humans during the process of spatial decision-making.
10.07.2018	How to test research hypotheses with geographic prisms	Werner Kuhn	University of California, Santa Barbara
The so-called "spatial turn" in the humanities and social sciences can largely be attributed to a fundamental observation that Barney Warf and Santa Arias stated in their 2008 book: “Geography matters, not for the simplistic and overly-used reason that everything happens in space, but because where things happen is critical to knowing how and why they happen.” In my presentation, I ask whether and how this observation can be exploited for scientific hypotheses testing. An analysis of case studies where some form of spatial analysis was beneficially used in hypothesis testing, leads me to conclude that a surprisingly simple recipe can indeed be stated to test hypotheses spatially. The preliminary results can be seen as an operationalization of the "geographic prism” metaphor that has been used informally in the literature.
17.07.2018	Institute Meeting
No GI Forum will take place.

GI-Forum Winter 17/18

Date	Topic	Presenter	Affiliation
10.10.2017	Institute Meeting
Sorry, no abstract available yet.
17.10.2017	Remote Sensing in Conflict Analysis and Human Rights Fact-Finding	Christian Knoth	Institute for Geoinformatics, WWU Münster
Timely and detailed information on violent conflicts is essential to assess their impact on civilians and to document human rights issues, such as large-scale destruction and displacement. Remote sensing provides unique capabilities to collect such information and to overcome limitations of ground-based conflict documentation, especially in remote and insecure areas. This presentation will briefly introduce different application fields of remote sensing in a human rights context. I will then give an example by describing my recent work on the use of Geographic Object-Based Image Analysis (GEOBIA) for the detection of conflict damage in bi-temporal remote sensing images of Darfur, Sudan. I will also provide an outlook on how human rights fact-finding based on remote sensing may benefit from an increased availability of open or low-cost data as well as Free and Open-Source Software (FOSS). In addition, some technical and non-technical issues regarding the use of remote sensing and geospatial technologies in this politically precarious field will be discussed.
24.10.2017	Modelling the world with PCRaster	Oliver Schmitz	Utrecht University
Numerical simulation models are essential tools in understanding, analysing and predicting change of environmental systems. Modelling languages aid environmental scientists in the construction of these simulation models. The PCRaster team at Utrecht University (http://pcraster.geo.uu.nl/) develops concepts for spatio-temporal model construction and implements these in their environmental modelling language. The presentation will give a brief introduction to the PCRaster modelling platform and its Python framework work for model construction and execution. We show results from modelling studies, amongst others coming from global hydrology or geo-health. Furthermore, two of our current research themes will be illustrated: First, a better model representation of human-natural systems requires the integration of the field-based and agent-based modelling paradigms. We currently develop and apply a unified data model capable of storing field and agent data, the data model will be basis of a new environmental modelling language. Second, approaches such as hyper-resolution modelling require a better exploitation of available hardware resources. We discuss challenges of a generic modelling platform providing parallel algorithms suitable for execution on shared-memory and distributed-memory systems.
31.10.2017	Reformation day (no GI Forum)
Sorry, no abstract available yet.
07.11.2017	Measuring and Modelling Spatial Cognition in the City	Ed Manley	CASA, University College London
Despite a wealth of advances in behavioural psychology, geography, and neuroscience, conventional measures and models of urban transportation typically ignore the interaction between spatial cognition and travel decision-making. This talk will cover research aiming to bridge this gap, covering the application of new forms of mobility data, novel models of spatial decision-making, and agent-based simulation. Specifically, the talk will describe the analyses of route choices made during over 700,000 vehicular journeys in London, highlighting how travellers deviate from optimal paths between origin and destination towards particular features on the road network. This research supports a hypothesis that route choice decision-making takes a heuristic rather than optimising form within this context, and computational models are subsequently introduced to reflect these behaviours. These models of decision-making are integrated within a large-scale agent-based simulation of traffic in London, United Kingdom, demonstrating an improvement on existing methods.
14.11.2017	Geo-Information for analysing and governing cities	Karin Pfeffer	ITC Twente
Geo-Information, derived from a variety of data sources and by means of different methods and tools, is increasingly used in urban analyses and becoming embedded in urban governance processes. Examples are the spatial mapping and analysis of geocoded register or census data to investigate spatial patterns of socio-economic characteristics, processing remote sensing imagery to map urban dynamics, or GIS-based monitoring applications to follow key indicators such as air pollution levels or housing price developments. Drawing on examples from past research in different geographic contexts, this talk will present: 1) an overview of urban deprivation mapping in Indian cities (Baud et al. 2010; Kuffer et al. 2016), 2) the use of different information sources to produce a typology of transit oriented development applied to Beijing, China (Lyu et al. 2016), 3) a GIS-based monitoring system to monitor spatial concentrations of urban phenomena in the city of Amsterdam (Veldhuizen & Pfeffer, 2016), and 4) GIS practices in local governments in selected cities in the Global South (Baud et al. 2014). The talk concludes with discussing the potential and challenges of using Geo-Information in urban analysis and governance in different geographic contexts (Pfeffer & Verrest, 2016).
21.11.2017	Sensitivity analysis and verification of Agent Based Models	Arend Ligtenberg	Wageningen University
Agent Based Models (ABM) are increasingly used as simulation tool to model social-ecological system (SES) such as cities or agricultural systems. SES are often regarded as complex adaptive systems that show non-linear behaviour and lack clearly identifiable causation due to large numbers of entities in the system (humans, animals). Despite the promising features of ABM to represent such systems there are a number of major issues. First, collecting data to identify behaviour of the represented system is cumbersome. Second it is difficult to assess how an ABM is influenced by its parameters and input variables. Common used sensitivity methods are not directly applicable to ABM. In this talk I will present 1) a project in the Mekong Delta of Vietnam where we combined ABM with Role Playing Games (RPG) to define and verify behaviour rules of small holder shrimp farmers and 2) a research project were sensitivity methods have been developed for ABM that help to identify tipping points and resilience in SES simulations.
28.11.2017	Institute Meeting
Sorry, no abstract available yet.
05.12.2017	A Primer on Structure and Laws in GIScience	Franz-Benjamin Mocnik	Heidelberg University
Structures have turned out to often be invariant under paradigm shifts and are thus an integral part of many theories. However, in GIScience most research focusses on technological and ontological foundations, while laws and structurally stable results are still rare. Tobler's first law of geography is one of the widely known exceptions, though even that law appears to be much less rigorous than comparable counterparts from the natural sciences. In my talk, I present three different examples from the field of GIScience that would benefit from a more structural point of view. First, I discuss how maps are structurally different from texts, which is why they can often not serve for the purpose of telling stories in the same way as text can render complex narratives. Secondly, a structural understanding of the concepts of data quality and fitness for purpose is imparted. Both concepts have been explored for decades, but theoretical foundations of how different quality indicators relate to each other are yet to be discovered. Finally, the example of geographical data is discussed in a more general approach. The guiding question thereby is: how are geographical data shaped by geographic space, and how can the characteristics of geographical data be traced structurally?
12.12.2017	A model-based assessment of the environmental impact of land-use change in Southern Amazonia	Rüdiger Schaldach	Kassel University
During the past decades the expansion of agricultural land has led to a massive conversion of rainforest and Cerrado ecosystems in Southern Amazonia (Brazil), thus degrading biodiversity and altering environmental processes. Although, in the last couple of years the rate of land conversion could be significantly reduced, e.g. by agricultural intensification and environmental protection policies, deforestation rates are on the rise again due to increasing global demands for food, bioenergy and biomaterials, going hand in hand with only weak law enforcement. This presentation introduces a model framework that has been developed with the objective to improve the information basis available to scientists and policy makers, regarding the effects of global change and national environmental policies on future agricultural development and the resulting land-use change. Results from a scenario exercise portray different plausible trajectories of land-use change until 2030 and their impact on climate, water, soil and natural habitats.
19.12.2017	How Interactive Maps Work	Robert Roth	University of Wisonsin - Madison / University of Twente
Advances in personal computing and information technologies have fundamentally transformed how maps are produced and consumed, as many maps now are highly interactive and delivered online or through mobile applications. Increasingly, students need to learn how to encode geographic information based on sound cartographic principles as well as how to code a useful and usable interface for exploring the resulting maps. While much of the empirical research in cartography over the past half century has addressed representation design, evaluating the graphic symbols employed to communicate meaning in geographic information, relatively few empirical studies in cartography approach interaction design, research that is needed to understand how to successfully create digital mapping interfaces that meet user needs and promote geographic understanding. My research grapples with the central question “How do interactive maps work?” In this presentation, I first make the case for cartography in the ever-widening discipline of GIScience by highlighting a series of recent developments in the United States. I then provide an overview of my recent contributions to interactive map design, introducing a composite taxonomy of interaction primitives—the basic building blocks of interaction design that mirror the visual variables in representation design—and walking through several case studies that demonstrate how the interaction primitives can be used to advance both cartographic science and practice. I conclude by discussing how I have refreshed the cartography curriculum at UW-Madison to meet the demand for a new generation of interactive cartographers.
09.01.2018	Open Source business in practice: Earth observation and more	Markus Neteler	Mundialis.de
In my talk I'll present some of our current projects at mundialis, a "free data with free software" startup business. mundialis (https://www.mundialis.de) specializes in the evaluation of remote sensing data and the processing of massive geodata for different purposes. We rely on Free and Open Source GIS in order to create tailor-made solutions for our customers. One of our main focuses lies on the Sentinel satellite data from the European Copernicus program. Time series of these satellites are processed on demand and offered as OGC web services (with added value, i.e., atmospheric correction, estimation of indices, etc.). Satellite images time series provide information on changes in land use, soil moisture, surface temperature or vegetation phenology. In my presentation I will showcase several ongoing and completed projects for WHO, ESA, industry and research fields with focus on geospatial and EO data analysis and high performance computing platforms. The software stack of mundialis includes GRASS GIS, GDAL, QGIS, GeoServer, and SHOGun which enable us to realize applications like "maps mundialis" (Sentinel-2 frontend, https://maps.mundialis.de), "EO-me" (Earth Observation Metadata Enhancer, https://eome.mundialis.de), and a long-term landscape change analysis based on Landsat and openNRW data. The growth of mundialis over the past two years highlights the fact that enterprises can be based on and contribute to Free and Open Source software by applying innovative business models.
16.01.2018	Agent-based modelling of individuals’ activities in space and time: approaches for large-scale simulation	Theo Arentze	Eindhoven University of Technology
Activity-based models constitute the core of large-scale micro-simulation systems that hold many promises for policy evaluation in urban and transportation planning. Using micro-simulation, the new models offer a wealth of information for policy making. However, existing models predict behavior cross-sectionally for a random day of the life of simulated individuals. Although this approach is convenient, it is too restrictive for taking dynamic behavior and bounded-rationality into account. Individuals are reluctant to change existing habits and respond emotionally to losses and gains compared to a reference situation. The field has recognized the need for extending the time-frame of activity-based models and make the transition from static to dynamic modeling of behavior. First steps in this direction have recently been made and are supported by new GPS-based survey technologies which allow collection of activity-travel behavior data for longer periods of time. Although dynamic models offer appealing new perspectives, they also raise new issues when it comes to implementation in large-scale simulation systems and application in practice. In this presentation, I review approaches in the area of dynamic activity-based models for large-scale micro-simulation and discuss theoretical and computational issues as well as first experiences.
23.01.2018	Modelling and Mapping Earth Surface Processes using Statistical and Machine Learning Techniques	Alexander Brenning	Friedrich Schiller University Jena
Novel analysis and prediction methods developed at the intersection between computer science and statistics have been gaining popularity in recent years in Earth surface process modelling and mapping using, for instance, geomorphometric and remotely-sensed predictor variables. These techniques promise to improve predictive performances by overcoming the limitations of statistical models, making better use of highly correlated and interacting predictors such as terrain attributes calculated at different resolutions, or image filter banks. However, these possible improvements come at the cost of potential overfitting and limited interpretability of models. This talk discusses the potential and pitfalls of novel and 'traditional' methods in case studies such as landslide susceptibility modelling and environmental remote sensing.
30.01.2018	Institute Meeting
Sorry, no abstract available yet.

GI-Forum Archive (SS 12 - SS 17)