BGR Bundesanstalt für Geowissenschaften und Rohstoffe

The climate and environmental sciences have recently undergone a rapid transformation from a data-poor to a data-rich environment. In particular, massive amount of data about Earth and its environment is now continuously being generated by a large number of Earth observing satellites as well as physics-based earth system models running on large-scale computational platforms. These massive and information-rich datasets offer huge potential for understanding how the Earth's climate and ecosystem have been changing and how they are being impacted by humans actions. This talk will discuss various challenges involved in analyzing these massive data sets as well as opportunities they present for both advancing machine learning as well as the science of climate change in the context of monitoring the state of the tropical forests and surface water on a global scale.

Vipin Kumar is a Regents Professor at the University of Minnesota, where he holds the William Norris Endowed Chair in the Department of Computer Science and Engineering. Kumar's research spans data mining, high-performance computing, and their applications in Climate/Ecosystems and health care. He has authored over 300 research articles, and has coedited or coauthored 10 books including two text books ``Introduction to Parallel Computing'' and ``Introduction to Data Mining'', that are used world-wide and have been translated into many languages. Kumar's current major research focus is on bringing the power of big data and machine learning to understand the impact of human induced changes on the Earth and its environment.

Although the idea of AI has been around for 70+ years, it was just recently that the media, the public, companies, and policy makers began to take notice. This talk tries to answer why this is. We look at what has happened that caused recent breakthroughs in AI, what AI really is, and what it can accomplish.

Christian studied computer science and physics in Bielefeld where he also obtained a PhD in computer science; he then worked as a postdoctoral researcher in Toronto and was a senior researcher at Telekom laboratories before he was appointed as a professor in Bonn. His research addresses theory and practice of artificial intelligence and machine learning and he advises companies, institutions and policy makers on these topics.

Ed Parsons is the Geospatial Technologist of Google, with responsibility for evangelising Google’s mission to organise the world’s information using geography. In this role he maintains links with Governments, Universities, Research and Standards Organisations which are involved in the development of Geospatial Technology.
He is a member of the Board of Directors of the Open Geospatial Consortium and was co-chair of the W3C/OGC Spatial Data on the Web Working Group.
He is a Visiting Professor at University College London and has been an industry advisor to a number of international universities.

Earth observation has irreversibly arrived in the Big Data era, e.g. with ESA’s Sentinel satellites and with the blooming of NewSpace companies. This requires not only new technological approaches to manage and process large amounts of data, but also new analysis methods. Here, methods of data science and artificial intelligence (AI), such as machine learning, become indispensable.
In this Keynote, explorative signal processing and machine learning algorithms, such as compressive sensing and deep learning, will be shown to significantly improve information retrieval from remote sensing data, and consequently lead to breakthroughs in geoscientific and environmental research. In particular, by the fusion of petabytes of EO data from satellite to social media, fermented with tailored and sophisticated data science algorithms, it is now possible to tackle unprecedented, large-scale, influential challenges, such as the mapping of global urbanization — one of the most important megatrends of global changes.

Xiaoxiang Zhu is the Professor for Signal Processing in Earth Observation at TU Munich, the head of the department EO Data Science at German Aerospace Center, the co-spokeswoman of the Munich Data Science Research School (MUDS), and the head of the Helmholtz Artificial Intelligence (HAICU) – Research Field "Aeronautics, Space and Transport". Her main research interests are remote sensing and Earth observation, signal processing, machine learning and data science, with a special application focus on global urban mapping.

The current narrative states that AI/ML is crucial to innovation but Europe is lagging behind and needs to re-establish “technological sovereignty”; that machine learning requires loads of training data; that if we federate, open and share data currently disperse among companies and research organisations we will be able to make up for the lack of large data holders and compete with “big tech”.
The reality is more nuanced and filled with tradeoffs. We need lots of data, but most importantly we need high quality data, well coded, which is costly to produce. Open data is one important option, but there are more. Data sharing is costly and difficult, companies and researchers are reluctant, but it has worked upon specific conditions – it is much more than just pooling data. We need open data, we need the so-called data spaces, but we also need large centralized data holders – large research and commercial companies.
Most of all, we need to use the full range of options to have the data we need – not driven by ideology but by the goal of scientific excellence.

David Osimo is director of research in The Lisbon Council, a Brussels-based think tank set up in 2003 to intellectually accompany the Lisbon Agenda, Europe’s original growth and jobs programme. David coordinates the research activity on issues such as the data economy, digital government, start-ups and scale-up policy as well as pioneering novel collaborative methods for policy research. He previously served as director and co-founder of two startups, as scientific officer in the European Commission Joint Research Centre, and advised the United Nations and the Organisation for Economic Co-operation and Development. He holds degrees from University of Milan and University of Cardiff.

For over a decade TomTom has been creating consumer devices for navigation routing people from A to B as fast as possible. One of the key components in routing is the availability of a high-quality map. While initially maps were being produced in a very laborious way involving a significant amount of manual work, map production is nowadays becoming more and more automatized. This talk shows the advantage of using Ai to automate the map-making process. Another key ingredient to providing a great routing experience are connected services. TomTom devices were one of the first to be “connected” and were able to inform drivers about the traffic ahead to even avoid congestions. This talk illustrates how Ai is used to infer traffic incidents based on GPS data. It concludes with a few more involved examples underlining TomToms transition from a consumer device company, to an AI company with automated map creation and autonomous driving in prime focus.

Soeren Sonnenburg has a master in computer science. In his academic career, he has been doing research in machine learning for over a decade. He did contribute to both theory, software and applications to Bioinformatics at the Fraunhofer and Max-Planck Society. He received his PhD in Machine Learning from TU-Berlin with honors. Both his master and Phd thesis were honored with nominations or prizes (Hugo Geiger Preis; Nomination GI- Dissertationspreis; Nomination ERCIM Cor Baayen Award as one of two Germans). He has established the machine learning open source software track in the renowned journal of machine learning research. Currently he is working at TomTom where he also won an innovation prize and is now leading the TomTom AI geospatial research lab (TAIGR) as its director.

Artificial intelligence (AI) and machine learning are so ubiquitous in the media these days that they have garnered a healthy dose of skepticism from the public, in many cases deservedly so. Machine learning comprises computer programs that are capable of solving classification or prediction problems by making inferences and decisions from a dataset without human intervention. We will explore the various applications and benefits of machine learning to the mining industry.

Guy Desharnais Ph.D., P.Geo, has worked in mineral exploration, resource estimation consulting and is currently working as the Director or Mineral Resource Evaluation with Osisko Gold Royalties. He led the team which won the Integra Gold Rush Challenge in 2016 (500k$) which applied a combination of geology, virtual reality, weight of evidence and machine learning techniques to identify the most prospective exploration targets. He was named CIM distinguished lecturer in 2017 where he shared his knowledge on Metallurgical Sample Selection, Application of Machine Learning to Exploration Targeting, and Resource Over-Estimation.

The application of ML technology to subsurface prediction shows that integrated and interdisciplinary approaches lead to deeper understanding of the subsurface and provide a framework for creative solutions and improved decision-making.
In short, modern machine learning methods make practical data driven petrophysics, elastic and rock property prediction and seismic interpretation on a scale not previously considered feasible. When placed within a framework for uncertainty estimation along side automated interpretation, we can start to make the transformative decision-making changes that future data-driven exploration and production workflow demands from artificial intelligence technology.
The talk will present results from work in the Norwegian Continental Shelf with focus on deep-learning assisted seismic interpretation and rock and fluid property prediction with uncertainty from well scale through to seismic scale. A range of different ML techniques are used, drawing on both regional and local data to arrive at both extensive seismic scale predictions and targeted reservoir level estimates within a robust framework estimation framework.

Eirik is cofounder and CEO of Earth Science Analytics. He has 20+ years’ experience from the E&P industry. He has held various technical and managerial roles in oil companies including Statoil, and 4 years as Exploration Manager in Rocksource. He has worked with exploration, field development, and production on the Norwegian Continental Shelf as well as internationally. He holds a MSC in Petroleum Geology and a PhD in sedimentology from the University of Bergen, and is now laser focused on implementation of AI and data-driven analytics in petroleum geoscience.

Recently, time series deep learning (DL) has shown unique capability to accurately predict hydrologic states and fluxes. Using our applications in soil moisture and streamflow predictions and literature work in groundwater, we show that the power of long short-term memory (LSTM) for mimicking hydrologic dynamical systems for the purpose of both long-term projection and short-term forecast. These new DL capabilities are distinct in procedures from existing approaches but could accomplish the same goals with high accuracy, easy implementation and with the flexibility to assimilate various forms of data. Physically-informed machine learning has made significant strides in parameter inversion and field synthesis in groundwater simulations. It is argued here that hydrologic DL opens up an alternative and transformative avenue toward operational predictions and hydrologic knowledge discovery. However, the DL methodology is not without its caveats and misuses.

Chaopeng Shen received the Ph.D. degree in environmental engineering from Michigan State University, East Lansing, MI, USA, in 2009. He developed the hydrologic model process-based adaptive watershed simulator. He was a Post-Doctoral Research Associate with the Lawrence Berkeley National Laboratory, Berkeley, CA, USA, from 2011 to 2012. He is currently an Associate Professor in Civil Engineering at Pennsylvania State University, University Park, PA, USA. His research interests are in hydrology including interactions between water and other subsystems, big data machine learning in hydrology, catchment-floodplain-lake systems and multiscale modeling. Dr. Shen is currently an Associate Editor of the Water Resources Research.

Earth observation (EO) data archives are explosively growing as a result of advances in satellite systems. As an example, remote sensing (RS) images acquired by ESA’s Sentinel satellites (which are a part of EU’s Copernicus program) reach the scale of more than 10 TB per day. The “big EO data” is a great source for information discovery and extraction for monitoring Earth from above. Thus, accurate and scalable techniques for RS image understanding, search and retrieval have recently emerged. In this talk, a general overview on scientific and practical problems related to RS image characterization, indexing and search from massive archives will be initially discussed. Then, our recent developments that can overcome these problems will be presented. A particular attention will be given to our deep hashing network that learns a semantic-based metric space, while simultaneously producing binary hash codes for scalable and accurate content-based indexing and retrieval of RS images. Finally, the BigEarthNet benchmark archive, which is one of the largest Sentinel-2 benchmark archive to drive the deep learning studies in RS, will be introduced.

Begüm Demir is a Professor and Head of the Remote Sensing Image Analysis (RSiM) group at the Faculty of Electrical Engineering and Computer Science, Technische Universität Berlin (TU Berlin) since 2018. Before starting at TU Berlin, she was an Associate Professor at the Department of Computer Science and Information Engineering, University of Trento, Italy. She received the Ph.D. degree in 2010 in Electronic and Telecommunication Engineering from Kocaeli University, Turkey. She performs research on developing innovative methods for addressing a wide range of scientific problems in the area of remote sensing for Earth observation. In order to address challenging topics in this field, her research activities lie at the intersection of remote sensing, machine learning, signal/image processing and management of big data in EO.

Catastrophic natural events continue to increase in frequency and severity. Munich Re applies advanced machine learning methods to streamline the property claims process before and after hurricane events. This is achieved by providing early damage predictions before and very detailed damage detection after the event.

M.Sc. in Statistics at LMU Munich, Experience within the computer vision domain on both damage prediction and detection in property insurance before and after natural catastrophes and risk assessment for enhanced underwriting based on aerial imagery.
In addition, Munich Re facilitates underwriting and risk assessment by feature extraction, based on high resolution aerial imagery.