UNIS Digital

Place: Møysalen
Time: 13:15
Presenter: Edward Boyda, Earth Genome and AI-Journalism Resource Center, Oslo Metropolitan University.

Using new geospatial AI to uncover environmental crime

In the last few years, geospatial AI foundation models have opened dramatic new possibilities for remote environmental monitoring, for experts and non-experts alike. Few- and zero-shot learning with foundation models translate to real-time satellite image analysis and web-based Earth-search interfaces. We will talk through some recent technical developments and discuss efforts to use this new technology to report on narcotrafficking, illicit logging, and illegal open-pit mining, which pose serious threats to forest ecosystems and nearby communities.

Edward Boyda is a physicist and environmental data scientist and journalist. He runs satellite-based investigations for environmental and human rights reporting with the California non-profit Earth Genome and the AI-Journalism Resource Center at Oslo Metropolitan University. He helped build Amazon Mining Watch. Previously, he was a professor of physics at Saint Mary’s College of California, with interests in particle physics and quantum computing.

The seminar is open for all UNIS staff and students.

In collaboration Oslo Metropolitan University, UNIS will arrange set of seminars and workshop introducing UNIS researchers and students to machine learning for scientific data analysis.

The seminars will introduce basic concepts in machine learning and demonstrate the potential and possibility with machine learning. This will be followed by dedicated interactive workshops where UNIS students and researchers can get help to apply machine learning techniques to analyze their own data. Bring your own laptop and data for the workshops.

The seminars and workshops will be held in the Kapp Schoulz (B 127) classroom. A certificate will be issued to participants completing one or more of the workshops.

Preliminary program

Below is a preliminary program. Detailed content will depend somewhat on the interests from the registered users. During the workshops (27-28 Nov) we aim to use various UNIS data supplied by UNIS researchers to demonstrate the techniques.

Wednesday 26 Nov – Seminar: Introduction to machine learning

13:15: Introduction: What can you do with machine learning?

14:15: Machine learning basics.  Topics covered: Data preparation, linear regression, common pitfalls in machine learning, supervised versus unsupervised learning.

Thursday 27 Nov – Workshop on image processing

Workshop consisting of brief lectures introducing the concepts, followed by hands-on workshop with prepared UNIS data or your own images data sets.

09:15 – 12:00 Image classification basics.

13:15 – 16:00 Image classification intermediate.

Friday 28 Nov – Workshop on time series analysis

09:15 – 12:00: Time series classification

13:15 – 16:00: Time series prediction

Preliminary schedule and registration (closed, fully booked).

Place: Møysalen
Presenter: Luke Marsden, UNIS & MetNo, Norway

How to prepare and publish FAIR data

UNIS own data manager will give a seminar on how to prepare and publish observational data according to the Findable, Accessible, Interoperable and Reusable (FAIR) principles.

The seminar will be followed by a Q & A session where you can get help and tips about specific data sets.

The slides from this session are available here.

Place : Kapp Wijk, UNIS, Svalbard
Presenter: Anna Allen, University of Cambridge, UK

Real world use of machine learning : Weather forecasting

Machine learning is transforming the way we analyse and act on Earth system data. This seminar will provide an overview of state-of-the-art machine learning tools driving advances in the field, before focusing on two topical ideas through the lens of weather forecasting. We will first introduce the Aurora project and the concept of foundation modelling. Diverse pretraining on large Earth system datasets enables Aurora to outperform full operational dynamical forecasting systems for high-resolution weather, wave, and air-quality forecasts for the first time. We then turn to end-to-end learning with the Aardvark Weather model, demonstrating that forecasts generated directly from observations without relying on reanalysis can achieve skillful performance at 0–10 day lead times, even surpassing operational NWP in certain scenarios. A case study forecasting 2 m air temperature in Nordenskiöld Land additionally illustrates how AI-guided sensor deployment can improve high-resolution local forecasts in a region with high uncertainty in training data. We will conclude with a discussion of potential opportunities for applying these approaches more broadly across different applications in Svalbard.

About the speaker:
Anna Allen is a postdoctoral research fellow at the University of Cambridge, UK, focusing on applying machine learning to Earth system forecasting. She recently completed a PhD in Computer Science at Cambridge, after earning a Master’s in Meteorology from the University of Melbourne, Australia. She additionally works with the UN Environment Programme leading the development of AI models for detecting and mitigating greenhouse gas super-emitters.

Place : Kapp Schoultz, UNIS, Svalbard

CF-NetCDF files can be used to publish FAIR-compliant data related to:

• Measurements of physical environment
• Time series, depth profiles, multiple dimensions
• Outputs from models
• Point clouds
• Satellite data
• Flow cytometry
• …

Come to this workshop and learn how to create a CF-NetCDF file using R.

Feel free to bring your own data to work on, or you can create a similar dummy dataset during the session to use as a template for later.

Bring a laptop!

Place : Kapp Schoultz, UNIS, Svalbard

CF-NetCDF files can be used to publish FAIR-compliant data related to:

• Measurements of physical environment
• Time series, depth profiles, multiple dimensions
• Outputs from models
• Point clouds
• Satellite data
• Flow cytometry
• …

Come to this workshop and learn how to create a CF-NetCDF file using Python.

Feel free to bring your own data to work on, or you can create a similar dummy dataset during the session to use as a template for later.

Bring a laptop!

Place : Kapp Schoultz, UNIS, Svalbard

Learn to access and use key aggregated global datasets that can compliment your research. Choose which data you want to work with and how.

• Sea surface water temperatures
• Sea ice concentration, extent, drift, emissivity
• Bathymetry/topography
• Biodiversity data
• Surface temperature anomalies
• More available on request (if requests made in good time)

Bring a laptop!

Place : Kapp Schoultz, UNIS, Svalbard

This workshop will include

• What the SIOS data access portal is, and what the purpose is
• What data contributes to it (and what does not)
• An exercise showing you how to use the access portal, showcasing some of it’s most useful features
• How to use it to access data you are interested in
• How to contribute data to the SIOS access portal

Bring a laptop!

Place : Kapp Schoultz, UNIS, Svalbard

It’s time to face the harsh truth; our current methods of managing scientific data are a mess. How can we make data meaningfully useful at scale, to anyone who is curious? Merely open data are often only theoretically findable and accessible in public archives—scattered, inconsistent, and difficult to use.

Yet across the world, scientists are measuring the same key variables over and over again. Imagine if we could automatically aggregate these datasets into powerful, unified data products that anyone could use. Imagine if we had the tools and software to easily find, analyse, process, and visualise any of these datasets. This is not a distant dream—it is possible, but only if we transform how we manage scientific data. Change is coming. Welcome to the future of scientific data.

Place : Radisson Blu Hotel, Longyearbyen, Svalbard

The digital season 2025 starts off with a workshop on how to publish data. For more information, see 

Workshop: Publishing Your Data for the SIOS Data Access Portal (SIOS/UNIS/HarSval) | sios-svalbard.org

NB! Note that a registration is required. Registration deadline is 20 Dec 2024.

Place : Kapp Mitra

UNIS employs a number of autonomous devices and remote controlled devices, like drones, gliders, weather stations, and other sensors. Many of these devices are connected to a wider local or global network, like e.g., the UNIS data network or the Internet via various protocols. In this seminar, and in the spirit of the ongoing risk analysis process at UNIS, João Pedro Dias will provide an overview of various devices, data transfer protocols used and potential security threats with focus on the Internet-of-Things ecosystem.

Link to presentation:
http://jpdias.me/assets/talks/UNIS_Cybersecurity_and_Internet-of-Things_2024.pdf

About the speaker:
João Pedro Dias is part researcher on the thin line between hardware and software, and part Software Engineer. He earned his Ph.D. in Informatics Engineering from the Faculty of Engineering, University of Porto (FEUP) in 2022. He maintains a Software Engineer position as a day-to-day job at Kuehne+Nagel. Since 2017, he has also been an Invited Assistant Professor at FEUP, where he teaches courses in Software Engineering and Operating Systems, and has been a co-supervisor for Master students and contributed to various research projects. João’s research focuses on Internet-of-Things systems, software engineering, software security and privacy. In his free time, he enjoys experimenting with Software-defined Radio, building web applications and reverse-engineering hardware.

Place : Lassegrotta 

This presentation will show how ChatGPT can make life easier for students and researchers. We will discuss what to use it for and what not to use it for, and how to get the best out it. We’ll explore how to use ChatGPT to troubleshoot and even generate code and explain tricky concepts. For writing, ChatGPT can act as a smart editor, helping polish research papers, correct typos and improve clarity. ChatGPT can be used to restructure information from one format to another. By the end, you’ll see how ChatGPT can save time and help you through the most tedious of academic tasks so you can focus on the what you enjoy the most!

Place : Møysalen

The Findable, Accessible, Interoperable, Reusable (FAIR) data principles define a set of criteria for modern publishing practices and the sharing of scientific data. UNIS aims to implement these principles and make all relevant observational data accessible through the Svalbard Integrated Arctic Earth Observing System (SIOS). 

In this seminar, Luke Marsden, UNIS Data Manager, will provide an introduction to FAIR principles, and give an overview of useful tools to assist individuals in working with FAIR data. Additionally, he will discuss how to integrate observational data into the SIOS data management system.

Place: Kapp Mitra

Warmer temperatures increase permafrost thaw, affecting infrastructure in town. Permafrost thaw in combination with more liquid precipitation increases the potential for landsliding. Therefore, we are building a platform to monitor permafrost and meteorological changes. What does this network look like? How are all these data managed? Who can use this platform? How is the data presented? And why is this platform useful for Longyearbyen? Come and find out more about this innovative network of instruments and how we manage our data.

Place: Kapp Schoultz

At KHO we take images of the sky automatically throughout the dark season, or whenever the sun is well below the horizon. That results roughly 1 million images per year. Everybody wants to know when the images contain aurora, and it seems so simple to say that we can just ask an AI tool to flag all images, which contain green light. This presentation we will explain why the problem is harder than it sounds, what has been done to mitigate it and how well the recent attempts to employ convolutional neutral networks handle this task. If nothing else, come and see our pretty data.

Place: Lassegrotta

The “LLMs in Education” group, including STEM educators from the University of Bergen, Norway, conducted two workshops in 2023 to explore the use of large language models (LLMs) like ChatGPT and similar in education. In one, we analyzed student responses from two surveys, specifically targeting responses to the question “How do you think these tools should be used in education, in a way that is fair and that supports your learning?”

From these responses, two broad themes emerged: questions and concerns about the tools themselves, and questions and concerns related to acceptable use, fairness, and ethics. In our discussion, I will share some of these concerns, and we can discuss UNIS-appropriate responses to guide our use of LLMs in our teaching.

Place: Lassegrotta

SvalSat is a world-leading supplier of satellite services. The installation of more than 100 antennas at Platåberget just outside Longyearbyen tracks more than 3,500 passes each day. SvalSat serves as a ground station for a number of civilian polar orbiting satellites, including Earth-observing satellites for meteorology, sea surface measurements, geodesy, as well as satellites for navigation and ship traffic monitoring. Every day, large amounts of data are downloaded and transferred to their owners all over the world.  In this seminar, the station’s director, Ole Kristian Kokvik will give an overview of the installations and activities at SvalSat.

Place: Lassegrotta

The Findable, Accessible, Interoperable, Reusable (FAIR) data principles define a set of criteria for modern publishing practices and the sharing of scientific data. 

UNIS aims to implement these principles and make all relevant observational data accessible through the Svalbard Integrated Arctic Earth Observing System (SIOS). 

In this seminar, Luke Marsden, Data Manager at the Nansen Legacy, will provide an introduction to FAIR principles, and dive an overview of useful tools to assist individuals in working with FAIR data. Additionally, he will discuss how to integrate observational data into the SIOS data management system.

Following the seminar, Luke will give dedicated workshops for each department in Van Mijen. These workshops will delve into details surrounding data formats, provide assistance with metadata, and offer guidance on converting data to suitable formats. Users are encouraged to bring their laptops along with sample data for these hands-on workshops.