ADS Drinks & Data: AI solutions for Climate Change | Part 1

Join us for this in-person Meetup on the 21st of September at 16:00! 👏

Artificial Intelligence has the potential to tackle the biggest challenge on the planet, Climate Change. What could the role of AI and Data Science be in this huge problem? How can we address this challenge, with AI technology?

Location: VU NU Gebouw, room: NU-5A47. If you enter the building, take the escalater to the 5th floor. In case you can't find it call 06-43423381

Programme
15:55 Walk-in
16:00 Introduction & Welcome
16:05 Talk #1: Sandra Merten (Elsevier)
16:20 Discussion
16:25 Talk #2: Chiem van Straaten (XAIDA, KNMI)
16:40 Discussion
16:45 Talk 3: Nick Schutgens (Earth & Climate, VU)
17:00 Discussion
17:05 Tak #4: Jannes van Ingen (VU IVM)
17:20 Discussion
17:25 Networking
18:00 End!

Talk #1 by Sandra Merten (Elsevier)
Data-driven approaches to decarbonization
To limit global warming to 1.5° Celsius above pre-industrial levels as set out in the Paris Agreement, greenhouse gas (GHG) emissions should peak no later than 2025 and fall to net zero by 2050 according to the Intergovernmental Panel on Climate Change (IPCC). The reduction of CO2 emissions plays a critical role in reducing GHG emissions, however, the United Nations 2022 Sustainable Development Goals Report found that energy-related CO2 emissions increased by 6% in 2021, reaching highest level ever.

There are several aspects to decarbonization, including avoiding emissions (e.g., through design and operation choices), switching to renewable fuels and energy, as well as sequestration of CO2 emissions that are hard to avoid (e.g., geological carbon storage). This presentation will show examples of data-driven approaches to decarbonization, including the application of taxonomy-based data retrieval from a geospatially enriched scientific database combined with spatial analytics to unlock patterns and insights from the scientific literature to systematically predict areas with geological carbon storage potential.

Talk #2 by Chiem van Straaten (KNMI, VU)
Understanding climate extremes through AI
As extreme summers become the rule instead of the exception, a two-fold call is heard. First is the call for better and earlier predictions, such that society can prepare itself. Second is the call for insight into the role of climate change, such that its current effects can be clearly communicated and such that future risks can b better quantified.
I will show how AI is used to disentangle the complex processes behind extremes. Strong challenges must be met. Among these are the high dimensionality of the earth system, the lack of labelled data, and the presence of fluctuations on all spatial and temporal scales. A case study of heatwave prediction is given.

Talk #3 by Nick Schutgens (Earth Sciences, VU)
Air quality problems
Air quality is (globally seen) one of the biggest causes of death. In order to involve citizens more in the problem, we try to predict air quality at street level. Unfortunately, it is very CPU intensive to do that using physical/chemical models. That is why we only use these models to predict air quality on a larger scale (several kilometres). Subsequently, super-resolution techniques from Machine Learning can be used to make predictions on a smaller scale (eg street level). This talk is about a first attempt at using stacked CNN and U-net for super-resolution to support air quality problems.

Talk #4 Jannes van Ingen (VU IVM)
Reducing weather-related risks in the energy and agricultural sector
As the effects of climate change become more pronounced all over the globe, the demand for longer range weather forecasts grows. The energy sector is becoming highly weather dependent through the growth in renewables and the effect of the weather on energy demand. I will explain how we forecast on these long lead times and aim to reduce weather-related risks in the energy and agricultural sector. Although AI is often regarded as a ‘black box’, I will show how AI is effectively used to improve accuracy, gain trust, and directly predict impacts.