How to use the power of AI to reduce the impact of climate change on Switzerland

Page 1

How to use the power of AI to reduce the impact of climate change on Switzerland

Recommendations for the Swiss society and economy to become more resilient against the impact from a radically changing climate

Make key technologies broadly available and apply key methodologies in the field of climate and AI to overcome challenges.

technologies

Key methodologies

Introduction

Climate change affects us all, either by direct impacts, such as infrastructure damage and supply chain disruptions, or indirectly by transition risks that arise as we journey towards a low-carbon society. Besides the direct physical impacts of climate change on Switzerland, our economy is closely intertwined with and dependent on the global economy. Thus, it is equally exposed and vulnerable through supply chains, resource availability and customer availability. The process of reducing greenhouse gas emissions needs to be accelerated. Adaptation to physical and transition risks should become an urgent priority for everyone, from individuals to start-ups, and small and medium enterprises to multinational and global companies.

Besides tremendous challenges, this also creates opportunities for interdisciplinary and transdisciplinary cooperation and research, the roll-out of breakthrough technologies and innovative business models and products.

New Earth observation data and Earth system models in combination with AI tools are capable of delivering highly accurate weather forecasts up to 40,000 times faster than conventional models. Consequently, climate impact predictions can be created globally at unprecedented resolutions, thus providing invaluable tools for climate resilience. However, running such Al models requires suitable cloud computing and storage resources, in-depth expertise in climatology and Al, software deployment, integration and life cycle principles and capabilities.

Although these all exist in the respective communities, it is only gradually becoming possible to break them down cost-efficiently to a level that is relevant to companies. However, these developments have not yet reached all those involved and there are considerable dependencies – on a few large service providers and across international and political borders.

Study overview: impact and beneficiaries of key technologies and methodologies to reduce climate change impact. Impact on climate and transition risk alert – assess – mitigate – adapt Key
Economy Nature Government Society
Earth observation data AI – supercomputers
and trustworthy AI AI workflow integration
Open science: data and code
Beneficiaries
Responsible
AI for Earth system models

Physical and transitional climate risks

Switzerland is particularly exposed to climate change: the mean tempearture today is 2.8°C above the pre-industrial average (1871–1900) compared to 1.3°C worldwide. Also, a clear increase in the frequency and intensity of certain extreme events (e.g. heatwaves and heavy precipitation) has been observed.

Climate change causes two categories of risk: physical risk is associated with physical impacts e.g. due to climatic extremes, whereas transition risk is related to the transition to an economy with lower carbon emissions. In Switzerland’s long-term climate strategy, the Swiss Federal Council estimates that inaction would involve an annual physical cost of up to CHF 38 billion by 2050.

Assessing and managing physical risks in Switzerland

Temperature and precipitation extremes are expected to become more frequent and intense, increasing risks such as heat stress; droughts; floods and slope instabilities; degradation of water, soil and air quality; spread of pests and diseases; and invasion by non-endemic species. This affects almost all sectors, e.g. health (heat waves), agriculture and forestry (droughts and spread of pests), building and infrastructure (extreme events) or industry (supply chain interruption or reduced workforce productivity).

Short-term risk management requires early warning systems and rapid information dissemination during and after events. In the medium and long term, Swiss society will have to adapt by adopting behavioural, structural, regulatory and technical measures. Accurate modelling of future physical climate risks is key to guiding this process of adaptation. Despite tremendous advances in climate science, long-term climate projections are still subject to considerable uncertainty at the cantonal and municipal level, and accurately modelling exposure and vulnerability remains a challenge. Thus, there is an urgent need for efforts to develop novel modelling techniques and for flexible decision making processes that are robust enough to withstand a broad range of uncertainties.

The effect of global warming: frequency and intensity of extreme temperature events. Source: IPCC AR6 WGI Fig. SPM.6

Managing transition risks –towards a low-carbon economy

The magnitude of future climate change depends on cumulative worldwide emissions of greenhouse gases. Thus, immediate and effective reduction is imperative. The associated transition to a low-carbon economy induces various risks, including legal, administrative, reputational, technological and market risks, all of which require management.

The legal requirements for detailed corporate carbon emission reporting present significant organisational challenges. Corporate entities must disclose their environmental, social and governance (ESG) performance, align their reporting standards or adopt the principles developed by the Taskforce on Nature-related Financial Disclosures (TNFD). Assessing organisations’ carbon footprint facilitates targeted reduction strategies and sustainable business practices and is key to managing their future environmental impact.

Once 1850–1900 now likely occurs 4.8 times (2.3–6.4) +1.2°C hotter +2.0°C hotter +2.7°C hotter +5.3°C hotter Present 1°C will likely occur 8.6 times (4.3–10.7) 1.5°C will likely occur 13.9 times (6.9–16.6) 2°C Future global warming levels will likely occur 32.9 times (27.0–41.4) 4°C 0°C +1°C +2°C +3°C +4°C +5°C +6°C INTENSITY increase FREQUENCY
per 50 years

Key enabling technologies and methodologies at scale

AI for Earth system models: Deriving and organising information from data (knowledge extraction) has been challenging and expensive in the past. With the advent of AI tools such as deep learning and next-generation Earth system models, insight extraction from Earth observation data and high-fidelity weather forecasts at a global scale have become feasible. These developments support critical business decisions in real-time or long-term risk assessments and create transparency in response to governmental and legal requirements within and across industries.

Earth observation data has become abundant and publicly available, with open data currently totalling 807 Petabytes (PB) and increasing annually by more than 100 PB. The spatial details in commercial satellite data make it possible to recognise objects the size of an A4 sheet of paper from space – every day and everywhere on Earth. Combined with proprietary business data, such as supply chain details from production facilities, logistics and distribution, climate-related risks for companies can be identified and managed at unprecedented spatial and temporal resolution.

AI workflow integration: Integrating data into AI workflows is challenging. There is a lack of analysis-ready datasets, formats and it is also necessary to consider the degree of veracity and ensure interoperability between systems. Moreover, the entire workflow should be compliant with FAIR principles. Due to limitations in reusability, modularity and documentation, AI tools and code from research are often not adopted by industry. Thus, while recognising AI’s potential to reduce the impacts of climate change, only 43% of public and private-sector leaders have a vision of how to put it to use. Even with a strategy in place, the cost of the associated computing often remains a major stumbling block. Furthermore, ecologically and economically viable, trustworthy and socially accepted integration is essential. Co-design approaches involving domain experts, stakeholders from affected communities and policymakers are a proven basis for this when combined with best practices and open standards.

AI – supercomputers: Sharing (computing) resources across organisations is the most affordable way of accessing them, given their limited availability and the cost of developing cutting-edge AI models, which can often total millions of Swiss Francs. Government institutions such as the CSCS Swiss National Supercomputing Centre as well as commercial services provide access to such infrastructure at a fraction of ownership cost. It is thus essential to put these resources at the fingertips of all stakeholders so they can run client-specific or open climate applications both efficiently and cost-effectively.

Now is the time to leverage these breakthrough technologies for sustainability applications. This requires increased collaboration between private, public and academic partners that is amplified through public-private partnerships and interdisciplinary competence centres with a culture of open innovation

Open science: Open source and open science ecosystems help accelerate research and assist AI tools in penetrating markets and industry. For instance, applications using maps by the community-driven OpenStreetMap mapping project generate global annual revenue in excess of CHF 1 billion. Many other examples show that viable and profitable business models can be developed around open-source technology.

Recommendations for decision makers

We recommend that the decision makers below take the following seven actions:

1 5

2

Build capacity for Swiss actors in the field of AI for climate and sustainability

Reinforce the capacity of national competence centres and research bodies to support AI for climate and sustainability.

➜ has an impact on all the case studies below

Ensure access to and involvement in international programmes

Negotiate participation in European and international initiatives supporting AI for climate and sustainability.

➜ has a particular impact on case study B

Accelerate the translation of research results in market impact

Strengthen and establish interdisciplinary and transdisciplinary collaborations as well as Public-Private Partnerships (PPPs).

6

3 7

Implement the principles of open science

Reinforce open data and open source principles (e.g. EMBAG), including data parsimony.

➜ has an impact on all the case studies below

4

Promote scalable and reusable code and ML model base

Provide resources to enable collaboration between environmental scientists and software engineers.

➜ has a particular impact on case study B

➜ has a particular impact on case study A

Implement responsible AI applications

Conduct technology impact assessments based on the UN SDG Agenda.

➜ has an impact on all the case studies below

Foster a quantitative understanding of the implications of climate change

Conduct data-driven studies on climate-related impacts on all Swiss stakeholders.

➜ has an impact on all the case studies below

Academia Federal government and governmental bodies Funding organisations Private entities

Case studies with AI and Earth observation data as key enablers

Case study A: Reporting carbon footprints

Greenhouse gas (GHG) emissions are the cause of climate change. A significant reduction in emissions is therefore essential in limiting global warming to well below 2°C. Since January 2024, large Swiss companies have been legally required to report their direct and indirect greenhouse gas emissions and to disclose emission reduction targets and their strategy for achieving them.

However, reporting on greenhouse gas emissions is challenging and labour-intensive. It requires the collection of extensive data from hundreds, if not thousands, of products across the entire supply chain, including raw material sourcing, production,

transport, use and recycling. AI has been shown to significantly streamline data collection and analysis and improve the accuracy and efficiency of emissions calculations. Machine learning algorithms can recognise similarities between products and efficiently estimate product-level GHG emissions for new or less documented items. In addition, AI can facilitate analyses that allow companies to identify opportunities to reduce emissions and optimise their supply chains for sustainability.

AI-supported reporting on greenhouse gas emissions can also improve transparency for consumers by providing real-time information on the environmental impact of a product. This feature can be integrated into online shops, for example, and serve as a decision-making aid for environmentally

➜ This case study would benefit from recommended actions 1, 3, 5, 6 and 7

© iStock / Creativemarc / m.malinika

Case study B: Greenhouse gas emission monitoring

Governments who signed the Paris Agreement on Climate Change are obliged to report their greenhouse gas (GHG) emissions to help evaluate the world’s progress towards net zero emissions. At present, countries mainly report their emissions by means of emission inventories compiled from socioeconomic and environmental statistics. As with product carbon footprint reporting (Case study A), compiling the data required to produce GHG inventories is a resource-intensive, time-consuming process that involves a high level of uncertainty.

To aid countries, environmental monitoring systems are currently being developed that will provide standardised, reliable and timely global information on anthropogenic greenhouse gas concentrations and emissions. These systems use measurements from ground, aircraft and satellite sensors. AI models then analyse this data to determine the emissions from countries, power plants and industrial facilities. Such models can process the vast amount of data accurately and in a timely manner to identify GHG emitters and estimate GHG emissions from GHG concentration measurements. This means that super-emitters can be identified quickly and potentially mitigated.

Dataflow of a GHG monitoring system based on the Copernicus GHG emission monitoring service. The example shows a methane leak from a gas processing facility being transported by southeasterly winds.

Google Earth 2019, Empa 2024.

This case study would benefit from recommended actions 1, 2, 3, 4, 6 and

Uptake ➜ Socio-economic
data Satellite observations In situ measurements Emission reports Economic statistics ➜ Data integration Coupled data assimilation system Atmospheric transport and emission models ➜ Output Decision support National uptake Market uptake User support Consolidated hotspot emissions Consolidated country/ regional emissions
7
and enviromental
©
AI

Case study C: Predict local heat hotspots for early warning

Which areas of a city are particularly affected by heat? Where should measures be taken to prevent or mitigate the formation of heat hotspots? Is a certain retirement home located in an area that is particularly affected? To answer these and similar questions, high-resolution temperature maps provide valuable assistance to urban planners, city authorities and inhabitants.

An AI model uses a combination of open government data (e.g. 3D building data), freely available satellite data and temperature measurements from citizen weather stations to predict the temperature at high spatial and temporal resolution (10 × 10 m), in contrast to classical weather models which only have a resolution of one to a few kilometres. This approach can be applied efficiently to cities worldwide. Incorporating Earth system models will also make it possible to estimate future heat hotspots for different greenhouse gas emission scenarios.

➜ This case study would benefit from recommended actions 1, 3, 6 and 7

Summary: Swiss companies could use the results of all these case studies, as could administrative staff and political decision-makers, as a basis for their climate protection and adaptation activities. It would therefore be important to conduct such studies and make them publicly available.

➜ This is emphasised by recommended actions 1 and 7

Citizen weather stations 3D building data for Zurich
AI
t
Satelite data Zurich High-resolution temperature maps © Zumwald, M., et al. (2021). Mapping urban temperature using crowd-sensing data and machine learning.

Background information on this SATW project

SATW promotes projects in the field of artificial intelligence and access to data

As part of activities associated with its Artificial Intelligence and Energy and Environment focus topics, SATW promotes projects in the field of artificial intelligence and identifies technologies that are important for climate neutrality and security of supply. This fact sheet and the accompanying white paper are the result of collaboration with IBM Research Europe - Zurich, Switzerland.

The project was launched with a kick-off workshop on 15 June 2023 attended by a broad community that included representatives of research, industry and government. During the workshop, the 50 or so participants identified important factors that require consideration in the context of the timely uptake of responsible and trustworthy AI for the purpose of creating a more sustainable and resilient future for everyone. In the following weeks and months, working groups of workshop participants and additional contributors discussed these topics in more depth and described them in greater detail. An editorial board (authors and project lead below) coordinated the whole content creation process and more.

The statements made in this factsheet are a précis of the more extensive white paper study “How to use the power of AI to reduce the impact of climate change on Switzerland” conducted by over 70 domain experts and scientists from over 30 renowned Swiss academic, government and industrial institutions. It can be accessed via the QR code on the right.

Publishing details

White paper authors: (in alphabetical order): Jennifer Susan Adams, Christos Altantzis, Tom Beucler, Julia Bingler, Samuel Brown, Thomas Brunschwiler, Chiara Colesanti-Senni, Frank de Morsier, Márcio dos Reis Martins, Andreas M. Fischer, Stefan Frei, Remo M. Frey, Alisa Freyre, Gregory Giuliani, Marvin Höge, Apolonio Huerta, Vincent Humphrey, Stefan Keller, Romeo Kienzler, Erwan Koch, Jonathan Koh, Agnieszka Kosinska, Sven Kotlarski, Manuel Kugler, Gerrit Kuhlmann, Markus Leippold, Reik Leiterer, David Leutwyler, Olivia Martius, Adrien Michel, Veruska Muccione, Michal Muszynski, Urs Neu, Diana-Denisa Rodila, Claudia Röösli, Andreas Scheidegger, Tobias Schimanski, Konrad Schindler, Christian Spindler, Tanja Stanelle, Benjamin Stocker, Devis Tuia, Saeid Ashraf Vaghefi, Jan Dirk Wegner, Jonas Weiss, Matthew Whellens, Mira Wolf-Bauwens, Hendrik Wulf, Biagio Zaffora, Marius Zumwald

Authors: Thomas Brunschwiler (IBM), Erwan Koch (UNIL), Reik Leiterer (Data Innovation Alliance), Tanja Stanelle (EBP), Jonas Weiss (IBM)

Project lead: Christian Holzner, Manuel Kugler

Proofreading: Caspar Türler, Ester Elices, John Gleaves

Illustrations: Adobe Stock

Translation: weiss traductions genossenschaft

Design: Andy Braun

May 2024

Swiss Academy of Engineering Sciences SATW St. Annagasse 18 | 8001 Zürich | 044 226 50 11 | info@satw.ch | www.satw.ch
Link to the white paper

Turn static files into dynamic content formats.

Create a flipbook
Issuu converts static files into: digital portfolios, online yearbooks, online catalogs, digital photo albums and more. Sign up and create your flipbook.