Agenda Day 2: Sustainability in AI
Morning Sessions – Sustainability in AI
Abstract:
Selected applications using big data, high-performance computing and artificial intelligence to mitigate and adapt to climate change will be given. Topics discussed could include the monitoring and analysis of data to improve climate models, improved weather predictions to enable better storm preparedness, energy demand forecasting, and demand. These tools will be used to coordinate complex electricity systems based on ever-increasing intermittent renewable energy and the need for complex flows of electrical power due to an increase in distributed power generation and energy storage. Weather and climate predictions using AI will be needed to develop more reliant water management systems, improved disaster response and management. Computational tools are needed to enhance energy efficiency in buildings, industry, transportation and agriculture. Finally, applications in health care that range from better early detection, diagnosis prognosis in individual treatments and public health measures to deal the next pandemic.
Abstract:
Prof. Hoefler will explore the potential of machine learning for climate simulation and research. Drawing from ideas presented at the Earth Virtualization Engines summit in Berlin, he will discuss research on ensemble prediction, bias correction of simulation outputs, and the extreme compression of high-resolution data aimed at achieving affordable km-scale ensemble simulations.
His ensemble spread prediction and bias correction network, applied to global data, demonstrates a relative improvement in ensemble forecast skill (CRPS) of over 14%, with even greater gains observed for extreme weather events. Additionally, his post-processing technique requires fewer trajectories to achieve results comparable to the full ensemble.
Prof. Hoefler’s ML-based compression method achieves data reduction from 300x to over 3,000x, outperforming the state-of-the-art compressor SZ3 in weighted RMSE and MAE while preserving significant atmospheric structures without artifacts. When used as a 790x compressed data loader for the WeatherBench forecasting model, the RMSE increases by less than 2%. This compression democratizes access to high-resolution climate data, paving the way for new research avenues. He will conclude with ongoing research directions and opportunities for integrating various machine learning techniques to enhance km-scale global climate simulations.
Abstract:
Is it possible for a nonhuman to participate in the economy, or is that privilege now and forever restricted to humans and the machines that support financial trading? If participation is possible, what new financial instruments could be created to align the interests of other species with those of the economy? Tehanu offers a world-first contribution to this question: another species holding and spending money according to AI-inferred interests. Founded by Jonathan Ledgard and Patrick McSharry, Tehanu, in collaboration with the government of Rwanda, has recently enabled a family of mountain gorillas to have secure digital identities (Know Your Gorilla = KYC) and to send mobile money payments to custodians for verified pro-nature services. This is presented as a first step in a multigenerational journey toward better and richer cohabitation on Earth. It is suggested that AI will become an evolutionarily significant co-pilot for the stewardship of other species. Furthermore, it is argued that Tehanu’s data flows can pave the way for new asset classes like “nature as infrastructure” and help to reduce volatility in a future shaped by climate change and extinction risks.
Afternoon Breakout Sessions – Sustainability in AI
Climate Risk Modeling and AI in Financial Markets
Digital Economy and Climate Change
AI for Climate Monitoring, Prediction, Adaptation
Abstract:
This presentation introduces a framework for pricing climate resilience in infrastructure projects by evaluating the associated debt and equity investments. The framework illustrates how to translate asset-specific climate risks into tangible financial risks, thereby enhancing decision-making for resilient infrastructure investment.
Abstract:
Emissions and economic variables are related to each other. Understanding that relationship is a fundamental step for several objectives, primarily to create roadmaps to emission reduction which minimize the economic impact and to produce baseline emission targets across countries taking into account the specific characteristics of each country. In this presentation we will consider a data-driven approach to such targets. The discussion will be based on association arguments, but causality will be discussed as a work in progress.
Abstract:
Artificial Intelligence (AI), blockchain, and digital currencies are transforming the digital economy by strengthening connections among stakeholders and generating sustainable value. AI enhances efficiency and innovation through data-driven insights, with potential use cases such as empowering SMEs to optimize decision-making, automate processes, and enhance competitiveness. Blockchain, with its decentralized and transparent nature, is revolutionizing supply chains by providing real-time traceability, ensuring secure transactions, and reducing inefficiencies. Digital currencies promote financial inclusion and streamline cross-border transactions, reshape the approach to business and fostering innovation and sustainability in the digital economy.
Abstract:
Digital finance is emerging as a key catalyst for the adoption of green finance. However current literature tends to offer fragmented perspectives rather than a holistic framework. In practice the primary challenges for green digital finance are regulation adoption and data integration. While traditional financial institutions have been slow to adapt fintech startups are already embedding sustainability at the core of their business models. Future research must take an interdisciplinary approach exploring the intersection of strategy organizational development and technological innovation to address this evolving landscape more comprehensively.
Abstract:
Digital twins (DTs) - virtual models that mirror physical assets, systems, or processes - transform numerous industries, from manufacturing to healthcare to logistics to art. By simulating and optimizing real-world operations, these models unlock unprecedented efficiencies, drive innovation, and enable predictive insights. DTs have two key enablers: Distributed Ledger Technology (DLT) and Artificial intelligence (AI). Interoperable DLT provides secure, transparent, decentralized platforms for managing and sharing digital twin data across networks, ensuring trust, ownership, and traceability. At the same time, AI drives the intelligence behind digital twins, enabling the real-time analysis of vast datasets generated by digital twins, offering predictive insights, optimizing performance, and enhancing decision-making. When combined, DLT and AI empower DTs to foster innovation and unlock new economic opportunities in tomorrow’s interconnected, data-driven markets.
Abstract:
The Arabian Peninsula faces significant water stress due to its arid climate, rapid urbanization, and population growth, which strain its limited renewable water resources. Accurate monitoring of precipitation and terrestrial water storage is critical to address water deficits and improve resource management, but sparse observational networks and complex land-atmosphere interactions pose challenges. This presentation will explore these challenges and propose solutions through data fusion and machine learning. It will also highlight the need for advanced forecasting systems and response strategies to address the increasing frequency of extreme weather events, emphasizing the role of AI-driven and physics-based models in effective water management.
Abstract:
Marouane Temimi's talk explores how AI and Machine Learning (AI/ML) can enhance the modeling and prediction of extreme weather events to build resilient urban communities. Traditional physics-based methods have limitations, but AI/ML offers new ways to process large datasets and improve weather predictions. These technologies, combined with advanced sensing tools, enable better integration of data from diverse sources like in situ and satellite sensors, and even leverage citizen science to gain valuable insights.
Prof. Temimi will present case studies demonstrating how AI/ML has improved analysis, such as satellite data segmentation and processing imagery from ground cameras. These examples show AI/ML’s potential to support more accurate weather forecasts and strengthen urban resilience against extreme events.
Abstract:
The Center for Western Weather and Water Extremes (CW3E) mission is to provide 21st Century water cycle science, technology, and outreach to support effective policies and practices that address the impacts of extreme weather and water events on the environment, people, and the economy of Western North America. Luca Delle Monache will discuss how CW3E develops predictive models using physics-based, deep learning, and data-driven approaches to deliver accurate estimates of precipitation and other atmospheric variables. These models support water managers in the Western US, particularly through the Forecast Informed Reservoir Operations Program, helping to optimize water supply and minimize flooding risks as part of climate adaptation strategies. He will also assess the skill of CW3E’s models and compare them to leading prediction systems like the US National Weather Service’s Global Forecast System and the European Centre for Medium-Range Weather Forecasts Ensemble Prediction System.