Kolloquium SoSe 2024

Dr. Christopher Kadow

DKRZ - Deutsches Klimarechenzentrum

Wann: Mo, 03.06.2024, 16:15 Uhr

Wo: Konrad-Zuse-Hörsaal (Raum B-201)

Thema

The technology that deletes photobombs can do climate research? The chat bot that writes poetry can do climate analysis? Climate Informatics!

Sprache: English

Abstract

Climate Informatics - an interdisciplinary field that combines climate science and data science to analyze, model, and understand climate-related data. It involves the application of computational, statistical, and machine learning techniques to study and predict climate patterns, assess climate impacts, and support decision-making in climate policy. By leveraging large datasets and advanced algorithms, climate informatics aims to improve the accuracy of climate models, enhance climate predictions, and provide insights into climate variability and change - Machine Learning for Climate Science.

Climate change research today relies on climate information from the past. Historical climate records of temperature observations form global gridded datasets that are examined, for example, in IPCC reports. However, the datasets combining measurement records are sparse in the past. Even today, they contain missing values. We found that recently successful image inpainting technologies, such as those found on smartphones to get rid of unwanted objects or people in photos, are useful here. The derived AI networks are able to reconstruct artificially cropped versions in the grid space for any given month using the missing values observation mask. So herewith we have found with AI a technique that gives us data from the past that we never measured with instruments. Other important datasets used in the Assessment Report 6 of the IPCC to study climate change, as well as advanced applications such as downscaling in atmosphere and ocean, a hybrid (AI&ESM) data assimilation approach within ICON, or precipitation in broken radar fields are shown in this presentation. Deep learning techniques include U-Nets, diffusion and vision transformer models.

Climate research, including the study mentioned in the previous paragraph, often requires substantial technical expertise. This involves managing data standards, various file formats, software engineering, and high-performance computing. Translating scientific questions into code that can answer them demands significant effort. The question is, why? Data analysis platforms like Freva (Kadow et al. 2021, e.g., gems.dkrz.de) aim to enhance user convenience, yet programming expertise is still required. In this context, we introduce a large language model setup and chat bot interface based on GPT-4/ChatGPT, which enables climate analysis without technical obstacles, including language barriers. Not yet, we are dealing with climate LLMs for this purpose. Dedicated natural language processing methodologies could bring this to a next level. This approach is tailored to the needs of the broader climate community, which deals with small and fast analysis to massive data sets from kilometer-scale modeling and requires a processing environment utilizing modern technologies, but addressing still society after all - such as those in the Earth Virtualization Engines (EVE).

Bio

Dr. Christopher Kadow has been Head of the Data Analysis Department at the German Climate Computing Center (DKRZ) in Hamburg since 2023. From 2019 to 2023, he headed the junior research group for climate informatics and technologies at the DKRZ, having previously spent years at the Freie University of Berlin, where he also completed his doctorate in climate prediction and data analysis platform. His research and service activities, including various DFG, BMBF, BMWK and EU projects, focus mainly on IT interfaces in the field of Earth system modeling and the application of machine learning and artificial intelligence.