Special Session     Thu-31 Jul     AM1   08:30 – 10:00     MR1

SS01: AI in Weather and Climate Prediction: Progress, Challenges, and Outlooks


Responsive image

Invited Talk
Application of Deep Learning to Atmospheric Model Physics Parameterization
Alessandro DAMIANI, Center for Climate Change Adaptation, National Institute for Environmental Studies

Biography

Dr. Alessandro Damiani is a Researcher at Center for Climate Change Adaptation, National Institute for Environmental Studies in Japan. He received his D.Sci. from Università di Siena in 2007. As a researcher in transdisciplinary science, he works at the intersection of remote sensing, climate, and environmental studies. Passionate about data science, his current focus is on regional climate downscaling using machine learning techniques. His recent work involves high-resolution modeling of solar radiation, precipitation, and snow depth using satellite observations, reanalysis, and climate simulations. The primary goals are to improve the physical realism and spatial coherence of downscaled fields to support scenario analysis and climate adaptation studies. He also maintains research interests in the climate of the Polar Regions and Southern Hemisphere, as well as in air pollution and its environmental impacts.


Challenges in projecting future precipitation and snow changes at kilometer scale for adaptation using CNNs

Alessandro DAMIANI#+, Noriko ISHIZAKI
National Institute for Environmental Studies, Japan

High-resolution climate projections are crucial for regional adaptation to climate change. However, the coarse resolution and biases of general circulation models limit their applicability. Statistical downscaling offers a computationally efficient alternative to dynamical downscaling, enabling corrections and large-ensemble analyses.We present a fast convolutional neural network (CNN)-based approach that leverages large-scale predictors to estimate high-resolution climate variables. Trained on observational data, the base model is further adapted to specific climate variables using physically consistent stochastic approaches, transfer learning, or binary classifications. It is then applied to climate simulations to enhance spatial detail and accuracy. After discussing challenges in reproducing precipitation and the ongoing activities, we focus on future snow cover changes. We present recent results on downscaling snow depth (SD) from an ensemble of Coupled Model Intercomparison Project Phase 6 models. Our model captures physically plausible relationships, facilitates high-resolution SD assessments, and produces results consistent with regional climate models. Additionally, it offers insights into future snow impacts on winter tourism and water resources, highlighting its value for adaptation studies.





AOGS Secretariat

Website: www.meetmatt.net
Email: meetmatt@meetmatt.net

Information & Support

Exhibition & Sponsorship: geomeet@asiaoceania.org
Scientific Program & Help Desk: info@asiaoceania.org
Society Business, Feedback & Complaints: admin@asiaoceania.org

Website: www.asiaoceania.org
Tel: (65) 6472 3108
Add: 1 Commonwealth Lane #06-23
ONE COMMONWEALTH, Singapore 149544

Supported By


Copyright © 2025. All Rights Reserved. Conference Managed by Meeting Matters International