Leader
Italo Epicoco

The Research Unit aims to explore and develop cutting-edge Machine and Deep Learning techniques, along with their applications in climate science. 

The activities focus on: (i) investigating neural networks to develop foundation models for the ocean and atmosphere at weather and climate time scales; (ii) leveraging generative AI models for downscaling and data assimilation; (iii) analyzing the predictive capabilities of neural networks for extreme events such as windstorms, cyclones, wildfires, droughts, floods, and heat waves; and (iv) exploring the application of ML/DL techniques to assess the impact of climate change on society, the economy, agriculture, coastal regions, vegetation, land use, and marine ecosystems.

Leader
Donatello Elia

The main goal of the Research Unit concerns the design and implementation of Data Science open source solutions and strategies for addressing efficient access, analysis and mining of scientific data in the climate change and environmental domains. 

In particular, the activities focus on (i) the management of scientific data in the context of major international initiatives and research infrastructures like the ENES Climate Data Infrastructure, the Earth System Grid Federation, and the European Open Science Cloud; (ii) the definition of novel storage models and approaches to enable efficient organization, storage and access to climate data (including parallel I/O and cataloguing); (iii) the design and deployment of advanced Data Science environments for climate scientists in the frame of HPC/Cloud infrastructures, by leveraging High Performance Data Analytics solutions as well as machine learning frameworks to accelerate scientific discovery; (iv) the development of workflow automation tools and applications (e.g., digital twins) designed to support large-scale weather and climate use cases on top of HPC machines while targeting FAIR principles; (v) the set-up of innovative platforms and services to support end-to-end data management, including data collection from multiple sources (e.g., IoT, drones, EO, ESMs), advanced analytics (e.g., image processing, data mining) and results visualization (e.g., interactive dashboards).

Leader
Francesca Mele

This research unit’s objectives focus on analyzing and optimizing the main computational kernels featured in the models used at CMCC, specifically on hybrid multi-core CPU and GPU architectures. This includes assessing the impacts of optimized compilers, numerical libraries, and new parallel paradigms on the accuracy, reproducibility, and computational performance of numerical models. Additionally, the activities explore the impact of GPU-accelerated and exascale computing architectures on the numerical algorithms used in the primary climate models studied at CMCC. 

Specifically, the following aspects are analyzed in the context of GPU and exascale technologies: (i) leveraging advanced parallel algorithmic structures and specialized GPU hardware to mitigate communication overheads in “dynamical cores”; (ii) optimizing the management of large-scale “multi-model” and “multi-emission” ensemble experiments on hybrid systems; (iii) optimizing memory hierarchies and data locality across GPU and CPU components; (iv) rationalizing I/O operations; (v) employing new parallel communication paradigms and efficient synchronization mechanisms tailored to GPU-accelerated tasks.