There are revolutions that make no noise. No visible upheaval, just a richer flow of data, a rewritten code to be more efficient, a simulation that runs faster — invisible, yet decisive. This is the transformation underway at the heart of hydrology, where the power of High-Performance Computing (HPC) meets hydrological models in a technological alliance that will progressively change the way we understand and model the water cycle. This is also where the IT-WATER project fits in, with one goal: to provide a consistent, high-resolution representation of the water cycle across the national territory — for yesterday, for today, and above all, for tomorrow.
As part of the project funded by the National Recovery and Resilience Plan (PNRR) and led by CIMA Research Foundation in collaboration with Fadeout Software Srl and the ItaliaMeteo Agency, a significant step forward is being taken in hydrological modeling: bringing the operational models, developed by CIMA Research Foundation for the Italian Civil Protection Department, onto High-Performance Computing (HPC) infrastructures such as those of CINECA. It is a methodological and technological transition that requires a profound rethinking of model architectures but enables a fundamental leap in scale for water cycle simulation.
A matter of computational power
In the language of computational science, High-Performance Computing (HPC) refers to the set of technologies that allow for the execution of complex, highly intensive operations in very short times. It’s not just about faster computers, but about distributed systems 1 capable of managing millions of operations simultaneously, from multiple users and across multiple parallel processes.
In such environments, workloads must be optimized, scheduled, and controlled. Every simulation has a priority, an assigned runtime, an execution queue. This imposes a deep transformation on scientific models: it’s not enough to port a code to a new system — the architecture must be rethought, procedures modularized, and stability and scalability ensured.
This is the challenge that CIMA Research Foundation focused on within the IT-WATER project: preparing its models to interact with supercomputing efficiently and with future reusability in mind.
This transition was also made possible thanks to the role of the ItaliaMeteo Agency, a strategic partner of the project, which facilitated technical and operational dialogue with the CINECA consortium and helped ensure alignment between modeling needs, computational capabilities, and the operational goals of the system. The collaboration with ItaliaMeteo helped to enhance IT-WATER’s integrated approach, fostering effective interaction between scientific modeling, computing infrastructures, and applications of national interest.
Inside the water cycle, at the national scale
The water cycle is one of the most complex systems to represent scientifically. Rain, snow, evaporation, infiltration, surface runoff, interaction with soils: every component is influenced by climatic, environmental, and morphological variables in a system (the Earth) that we do not fully understand and with which we interact in complex, multi-scale ways. Hydrological models have been trying for decades to describe this system, and CIMA Research Foundation has developed two operational tools in this context: the cryospheric model S3M and the hydrological model Continuum, both already operational on a national scale.
“Until recently, these models were executed on traditional, even very powerful, computers, but with intrinsic limitations in terms of data volume and execution time,” explains Francesco Avanzi, from the Hydrology and Hydraulics Department of CIMA Research Foundation. “Today, we can talk about parallel simulations, on a national scale, leveraging the CINECA supercomputing infrastructure. This allows us to work on long time horizons, from 1980 to 2070, and with a spatial resolution of up to 200–500 meters. It’s a new methodological paradigm.”
A step forward based on a complex reengineering effort 2 . The entire structure of the models and accessory tools, such as the Python package shybox developed within the IT-WATER context, has been rethought to operate in distributed environments. “To truly harness the potential of supercomputing, we had to thoroughly rethink how our models are executed,” explains Fabio Delogu, also a researcher in the Hydrology and Hydraulics Division at Fondazione CIMA. “We adopted a container architecture 3, which not only allowed the models to run on HPC systems but also improved their modularity and reusability.”
“In fact, it’s not just about speed: the modularity of containers allows us to build more reusable models, which can also be executed independently by other research groups simply by transferring ‘images’ of the system to CINECA. This openness represents a significant shift for the scientific community. Models are no longer closed tools, tied to local configurations, but integrable components, ready to be shared, adapted, and used in different contexts,” adds Alberto Tasso of Fadeout Software Srl.
Simulating the future: climate scenarios and water resource management
Simulating the future starts with understanding the past. Thanks to the new HPC framework, the IT-WATER project enables the generation of historical and future climate scenarios for water resources across the entire Italian territory. We’re talking about more than forty years of history, and just as many years of projections, for every hydrological basin in the country. But it’s not just about data: the goal is to produce useful tools for the operational and strategic management of water resources, to provide institutions, authorities, and citizens with scientific tools to understand water availability across Italy.
“We’re working on deriving hydrological impact indicators related to drought,” says Avanzi. “It’s a crucial aspect of adapting to climate change. Having models that can realistically reproduce the behavior of water, in its various forms, helps support decision-making and planning.”
An infrastructure for the future
IT-WATER is also an investment in the future of operational hydrology in Italy. The combination of scientific modeling, supercomputing, and open infrastructures forms a platform ready to be extended and maintained over time. The results achieved today will be the foundation for new applications, new simulations, and new research questions.
- A distributed system is an architecture in which computational work is not performed by a single computer, but is divided among multiple computing units, often located on different nodes of a network. These nodes collaborate to simultaneously execute different parts of a complex process; in this way, it is possible to speed up processing and handle large amounts of data. ↩︎
- In the scientific and technological field, reengineering refers to the process of rethinking and deeply restructuring an existing system, with the aim of improving its performance, efficiency, and adaptability to new contexts. In the case of CIMA Research Foundation’s hydrological models, reengineering meant modifying the code structure supporting the model, data management, and execution modes to make them compatible with the supercomputing environment. It is therefore not just a matter of updating software, but of adapting its architecture — algorithms, input/output procedures, memory management, and parallelization — so that it can operate stably and efficiently on complex infrastructures such as HPC systems. ↩︎
- Container architecture is an approach to software execution that allows models, libraries, and configurations to be packaged within isolated and portable environments, known as “containers”. These environments ensure consistency and reproducibility of execution across different infrastructures, including high-performance computing systems, making the models more modular, scalable, and easily reusable in the scientific domain. ↩︎