When AI Sees What Humans Can’t: Predicting Colorectal Cancer Risk With AI

Imagine an agricultural analyst trying to anticipate how droughts, heatwaves or extreme rainfall might impact next season’s crops. Existing climate datasets - from CMIP6, EURO-CORDEX, and DestinE - offer powerful insights, but their resolution is too coarse for precise sector-specific applications in agriculture and insurance. What if AI could sharpen the view?

This RI-SCALE use case leverages Convolutional Neural Networks (CNNs) - AI models designed for 2D pattern recognition - to create application-ready high-resolution maps. AI-based downscaling techniques are used to create and validate high-resolution climate datasets (O(5km)m) from coarse resolution simulations (O(100)km) and assess their usability and value in real-world agriculture and insurance scenarios.

The result? More localised climate risk assessments and new generation of climate products that can support business models across the value chain, from Data-as-a-Service offering to operational climate analytics.

RI-SCALE unites Europe’s leading climate datasets with powerful AI-ready compute to make high-resolution climate mapping possible. Through its Data Exploitation Platform (DEP), researchers can access massive ESGF, DestinE, and Copernicus datasets directly in place and use scalable GPU resources to train and run CNN-based downscaling models. With this unified environment, with shared tools and sector-driven validation, RI-SCALE turns climate projections into field-level insights for operational use in agriculture and risk analysis. The DEP takes advantage of itwinai, HPC-powered ML workflows platform, to combine four super-resolution convolutional neural networks. Each of these networks performs downscaling by a factor of two. When the trained pipeline is assembled in series through itwinai, impressive downscaling from 100km to 5km resolution can be performed with ease.

This use case demonstrates how RI-SCALE bridges the gap between climate simulations  and diverse users of climate products transforming environmental research infrastructures into drivers of AI-enabled innovation for many industries.

Once validated, these ultra-detailed climate datasets can open new opportunities of exploitation across the value chain and support better risk-based crop planning, more accurate insurance products and new opportunities for Data-as-a-Service models across the environmental domain.