Climate Futures

About Climate Futures

Climate Futures is a Centre for Research-based Innovation (SFI), funded by the Research Council of Norway. The Centre develops climate forecasts up to ten years in advance, with the aim of improving the management of climate-related risk.

The climate crisis is considered one of the greatest challenges of our time, yet the understanding of climate risk remains insufficient. Climate change leads to increasingly unpredictable weather events, with significant consequences to biodiversity, settlements, the economy and food supply. Addressing the enormity of these challenges requires collaboration across sectors.

Climate Futures brings together around 40 partners from research, the public sector and industry to develop knowledge, methods and practices for managing climate risk.

NR contributes primarily through methodological development in statistics and machine learning. This includes the development of robust models for hydrological processes, downscaling weather data to local conditions, and methods for combining numerical weather models to produce more accurate forecasts.

Modelling streamflow in a changing climate

A central question in Climate Futures is how river streamflow will evolve as a result of climate change.

The relationship between precipitation, temperature and streamflow is modelled for both monitored and unmonitored river basins. These models are then applied together with scenarios from international climate models to analyse how streamflow may develop over time.

When models are estimated separately for each basin and month, they may become overfitted to historical data and produce unrealistic results. To address this, key factors are estimated jointly using neural networks, while preserving an interpretable model structure. This results in more consistent and scalable models across catchments and climate zones.

The work includes both traditional hydrological approaches and data-driven models, including Long Short-Term Memory (LSTM) models for time series. Comparing different model types provides insight into how data availability, time-series length and observational coverage influence predictive performance.

Such knowledge is particularly relevant for forecasting, power planning and the operation of hydropower systems.

Downscaling weather and climate data

Global weather and climate datasets often have insufficient spatial resolution for use in local analyses.

At NR, methodological development focuses on statistical approaches for downscaling temperature and precipitation from global datasets to station level. These methods preserve both natural variability and uncertainty in the climate system, providing a stronger foundation for analysing local water resources and energy systems.

Weather forecasting and event prediction

Combining multiple numerical weather prediction models can produce more accurate forecasts than individual models alone. At the same time, such combinations require systematic methods and rigorous evaluation, particularly when the aim is to predict the timing of a specific event.

We develop and evaluate methods for combining weather forecasts, with a particular focus on time-to-event prediction. Improved time-to-event forecasts strengthen decision-making in climate-sensitive sectors. One example is estimating the timing of the first frost, which often marks the end of the growing season and may lead to significant crop losses if it occurs unexpectedly.

To learn more about Climate Futures and our role in the centre, get in touch.