Introduction

This dataset was generated to support a model-based assessment of how variability in hydropower inflows affects the long-term European energy system, with a focus on the power sector. It provides machine learning–enhanced daily hydropower inflow time series [MWh] representative of each country's run-of-river (ROR) generation profile.

The inflow estimates are based on river discharge data from the SMHI European hydrological model E-HYPE and cover the following climate scenarios: RCP2.6, RCP4.5, and RCP8.5.

The dataset is provided at the national level for the following countries:
Austria, Belgium, Bulgaria, Croatia, Czech Republic, Finland, France, Germany, Hungary, Italy, Latvia, Lithuania, Norway, Poland, Portugal, Romania, Serbia, Slovakia, Slovenia, Spain, Sweden, and Switzerland.

• Time period: 2020-01-01 to 2071-01-01

• Temporal resolution: Daily

EURO-CORDEX Ensemble Members:

• ICHEC-EC-EARTH_r12i1p1_CCLM4-8-17

• ICHEC-EC-EARTH_r12i1p1_RACMO22E

• ICHEC-EC-EARTH_r12i1p1_RCA4

• ICHEC-EC-EARTH_r3i1p1_HIRHAM5

• MOHC-HadGEM2-ES_r1i1p1_RACMO22E

• MOHC-HadGEM2-ES_r1i1p1_RCA4

• MPI-M-MPI-ESM-LR_r1i1p1_RCA4

• MPI-M-MPI-ESM-LR_r1i1p1_REMO2009

• MPI-M-MPI-ESM-LR_r2i1p1_REMO2009

This dataset was produced as part of the CLINT project (see below).

Data and methodology

A supervised machine learning regression model (XGBoost) was used to learn the relationship between river discharge (input features) and ROR hydropower generation (target variable) at the country level.

• Feature data: Daily river discharge for 35,408 sub-basins across Europe, derived from the E-HYPE model. The training period spans January 1, 2015, to January 31, 2023. Climate projections span 2025–2100 under three RCPs.

• Target data: Hourly ROR generation in MWh, aggregated to the national level, sourced from the ENTSO-E Transparency Platform for the period 2015–2023.

During model training, the discharge data for all sub-basins within a country were aggregated and used as features. The national ROR generation served as the target. XGBoost's feature importance metric was used to identify the sub-basins most influential in predicting hydropower generation for each country.

A detailed description of the methodology and data processing steps is available in the public deliverable:
D6.2 – Preliminary report on AI-enhanced Climate Services for extreme impacts.

CLINT project

CLINT – CLImate INTelligence: Extreme events detection, attribution and adaptation design using machine learning
Grant Agreement ID: 101003876
Project duration: July 1, 2021 – October 30, 2025