Recent research demonstrates how automated calibration techniques are advancing hydrodynamic modelling, delivering faster, more reliable results to support informed decision-making in water management.

Hydrodynamic models are essential tools for simulating how water moves through environments such as rivers, estuaries and coastal areas. They are critical for decision-making in areas including flood protection, navigation, coastal development and environmental conservation. However, building reliable models requires inputting a wide range of parameters — many of which are difficult, costly or even impossible to measure directly across time and space.

‘Parameters like bed roughness or turbulence coefficients can vary greatly across landscapes and water bodies’, explains Clemens Cremer, DHI’s GenAI Lead and Expert in Data & Modelling. ‘It is not feasible, be it economically or for technical reasons, to measure all of them, their spatial resolution or temporal variability. Calibration allows us to fine-tune the models by adjusting these parameters in a physically reasonable range until the outputs align with observed data, such as water levels or flow speeds.’

The limits of manual calibration
Traditionally, this calibration has relied on manual trial-and-error by experienced modellers. While effective, it is time-consuming, can be subjective and increasingly impractical as models become more complex. For example, calibrating a model with 11 different roughness zones, as in DHI’s Elbe River case study, means tracking thousands of possible parameter combinations and their interactions.

This is where automatic model calibration is making a breakthrough.

DHI’s study shows that automated calibration techniques can systematically explore vast combinations of parameters, identifying the optimal settings to produce accurate, consistent model predictions. These methods can shift the focus of the modeller from tedious assessment of model results and adjustment of parameters for each model-run to more thorough thought about right metrics (what defines a ‘good’ model for a specific purpose) prior to starting the calibration. As an added benefit, autocalibration might obtain solutions human experts might miss — especially in high-dimensional parameter spaces.

Efficiency, flexibility and smarter insights
Clemens adds, ‘The benefits of automated calibration extend beyond just efficiency. Automated calibration can work with diverse data sources, from traditional water level gauges to satellite measurements, making it possible to create accurate models even in regions with limited in-situ monitoring infrastructure. 

‘They can also determine parameter sensitivity and provide insights into tradeoffs when multiple objectives (e.g. optimising for error in space or in time) are pursued. This makes it easier to align model outputs with client goals and regulatory standards. That’s a game changer for decision-makers in water environments.’

Join the conversation at IAHR 2025
Clemens Cremer will be presenting his research at the IAHR World Congress 2025, where DHI's representatives will be available to share insights, discuss ongoing projects and highlight recent technological developments. Find us at the Singapore EXPO, Hall 2A, Booth C4.

Follow DHI’s LinkedIn page for updates and coverage from the event.