Work Package 5 – Hydrology

At the national scale, this work focused on two overarching questions: i )What happens when we join-up data and modelling nationally? and ii) What would this deliver in terms of national capability? The approach was to bring together, for the first time, national datasets, models, and uncertainty analysis approaches into cloud-computing environments to explore and benchmark the current predictive capability for hydrology nationally. The challenge was to manage and visualise these complex results to synthesise and improve scientific understanding. Key to this was providing a framework to evaluate which model structures emerged as the best predictors of stream flow across a diverse range of catchment characteristics.

The solution

WP5h_flyer_coverThis work has provided the first UK-wide benchmarking of hydrological predictions using multiple-model structures and including prediction uncertainty. This allows the quantification and visualisation of the uncertainty within model predictions for both high flow and low flow simulations, and how this varies for different catchments across the UK and at different times of the year.

These results were further developed into a web-based portal to aid visual interpretation of the results, and to show how such capability could be made available at national scale using cloud-computing resources.

Benefits for the future

The benefits of this work include –

  • Improving model diagnostics for streamflow predictions by explicitly taking into account the uncertainties in the observed data products and hence data quality. This will allow better models to be built in the future.
  • Identifying links between model structures and parameters to catchment characteristics than can aid regionalisation and predictions of stream flow in ‘ungauged’ basins.
  • Quantifying the prediction uncertainties for different flow magnitudes and different times of the year to effectively benchmark our national stream flow modelling capability, and to target catchment typologies that need further model refinement to improve such predictions.
Evaluation of four separate model structures from an uncertainty analysis framework reveals that best sampled individual model performance is different in different regions and no one structure is dominant across the diversity of UK catchment typologies. Red is “good”, blue is “poor” model performance.

Evaluation of four separate model structures from an uncertainty analysis framework reveals that best sampled individual model performance is different in different regions and no one structure is dominant across the diversity of UK catchment typologies.
Red is “good”, blue is “poor” model performance.

 

WP5h_diag_02

Super ensembles (from multiple model structures) to quantify seasonal predictions of stream flow across the UK. Such results can be used to understand where model dynamics need further improvement (i.e. summer low flows in the South East).

Where to next?

A next step is to dynamically couple hydrology and biogeochemistry across the geoclimatic landscape of the UK to improve predictions of water quantity and quality, for scientific and policy support. This would provide an improved national simulation framework for land use, mitigation measures, and climate change scenarios (that include prediction uncertainties) to be quantified .