Task 3. How well can we model the flood record distributed across Cumbria?
The first modelling task was to produce credible models of the observed records of river discharge across the three catchments; so that partners and wider stakeholders have trust in the way that we have captured and reduced modelling uncertainties (see e.g., Beven and Lane, 2022 and Beven et al., 2022 e14703).
Specific tasks:
To demonstrate that Dynamic-Topmodel is able to simulate historic sequences of observations in a physically reasonable way we simulated observed sequences of flood events (including low-flows) at distributed 2m resolution across the 2287 km square Eden, 663 km square Derwent and 209 km square Kent catchments in Cumbria. Dynamic-Topmodel's parsimony permits millions of model simulations to be run on Lancaster University's high-performance computer platform. This high number of simulations allows uncertainty estimation based on the GLUE limits of acceptability framework. The acceptable simulations have been identified by rejection of simulations that are deemed not physically reasonable given multiple-criteria developed from local observations (Beven et al., 2022 e14703).
At an early stage in the Q-NFM project, we saw that we needed a radical improvement to the way the Dynamic-Topmodel solves surface-subsurface water routing quickly enough, even when using Lancaster's supercomputer. This involved Q-NFM developing a new version of Dynamic Topmodel involving a considerably faster implementation. This new version of Dynamic Topmodel (funded by NERC Q-NFM) - dynatop v0.2.3 we made publically available.
We further modified the catchment modelling approach by using Dynamic Topmodel only to route rainfall through hillslopes to segements of perennial channels, and then using a different model to route the water along the channels and in/out of floodplains, in a hybrid approach. For the channel and floodplain routing we used either JFLOW or HEC-RAS. The first Q-NFM study to adopt this approach was Hankin et al. (2019).
Each of the acceptable simulations identified in this way (and associated parameter sets) then became the baseline ensemble for each of the catchment areas studied and were subsequently modified to represent NFM interventions in Task 4. As this detailed constraint identifies an ensemble of physically meaningful simulations, modifications to the parameters representing NFM interventions should be credible to the international academic community, local stakeholders within the 3 test basins, and nationally. These modifications were made by parameter shifts as detailed in Task 4.
