“Model rejection is a good thing” according to Keith Beven at the BHS Symposium

Duncan Faulkner, Katie Burnham and Barry Hankin talk us through their experiences at the British Hydrological Society’s biennial symposium that took place last week with the theme of ‘Hydrology: Advances in theory and practice’.

Duncan Faulkner – Head of Hydrology

JBA Jon Whitmore BHS poster 2018
Jon Whitmore BHS poster 2018

Every two years the British Hydrological Society holds a symposium: an opportunity for hydrologists from the UK and beyond to get together and talk about the things they love most: too much water, too little water, why we need more measurements and how wrong someone else’s model is.

Attracting a good cross-section, attendees include academics and practitioners from both the public and private sectors. This year’s packed programme, which was cut from three days to two, meant there was no time for field visits this time round. Not even to the Regents Park lake where I watched a moorhen carve a path through the pondweed like an icebreaker ploughing through the Arctic ocean.

Model rejection is a good thing

A memorable moment for me was Keith Beven’s opening address on how to make advances in hydrological modelling. Keith’s answer was that we need to learn that model rejection is a good thing; that’s the way that we can advance our science. We need to improve our representation of hydrological processes so that we get the right results for the right reasons.

That really resonated with me, having recently seen some calibration results from a model that simplified runoff processes into overland flow only. It calibrated surprisingly well, but I was left wondering if this representation would be suitable for its intended purpose: testing options for Natural Flood Management (NFM).

Throughout the event it was great to hear such a range of ideas and approaches. We had everything from the latest hydrometric measurement techniques (using cosmic rays, mobile phones and drones) to good old-fashioned hydrological common sense. Dave Cameron, Principal Analyst at JBA, described how a bucket can provide a much lower-cost method of flow measurement using the example of two small ditches from which a Scottish distillery was planning to abstract water.

Learn from your mistakes

Keith’s theme of advances in hydrological modelling re-emerged towards the end, with David Hannah suggesting that we need a ‘Journal of Hydrological Failures’ where we can publish null results and rejected ideas.

That made sense to me. Falsification is the way to go. I have learned more from my failures than my successes over the years. This is what I tried to bring across during my talk on my experiences of spending 11 days in the witness box during a prolonged court case in Dublin about the role of hydroelectric dams in flooding of Cork.

It’s easier to identify lessons; harder to learn them.

Ewan Larcombe, who has been flooded three times from the River Thames, tore into the Environment Agency who he said had not learned from past failures and reviews.

David Boorman from the Centre for Ecology and Hydrology (CEH) asked whether hydrologists have embraced post-truth. I wonder whether our customers, clients or stakeholders want rejected hypotheses, nuanced answers and statements of uncertainty. Or are they hunting for ‘biddable hydrologists’ who won’t get in the way of decision-making in organisations dominated by a managerial culture with a focus on ‘delivering outcomes’.

Water-related needs

In the midst of all this soul-searching and some fancy statistical footwork by other presenters, I was glad to hear that the prize for the best presentation went to Olivia Cooke from the University of Bath, whose research deals with storm water runoff from South African slums.

It was good to be reminded of the immense water-related needs and vulnerability in less fortunate parts of the world. I can’t help feeling that hydrologists in countries like the UK should direct more of their effort to addressing those sorts of problems.

Katie Burnham – Assistant Environmental Modeller

Katie Burnham BHS habitat modelling JHab talk 2018

Overall it was some really great two days. This being my first time attending the BHS symposium I felt welcomed by all, a credit to all the people who were involved in making the symposium happen.

The quality of the talks across both days was high. I managed to get to see all bar one of the JBA speakers.

Read the conference abstracts on the British Hydrological Society website. 

How do you use your model results?

Day one started off with a really great talk from Keith Beven on ‘How to make advances in hydrological modelling’. He highlighted some great points on model calibration and validation that should make everyone think twice about the way in which we use our model results.

To end the final session of the day, David Gilvear gave a great talk on behalf of the work he has been involved with around the ‘Dispersal of augmented gravel, to increase a salmonid dispersal of augmented gravel in a boulder-bed river draining Dartmoor’. It was really interesting to hear how they were using Radio-frequency identification (RFID) tagged particles to monitor the sediment augmentation, which is something that could be used to validate sediment transport modelling.

Refreshing presentations

Both Duncan and Barry did a great job of presenting on day two.

Barry’s presentation looked broadly at the ways we can improve our hydraulic modelling methods. Really clear and well explained, the presentation covered a breadth of work JBA has been involved with.

It was great to see the effort involved to really capture small scale hydrological changes at the catchment level through their fieldwork up at Tebay Ghyll. Hearing about how they have been incorporating Dynamic TOPMODEL into both JFLOW and HEC-RAS-2D was also very interesting to learn.

As for Duncan’s presentation, it was refreshing to have a talk that was light hearted and not technically driven. He told a great story about his experiences working as a technical expert on the Cork 2009 flooding court case. It really gave an impression of his vast experience and showed how this type of project differs from the average day to day of a consultant.

Two interesting back to back talks on NFM followed:

  • Mark Huband – ‘Estimating effectiveness of Natural Flood Management – keeping it simple’
  • Andrew Black – ‘Landscape-Scale Change and Hydrological Response: Devils in the Detail’.

Both took quite a different approach, with Mark focusing on the validating approaches from a Finchingfield case study through modelling results, and Andrew using pre and post change monitoring from the Eddleston catchment.

Barry Hankin – Head of Environmental Modelling

Barry Hankin presenting at BHS conference 2018

As Duncan and Katie have mentioned, Keith was on great form with his opening presentation. He has long recommended advancing hydrology through ‘Popperian-rejection’ and so I thought his theme on ‘models of everywhere’ and working with communities to improve them, was worth exploring more.

It would be interesting to hear his thoughts on the challenges to what quality scores we should put on different sources of information. Quite a lot of text books think the future will be much more spatial and therefore GIS based (not necessarily pattern-matching big-data), so we do need to think of better ways of proving distributed models with different qualities
of information.

This merges into Dave Boorman’s talk of post-truth, although I’m surprised Keith did not say his whole GLUE approach provides a framework for handing partial truths. A whole branch of maths was developed called fuzzy logic to handle partial-truths, which I think our own JHab uses.

In my talk I combined Dynamic Topmodel with JFlow and HEC-RAS-2D for the first time, to get the best of both worlds for hillslope hydrology and hydraulics – which captured more of those processes that Duncan mentions. Exploring this in an uncertainty framework, I demonstrated why more micro-catchments like those in the Q-NFM project are needed to tie down uncertainties in hydrological parameters (….and reject some bad models).

Want to know more?

Email Duncan Faulkner for more information on our work in hydrology. You can also find out more on our flood and water management web pages.

Leave a Reply