Modelling water
In the second week of training for the flood risk project we learnt how to model the passage of a river through a landscape. Using a sophisticated piece of software we were able to play around with different parameters that changed the shape and flow of the river. Playing god in silicon we flooded and saved a virtual valley.
Complex computer models lie at the heart of flood risk mapping in the UK. They are central to efforts to predict future inundations, to decisions on where to site defences and, perhaps most controversially, as to whether or not your house is eligible for insurance. There are a number of different models on the market, which have been developed by academics and consultants and there is a fair degree of competition between them.
In an extended series of sessions, Stuart Lane – the hydrologist on the team – introduced us social scientists to the mathematics of hydrological modelling. We learnt how to turn fluxes of water into equations, comprising Greek symbols and other elegant hieroglyphics. We worked our way through Newton’s laws and were regaled with the specific material properties of H2O. Standing on the shoulders of past hydrological heroes (and heroines) we formulated differential equations that claimed to simulate the conservation of mass and momentum, that accounted for turbulence in a body of water and acknowledged the effects of the roughness of a river’s channel.
Water appears to be a fundamentally complex material that has different dynamics over three dimensions and through a variety of spatial and temporal scales. The secret to modelling is to simplify this complexity. The model we used reduced three dimensions to one and lumped the range of variables that effect the roughness of a channel into the ‘Manning’s coefficient’ – a constant devised by an Irish water engineer at the end of the nineteenth century.
It is useful to understand models as conceptual representations of reality. The Greek and hieroglyphics can be understood as an hypothesis of how the landscape works – like a photograph or painting, they depict the modeller’s view of the form and dynamics of a place. However, unlike a painting, the representations they embody can be tested empirically. To test a model you need to compare its predictions against data gathered from the field on river form and dynamics and past flood events. If the data match what is predicted then you can have some confidence in the model, if not then you need to re-examine your equations.
In practice the standards for verification are messy and contested. Modelling emerges more as a process of tinkering, or moving back and forth between data and representation, cobbling together equations and varying constants to reach a likeness. I pictured a scene from a hair-brained Heath Robinson drawing, where the pre-War car-boot sale miscellanea he favoured has been replaced with other people’s χ’s, γ’s and λ’s. In place of the elegant continental heuristics of French hydrological theorists, we have the can-do fudging of British engineering. Equifinality – or the existence of mutually incommensurable solutions to models – is a common phenomenon and its resolution draws on tacit experience not easily formulated into equations.
Having flown through the theory we shifted to the computer and had our first encounter with a software encoding of what we had just covered. A consultant from a modelling company introduced as to HEC-RAS – a free hydrological model that has been developed by the US Army Corps of Engineers. For those of us familiar with Windows it had a fairly intuitive interface and we soon had the water flowing and we worked our way through two exercises. The first of which was to calibrate the model encoded in the software with some real data. This involved shifting constants to try and match the model to what had been observed – we tweaked and tinkered until the lines of our graphs converged.
The second exercise was more satisfying and involved simulating different interventions to try and prevent a bridge overcharging on a fictional Beaver Creek in Oregon. Bridge overcharging occurs when water from a flooding river builds up behind a structure until it breaks over the top. This is dangerous and often causes structural damage. The options on the front page of the software guided us towards some hard engineering solutions and we were soon digging ponds, building embankments and creating offline storage space for excess water upstream of the bridge. I dug a giant pond on the floodplain and running simulations I could follow a surge of water down valley and identify whether or not the bridge was saved. This provided a fairly graphic visualisation – no one ran for the hills and no arks were built – but you got a sense of the strength and visceral surging of a flood. The model was clearly a powerful tool that could give planners a great deal of confidence in planning building, land management and interventions.
This exposure to modelling was a thought-provoking experience and there were a number of issues that struck me. The first, and perhaps most uplifting of these was to understand the esoteric language of hydrological equations not as the exclusive code of an elite group of experts but as a finely developed expression of a whole new way of seeing and thinking about landscapes. In some ways we can understand this ‘thinking like a river’ in Greek and hieroglyphics as an expression of a deep-seated and sincere ethical sensibility. This sensibility is grounded in a well-honed sense of curiosity and a desire to witness and convey the complex dynamics of nonhuman processes. It represents the apotheosis of a way of learning to be affected by a landscape and of tuning into to its ebbs and flows.
The second issue which stuck me related to the separation I felt at the computer from the real landscape I was supposed to be modelling and altering. Playing god from an aerial viewpoint I felt enormous power and irresponsibility while digging lakes, flooding farmland and inevitably altering people’s lives. The software successfully enabled the God-trick of mapping, which divorces the observer from the lived reality of the places and peoples they are overseeing. Equipped with this device a modeler could have enormous power to change the risks and opportunities provided by a landscape, not to mention its form and habitability for things other than humans.
In some ways such encoded models can be understood as what the French sociologist of science Bruno Latour has termed black boxes – technologies whose internal logics are effaced once they have been constructed – which have real material effects on the worlds they are applied to. Stuart was refreshingly open, humble and uncertain about the models he introduced us to. He provided a rare and generous insight into these black boxes and in shedding light on their internal operations and practical operation he opened a valuable interdisciplinary space for new and more democratic ways of engaging with flood risk modeling. From here we aim to follow models and modelling in practice to find ways of opening them out to those they affect.
