Some of Prof Tom Coulthard‘s and my own research has just been published as a discussion paper in the European Geoscience Union’s Journal – Earth Surface Dynamics. It’s my first proper open-source paper, so this release is not yet peer reviewed but will be reviewed in the same way with anonymous reviewers. However, it is also open to anyone to make comments (but these are public, so no hiding). After review, and edits, hopefully it will be published fully later in the year.
© Copyright George Tod and licensed for reuse under this Creative Commons Licence.
Here is where I try to write a ‘plain English’ summary of the work and the backstory. The work was conducted as part of the Natural Environment Research Council funded project, Flash Flooding from Intense Rainfall. The project hopes to improve our ability to forecast the intense, rapidly forming, but small and short-lived thunderstorms which can trigger flash flooding in the right conditions. We want to be able to predict their occurrence better and also understand the conditions required for flash flooding. We (Tom and I), in particular, look at the erosion and deposition which occur during the flash floods.
The computer model we use (CAESAR-Lisflood) was only able to use an input of rainfall which is averaged out over the whole area covered. These areas can be quite large, and as you probably know, if it’s raining in one part of the town you live, it might not be raining over another part. With the storms we are looking at they exist at a scale often much smaller than a whole river catchment, so that intensity is smoothed out by the model. This will likely reduce local river flows (in the model) and consequently reduce the amount of material (rocks, stones, mud etc) moved around (in the model).
Clearly, we needed to add the ability to represent rainfall in much greater detail, so I came up with a plan and arranged to meet with Tom to discuss how I was going to build this into to the computer code. I sat down with Tom and told him my plan, and in typical Tom fashion he tells me “I’ve already done this, I’ll send you the code”.
This single sentence saved me several months of coding and debugging and banging my head on my desk.
We used rainfall records taken from the MetOffice’s archive based on weather RADAR measurements. For the River Swale catchment (the catchment of choice for testing CAESAR-Lisflood), this data was available in grid squares of 5 km x 5 km, and recordings every 15 minutes. We wanted to test how the model reacts to the same rainfall data but applied in different resolutions, so we averaged out this data to various resolutions, both spatially (5 km, 10 km, 20 km and full catchment), and temporally (15 min through to 24 hours).
Incredibly, it made a big difference, with the best resolution (5 km every 15 minutes) moving over twice as much material as the worst (Full catchment every 24 hours) in some cases! We then looked at the longer term impacts by repeating our rainfall record (but jumbling up the locations at the end of each ten year cycle) for 1000 years (in the model). This showed that using the best resolution rainfall instead of the worst predicted more erosion in upland areas, and more deposition in lowland areas – this has implications for studies looking at the long term development of landscape that often use averaged rainfall records which miss out this detail.
This is because of the relationship between the discharge of a river (the amount of water flowing past a point in a specified time) and the amount of material moved is disproportionate. We called it ‘non-linear’, in that a small increase in the discharge results in a big increase in material moved – by representing the rainfall in greater detail, the model focusses it over a smaller area for a shorter amount of time, increasing the discharge in that section of the river.
The research also highlights the need to consider how our rainfall is likely to change with climate change. Often, only the overall change in volume of rainfall is considered but if this is in the form of frontal rain which covers large areas over long periods, the rain is low intensity and will unlikely cause flash flooding or move much material. If we are to expect an increase in the intense thunderstorms then we can expect our rivers to become more active in the future – the implications of which are as yet unknown.
The paper is free to read, so does not require a subscription, and can be viewed here.