Last year Keirnan Fowler and colleagues published a paper on simulating runoff under changing climatic conditions in Water Resources Research. The paper provides an interesting perspective on the ability to model climate change with the current generation of hydrological models and calibration techniques. We decided to ask Keirnan a few questions about himself and the paper.
Q: Where are you from, where are you based, and what are your research interests?
A: I’m from Melbourne, Australia and my PhD is based at the University of Melbourne. I’m interested in the way river catchments respond to changes in climate, and also in how we as a society make decisions amid uncertainty due to climate change and climate variability.
Q: What is the take home message of your paper?
A: In hydrology, models used to convert rainfall to runoff (rainfall-runoff models) often perform poorly when climatic conditions are different (eg. less rainfall) to the conditions they were originally calibrated to. Among hydrologists, this problem has led to questions such as, ‘what might be wrong with these models?’ and ‘how should we fix them?’. The key contribution of this paper was to demonstrate that existing rainfall runoff models are in fact more capable than previously thought. We showed that much of the apparent poor performance is actually due to poor or insufficient model calibration techniques, rather than the models themselves.
So the two take home messages would be ‘calibrate with care’ and ‘don’t judge your model too hastily’.
Q: You decided to start your PhD after working in industry for several years. Can you explain what motivated you to start a PhD, and does this relate to what is presented in your paper?
A: I used to work as a consulting hydrologist, a job that I really enjoyed. I had a number of role-models in consulting who had done a PhD, and the consulting projects I found most fulfilling were those that had a research element, so it was natural to consider a PhD. I was keen to contribute to cross-over between research and industry, as it seems an good formula for relevant research.
How does this relate to the paper? Well, I worked with many of these models as a consultant, and I had already noted their apparent problems. Indeed, the problems sent one of my projects way over budget! I went into the PhD thinking ‘surely someone has looked into this’. On reviewing the literature for myself, it was clear that a lot of people had noted the problem, but there was plenty of room for diagnosing the contributing factors.
Q: In the paper you identify the suitability of models and their parameter sets to simulate changing hydrologic conditions of the past. Is it possible to use your work for better understanding if models are suitable to simulate the response to climatic conditions that have not yet been observed in the past?
A: Yes, although the paper is only the first step. My home country of Australia has recently had some really severe droughts, and I’m using these droughts as a ‘test case’ for climate change, hoping to learn how to make our modelling more robust. Ultimately, the ‘end-user’ for my PhD research would be a modeller faced with climatic conditions beyond what has been observed in the past (particularly, conditions drier than the past), and thus, a shortage of relevant calibration data. The idea is to develop methods that are tailored to this kind of data-poor situation.
Q: In your paper, you have focused on simple conceptual hydrological models to simulate hydrologic change. Is your work also relevant for models with many more parameters (e.g. landsurface models/distributed models)?
A: Yes, it is. Regardless of which model we use, we somehow have to choose which set of parameters (or ensemble of such sets) is the ‘best’. A key lesson from my paper was that, in data poor environments like the one I just described, we may sometimes fail to choose the parameter sets that perform best when extrapolated to different climatic conditions. The resulting poor performance gives us an unfairly negative view of the capabilities of the model. Unfortunately, this problem isn’t helped by more parameters, so yes, it’s highly relevant to distributed and landsurface models.
[END] – Interviewed by Wouter Berghuijs
This interview is part of the new YHS Research “Hylights” series to showcase interesting and outstanding work by early career scientists. Selection criteria are not set in stone, but reasons to select work can include e.g. novelty and relevance of findings, fun of reading, unique collaborations, media coverage and generated controversy. Selected work will be provided with a short layman summary, and a short written or video interview with the (first) author(s). Tips can be sent to young email@example.com or firstname.lastname@example.org.
Reference paper: Fowler, K. J. A., M. C. Peel, A. W. Western, L. Zhang, and T. J. Peterson (2016), Simulating runoff under changing climatic conditions: Revisiting an apparent deficiency of conceptual rainfall-runoff models, Water Resour. Res., 52, 1820–1846