An international snow model study has shown how scientific and human errors can limit progress.
The intercomparison study – led by a team at the University in collaboration with twenty-one modelling groups from around the world – identified that lack of scientific advancement since the last snow model intercomparison was partly due to mistakes made by modellers.
Model intercomparisons are projects that allow models from all over the world to be compared against measured data and against themselves in order to drive progress in all fields of climate modelling.
Cecile Menard, Lead author from the School said:
“The role of human errors in snow modelling – or in other fields of Earth System modelling – has never been the focus of a scientific publication because acknowledging errors is not part of our scientific practice or culture.”
The authors introduced the concept of the “False Hypothesis”, an analogy with the “False Protagonist” used in films and literature, in which the character assumed by the audience to be the main protagonist is killed off half way through the story (the paper cites Ned Stark in Games of Thrones and Marion Crane in Psycho as famous examples).
In the study, the “false hypothesis” is the authors’ working hypothesis which was inherited from previous snow model intercomparison projects but failed.
By “killing it off” at the start of the paper, the aim of the study is revised and it becomes an investigation into why the working hypothesis failed. Using quantitative and qualitative, they show that this failure is not due to models but to modellers.
The researchers say human mistakes, like in many other fields, are expected (“To err is human”): the models used to describe and to understand snow processes are complicated models with tens of thousands of lines of code that describe physical processes but also procedures for the code to be run smoothly.
It is the first study explicitly to address this issue and to quantify the errors. Other limitations, such as the type of data that can be collected, are preventing snow models from evolving as quickly as other models in scientific fields relying on technology.
The research concludes that new data is needed for a new generation of models.
The research has been published in Bulletin of the American Meteorological Society (BAMS) and is supported by the Natural Environmental Research Council.