On the life and death of hydrologic models
by Marie-Amélie Boucher, a HEPEX 2015 Guest Columnist
During the 1950-2000 period, a very large quantity of hydrologic models of all varieties were created due to a rapid increase of computational capacities (Roche et al. 2013).
Now, on the one hand, although the creation of entirely new models seems to have slowed, it has not entirely stopped. On the other hand, some models developed during this 1950-2000 period are slowly dying because of the retirement of their original developers.
Are they worth saving or not? Are there already too many hydrologic models out there? Can obsolete hydrologic models be adapted or transformed to better suit new operational and research needs?
Are there too many hydrologic models?
In 1999, the U.S. Bureau of Reclamation, in collaboration with universities and other partners, published the « The Hydrologic Modelling Inventory ». The aim is to centralize all the information regarding a very large quantity of hydrologic models in one list published on the web, all sorted according to different categories.
However, I found that the models I use most often are not included in the inventory so I guess that many other hydrologic models are still missing from that already impressive list. Of course, as HEPEX participants, we should be happy to see such diversity. If you want to perform multi-model ensemble forecasts, you will be well served. Although I am not fit to judge that, I can’t help but wondering: are there already too many hydrologic models out there?
I guess that every creator of a hydrologic model has good reasons to advocate that his or her hydrologic model is the best, or at least the best for this particular context. There already exists some comparative studies between hydrologic models (see for example this website, or this one and this article). However, as you can see, these studies most often compare a very small subset of models compared to the extent of all possibilities.
While some models die, some are born
Many hydrologic models depend on one person or a small group of persons for maintenance, update and support. In this post, it is explained that HBV will survive the retirement of its creator. This is not always the case. For instance, the distributed model CEQUEAU was abandoned for ten years or so after its creator, Pr. Guy Morin, retired from the Institut National de la Recherche Scientifique.
So, after some time of being un-supported, the model became somewhat obsolete because it didn’t follow the improvements of hydrologic sciences. Most of all, this particular model, CEQUEAU, is the operational model for Rio-Tinto-Alcan and the time came to admit that the model in its current state could not fulfill their new needs. I guess we can say that an obsolete model that is no longer maintained and improved, which is also abandoned by its users, is a dead model.
Recycling hydrologic models: the Model Stewardship Initiative
What if some obsolete hydrologic models could be recycled? This is what is happening right now to CEQUEAU. The model was « adopted » by a new team (including myself) and we are currently working on modernizing it, hoping that it will become especially appropriate for ensemble forecasting.
Last year, the Canadian Society for hydrologic Sciences (CSHS) decided to begin implementing a « Model Stewardship initiative ». The first goal of this initiative resembles the Hydrologic Modelling Inventory discussed above: centralizing the information regarding many hydrologic models all in one place so this information can be accessed easily.
The second goal might be even more interesting though. That is to identify hydrologic models that are to be abandoned in the near future, for one reason or another, and to make them available for « adoption » by a new person or new team. Then, people who have new ideas to improve those models and are willing to take over could ensure their continuation.
It is possible, however, that not all dying models are worth saving. Even if some diversity among modelling options is desirable, maybe there are already enough hydrologic models out there. It is also possible that some of the new ideas in hydrology are not entirely compatible with existing models or that adapting an existing model would need more work than building a new one. Hopefully the Model Stewardship Initiative will help answer these questions in the future.
Marie-Amélie Boucher is a HEPEX 2015 Guest Columnist.
January 19, 2015 at 06:52
Among concerns are those that echo from George Santayana in “Reason in Common Sense”, namely “Those who cannot remember the past are condemned to repeat it”; one hopes we advance with newer hydrologic models, especially with more complete physics, but without the detailed data that is necessary to drive the these models, how much can we really progress? Are we really progressing in terms of the accuracy of prediction at a basin outlet, let alone interior predictions of a distributed model and other physical quantities besides streamflow?
January 19, 2015 at 16:19
I confess I have never read « Reason in Common Sense » but you got me curious so I found it and read some parts of it.
The idea behind the Model Stewardship Initiative is precisely to avoid forgetting the past (older hydrological models) and repeating it (new models made from scratch instead of building on ones that already exist). But then « Reason in Common Sense » also mentions that « Not all readaptation, however, is progress » and that there are limits to adaptation. So again, maybe not all models are worth preserving.
Of course I agree with you that complexity in a model does not guarantee its accuracy and that complex models require data that are often not available. I also think that there is more than just hydrological models to consider if we want to achieve more accurate streamflow/volume forecasts. The improvement of atmospheric models, for example, has a large impact on our ability to make hydrologic forecasts.