HBV – The most famous hydrological model of all? An interview with its father: Sten Bergström

Contributed by Fredrik Wetterhall

Numerous are the hydrologists that has come in contact with the HBV (Hydrologiska byråns vattenbalansavdelning) model, either through their own hands-on experience in applying it themselves or through the vast literature on the model and its applications. The HBV model, originally developed by Sten Bergström at the Swedish meteorological and Hydrological Institute (SMHI), is one of the most used hydrological models in the world.

Next year, Sten will retire from SMHI, and to mark that occasion Sten has collected information on where the model has been applied around the world (see: HBV_publications.pdf ). The reference that is most often used as the origin of the model is Bergström (1976), however the earliest versions of the model dates back to 1972. In its more than 40 years of existence, it has been used in more than 1400 publications, applied in 95 countries and used in almost 50 PhD theses.

I sat down with Sten to find out more about him and how the model came about.

Fredrik: How did you start your hydrological career?

Sten: I graduated in Civil Engineering in 1971 at Lund Institute of Technology (LTH) in south Sweden and came directly to SMHI after that. The work on the HBV model was one of my first tasks. In 1976 I defended my Ph. D. thesis on this work, at the Department of Water Resources Engineering of LTH, but I worked full time at SMHI during the development of the model.

Fredrik: What was the first motivation of creating a hydrological model?

Sten: The first objective was to develop a model of the water balance based on data collected in Swedish research basins, as a contribution to the International Hydrological Decade (IHP). But soon the hydropower industry got interested. They needed a forecasting tool for the operation of the hydropower system that was about to be completed at that time in Sweden. For the same reason the model soon became popular in Norway, where major development contributions were delivered.

Fredrik: HBV was one of the first hydrological models but there were other models around in those days, from where or from whom did you get most inspiration?

Sten: The Stanford IV and the SSARR models were available and tested in Sweden. But we were looking for a less complex approach, with data demands that could be met by our networks of observations. The single most important source of inspiration was the famous paper by Nash and Sutcliffe from 1970 and their proposed strategy for model development. As a young hydrologist I also had the favor of spending some time with Eamonn Nash in Galway and in Stockholm 1972. The development of the snow-routine of the HBV model was to a high degree inspired by the SSARR model.

Fredrik: The name HBV, what does the acronym stand for and how did it come about?

Sten: The origin of the acronym is simple. In the 1970s the hydrological research at SMHI was carried out at a small department called “Hydrologiska byråns vattenbalansavdelning (HBV)”. In English: The Water Balance Department of the Hydrological Bureau. When we were about to publish the first results the head of that department, Arne Forsman, and myself realized that we needed to give the model a name. So it became the HBV model (“HBV-modellen” in Swedish). At that time we could not imagine the impact of this decision.

Fredrik: The computing power in those days was quite modest in comparison with today, can you describe how the earliest HBV implementation was made?

Sten: As SMHI is both the national weather service and the hydrological service adequate computer resources where available already in 1971. A very important asset was also a plotter that was used to visualize the results. The first code was written in the Algol-Genius language and together with input data, parameter values and initial conditions it was fed into the computer via 5 channel paper tapes. If we were lucky up to three runs could be made in a day, in between the operational runs for the weather service.

In the beginning visual inspection was the most important calibration tool, supported by the Nash-Sutcliffe R2 criterion of fit. Relatively early we began to analyze the error function topography, but we had to be a bit careful as this increased the demand for computer time.

Fredrik: The HBV model has been and is still one of the most used hydrological models, what is the reason for its success?

Sten: I think that we found a sound balance between complexity and the information contained in most operational datasets. The early contact with the Scandinavian hydropower industry was important in this respect, as was our dedication to avoid overparameterization. The intense debate about runoff formation and field studies within IHD were also important. In the beginning the scientific value of the model was sometimes questioned, but today it seems that its structure is scientifically proven and considered to be sound.

The HBV model is also quite easy to communicate and it requires very modest computer resources. This made it easy to export the concept, in particular after the arrival of desk-top computers.

Fredrik: HBV is a representative example of a lumped (or semi-distributed) model with conceptual representations of the hydrological processes in contrast to distributed models where more of the processes are explicitly described. This divide into two “schools” still exists today and the debate is very much alive. What is your perspective on the benefits and drawbacks of a conceptual model?

Sten: In my opinion this debate is not always so fruitful. The important thing is to find the proper level of sophistication, related to available data and problem to be solved. In 1996, when we launched the HBV-96 version of the model, we deliberately called it “distributed” in an attempt to bridge the gap between these two concepts. I am not sure that we succeeded. The draw-back of the conceptual approach became most evident when hydrochemical modelling was attempted. This requires a more complete physical description of the hydrological system than the HBV concept can offer.

Fredrik: The HBV model has undergone a number of developments over the years and is still in use both in operational hydrology and research. Do you think that we still will be using HBV 40 years from now?

Sten: Looking at the international spread of the HBV model it seems that there is a demand for a model of this kind, which in a simple way links meteorology to water resources. I think that this basic concept will survive, even if it might be called something else in the future. Sometimes it seems that the modelling society is bit conservative too. So has, for example, the improved response function, with its reduced number of parameters in HBV-96, and a further improvement relating the response function to the soil moisture state, not become as accepted as the original model structure.

Fredrik: Finally, do you have any advice to young researchers or people that are starting their career in hydrological modelling?

Sten: First of all, identify the problem that you intend to solve. Ask yourself: Why am I doing this at all? Seek contact with potential users of your results. They can help you understand what needs to be done, and may also have valuable scientific ideas. Secondly, look at Mother Nature. Field work, or at least visits to the catchments, give you inspiration and maybe also a realistic attitude when coping with the complexity of nature. Finally, try to get an overview of what has been done before. You can’t read everything but try to find and read the most important papers. Contact with key personalities may also help as well as the attendance to key scientific meetings.

We thank Sten for the first 40 years of hydrological modelling with the HBV model and wish him a nice retirement, knowing that the model will continue doing hydrological modelling for many years to come. If you have further examples of applications of the model that is not on the list, please put it in a comment below or use the contact form.

References:

Applications of the HBV model HBV_publications.pdf

Bergström, S. (1976). Development and application of a conceptual runoff model for Scandinavian catchments. SMHI, Reports RHO, No. 7, Norrköping.

Bergström, S., and Forsman, A. (1973) Development of a conceptual deterministic rainfall-runoff model. Nordic Hydrology, Vol. 4, No. 3. 147-170.

Nash, J.E. and Sutcliffe, J.V. (1970). River flow forecasting through conceptual models part I – A discussion of principles. Journal of Hydrology 10: 282–290.

Do you know of more publications that are not on the list? Please let us know: