The HYPE code was released as open source in 2011 under the Lesser GNU Public License. It is version managed at SMHI and different versions can be downloaded from sourceforge. The HYPE code is structured so that the algorithms for hydrology is embedded in the HYdrological Simulation System (Fig. 1), which is the infrastructural part of the model source code. HYSS handles the simulation instructions, provides the hydrological model with variables containing data on the model set-up, time and current forcing data (i.e. precipitation and air temperature) and writes the result files. It also provides variables for model state, model parameters and output, which are used and set by the model. In this way, the hydrological algorithms are separated from the model infrastructure in its own module. HYSS can thus be coupled to several different hydrological modules with different model structures and process descriptions.
With time the HYPE model proved to be efficient also for hydrological forecasting and is currently used operationally in the SMHI flood warning service providing hydrological short-term and seasonal forecasts. Extending the HYPE model to new purposes and geographical domains have resulted in an increased number of lines in the model source code, as well as numbers of FORTRAN files (Fig. 2). Each year 5-10 new model versions are released.
The main motivation for developing the HYPE code was to harmonise the simulation of water balance and flow paths with substance transport and water quality. Previous model concepts at SMHI was using the HBV model (Bergström et al., 1976; Lindström et al., 1997) for nutrient modelling (e.g. Arheimer and Brandt, 1998; Arheimer and Wittgren, 2002; Arheimer et al. 2005), resulting in a very complex and ineffective code, which was trying to reconstruct flow paths from the more lumped HBV concept.
Below are some results from tests run for new HYPE versions, when executing the model on a laptop with a Linux server. None of these machines were configured for testing runtime of HYPE (i.e. other applications may run at the same time and the timing can also depend on the machine).
The run time is defined by:
- the size of the model (number of subbasins, number of soil-land use-classes)
- the simulation time
- if substances are simulated in addition to flow
- the number of outputs you want also have an influence.
The runtime (HYPE ver. 4.13.1) in seconds per subbasin, class and simulation years varies between 0.0002-0.0005 when running five different models of different size (with nutrients) on a laptop:
|4||36693||60||1||376-490 (depending on output amount)|
NOTE: The global model run at a Linux cluster (using nodes of 8 processors and 16 threads) with calculations in approximately 1 800 000 hydrological response units (HRUs) and 130 000 catchments covering the worlds land surface, except for Antarctica. The model runs in parallel in 32 hydrologically-independent geographical domains with a run time of about 3 hours for 30-year daily simulations.