Important model selection criteria in watershed modelling
The efficiency of the results obtained by a hydrological model depends on many factorssuch as the method of distribution and resolution of input data, calibration method of the model, type of model, size and topography of study area etc.
So, before choosing any model for flow simulation, one must find out wisely that which software or model would give better results for the particular application.
Following are the criteria for model selection:
a) Model availability:
A model should be easily available for use either open source or through agreements along with an active developer community and the possibilities of adaptation to specific purposes. Open-source models also allow users to perform modifications.
b) Type of Model:
Type of model to be used can be decided based on the catchment size and relief.
c) Input data requirement:
The model selection depends largely upon the available input data. If the distributed or gridded data is not available, the lumped model is the only option. Whereas some models have large data requirement which leads to difficulty in model calibration and validation.
d) Spatial and temporal distribution:
The required resolution of the data depends on the predominant rainfall which is causing a flood in the area. If the flood is caused by seasonal rainfall than the daily data will be required, if it was because of frontal or advective rainfall then the daily or hourly rainfall data required. On the other hand, if it is the convective rainfall,then the hourly or sub-hourly rainfall is required. Require resolution of the data also depends on the purpose of modelling.
e) Model calibration process:
Model calibration can be done either manually or automatically. Manual calibration is usually time-consuming and requires experienced users. If the option for automatic calibration is provided in the model, one of the factors for model selection is that there are a limited number of calibration parameters. It is better if there is an option of combining the manual and the automatic calibration.
f) Additional features:
Depending upon the purpose of modelling,the model should have options for choosing additional routines such as snow routine, soil routine, storage routine, response function and routing etc.
g) User community/assistance available:
The user community should be large enough to provide sufficient assistance without having waited long in case one gets stuck in the process. There should be a good platform to share problems, where one can get the solutions quickly because the developers may not be always available to assist quickly.