Main outlet vs Multisite Calibration

[Emily Blue]

Hey. 

I was wondering if calibrating the model only at the main watershed outlet is generally sufficient, or if there are cases where multisite calibration (for example, including upstream gauges) is recommended.

More specifically, are there any practical criteria you use to decide when adding upstream gauges is necessary

Best,

Emily. 

[Maarten J. Waterloo]

Hello Emily,

Good to give some attention to this subject. I would say that in general calibration on multiple sub-catchment and the main outlet would make your model more "realistic" as it takes spatial variation in hydrological processes into account. To decide on multisite calibration you should take a number of factors into consideration, such as:

- Heterogeneity in topography, climate, soil and land use/management in the area to be modelled;

- The availability of good-quality calibration data for sub-catchments in the modelled area;

The above factors are often related to the catchment size, with larger catchments usually showing higher heterogeneity in land characteristics and having a higher need for multisite calibration. There is always the problem of equifinality in calibration, where very different sets of calibration parameter values result in the same output at the main outlet. This could mean that too high discharge from a sub-catchment may be compensated by too low discharge from another sub-catchment. Using multisite calibration may then provide you with better sub-catchment discharges than if you would have relied on a single main outlet calibration.

One area of interest is often estimation of spatial variation in groundwater recharge. In this case spatial variation in the geology (bedrock permeability) can then be a reason to change the deep aquifer recharge fractions (low for low permeability areas (e.g. massive granite rock) and higher for other areas (deep sedimentary deposits)) and use this knowledge in the calibration of the model. There are many more examples where sub-catchment can have different responses in the model catchment.

If you do have data to perform multisite calibration, then go for it, unless your catchment area is very small or uniform. In the areas where I work there is usually very limited to no calibration (discharge) data available and we resort to using public actual ET data for sub-catchment calibration, but the ET datasets (e.g. WAPOR, MODIS) do provide very different estimates, with actual annual evaporation sometimes exceeding precipitation in areas where there is no irrigation water input. 

I hope that this helps, do contact me if you want to discuss more...

Regards,

Maarten Waterloo

[SWAT+ Toolbox]

Thanks Maarten, this is very useful!