Multi-site calibration against individual site calibration

[Heou]

Hi All

I have performed a multi-site calibration using three outlets (one upstream, the second in the middle and the third downstream). 

Among the results of the three sites, I got the following figure for the upsteam outlet.

Based on this figure, I wanted to focus on the upstream catchment first and see if I can improve the model for the upstream catchment from the best parameters that I obtained.

When I try to calibrate only the upstream catchment with the same best parameter set, I get the following result.

I couldn't understand why this happens. Any help will be much appreciated.

Thanks in advance.

Heou

[Steven Jepsen]

Hi Heou,

What do you mean by "When I try to calibrate only the upstream catchment with the same best parameter set..."? Using the "same parameter set" might make sense, but using the "best parameter set" for a new calibration does not make any sense to me.

The multi-site calibration at the upstream outlet doesn't look too bad. Maybe increasing the ET would help? Nice extended recession limbs! Something is way off with the separate calibration to the upstream site. Maybe check the amount of surface runoff versus baseflow. The observed hydrograph looks consistent with abundant baseflow. 

Steve

[Heou]

Hi Steve,

Thanks for your reply and suggestion. Sorry for using the best parameter set previously as these are unique values. It was actually the intervals of the same parameter set I used for multi-site calibration.

For the separate calibration of the upstream, I checked and the model produces high amount of surface runoff. I tried to adjust the baseflow and surface runoff parameter but the results are similar. I am not sure if it has something to do with soil and crop properties as that upstream catchment is dominated by arable land. 

Any further suggestion will be welcome.

Thanks

Heou

[Steven Jepsen]

Hi Heou,

Yeah, it kind of looks like CN is set very high and that a lot of small rain or snowmelt events are flashing off to streamflow.

About the only thing I'd suggest is to make sure you haven't inadvertently changed any parameters in the active project folder from modifications to a previous (inactive) set of parameters. This can happen if you are NOT using parallel processing and if you HAVE changed the parameters that are being adjusted in a single SWAT-CUP project. That is at least true in the version of SWAT-CUP I use. For example, if for a while you are changing CN and then decide to hold CN fixed, the input *.mgt files in the active folder will have CN set to the last simulation of the last iteration where CN was varied (if that makes any sense). Again, this is not an issue with parallel processing--if I'm not mistaken.

Steve

[Heou]

Hi Steve,

Many thanks for your suggestions. That makes sense. I'm actually not using parallel processing and it is possible that I may have changed something inadvertently. I am going to generate a new project folder and give it a try.

Kind regards

Maleki

[2w2e]

looks like the time step had changed!

[Heou]

Hi 

Thanks for your feedback. The time step has not changed, it is the same, I have tried to work on the ET parameters and method as well as CN2 and could only reduce the pics and a bit of overestimation for the low flows and the rising limb.

Heou

[Steven Jepsen]

Heou,

I wonder if it would help to plot the modeled snowmelt + rainfall hydrograph and compare it to the observed hydrograph, noting that land acts as a type of filter between these two things. This might help ID a problem such as simulating rainfall when it should be snowfall, which could erroneously produce many flashy runoff events.

Steve

[Heou]

Hi Steve,

Thanks for your suggestion, I did plot the snowmelt, ET, PERC, LATQ and SW and found issues with percolation, lateral flow and deep aquifer recharge. This helped to improve the simulated hydrograph. 

Regards

Heou