Hi Bhumika,

I just checked it for my current project, a 144 km² catchment in Central Germany. The model was calibrated using variable storage. Changing to Muskingum led to a streamflow reduction of 75%! Such a big difference in streamflow volumes is very strange. The routing method should rather have an effect on the timing than on total volumes. Of course ET and transmission losses change, but I would have expected a much smaller effect.

I never did a systematic comparison of the two methods. I prefer using variable storage to avoid touching the additional calibration parameters (MSK_CO1, MSK_CO2, MSK_X). Did you calibrate your model for any of the two methods before comparison? Probably you would get similar results after specifically calibrating the model for each method. But still, I find this huge difference (for the same parameterization) very strange. Would be interesting to get an explanation for this.

Best,

Michael

Thanks a lot for checking this in your existing model!!!

Me and my colleague created a test case for one basin only (without any external storage structure or infiltration/Transmission loss). We extracted the streamflow before and after performing routing (from the fig.fig file and from the code)

The variable storage method gives almost the same streamflow volume before and after routing (look reasonable)

The Muskingum gives less streamflow volume after performing routing (up to 40%). Strange!

We are reviewing the theoretical documentation and the code to see why.

Did you calibrate your model for any of the two methods before comparison?

No we didn't try that.

Probably you would get similar results after specifically calibrating the model for each method.

Not sure about that!!!

Currently we are using default routing parameters.

I and my colleague have reviewed the documentation and the code. As we understood:

1) SWAT does not performed stream routing with Variable Storage Method

2) In the Muskingum, to maintain numerical stability, sud-daily time step is used. However, evaporation in channel for each time step is calculated as evaporation for the whole day. This causes evaporation in channel so high -> therefore, when Muskingum method is used, much lower simulated streamflow volume is observed (compared with Variable Storage method or inflow to stream before routing)

The two files related to stream routing are "rtday.f" and "rtmusk.f"

In an extreme case, e.g., arid and semi-arid area, if Muskingum is used, the error could be extremely high (in our case, up to about 70% total streamflow is evaporated) because high evaporation in channel.

Just to finish my answer:

The above issue might not exist in all SWAT version (just found it in version 664). To check whether your version is affected:

1. With variabale storage (IRTE = 0): In the "fig.fig" file print out streamflow before and after routing (as shown in the attached SWAT example). Compare the plotted time series before and after routing, if they are the same (or slightly less, because SWAT simulates evaporation in stream), it shows that that SWAT version does not perform stream routing

2. With Muskingum (IRTE = 1): If the simulated streamflow volume (or sumQ) after routing is much less than before routing , it means that that simulated in-stream evaporation is too much.

Attached please find the example SWAT project (only 1 basin without any structure like reservoir, ponds, etc) and the SWAT source code (version 664 unmodified) downloaded from SWAT website

Thanks a lot for sharing the model's TxtInOut files!!!

I am already late to answer this question.

I have attented a SWAT workshop last week, there Prof Srinivasan told that if you have meander in your river then choose Muskingum method.

Might be this information help you to choose which routing method you will use in your model.

Best

Jakir

thanks a lot Jakir!!!