Sediment calibration: channel deposition/degradation and SED_IN vs SED_OUT (sub-daily)

[Bendik Hansen]

Hi,

I am (still) trying to calibrate my model for sediment concentration in my river (hourly). I am having an issue with sediment getting deposited (and never becoming re-entrained) in my channels, and I can't seem to get rid of it no matter how I tweak my parameters. My questions/issues are:

1) SED_IN is much larger than SED_OUT (in output.rch) for a given reach, and the difference does not even out over time. It is not just that the sediments temporarily deposit and get re-entrained, they just deposit and stay there (see sedinout.PNG). 

2) If I could solve (1), I think my model would have potential. One of the issues with the current simulation is seen when zooming in on the sim/obs graph (sed_simobs.PNG). In the observed data there is often a peak, and then a period after the peak with high concentrations (presumably the sediment that was deposited in the river being resuspended?). In the simulation there is only a peak and then the concentration flattens out almost immediately. 

3) I tried enabling channel degradation, suspecting that deposited sediment was considered part of the channel bed. I tweaked the following parameters in the .rte files, but nothing changed in my simulations (which is odd regardless of the other problems...):

• ch_bed_kd
• ch_cov1
• ch_bed_bd
• ch_bed_d50
• ch_bed_tc

Other things:

• I have tried every combination of SPCON and SPEXP.bsn, even unrealistic ones. Nothing makes SED_IN match SED_OUT. 
• I have tried changing the parameters for subdaily erosion. This changes my sediment yield and concentration significantly, but that is not what I'm after since my sediment yield is already more or less where I want it. What I need to change is the channel sediment routing. 
• I see there are different channel erosion and sediment routing models that can be used, but the documentation is not very good. I don't know which parameters I have to tweak if I change the channel sediment routing method in basins.bsn or .rte.
  
• I tried increasing stream power by setting ch_w2.rte (channel width) to 1/3rd of its original value. I also tried setting it to 1/100th. Neither had a significant impact on sediment routing in the rivers.

I am using SWAT rev664, since there were issues with hourly simulations in rev670 (also reported by other users). 

I attached basins.bsn and a random .rte file to show the parameters I use. 

Any suggestions?

Best regards,

Bendik

[Jim Almendinger]

Bendik --

I've puzzled over your results for the last week or so.  And I haven't used the sub-daily time step, but I presume algorithms function similarly to the daily time step.  I don't have good answers for you.

(1) At least your sed_in and sed_out loads are correlated with flows, which makes sense.  But why is there so much deposition (sed_in > sed_out), when spcon is so high (0.1)?  Usually by the time I set spcon to 0.01 or so, that's large enough to stop significant deposition.  All I can suggest is to artificially increase the slope of each reach, to increase velocities and sediment transport capacity.  (But slopes are "real" data, and I prefer to alter "parameters").  Or perhaps changing the peak-flow parameters could do the same thing.  But you'll have to beware of uncontrolled channel erosion, by keeping the erosivity and cover factors low (basically zero).  And then all channel interaction will be blocked out, if nothing deposits or erodes.  

  I don't know how this works on a subdaily timestep, or with the non-Bagnolds options , or with channel degradation turned on.  

(2) I think your concentration data imply mechanisms not well modeled by SWAT.  My working hypothesis is that there are grain-size differences under different flow regimes that are mobilized at different times -- which I think you share.  One idea is that during non-flood periods, finer grains can accumulate in the channel or backwaters, and these are easily mobilized at the beginning of the next flood event.  But you have substantial concentrations during low flow periods following major runoff events.  I wonder if these are bank-failure events, which are highly unpredictable (I presume) in detail, and really not modeled in SWAT's algorithms.  Even so, because the flows are so low, the loads are small despite the higher concentrations.  So if loads are the primary concern, then maybe these concentrations are not that critical.  Finally, is there a chance these low-flow particles in high concentration are algae from impoundments?  They may not add a lot of mass compared to inorganic grains but if your estimates of sed concentration were based on turbidity, then algae could be giving you misleadingly large sediment loads (I suspect this is idea is not very likely... just thought I'd include it just in case...).  

Well, you have good and interesting data, if that's any consolation.  

Good luck, and thanks for the detailed look as what SWAT can (or cannot) do.  

-- Jim

---

James E. Almendinger, PhD Director, St. Croix Watershed Research Station 16910 152nd St N, Marine on St. Croix, MN  55047 e: jalmen...@smm.org o: (651) 433-5953 x 11

--
You received this message because you are subscribed to the Google Groups "SWAT-user" group.
To unsubscribe from this group and stop receiving emails from it, send an email to swatuser+u...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/swatuser/0294183b-bf7b-4290-8f0d-2be0198d70cf%40googlegroups.com.

[Natalja C.]

Hello,

So just to make sure (and maybe get a better idea of the problem) do you use the SS (and not the TSS) for your observed data? SWAT produces sediment outputs of the mineral particles only. In our rivers, the organic part of the suspended material may reach up to 80%, which means that I have to either calculate the mineral part (if I know the relationship), or ask for specific SS data (they burn the samples to determine the mineral part, I think). If you use TSS, then for sure your observed data will be much larger than what SWAT produces.

Second, you mentioned that you made the river widths smaller (by multiplying by 1/3), but is that ACTUALLY correct? In my case, what ArcSWAT produces is totally wrong, so I have to correct channel widths every time. And the relationship is not linear. So my advise would be to check are the stream widths are ACTUALLY like this. 

Also, did you try to adjust the ch_n2?

And finally, (what blew me away at one point), maybe there is some sediment addition in the LAT_SED? I know, that usually people just ignore this issue, but in my case, a very low concentration of sediment in the gw that made the difference. But, of-course, you have to have some means to verify the value and make sure that this is actually what is going on in your case.

Best of luck,

Natalja

[Bendik Hansen]

Thanks for both of your replies. I will get back to you later, right now I have some deadlines to beat on other projects! 

Best regards,

Bendik

[Bendik Hansen]

Hi again, now I have time to work on this once more. 

Jim:

1) I will try changing the slope just to see if I can force the model to block out all channel interaction, but I have my doubts (since changing the river width to increase velocity did not work). I don't really consider this a viable option for the model, but just a way to see if I can get rid of the apparent deposition.

2) I agree, we already expected landslides to play a major role in sediment contribution to the river (see the attached picture). I am not so worried about my inability to reproduce the observed values as the inability to get sed_in to match sed_out. It seems like it should be theoretically possible with your suggestions, but I'm not seeing the expected results. The observed sediment concentrations are from ADCP measurements, and I'm pretty certain the vast majority is inorganic (once again see the picture). I will double check his though.

Natalja: 

My measurements are from ADCP so I suppose it's SS, I will investigate how much of this might be organic, but it seems unlikely that it's a very high fraction (the attached picture shows the extreme erosion in the catchment).

You're right, changing river widths is not necessarily accurate, my main concern so far has not been to reproduce reality as much as getting the main processes the way I want them (which starts with my sed_in/sed_out issue that just won't be solved). Now it's almost a matter of principle, it should be possible to exclude channel deposition/erosion, but nothing I try seems to work. I was just starting to play with ch_n2 when I had to change my focus to some other projects, I will also check that now. I will check what is going on with lat_sed in the model, but unfortunately I don't have much data for the catchment beyond climate inputs and observed discharge and sediment concentration in the river. 

I can't find much literature on sub-daily sediment simulations in SWAT, and it seems like most people I talk to here have not used SWAT for that. It might be that it is not possible to simulate my catchment due to unsimulated processes like landslides and bank failures, but even if that weren't the case I'd still have the same weird issues concerning sed_in vs sed_out. I must say I'm becoming very sceptical to the sub-daily sediment processes in SWAT. I might be missing something obvious since I'm not able to understand the source code and I'm not an expert on sediments...

Thank you for your suggestions and feedback.

Best regards,

Bendik

[RP L]

Hi

In my model, the SED_IN value and SED_OUT value of some rivers are completely equal, 

but some are not equal.

when SED_IN is greater than SED_OUT, the channel should be deposited，and  CH_DEP is a positive number

When SED_IN is equal to SED_OUT, the channel should be in balance of erosion and deposition，and  CH_DEP is equal to 0.

When SED_IN is less than SED_OUT, the channel should be erosion, and CH_DEP is a negative number.Is that correct ？

However, in the result file,  CH_DEP is always greater than or equal to 0, and even when SED_IN is greater than SED_OUT.

In the sed file, why are the values of CH_BNK, CH_BED and FP_DEP always 0？

I am looking forward to your reply.

[yinan ning]

Dear all,

I hope this message finds you all well!

There is also the same problem in my watershed, channel deposition is very high in some channels and then the sed_out much lower than the sed_in. 

Does anyone solve this problem and have any suggestions on it? 

Appreciate any reply from you!

Best regards,

Yinan

[yinan ning]

Dear Bendik Hansen,

I hope this email finds you well, and I am so sorry to bother you. My name is Yinan Ning, a PhD candidate at Wageningen University, and I am reaching out to you regarding an issue I encountered while running the SWAT model.

I recently faced a significant disparity between sed_in and sed_out in my SWAT model simulations. Concurrently, I observed a considerable amount of sediment deposition in the channel (I attached my simulation results below). While attempting to seek a resolution to this problem, I  found you also encountered a similar issue in 2019 in the SWAT User Google Group. However, it appears that no feasible solution was provided at that time.

I am curious to know if you have since managed to resolve this issue. If so, I would greatly appreciate it if you could share your insights or any potential strategies you employed to address this discrepancy. Any guidance or advice you could offer would be immensely helpful in troubleshooting and resolving this issue in my own simulations.

Thank you very much for your time and help in advance. I look forward to hearing from you soon.

Best regards,

Yinan

[ONC]

Hi,

The desparity between SED_IN vs SED_OUT has to do with the maximun sediment transport capacity (MSTC) equations that you are using (CH_EQN parameter to be clear), if you check the SWAT2009 Theorical Documentation, you will find that in every channel, the max amount of sediment that can be transported is determined by that equation. So it seems that in your study area, the  MSTC equation   is very low, that is why deposition is so high. 

I strongly suggest you to read  swat2009-theory.pdf (tamu.edu), review some papers that discuss the differents  MSTC equations (there are 5*), and then see the best fit. By experience, I would suggest you to use the default one but change key values using SWAT-CUP.

Regards,

Oscar Cabezas-Nivin

[César Carvallo]

Dear Yinan,

I saw your reply to the post. I am facing the same issue and would like to know if you were able to resolve it. I’ve been searching for similar cases but haven’t found a solution yet. I hope you see this message.  

Best regards,

Cesar