Lag and high recession flows in the simulations

[Heou Maleki]

Hi All,

I have posted the following issue in SWAT-CUP google group but haven't receive any suggestion yet. Sorry for crossposting.

I have been trying to fix the lag (which is one to two days appearing earlier) in my simulated flows as well as high recession flows without success. 

I have tried to play with the related parameters as well as the HRU_SLP, SLSUBBSN, LAT_TTIME but couldn't resove the problem. Below are the parameters I am calibrating. I have also attached the figure of the lastest calibration result. I am using multisite calibration and it th early peaks are observed for all sites.

 Parameter_Name           Fitted_Value     Min_value      Max_value

1:V__ALPHA_BF.gw        0.910000        0.900000       1.000000

2:V__GW_DELAY.gw        72.000000       30.000000      450.000000

3:V__GWQMN.gw           900.000000      0.000000       1000.000000

4:V__RCHRG_DP.gw        0.100000        0.000000       1.000000

5:V__GWQMN.gw           700.000000      0.000000       1000.000000

6:V__RCHRG_DP.gw        0.150000        0.000000       0.500000

7:R__CN2.mgt            -0.410000       -0.500000      -0.200000

8:V__GW_DELAY.gw        93.000000       30.000000      100.000000

9:V__ESCO.hru           0.930000        0.900000       1.000000

10:V__SURLAG.hru        0.505000        0.010000       1.000000

11:R__SOL_BD(..).sol    0.190000        0.100000       0.200000

Any suggestion on how to fix the lag and improve the recession flows would be much appreciated.

Thanks in advance.

Maleki 

[Steven Jepsen]

Hi Maleki,

Your results to me don't look all that bad. It looks like your model is predicting too much baseflow during some years. I've improved that in the past by reducing GW recharge (e.g., it is not necessarily a recession constant problem). This might be done by increasing lateral flow (interflow) and/or surface runoff, which in turn could be done by increasing CN_2 and/or increasing SOL_K. To understand why increasing SOL_K would increase lateral flow and reduce GW recharge, my best suggestion would be the governing eqn in the theory manual. Kind of non-intuitive. Those are two things to try.

Good luck,

Steve

[Heou Maleki]

Hi Steve,

Many thanks for your suggestions. Adjusting SOL_K and CN_2 helped to improve significatly the simulated flow hydrograph: better recession flow and reduced lag to one day which probably with further adjustment can be sorted.

Kind regards

Maleki

[Heou Maleki]

Hi Steve,

Sorry, I forgot to attach the figure in order to show the improvement after adjusting SOL_K and CN_2. I have attached it now.

I am now trying to adjust REVAPMN and GW_REVAP parameters.

Regards

Maleki

[Jim Almendinger]

Maleki --

If I'm looking at your plot correctly, I'd reduce CNs to reduce your modeled peaks.  I've had some luck in adjusting baseflow seasonality by adjusting GW_DELAY (larger values smooth out baseflow over the year).

Best,

-- Jim

--
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/a9a5b1fc-ad28-4926-b44c-3fdb2ee8c51bn%40googlegroups.com.

[Steven Jepsen]

Hi Maleki,

Thanks for posting your results. What is your streamflow volume bias? If it's excessive (> 10% overestimate?), you might try increasing ET (GW_REVAP, SOL_K, SOL_AWC). Also, you have redundant rows for parameter GWQMN--not sure what that would do to your calibration. Also, check the ALPHA_BF values you are using against the literature. I've found SWAT recommended values to be too high (for my study sites). I've used a range something like 0.03-0.3 days^-1, with a best fit typically near 0.1 days^-1. A value of 0.9 would reflect fast recession.

Steve

[Heou Maleki]

Hi Jim and Steve

Many thanks for your suggestions.

Steve: There is an overestimation of more than 10% (approximately 50% of PBIAS). I have redundant rows of paramter GWQMN because I am performing multisite calibration and wanted to define a different parameter set for some upstream subbasins that have different geology. I have set a value of ALPHA_BF between 0.9 and 1 in order to have a fast response from the shallow acquifer. I have already tried the range 0.01 - 0.3 previously which did improve the result.

As you suggested, I have SOL_AWC and GW_REVAP to the calibration and it reduced the flow peaks and lowed a bit the hydrograph. Please find attached the results. probably I will need to extend the range of some parameters like CN2 in order to increase a bit some peaks.

Jim: Trying the calibration with CH_N2 also lowers the peaks and  reduces the lag of severals peaks but smoothes completly some small peaks. Some people argue that CH_N2 should not be calibrated but am not sure why.

Regards

Maleki

[Steven Jepsen]

Hi Maleki,

Thanks for sharing. Your results are looking better to me. I'm not seeing the flow bias that I was seeing before.

I'm not sure you're doing the upstream calibration correctly. You would need to explicitly specify the subbasin numbers or HRU numbers of the upstream contributing areas in your par_inf.txt file, but the file contents in your original post doesn't reflect that. The correct formatting for this is in the SWAT-CUP manual.

Steve

[Heou Maleki]

Hi Steve

Yes I have specified the subbasins numbers in the par_inf.txt file. What I pasted before was copied from the Best_par.txt file in the upper part the reason why the subbasins number are not shown. In the same file, the subbasin numbers are shown with the best parameters as in the par_inf.txt file.

Thanks for your availability and support.

Kind regards

maleki

[Natalja C.]

Hi,

Is this daily flow?

There is no "SURLAG.hru" (at least in my models o.O). Try to correct it as SURALG.bsn. I remember it was responsible for the distribution of such peaks in several of my models. Try tweaking it. It is in hours, so put something from 1 to 24 and see the result :) And it is a fast experiment, you can do it manually as well.

If the flows are monthly.. well, worth the try at least :D

But it seems you are on a right path. Cheers!
Natalja

[Heou Maleki]

Hi Natalja

Thanks for your feedback. 

Yes it is daily flow. Yes you are right, I checked before and there is no SURLAG in .HRU but in .BSN. However both are available in SWAT-CUP. Previously I tried both and finally kept .HRU one. 

What I have noticed before was that setting SURLAG between 1 and 24 reduces some flow peaks without fixing the lag. Please find below the simulation I have just run with the latest parameter sets and SURLAG.bsn set to 1 to 24.

I will further investigate different set of SURLAG value and see.

Cheers

Maleki

[xiaofeng liu]

Hi, I also have a similar issue which shows one day peak earlier. Did you solve your issue? Thank you.