Sediment Calibration Issues

[Ashwin Ghimire]

Hi All,

I am trying to calibrate my model's Sed_out_tons against observed data and my sediment values are consistently underperforming against the observed.

SWAT error checker says my sediments are being trapped by the channel as the instream sediment change is -5.34 mg/ha compared to average upland sediment yield 4.67 mg/ha. But, no parametrization is fixing this.

My flow was calibrated first with nse 0.68 but the sediment at best is giving 0.10 no matter what i do. It does not seem sensitive to the parameter values.

This is what I calibrated my flow with:

1:R__CN2.mgt            -0.026600       -0.100000      0.100000
2:V__SURLAG.bsn         1.122500        0.500000       2.000000
3:R__SOL_AWC(..).sol    0.038950        -0.050000      0.100000
4:V__ESCO.hru           0.777450        0.750000       0.900000
5:V__ALPHA_BF.gw        0.237590        0.080000       0.250000
6:V__GW_DELAY.gw        33.450001       15.000000      45.000000
7:V__CH_N2.rte          0.070810        0.030000       0.100000
8:V__CH_K2.rte          2.270000        2.000000       12.000000
9:R__OV_N.hru           0.017000        -0.100000      0.100000
10:R__SLSUBBSN.hru      0.248900        0.200000       0.300000
11:R__HRU_SLP.hru       -0.075900       -0.100000      0.000000
12:V__SFTMP.bsn         -1.540000       -10.000000     10.000000
13:V__SMTMP.bsn         2.940000        -10.000000     10.000000
14:V__SMFMX.bsn         9.469999        0.000000       10.000000
15:V__SMFMN.bsn         8.330000        0.000000       10.000000
16:V__TIMP.bsn          0.255000        0.000000       1.000000
17:R__DDRAIN.mgt        0.096200        -0.100000      0.100000
18:R__TDRAIN.mgt        0.061400        -0.100000      0.100000
19:R__GDRAIN.mgt        0.097000        -0.100000      0.100000

For Sediment, even if i tweak the spcon and spexp to the maximum, I get no sensitivity. and this is the case for some other sediment parameters like adj_pkr and prf_bsn as well.

Basically this line graph is what is a representation of sediment in mg/L compared against observed data in mg/L.

[Ashwin Ghimire]

Hi All,

These are my SWAT Error Checker and Sediment calibration Line graphs.

As you can see my sediments are not even close to the observed data.

For parametrization, I was using these parameters for flow (NSE achieved: 0.68):

1:R__CN2.mgt            -0.026600       -0.100000      0.100000
2:V__SURLAG.bsn         1.122500        0.500000       2.000000
3:R__SOL_AWC(..).sol    0.038950        -0.050000      0.100000
4:V__ESCO.hru           0.777450        0.750000       0.900000
5:V__ALPHA_BF.gw        0.237590        0.080000       0.250000
6:V__GW_DELAY.gw        33.450001       15.000000      45.000000
7:V__CH_N2.rte          0.070810        0.030000       0.100000
8:V__CH_K2.rte          2.270000        2.000000       12.000000
9:R__OV_N.hru           0.017000        -0.100000      0.100000
10:R__SLSUBBSN.hru      0.248900        0.200000       0.300000
11:R__HRU_SLP.hru       -0.075900       -0.100000      0.000000
12:V__SFTMP.bsn         -1.540000       -10.000000     10.000000
13:V__SMTMP.bsn         2.940000        -10.000000     10.000000
14:V__SMFMX.bsn         9.469999        0.000000       10.000000
15:V__SMFMN.bsn         8.330000        0.000000       10.000000
16:V__TIMP.bsn          0.255000        0.000000       1.000000
17:R__DDRAIN.mgt        0.096200        -0.100000      0.100000
18:R__TDRAIN.mgt        0.061400        -0.100000      0.100000
19:R__GDRAIN.mgt        0.097000        -0.100000      0.100000

Even when I add 4/5 sediment parameters like spcon and spexp, the model does not seem sensitive to those parameters.

I think a lot of sediment is being trapped by the channel as the SWAT Error checker suggests. Can you tell me how should I move forward with this?

Thanks

[JoeKHL]

Hi Ashwin, 

Did you change spcon and spexp in the .bsn file or the .rte files? Are they all not sensitive after changing? 
You might want to directly compare the numbers from output files before and after parameters changes before moving to the calibration phase.

Best, 
Joe

[Ashwin Ghimire]

Yes, I did do that numerous times but to no gain. 
It does not really influence my graph much and neither does my nse change as needed.

The observed data is just out of reach