Hello Dear SWAT-CUP users / experts
so, to sketch my problem here,
I am working on a SWAT model of a nival/glacier regime Arctic catchment. A big part of a recharge is snow/ice melt. My main goal is to calibrate the discharges (FLOW_OUT). I do not care about sediments/nutrients here.
I finished creating the model, and I started to learn SWAT-CUP.
I already know the basics and with that knowledge I have chosen a set of parameters to calibrate the model. (and method: SUFI)
Here, I would want you to take a look at the parameters as well as on the dotty plots.
I have a little struggle with interpreting dotty plots results. I know that I must look for some trend in the case of parameter sensitivity. no trend = no sensitivity, right?
I am adding 3 screenshots here.
First (1st iteration), with parameter ranges of my own choice.
Second (2nd iteration), with first "new_pars" suggestions from calibration output.
Third (3rd iteration), with second "new_pars" suggestion after running 2nd iteration.
number of simulations = 10.
I know the number should be bigger, I did 100 simulations once and dotty plots looked pretty bad in my opinion...Pretty much structureless graphs with no trend.
Also, I have a question about the calibration and validation periods.
My SWAT model run period is 2004 - 2017 (3 years of warm up).
I have a field discharge data from 2007 - 2012 and from 2017.
Now, I don't really know how to handle the calibration and validation periods.
The screenshots I add here are from the 2007 - 2017 SWAT-CUP run. But now I'm thinking, maybe I should do the calibration period for 2007 - 2012 and validation period for 2017 only?
Please, let me know what you think. Thanks in advance.
list of the parameters I've chosen:
v__SHALLST.gw 0 1500
a__GW_DELAY.gw -10 10
v__ALPHA_BF.gw 0.5 1
v__GWQMN.gw 0 500
v__GW_REVAP.gw 0.02 0.2
v__REVAPMN.gw 0 400
v__CH_L2.rte 5 10
v__CH_D.rte 0.2 3
v__CH_N2.rte 0.05 0.1
v__CH_K2.rte 25 300
v__CH_COV1.rte 0 1
v__CH_COV2.rte 0.5 1
v__ESCO.hru 0.01 1
v__SNO_SUB.sub 0 200
v__SFTMP.bsn -15 1
v__SMTMP.bsn -0.01 0.01
v__SMFMX.bsn 0 20
v__SMFMN.bsn 0 10
v__TIMP.bsn 0 1
v__SNOCOVMX.bsn 0 500
v__SNO50COV.bsn 0.05 0.8
v__LAT_TTIME.hru 0 2
v__CN2.mgt 70 90
Ina mentioned all the points. try to increase your simulation numbers.
You can also look into the p-values for the global sensitivity. And to find out the most sensitive parameters you can use on-at-time sensitivity analysis.
Yassine,
1st of all thanks for the opinion.
when it comes to the CN2 parameter I was suggested by the input/output documentation and the table there.
It basically says that for bare soil, runoff curve number equals like 77 - 94 range.
I agree with the several types of land use statement....but in my case the biggest % of the catchment is bare soil (BARR) and sometimes it's BSVG (barren or sparsely vegetated). The rest of my catchment's land use is water and that's it basically.
I will do 500 sims the following night and let know about the results.
Manashi, thank you for the opinion too!
Cheers
Dear all,
I would like to attach results of my 500 simulations.
To be honest I do not see any trend in the dotty plots, except CH_K2. Maybe I do not know how to properly interpret these charts...
I am adding par_inf that I used and an Excel screenshot with a 95PPU chart (based on this tutorial by Liem: https://www.youtube.com/watch?v=rFEP6Y3XAEA).
p-factor seems low I guess.
Also, when I try to enter one-at-a-time sensitivity analysis, my SWAT-CUP gets stuck and eventually stops working. I have no idea why that happens.
Kind Regards, KJ
Dear KJ,
You are right only CH_K2 shows a pattern. Most of the cases you find same thing that most of the parameters do not show pattern.
For sensitivity analysis do not use OAT method rather choose only Glabal sensitivity analysis. What happened to you that what Prof Srinivasan mention in waorkshop, your computer will go crazy.
Your calibration result is excellent acconrding to Moriasi guideline. So, be happy with it. If you want to do more then you can separate your calibration for wet period and dry period. You can use Base Flow Separation program to separate your observe data and then use them for calibration.
Best
Jakir
- Badges
- Science topic
- Similar topics
- Geoscience
- Asia
- India