Sediment Calibration

[fatemeh babakhani]

Hello All

I am trying to calibrate a watershed based on Sediment Component (It's already Hydrologically Calibrated!).

 At the first step, I did sensitivity analysis and ended up with 7 sensitive parameters that have been shown in the attachment. Then for calibrating the model, I ran the SWATCUP several times with different parameters ranges! In all of my simulations, most of Peak Sediment Loads were out of 95ppu band!

* I was wondering if any of you have this problem! if yes, could you please tell me how you solved it?!

** p- factor is 0.66 (I believe for sediment calibration it can be less than 0.7.), r- factor is 0.93 which is good, I am just worried about NS which is 0.61 (that should be close to 1)! What do you think, do you think this numbers are acceptable for sediment calibration. 

All required information has been enclosed, if you need further information please let me know.

Than you so much,

Fatemeh

Fatemeh A. Babakhani

Graduate Research Asistant, Department of Civil and Environmental Engineering 

Wayne State University

Detroit, MI 48202 (313) 844 4902

[Abbaspour, Karim]

It is already amazing that you could calibrate sediment with such precision! And that, using non of the sediment parameters! 

Why not including some of those?

--
You received this message because you are subscribed to the Google Groups "SWAT-CUP" group.
To unsubscribe from this group and stop receiving emails from it, send an email to swat-cup+u...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

<RockWellDamWatershed.docx>

[fatemeh babakhani]

Thanks for your response!

I did sensitivity analysis for 23 parameters based on One-at-a- time method. But just 7 of them were sensitive! 

It is somehow weird to me also, that none of the CH_COV, CH_BNK_KD,... are sensitive!! But in my model SOL_ROCK, and LAT_SED, HRU_SLP are too sensitive parameters.

These parameters have been enclosed!

Thanks again,

Fatemeh

[Jim Almendinger]

What ch_eqn were you using?  Default is ch_eqn=0, and in that case, in recent releases of SWAT (rev635 and some earlier), it seems that the spcon, spexp, and ch_cov1-2 parameters are not activated.  It's unclear to me whether this is a bug or by design.  If you choose another sediment transport equation (ch_eqn=1, for the simplest), then the channel deposition and erosion parameters become effective. 

-- Jim

---

From: "fatemeh babakhani" <babakhan...@gmail.com>
To: "SWAT-CUP" <swat...@googlegroups.com>
Sent: Thursday, May 14, 2015 7:22:50 AM
Subject: Re: Sediment Calibration

---

Spam
Not spam
Forget previous vote

--

Dr. James E. Almendinger
St. Croix Watershed Research Station
Science Museum of Minnesota
16910 152nd St N
Marine on St. Croix, MN  55047
tel: 651-433-5953 ext 19

[fatemeh babakhani]

Dear Dr. Jim

Thanks for your response!

I didn't change the Sediment Default Equation in the SWAT. So in my model CH_EQN=0 (or simplified Bagnold Equation)! 

According to the SWAT USER Manual PRF, SPCON, SPEXP are parameters used only when CH_EQN is zero, so it seems in my model they should be active! (But I am not sure.)

Again I did sensitivity analysis for COV1 and COV2 based on one at a time method, but my results don't change, at all!! So I want to do sensitivity analysis for all parameters at the same time, and check t-factor to see what will happen.

Thanks,

Fatemeh 

[fatemeh babakhani]

Hello Dr. Abbaspour and Dr. Jim

I re-run the SWAT CUP but this time I added Sediment Parameters to the model! Results show p- factor is 0.73 , r- factor is 0.97,  it seems that the model is calibrated!! 

As I explained you in the last emails, when I was doing Sensitivity Analysis based on Once at a Time method, the model wasn't sensitive to the Sediment Parameters! But when I am doing Global Sensitivity it seems that the model is sensitive to the Sediment Parameters, although  P-Value for Sediment Parameters are still high in comparison to other parameters.

I believe my model is calibrated based on Sediment Component, my question is that which number I can choose for each of the parameters. For example based on my run USLE_P should be between 0.17 and 0.23, and SWATCUP suggested 0.22 as a Fitted Value. Could you please tell me if it is better I choose 0.22 as a Final Number and re-run ARCSWAT based on USLE_P=0.22.

Required information has been enclosed. If you need further information, please let me know.

Thank you so much.

Fatemeh A. Babakhani

Graduate Research Asistant, Department of Civil and Environmental Engineering 

Wayne State University

Detroit, MI 48202 (313) 844 4902

[Abbaspour, Karim]

The solution to your calibrated problem is the range of the parameters you obtained. You need to keep this range and propagate this range for all of your model uses.

You don’t need to put the parameters in ArcSWAT! Many people ask this and I don’t understand why they think that is what they should have to do!

ArcSWAT is an interface to enable some GIS calculations. SWAT is a Fortran program only concerned with the files in TxtInOut directory. These files are manipulated with SWAT-CUP. So after you build your project, there is really no more need for ArcSWAT for that project.

Also, please note that SWAT is a deterministic model. Meaning, it takes only one set of parameters and gives you one single set of responses. I have written before that deterministic approach is not acceptable for environmental projects and the uncertainty in the model prediction must be quantified. SWAT-CUP allows you to quantify this uncertainty as 95PPU by propagating the uncertainty in the parameters. Again, the parameter ranges are the solution to your calibration problem.

Karim  

 

-------------------------------------------------
Dr. K.C. Abbaspour
Eawag, Swiss Federal Institute for Aquatic Science and Technology
Ueberlandstr. 133, P.O. Box 611, 8600 Duebendorf, Switzerland
email: abba...@eawag.ch
phone: +41 58 856 5359
fax: +41 58 856 5375
http://www.eawag.ch/index_EN

[fatemeh babakhani]

Dear Dr. Karim

Thanks for the response!

Actually I am calibrating a watershed of the interest to estimate how much sediment traps behind dams, on the other hand, I am looking for SED_IN and SED_OUT in the Reservoirs. So I had decided to put parameters into SWAT to see how much Sediment traps behind dams.

As you said if I want to keep the range of parameters as inputs, so the solution should be the range of numbers. But when I looked at Output. rsv file in the SWATCUP folder I see just one specific number as a SED_IN, SED_OUT, ... for each month, while I expect to see the range of output! Could you please tell me which input data (Best Estimation or Range of Parameters) is corresponding to the Output. rsv file (SED_IN, SED_OUT,... parameters)?

My next question is that, if the model is calibrated based on the Sediment USGS gage located at the outlet of watershed, will all the entire watershed such as all dams be calibrated? 

I really appropriate your consideration and time.

Regards,

Fatemeh A. Babakhani

Graduate Research Asistant, Department of Civil and Environmental Engineering 

Wayne State University

Detroit, MI 48202 (313) 844 4902 

[Karim Abbaspour]

you need to distinguish between SWAT and SWAT-CUP. SWAT is a deterministic model, meaning it takes one set of parameters and gives you one set of results. Output.rsv is a SWAT output file. You cannot expect that to have a range of values. Same with Output.rch, which has only one set of values.  SWAT-CUP is a parameter estimation program that is stochastic and provides you with a range of parameters that give you acceptable solutions for your inverse problem. You need to propagate the parameter ranges by taking for example, 1000 samples or so and get 1000 set of SWAT output values, which serve as the solutions to your problem. Hence, the concept of model uncertainty. So, you don't have one solution to the problem, you have infinitely many, from which we quantify a band expressed as the  95PPU. So the "best solution" is not the solution to your problem.

If the model is calibrated based on the data at the outlet of the watershed, you cannot say much about what is happening at different locations inside the watershed. The sediment could be coming 10 parts from the east and 90 parts from the west, or 10 parts from the west and 90 parts from the east and you will still have 100 at the outlet.

You need to calibrated the model for different locations within the watershed to be able to say something about what is happening inside the watershed. This is the problem "conditionality" , which I have written a few times about.

Therefore, the model will have no idea what is happening at the dams unless you have some data to calibrate the dam also.

Hope this makes some sense.

Karim

 

-------------------------------------------------
Dr. K.C. Abbaspour

Eawag, Swiss Federal Institute for Aquatic Science and Technology
Ueberlandstr. 133, P.O. Box 611, 8600 Duebendorf, Switzerland
email: abba...@eawag.ch

phone: +41 44 823 5359
fax: +41 44 823 5375
http://www.eawag.ch/index_EN

---

Spam
Not spam
Forget previous vote

 

 

--

Dr. James E. Almendinger
St. Croix Watershed Research Station
Science Museum of Minnesota
16910 152nd St N
Marine on St. Croix, MN  55047
tel: 651-433-5953 ext 19

--
You received this message because you are subscribed to the Google Groups "SWAT-CUP" group.
To unsubscribe from this group and stop receiving emails from it, send an email to swat-cup+u...@ googlegroups.com.

For more options, visit https://groups.google.com/d/ optout.

[fatemeh babakhani]

Dear Dr. Abbaspour

 

Thank you so much! 

 

Your explanations helped me a lot to understand the difference between SWAT and SWATCUP performance, also I read one of your papers which is Estimating Uncertain Flow and Transport Parameters Using a Sequential Uncertainty Fitting Procedure, so now the concepts of 95PPU, SUFI_2, p-factor,... are clear to me.

 

But a part of your response is somehow confusing to me!! In the last email you said "You need to propagate the parameter ranges by taking for example, 1000 samples or so and get 1000 set of SWAT output values, which serve as the solutions to your problem". My question is that if after calibrating the model, I ended up with USLE_P parameter range between 0.2 to 0.5, should I take 1000 USLE_P between 0.2 to 0.5 and plug them into SWAT and run SWAT 1000 times for each USLE_P, to see for example Sediment Load Range that trap behind dams,... ? On the other hand, after finding Parameters Range, how can I find OUTPUT Range (like sediment load in other part of the Watershed) ?

I really appreciate your help,

Fatemeh 

On Wednesday, May 13, 2015 at 5:17:09 PM UTC-4, fatemeh babakhani wrote:

Hello All

I am trying to calibrate a watershed based on Sediment Component (It's already Hydrologically Calibrated!).

 At the first step, I did sensitivity analysis and ended up with 7 sensitive parameters that have been shown in the attachment. Then for calibrating the model, I ran the SWATCUP several times with different parameters ranges! In all of my simulations, most of Peak Sediment Loads were out of 95ppu band!

* I was wondering if any of you have this problem! if yes, could you please tell me how you solved it?!

** p- factor is 0.66 (I believe for sediment calibration it can be less than 0.7.), r- factor is 0.93 which is good, I am just worried about NS which is 0.61 (that should be close to 1)! What do you think, do you think this numbers are acceptable for sediment calibration. 

All required information has been enclosed, if you need further information please let me know.

Thank you so much,

[Abbaspour]

Well in the last iteration of the calibration process you have already propagated the parameter ranges and have your solution. If you want outputs in locations for which you have no observation, then you need to use the No_Observation option in swatcup to also extract them.

If you want to do a scenario and, for example change fertilizer regime, and see the impact of that on water quality, then 

you need to run your model with the parameter ranges a number of times (say 1000) and extract the  the variable of interest (say nitrate) using the No_Observation option and get the 95ppu and compare it with the previous case.  

Hope this is clear!

Karim

--

[vivian]

Dear Fatemeh,

Could you please help me? How did you determine which parameters are sensitive and which are not? I can't find the ranking number in the output files after performing a on-at-a-time sensitivity analysis..

Looking forward to your response!

[fatemeh babakhani]

Hello,

Please go to the Page 59 of User Manual, you will find out how to do sensitivity analysis. 

Hope it helps!

--
You received this message because you are subscribed to a topic in the Google Groups "SWAT-CUP" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/swat-cup/0qJBpIzm5cs/unsubscribe.
To unsubscribe from this group and all its topics, send an email to swat-cup+u...@googlegroups.com.

For more options, visit https://groups.google.com/d/optout.

--

 

Fatemeh A. Babakhani, Ph.D. Candidate
Graduate Research Assistant, Department of Civil and Environmental Engineering

Wayne State University
Detroit, MI  48202    (313) 844-4902

[Felipe Mamedio]

Hello,

I am trying to calibrate a watershed based on Sediment Component, but after the calibration all outputs to sediments are 0. I don't know what I'm doing wrong, because the SWATCUP performs all the steps but not calibrate sediment data. I used subdailly data, CH_EQN=0, Green-Ampt method in ARCSWAT. The outputs in SWATCUP are daily and i did some changes in IEVENT to the program come out daily data.

If someone can help me I would greatly appreciate it.

[Karim Abbaspour]

I don't know why you get 0 sediment. But you are wasting your time trying to simulated or perhaps calibrate sub-daily sediment!

Karim

[Felipe Mamedio]

"I don't know why you get 0 sediment. But you are wasting your time trying to simulated or perhaps calibrate sub-daily sediment!

Karim"

Hi Karim,

I put the subdaily to calibrate better hydrological using green-ampt. Because nash by CN was in 0.55, and with Green-Ampt went to 0.65. But the output i put daily. So I'm calibrating sediments with daily data, the same way that I calibrated the flow.

[Felipe Mamedio]

Hi people. Sorry about my english. I solved my problem with no results to sediments. I was using arcgis 10.1  and no results for sediments was generated. Then I installed the arcgis 10.2 and the results for sediments began to appear. Swat cup wasn't calibrating the sediment data, so I took the executable swat 2012 (in the following way: C:\SWAT\ArcSWAT), renamed as swat, and replace the file of the same name in the folder created for calibration with the swat cup.

[LEGEND LEGEND]

I used SWAT cup for calibration, Q of 3 stations and Sediment data of 2 stations and run swat cup. but got the simulation results for 3 number of Q station and only one number of sediment i.e. could not get the results for one sediment station. what may be the reason for this? NOTE SWAT CUP calibration process run successfully without any error. any  

[Sefa Nur Yeşilyurt]

Hello Everyone,

I hope this email finds you well.

I have set up a SWAT model and conducted calibration using SWAT-CUP. The calibration results are quite satisfactory, as shown below:

Variable           p-factor  r-factor  R2    NS     bR2      MSE       SSQR       PBIAS  KGE  RSR   MNS   VOL_FR  ---  Mean_sim(Mean_obs)   StdDev_sim(StdDev_obs)
FLOW_OUT_2         0.85      1.25      0.81  0.79   0.5745   1.1e+001  5.5e+000   -11.0 0.74  0.46  0.61  0.90            5.81(5.24)           5.74(7.29)

Additionally, I manually calibrated the same model. Interestingly, SWAT-CUP suggested different parameter values during calibration, but the results remained consistently good. I have the option to use either calibration approach.

However, I have encountered a challenge regarding satellite data integration. My aim is to compare satellite data (e.g., NASA POWER) with observation data. To achieve this, I input the satellite data into the model that was initially calibrated with observation data and re-ran the model using the same parameter intervals. Initially, the results were promising, but the parameter intervals were quite broad.

To refine the analysis, I narrowed the parameter ranges to align closely with the best values obtained in the initial SWAT-CUP calibration. Unfortunately, the results degraded significantly.

At this point, I have two key questions:

1. Can satellite data (e.g., NASA POWER) be directly used instead of observation data for calibration and simulation? While the monthly data from NASA POWER aligns closely with the observation data, there is significant deviation on a daily basis.
2. Am I approaching this problem correctly, or would you suggest an alternative methodology to better integrate satellite data into the SWAT model?

I would greatly appreciate your insights and recommendations on this matter.

Thank you for your time and support.

Best regards,

Sefa Nur

18 Aralık 2017 Pazartesi tarihinde saat 10:02:17 UTC+3 itibarıyla LEGEND LEGEND şunları yazdı: