RE: Medium-term Modelling using delft3d (stationary or non-stationary) - Forum - Delft3D-Old
Message Boards
- Home
- Coastal/Estuary
- RE: Medium-term Modelling using delft3d (stationary or non-stationary)
RE: Medium-term Modelling using delft3d (stationary or non-stationary)
RATHNASINHA JAYATHILAKA, modified 7 Years ago.
Medium-term Modelling using delft3d (stationary or non-stationary)
Youngling Posts: 8 Join Date: 12/14/14 Recent Posts
Hi,
I’m running Delft3d Wave-Flow model using schematized wave conditions (16 conditions). I need bits of advice regarding with the computational mode. I have run test cases for non-stationary and stationary mode for a particular wave condition (Offshore boundary hs=2.4m, wave dir=230, hp=11s, wind speed=8 m/s and wind dir=250).
Offshore sea boundary is ~35 Km from the coastline.
Case1) non-stationary, time points for wave computations as 00h 01h 02h 03h 4h 05h (time step= 10 min time interval=30 min) (write and use hotstart on)
Case2) non-stationary, time points for wave computations as 00h 02h 04h 06h 08h 10h (time step= 10 min time interval=30 min) (write and use hotstart on)
Case3) stationary (time interval=60 min)
Pls see attach. for results of case1,2 and 3
Q 1: it seems that both non-stationary cases showed same wave outputs even if I run different computational time i.e case 1 run 5 hours and case 2 run 10 hours. Can you pls advice me which time step, interval and computations should you recommend such a case study (suppose enough residential time and distance to be non-stationary)?
Q2: The model is running for 16 tide cycles (after each multiplied by morfac). If I used stationary mode then what would you recommend for time step? and write and use hotstart ?
Thanks,
Ruchii
Ps: it would be appreciated if you can give me some references regarding with above.
I’m running Delft3d Wave-Flow model using schematized wave conditions (16 conditions). I need bits of advice regarding with the computational mode. I have run test cases for non-stationary and stationary mode for a particular wave condition (Offshore boundary hs=2.4m, wave dir=230, hp=11s, wind speed=8 m/s and wind dir=250).
Offshore sea boundary is ~35 Km from the coastline.
Case1) non-stationary, time points for wave computations as 00h 01h 02h 03h 4h 05h (time step= 10 min time interval=30 min) (write and use hotstart on)
Case2) non-stationary, time points for wave computations as 00h 02h 04h 06h 08h 10h (time step= 10 min time interval=30 min) (write and use hotstart on)
Case3) stationary (time interval=60 min)
Pls see attach. for results of case1,2 and 3
Q 1: it seems that both non-stationary cases showed same wave outputs even if I run different computational time i.e case 1 run 5 hours and case 2 run 10 hours. Can you pls advice me which time step, interval and computations should you recommend such a case study (suppose enough residential time and distance to be non-stationary)?
Q2: The model is running for 16 tide cycles (after each multiplied by morfac). If I used stationary mode then what would you recommend for time step? and write and use hotstart ?
Thanks,
Ruchii
Ps: it would be appreciated if you can give me some references regarding with above.
Attachments:
Ben Williams, modified 7 Years ago.
RE: Medium-term Modelling using delft3d (stationary or non-stationary)
Padawan Posts: 59 Join Date: 3/23/11 Recent Posts
Hi Ruchira,
Others may correct me on this but:
* Unless you are considering waves generated or appreciably modified by wind between the boundary and your area of interest, stationary computations are the way to go. You might glean some idea about how important about stationary vs non-stationary wave simulations are by comparing predictions from your model with that measured by wave buoy at or near your site of interest. You could also try analysing the relationship between wind speed and direction and the wave hieght, period and direction you are applying at your model boundary.
* For coastal applications you don't normally need to worry about writing hotstart files for SWAN. Numerical convergence is usually achieved fairly rapidly if you have specified your directional and frequency resolution adequately.
* If you are simulating morphology, it usually pays to keep your model as simple as possible. Some find this a good approach as it forces you to think about exactly what process you are trying to capture in your model and how you should interpret the results.
* Setting the coupling interval between WAVE and FLOW model is usually a compromise between computational accuracy (in theory, more frequent coupling is better) and computational efficiency (less frequent coupling is better). If your site is tidal then as a very rough first guess hourly coupling is a practical and useful interval. You will need to check if increasing or decreasing the coupling interval has an influence on your final morphological prediction. If your site is very tidal (range or rapidly changing wave and current conditions) then this can be an indicator that you need to have more frequent coupling.
* Morphological models are typically in the order of 1km offshore vs 2-5km alongshore. Rather than the 160km you seem to be implying.
* Writing hotstart for FLOW model can be useful and I would suggest either 12.5 hours (i.e. tidal cycle) or 24 hours. But really you only need to write hotstart files if you think you will need to restart your simulation halfway through. Otherwise you can forget about them.
* It might not be that surprising that you are getting the same results for (Case 1) and (Case 2), as you are essentially applying a static wave condition at the boundary and applying a static wind conditions across the model domain. The difference between (Case 1 + 2) and (Case 3) might be one of numerical convergence, but it is hard to say with the information provided.
* References: There are rather numerous references that a google search will reveal. Papers and books contributed by Dano Roelvink are the place you should start.
Success,
Ben
Others may correct me on this but:
* Unless you are considering waves generated or appreciably modified by wind between the boundary and your area of interest, stationary computations are the way to go. You might glean some idea about how important about stationary vs non-stationary wave simulations are by comparing predictions from your model with that measured by wave buoy at or near your site of interest. You could also try analysing the relationship between wind speed and direction and the wave hieght, period and direction you are applying at your model boundary.
* For coastal applications you don't normally need to worry about writing hotstart files for SWAN. Numerical convergence is usually achieved fairly rapidly if you have specified your directional and frequency resolution adequately.
* If you are simulating morphology, it usually pays to keep your model as simple as possible. Some find this a good approach as it forces you to think about exactly what process you are trying to capture in your model and how you should interpret the results.
* Setting the coupling interval between WAVE and FLOW model is usually a compromise between computational accuracy (in theory, more frequent coupling is better) and computational efficiency (less frequent coupling is better). If your site is tidal then as a very rough first guess hourly coupling is a practical and useful interval. You will need to check if increasing or decreasing the coupling interval has an influence on your final morphological prediction. If your site is very tidal (range or rapidly changing wave and current conditions) then this can be an indicator that you need to have more frequent coupling.
* Morphological models are typically in the order of 1km offshore vs 2-5km alongshore. Rather than the 160km you seem to be implying.
* Writing hotstart for FLOW model can be useful and I would suggest either 12.5 hours (i.e. tidal cycle) or 24 hours. But really you only need to write hotstart files if you think you will need to restart your simulation halfway through. Otherwise you can forget about them.
* It might not be that surprising that you are getting the same results for (Case 1) and (Case 2), as you are essentially applying a static wave condition at the boundary and applying a static wind conditions across the model domain. The difference between (Case 1 + 2) and (Case 3) might be one of numerical convergence, but it is hard to say with the information provided.
* References: There are rather numerous references that a google search will reveal. Papers and books contributed by Dano Roelvink are the place you should start.
Success,
Ben
Ruchira JAYATHILAKA, modified 7 Years ago.
RE: Medium-term Modelling using delft3d (stationary or non-stationary)
Youngling Posts: 8 Join Date: 12/14/14 Recent Posts
Hello Mr. Ben,
Thank you for the information and the clarification on issues. I will compare predictions from my model with available wave buoy at near-shore to understand about stationary vs non-stationary wave simulations and update you.
I appreciate your quick reply.
Kind regards,
Ruchira
Thank you for the information and the clarification on issues. I will compare predictions from my model with available wave buoy at near-shore to understand about stationary vs non-stationary wave simulations and update you.
I appreciate your quick reply.
Kind regards,
Ruchira
Ruchira JAYATHILAKA, modified 7 Years ago.
RE: Medium-term Modelling using delft3d (stationary or non-stationary)
Youngling Posts: 8 Join Date: 12/14/14 Recent Posts
Hi Ban,
Thanks for your valuable instructions.
I have done D3D-Wave stationary wave transformation with space varying boundary condition (.bcw) and the wind as MeteoFile (.wnd). ERA-Interim was used as off-shore wave boundary and NCEP was used for wind filed.
Different simulations were carried out and the model was partly calibrated with JONSWAP friction value. The model result is compared with measured near-shore wave buoy (attached here).
Interesting to remark is that, for the measured wave buoy, an underestimation of wave height was observed during storm periods. I thought the fact that hindcasted waves are underestimated at the buoy location during the stormy periods points out to some principal limitations of delft3d-wave.
Can you pls advice me on what parameters should I focus (calibrate) to obtain more accurate model results during the storm days?
Cheers
Have a nice weekend.
-Ruchira
Thanks for your valuable instructions.
I have done D3D-Wave stationary wave transformation with space varying boundary condition (.bcw) and the wind as MeteoFile (.wnd). ERA-Interim was used as off-shore wave boundary and NCEP was used for wind filed.
Different simulations were carried out and the model was partly calibrated with JONSWAP friction value. The model result is compared with measured near-shore wave buoy (attached here).
Interesting to remark is that, for the measured wave buoy, an underestimation of wave height was observed during storm periods. I thought the fact that hindcasted waves are underestimated at the buoy location during the stormy periods points out to some principal limitations of delft3d-wave.
Can you pls advice me on what parameters should I focus (calibrate) to obtain more accurate model results during the storm days?
Cheers
Have a nice weekend.
-Ruchira
Attachments:
Ben Williams, modified 7 Years ago.
RE: Medium-term Modelling using delft3d (stationary or non-stationary) (Answer)
Padawan Posts: 59 Join Date: 3/23/11 Recent Posts
Hi Ruchira,
To me that looks like a good calibration and you should be proud.
If I am to understand what you have written, you have used a global/ regional wave model as boundary conditions to your 'coastal scale' model, and then calibrated certain numerical parameters to try and best reproduce integral wave parameters (Hm0, Tp, MWD etc.) measured at a wave buoy near your site of interest.
It is pretty difficult to get a global model to reproduce fine-scale wave processes, so in part your results show the quality of the ERA-interim model you have used as boundary conditions. However, this means that your under-prediction may not be an artefact of your Delft3D model but instead an artefact of the ERA-interm model at the location you have used as boundary conditions. If true, it will be hard for you to 'bump up' your wave height at the calibration point without adding energy - e.g. increasing wind speed a bit.
In terms of parameters in vary in the calibration, you have investigated JONSWAP. The other parameters one could try might be:
* Other bed friction formulations
* Wind growth processes
* Whitecapping formulation
* Depth-indiuced braking alpha and gamma values
* Accuracy/ resolutrion of your frequency spectrum
* Discretisation of your directional spectrum
* Presence of surge during storms
The wave height is most underestimated during the storm so I might try with increasing wind speeds first.
You don't show your bathymetry but if there are strong 2D features - such as canyons or offshore sandbanks - then these can act to strongly refract waves depending on thier incident direction and frequency. If these features are present then I would try increasing your directional and frequency resolution.
If your wave buoy is in shallow water then you might need to look at your wave breaking coefficeints in more detail.
If your calibration site has directional spectra available then this is often very useful for drilling down in to the details of how well your model is performing - is it getting the frequency shape correct, is it getting the peak and mean directions correct.
Lastly, models calibration are usually also quoted as error (such as Mean Absolute Error, Bias, Scatter Index, Correlation Coefficeint, Skill, Root Mean Square Error, etc.) to the measurements. If you are unable to improve your calibration then at least you will know to what accuracy your model results are.
Cheers,
Ben
To me that looks like a good calibration and you should be proud.
If I am to understand what you have written, you have used a global/ regional wave model as boundary conditions to your 'coastal scale' model, and then calibrated certain numerical parameters to try and best reproduce integral wave parameters (Hm0, Tp, MWD etc.) measured at a wave buoy near your site of interest.
It is pretty difficult to get a global model to reproduce fine-scale wave processes, so in part your results show the quality of the ERA-interim model you have used as boundary conditions. However, this means that your under-prediction may not be an artefact of your Delft3D model but instead an artefact of the ERA-interm model at the location you have used as boundary conditions. If true, it will be hard for you to 'bump up' your wave height at the calibration point without adding energy - e.g. increasing wind speed a bit.
In terms of parameters in vary in the calibration, you have investigated JONSWAP. The other parameters one could try might be:
* Other bed friction formulations
* Wind growth processes
* Whitecapping formulation
* Depth-indiuced braking alpha and gamma values
* Accuracy/ resolutrion of your frequency spectrum
* Discretisation of your directional spectrum
* Presence of surge during storms
The wave height is most underestimated during the storm so I might try with increasing wind speeds first.
You don't show your bathymetry but if there are strong 2D features - such as canyons or offshore sandbanks - then these can act to strongly refract waves depending on thier incident direction and frequency. If these features are present then I would try increasing your directional and frequency resolution.
If your wave buoy is in shallow water then you might need to look at your wave breaking coefficeints in more detail.
If your calibration site has directional spectra available then this is often very useful for drilling down in to the details of how well your model is performing - is it getting the frequency shape correct, is it getting the peak and mean directions correct.
Lastly, models calibration are usually also quoted as error (such as Mean Absolute Error, Bias, Scatter Index, Correlation Coefficeint, Skill, Root Mean Square Error, etc.) to the measurements. If you are unable to improve your calibration then at least you will know to what accuracy your model results are.
Cheers,
Ben
Ruchira JAYATHILAKA, modified 7 Years ago.
RE: Medium-term Modelling using delft3d (stationary or non-stationary)
Youngling Posts: 8 Join Date: 12/14/14 Recent Posts
Hello Mr Ben,
Many thanks for your detailed and prompt answers and it would be highly appreciated. I would like to emphasise the important of this forum and we learnt a lot from these conversations.
Ben, I would like to add nested grid for excising larger wave grid to understand the near-shore sediment transport. But as I understand, a combination of spatial varying meteo files with nesting grid gives following error (you also had the same experience, pls see link below).
*** MESSAGE: Delft3D-WAVE runs stand alone
Done reading input
*** MESSAGE: Using data from the following FLOW domain:
slkwav
*** ERROR: Reading file "com-XXXX.dat" or "com-XXXX.def"
http://oss.deltares.nl/web/delft3d/d-flow-flexible-mesh/-/message_boards/view_message/338470
Was there any update on this issue?
Please let me know any other way of handling such situation.
Thanks.
Cheers
Many thanks for your detailed and prompt answers and it would be highly appreciated. I would like to emphasise the important of this forum and we learnt a lot from these conversations.
Ben, I would like to add nested grid for excising larger wave grid to understand the near-shore sediment transport. But as I understand, a combination of spatial varying meteo files with nesting grid gives following error (you also had the same experience, pls see link below).
*** MESSAGE: Delft3D-WAVE runs stand alone
Done reading input
*** MESSAGE: Using data from the following FLOW domain:
slkwav
*** ERROR: Reading file "com-XXXX.dat" or "com-XXXX.def"
http://oss.deltares.nl/web/delft3d/d-flow-flexible-mesh/-/message_boards/view_message/338470
Was there any update on this issue?
Please let me know any other way of handling such situation.
Thanks.
Cheers
Ben Williams, modified 7 Years ago.
RE: Medium-term Modelling using delft3d (stationary or non-stationary)
Padawan Posts: 59 Join Date: 3/23/11 Recent Posts
Hi Ruchira,
Thanks for putting up the post to that email thread! I was unaware that there was continued discussion in the years after I posted, and there is some useful and relevant information there.
In the end in desperation I had to resort to using SWAN as a stand-alone wave model, and prepared the wind files in a format that it could read. My situation was different as I was only after nearshore wave parameters, not boundary conditions for sediment transport.
Unfortunately I'm not able to advise for your specific situation and I don't know if Deltares ever solved the problem.
If you do discover a solution, please post your findings here :-)
Best regards,
Ben
Thanks for putting up the post to that email thread! I was unaware that there was continued discussion in the years after I posted, and there is some useful and relevant information there.
In the end in desperation I had to resort to using SWAN as a stand-alone wave model, and prepared the wind files in a format that it could read. My situation was different as I was only after nearshore wave parameters, not boundary conditions for sediment transport.
Unfortunately I'm not able to advise for your specific situation and I don't know if Deltares ever solved the problem.
If you do discover a solution, please post your findings here :-)
Best regards,
Ben
Ruchira JAYATHILAKA, modified 7 Years ago.
RE: Medium-term Modelling using delft3d (stationary or non-stationary)
Youngling Posts: 8 Join Date: 12/14/14 Recent Posts
Hi Ben,
I have recently complied tagged version 6906. With this version i have resolved the above problem.
A two domain model (with a finer resolution domain nested within the coarse resolution domain) but with wind applied only to the coarse domain via MeteoFile keyword work without any error. The xwind and ywind files can be inputted as both .wnd or amu/amv formats.
Thanks.
Cheers
Ruchira
I have recently complied tagged version 6906. With this version i have resolved the above problem.
A two domain model (with a finer resolution domain nested within the coarse resolution domain) but with wind applied only to the coarse domain via MeteoFile keyword work without any error. The xwind and ywind files can be inputted as both .wnd or amu/amv formats.
Thanks.
Cheers
Ruchira
yasmine Lee, modified 1 Year ago.
RE: Medium-term Modelling using delft3d (stationary or non-stationary)
Youngling Posts: 2 Join Date: 3/4/23 Recent Posts
Based on the information provided, it appears that Delft3D FM is a powerful modelling package that can simulate a wide range of physical and chemical processes in coastal and estuarine systems. It can be used for research purposes, as well as operational tasks such as storm, surge, and algal bloom forecasting.One potential area of research or discussion could be the accuracy and reliability of the modelling results produced by Delft3D FM. As coastal systems are complex and subject to a variety of forces, it may be challenging to accurately simulate all of the relevant processes. Therefore, it could be beneficial to share experiences and compare results with other researchers and users of Delft3D FM.Additionally, it could be interesting to explore the potential applications of Delft3D FM in predicting the impacts of climate change on coastal and estuarine systems. As sea level rise and changes in storm patterns are expected to have significant effects on these environments, modelling tools like Delft3D FM could be valuable in predicting and mitigating these impacts.
Cheers,
Yasmin
party bus rental orange county
Cheers,
Yasmin
party bus rental orange county