Hi, 

What is the difference between the simulation crashing in Driel and the one in the Waal?

Several things can be happening. Structures with morphodynamics is a tricky thing. It can be that aggradation or degradation lead to an unreallistic flow condition at the structure yielding unreallistic bed level changes. The simulation crashing in Driel, does it crash right downstream of the structure? If that is the case, a practical solution to resolve this issue is to prevent both aggradation and degradation in the cells surrounding the structure, still computing the sediment transport rate. You can see this as the river manager continuously dredging or dumping sediment to keep the bed level at a certain elevation. It is not far from the actual management of the strucutres. Rijkswaterstaat would not allow large aggradation or degradation in the surroundings I would say. This can be done with <morphopol> flag in [sediment] block of the mdu-file (extended feature, not yet in the manual). The input is a polygon where changes are computed. If you check the repository of the RiverLab you will find already a polygon to use this feature. 

This does not explain the crash in the Waal. Where in the Waal? A key point in general is that aggradation cannot exceed the elevation of the points at which the main-channel width is found. This elevation should be more or less the groyne crests. Is aggradation prior to crashing excessive?

The drop in grain size, is it prior to crash? or is the bed level already unreallistic? 

Not sure what your maximum discharge is but be aware that fixing the water level downstream will cause a large backwaer effect. 

Hi Victor,

Thanks for your reaction.

The differences between the simulation crashing in Driel and in the Waal are:
1) The simulation crashing in Driel was done using a windows-pc and the simulation crashing in the Waal was done using a Linux-pc. Furthermore, the frequency of storing data in the HIS-file was increased from 1 time per year to 4 times a year.

The simulation that crashes near Driel indeed crashes just downstream of the weir. However, this happens within one year. The simulation in the Waal crashes just after the bifurcation with the Pannerdensch Kanaal.

I see a general pattern that for the first 70 years the Rijn, Waal, Pannerdensch Kanaal and the upper parts of the IJssel are degrading (with various speeds). Furthermore, the D50 in these rivers goes up and the discharge distribution near Pannerden shifts in favor of the Waal.

Most rivers have degradation in order of 0.5 meter after 70 years, but the Waal and Pannerdensch Kanaal show values that exceed 2.5 - 3 meter, which are way higher than described in literature.

After 70 years the grain sizes drop considerably and sediment transport of finer fractions rises. Furthermore, the bed starts to aggravate and within 5-10 years all degradation of the previous 70 years is made undone.

My maximum discharge is 8000 m3/s and I constructed a ‘representative’ hydrograph based on the discharge data from 1994-2019.
I have attached some gifs of my simulation here: https://imgur.com/a/14xDJMr

Kind regards,

Thorvald

A different behaviour in Linux than in Windows is a clear symptom of a strong instability. There are small differences in the compiler only visible when something goes really bad (e.g., handling of NaN or division by 0). I would propose that you fix the bed using Morphopol downstream of the structures. Also, check that the bed level does not exceed the maximum in a cross-section (i.e., the elevation of the highest point forming the main-channel width). 

Hi Victor,

I've searched for the 'Morphopol' statements, but I could not find them in any of the manuals,  nor on online forums or in other scripts. Do you have a template for these statements?

Furthermore, I've checked if the bed level exceeded the highest point in the main channel, but this is not the case.
Third, I've looked at the layers and sediment transports. I've made stacked plots of the layers of the different branches. In these plots I see that some moments the lowest layer comes up. Some examples are given in this link: https://imgur.com/a/aeSJflF. This behaviour seems strange for me, as I'd expect bed aggradation to happen in the upper layer(s), and not all the way down.

For the sediment transport, I see in my results that there happens to be a sudden increase in fine-graded sediment transport, which happens to come out of nowhere. It happens both in the upstream region (near Wesel) and at the bifurcation of the Rhine into the Waal and the Pannerdensch Kanaal. Do you have any idea how this can happen as at these points the layers thickness is close to zero meters?

Kind regards,
Thorvald

Hi, 

Indeed, as mentioned above, <morphopol> is not a general-available feature and it is not explained in the manual. You can still use it by adding a flag in the [sediment] block of the mdu-file. E.g.



very recently (only available in the latest dimr release, I implemented the flag <InMorphoPol>. This allows chosing whether inside the polygon the bed level is updated and outside is not (InMorphoPol=1) or viceversa (InMorphoPol=0). The default (which is the only option available before) is 1. 

Regarding your plots, are you sure it is correct? It seems that the thickness increases from bottom to top. It also seems that the active layer has a thickness of about 4 m. 

Hi Victor,

Many thanks for the MorphoPol feature! Now it is working! 


I made plots from the f004-grid and the f007 grid. It appears that only the X/Y-values of the nodes and edges are different and the branch/offset is equal for both models. Also, the .mdu-files appear to be similar. Clearly, X/Y-related features (such as the sediment calibration constants and sediment distribution are different.

Are there any other differences between the f004-case and f007-case that are not related  to X/Y-coordinates?

Thanks in advance!

Thorvald

Hi Thorvald,

Beste wensen! 

No, there should be no differences between the two simulations apart from the straightening of the domain (and of the spatial dependent input).

Regards, 

V

Dear Victor or other members of this forum, 

I have a problem that seems to be similar to Thorvald's problem as explained above. The model appears to behave as expected for approximately 90 years, but after that, the bed starts to aggrade, especially in the locations where it degraded before. This is accompanied by a rapid decline in the sediment grain size. I have attached a visualisation of the bed level difference and the sediment grain size here: https://imgur.com/a/FkD8Yvo
 
I am using the f_009 model with the Delft3D FM version 2022.02. As boundary conditions, I have used the boundary conditions for 1995-2019 as provided in the model, which are repeated four times to obtain the boundary conditions for 100 years. I have already implemented the morphopol feature in the mdu before running and the file is also saved in the folder where the mdu is stored. I have not changed anything to the morphopol feature compared to the one provided in the repository.

Also, I was wondering whether there is also a morphopol feature available for the curved version of the model (e.g. f_008 but with morphodynamic development switched on)? I tried to make a polygon around the cells around the weirs myself, but that does not seem to go right. I think that may have something to do with cutting off edges. If it is not available, could you give me a hint on how to make it myself? Should I follow certain edges?

Finally, I get a large number of this error for various cross-section definition ID:
** WARNING: Total area behind levee should be larger then flow area behind levee. Cross-Section Definition id: 0001.
Is that something I should worry about or is it no problem that this warning occurs? It happens with both the straight (f_009) and curved model (f_008 with morphodynamic development turned on).
 
Thank you in advance.
Kind regards, Birgit

First thing to check is how does the hydrodynamics look like when you see the large changes in grain size. Also, how did you join the hydrographs? Is there a sudden jump in discharge? After the hydrodynamics, have a look at the bed level. Are the changes in grain size accompanied by changes in bed level? Where do these changes occur? After structures? After the bifurcations?

It may be interesting to make plots of the longitudinal profile (contrary to top-view plots) in which you can easily visualize bed level and grain size at the same time. Even better if you plot the hydrograph on top. Have a look at the routines in the OET to plot 1D data. This is also explained in the webinar.

In essence, try to exactly locate where and when the large changes occur to see whether it is realistic or not.

Regarding software version. I understand from your post that you are using the release version and directly calling the executables. Is that correct? You can also use the executables I sent you with the option <Pure1D=2> in the curved model.

About <MorphoPol>. I made them for the straight model only, indeed. Not long ago I implemented that you can set whether the area inside the polygon is updated or not (see post above). You simply need to make a set of polygons that enclose the part of the grid that you don’t want to update. There are several ways to do it. Crude way is to simply write down the coordinates (in RD New, EPSG 28992) in a file (see the example file for the format). You can also draw polygons in Google Earth and convert them to pol-files using the OET (in particular, <EHY_convert>).

About the warning of the cross-sectional area. The FM model comes from the SOBEK 3 model. The cross-sections from the SOBEK 3 model come from applying WAQ2Prof, which is a procedure for deriving cross-sections out of a WAQUA model, which eventually comes from Baseline. For some reason, the result from WAQ2Prof is wrong. We are currently changing the procedure to derive cross-sections. In any case, the model is calibrated with these values.

Dear Victor, 

Thank you for your help. Unfortunately, I still have not found the solution to my problem. 

I ran the simulation again, this time using the f_004 model instead of the f_009 model, but this yields the same results. I made some 1D plots of the simulation, which I have uploaded here (4 images): https://imgur.com/a/0mED4V5 (they are not very neat yet, since I use them for a first analysis).

From the elevation profile of the Neder-Rijn and Lek it can be seen that the morphopol is indeed correctly implemented (there is no morphological development at the locations of the weirs). It can again clearly be seen that in the last +/- 10 years, the grain size drops and the bed elevation increases rapidly. The changes appear to enter the system at the upstream boundary and then quickly travel downstream. Also, the changes occur over the whole stretch of the river (with the most changes at the upstream end) and not just around structures. 

The hydrograph indeed now still shows a jump from one series to the next. However, this is around 75 years, rather than around 90 years. Furthermore, Thorvald has used the same hydrograph and he does not experience these sudden changes. I have also included the hydrograph here: https://imgur.com/a/0mED4V5. 

I am uncertain what causes this sudden change in behaviour around 90 years, do you have any idea what may cause it and what I can do to prevent this behaviour?
   
About the software version, I have indeed used the option <Pure1D=2> in the curved model. These results are however from the straight model since I was still working on the morphopol for the curved model. 

Furthermore, I have made a morphopol for the curved model. I am now running a simulation to check whether it indeed works. If it works, I will make sure to share it.

Kind regards, 
Birgit

Dear Victor and other members of the community, 

I have a small addition to my previous post: I have also made a plot of the bed load transport for each sediment fraction, which I also added to the Imgur link. it shows that there is a sudden increase in the bed load transport of the finest fraction. Since it is so sudden and no 'special' things are happening in the simulation at that point, I think it must be related to some numerical effect. Do you have any idea what may cause this?

Kind regards, Birgit

Do I understand correctly that a wave of fine sediment enters the model and travels the Boven-Rijn and when it reaches the bifurcation it causes unreallistic behaviour? That would be interesting... Note that there is a nodal point relation setting how much sediment and from which type goes to each branch as a function of the discharge ratio. The first thing I would do is to check if indeed there is a hump of fine sediment entering the model and try to understand why. A movie showing the hydrograph above with a dot at the current time and the bed level along the river as you show would help in clearly relating the input of sediment to an abnormal high or low discharge. 

Can you confirm the type of morphodynamic boundary condition applied? Check the dia-file to see what the model reads from the input.