Jump to content
GMS, SMS, and WMS User Forum


  • Posts

  • Joined

  • Last visited

Everything posted by Michal

  1. I would say that MF-USG is very different in terms of how the solution to the flow equation is aproached compared to the traditional structured modflow. However it is usually not too difficult to set up a MF-USG simulation in GMS. There are some tutorials out there that could help. I would recommend reading the MF-USG manual as well to grasp the internal differences. Regarding your second question. I believe there are other people here that could provide more accurate answer. The problem comes from the way MODFLOW solves the groundwater flow equation, that is by numerical approximation. During these approximations (in MF-2005) if head in a cell falls bellow the cell bottom the cell is assumed dry and is inactivated (becomes no-flow cell). However in subsequent approximation the cell could become wet and active again if certain criteria are met. For instace if head in the cell below (i,j,k-1) is 0.1 m above the cell (ijk) bottom. The 0.1 m would be the user specified threshold for conversion from dry to wet. This whole process brings instability into the solution and may cause convergence problems especialy if poor initial head values and low threshold values are specified. However MF-NWT adressed this issue in a brilliant way. It does not deactivate dry cells, but instead it uses upstream weighting approach to limit flow from dry cell to a neighboring partialy saturated cell. It also uses additional smoothing for conductance and storage functions which allows the use of newton numerical method for outer iterations. It therefore provides a continuous solution for all unconfined groundwater flow conditions and is very robust. However if you have steep aquifer your problem is different. Consider two neighboring cells of 10 m thickness. Lets say, under extremely steep conditions they share only 1 m thickness. However structured modflow will assume 10 m anyway and would not modify transmissivity accordingly. Such a solution would be incorrect. That is why I recommend using MF-USG in your case as mentioned earlier in my previous post. For literature I suggest reading the manuals published by USGS for modflow 2000, 2005, NWT and USG. There are plenty of further references there too.
  2. Hi, MODFLOW assumes flat continuous aquifer. You could probably overcome the drying/wetting problem by using Newton linearization implemented in MODFLOW-NWT. However, if there are large not negligible differences in elevation between two horizontaly adjacent cells MODFLOW results will be incorrect. Fortunately the finite volume formulation in MODFLOW-USG adressed this issue - it involves true interface area into the calculation. If I remember it right, this is supplied for each cell connection in the DISU file somewhere. If you use GMS, this should be done automaticaly for you. You could use the upstream weighting (LPF package) and newton linearization (SMS package) for a more robust solution. With this setup the model should converge to a correct solution.
  4. Ok, I see your dilemma, since the model was already calibrated. Try loading the .nam file via Read Solution command. It should have the same format as .mfn and it should load the binary output data referenced there. You could also try to run the model with the MF executable that worked through GMS via custom Run dialog.
  5. Hi T, I didnt understand what you do, so forget my last comment :). I dont know much about this topic. It would be nice if GMS would allow us to do "Interpolate to modflow layers" command with RIV package (to prepare the river bottom/elev dataset and head), but i think only Head in GHB package could be distributed this way. If you decide to model it with RIV, which would be my choice, you need to prepare the package outside GMS.
  6. It may be a long shot as I dont have much details about the model itself, but there might have been some changes in the default values for the solver between MF versions.
  7. If I understand it right, you have raster dataset of measured heads and you basicaly want your model to reproduce these as close as possible. Why not just convert the raster data to points a treat each point as observation point? Then you can formulate the inversion problem as usual.
  8. Hello, what kind of errors are there during the check? If you were able to load the simulation into gms, you should also be able to read the solution. Try right clicking on the simulation (MODFLOW in the project tree) and select Read Solution command. Then just select the modflow name file and the binary data should load.
  9. Michal


    You are welcome, I am happy it worked
  10. Michal


    Hi Kanchana, its realy hard to tell what may be causing the troubles. I guess it will be something in the reaction module related to the reaction rate. Try to run the simulation for lets say a minute with time steps in the order of seconds or fractions of a second. Also increasing grid resolution may improve the solution.
  11. Michal


    I guess it can. For instance if the reaction module causes abrupt changes in concentration of some specie. Does it run ok with MT3DMS? I mean without the reactions only as conservative transport?
  12. Michal


    Hi Kanchana, on the stress period dialog, try to increase the Max. trans. steps variable. It is the maximum number of transport steps allowed during each time step of a flow solution. Give it a high value like 1e6 and see what happens. While performing transport simulation with MT3D based model, time steps are further divided into transport steps. When you leave 0.0 as transport step size, the model calculates transport steps size itself according to certain criteria based on the flow solution. These stability criteria relates among others to the Courant number, you can find more about these in the MT3DMS manual (Zheng, Wang, 1999).
  13. Hi, I would like to see a functionality in GMS, that would make working with pumping data more easy. Let me make an example to explain what I mean. Consider we would like to import pumping schedule for a single well into a transient model with three periods with a length of 30 days each. The well Q for first period was -100, for second the well was off and for third it was -200. The pumping schedule extracted from a database has this format: well_name time Q well1 1 -100 well1 61 -200 I would like GMS to ask me during import, how do I want to treat the stress periods without any pumping data, in this case the second period. I would like to have an option to choose, that when there are no data for a given period GMS would add 0 for that period automaticaly, e. g. the well is off for that period. This would be realy useful, because otherwise one easily forgets to add the zeroes manualy and ends up with incorrect schedule.
  14. Hi Niklas, I have encountered similar problem with MODFLOW-NWT. After an update MF-NWT started to report head below cell bottom error and a model that run ok with previous version was terminated due to this. I solved it by increasing the head slightly as it was only a very small difference. My point is that MF developers added a check module that was not in the old version and you are probably strugling with something similar. MF2K5 being smarter in finding the errors then MF2K (or the SFR2 module). As the best option of fixing the errors is not feasible, you could try looking into MF2K5 change log and maybe try to use older version of MF2K5 that may not contain the check. Or you could disable the check module in the SFR2 source code and recompile MF2K5 yourself. It is a matter of compiling MF2K5 with ICHKSTRBOT_MODULE located in gwf2sfr7.f modified accordingly. Just be aware that you may get incorrect solution as the check has its purpose.
  15. Hi Carrie, you can uncheck the Enable saving of computed flows for all source/sink objects in the Output control.
  16. Hi speedy, I have also struggled with krigging setting not saved properly. But this was solved in GMS 10.3.4. I have recently succesfuly solved inverse problem with Aquaveo MODFLOW model using PEST. There was pp kriging set up for several parameters. Although it worked ok for me , I would still prefer GMS to use native PEST tools for pp interpolation and PEST structure files for storing interpolation setting as it would allow us to use native USGS modflow for calibration with pilot points very easily. Anyway, thank you for sharing your experience on the forum. It is always very helpful. Best Michal
  17. Hi, mt3d is trying to read the saturated thickness array from the flow-transport link file (.hff) on unit 10. I suggest you to check that the .hff file is written correctly by modflow. It must be allowed in the GMS output control dialog.
  18. Hi, I think the flow could be calculated as a difference between the head in the model cell and corresponding head at the boundary. Conductance is the proportionality constant. Therefore Q = C (h1 - h2). If you know the head at the boundary (in the neighboring aquifer) and in the model cell, conductance could be easily calculated from the above formula for each model cell and the flow through the boundary would be as specified for each cell.
  19. Thank you for sharing the solution. Nice one Have you used some sort of interpolation to obtain modflow compatible dataset from the raster data? Or is the grid and rasters identical in number of rows and columns?
  20. Hi, try preparing the transient dataset so, that for each polygon in GMS coverage, there is a corresponding timeserie with the polygon id assigned at each row of the dataset. It will be a large dataset, but GMS should handle it unless you run out of memory.
  21. I agree with Hisham on this. There is no such concept as multipart polygon in GMS, as far as I know.
  22. Hello, this is interesting. Maybe there are 553 multipart polygons in the shapefile and when converting to feature object GMS performs multipart -> singlepart conversion. http://desktop.arcgis.com/en/arcmap/10.3/tools/data-management-toolbox/multipart-to-singlepart.htm
  23. Hi Bruce, that would be very nice GMS feature. It could be even so, that head observations for different coverages would fall into different observation groups in the PEST control file. Also parameter grouping with custom naming would make thinks a lot easier. I can imagine, you could specify PEST parameter group in the Parameter dialog for each entry.
  24. Hi, I think it could be possible by interpolating transient TIN data (or maybe even rasters) into Recharge package using Interpolate to Modflow layers command. But I have not done it before. Let us know how you manage to handle it.
  25. Helo, try using MODFLOW-NWT instead of MODFLOW-2000. Its excelent in handling problems with cell drying/rewetting. In the LPF package set all layers to convertible. If insist on using MF-2000, maybe try different timestepping scheme to avoid overshoots.
  • Create New...