Leanne 0 Report post Posted August 26, 2009 Hello, We are building a 2D only model and are finding big differences in the Cum ME% between the _MB.csv and the _MB2D.csv. In the _MB.csv file the error remains within +/-1% apart from the start of the inflow hydrograph, but the _MB2D.csv has errors which range between -3% and -5% for most of the run following the start of the hydrograph. We are running the model on double precision due to the area being quite steep and with ground levels >100m. In previous model runs we had higher instabilities on the _MB file but these have been reduced by running in double precision, smoothing the inflows and downstream boundary, and improving the position of the inflows. Any suggestions for what may be the reason for the differences in the .csv files? Is the model healthy enough based on the _MB.csv or is there still a significant problem somewhere as the _MB2D.csv file suggests? We have looked at the MB1.dat file at various timesteps but can't find any errors much greater than 1%. Thanks Leanne Quote Share this post Link to post Share on other sites
Bill Syme 0 Report post Posted September 1, 2009 Hi Leanne Can you please email the .tlf and _MB.csv files to support@tuflow.com so that we can get a better idea of the model inputs, timestep, etc. FYI, healthy models should usually have a zero to +/-1% mass error. Cheers Bill Quote Share this post Link to post Share on other sites
katebradbrook 0 Report post Posted October 12, 2010 What was the conclusion on this? I have spotted the same issue in a model I am looking at.... Quote Share this post Link to post Share on other sites
Bill Syme 0 Report post Posted October 15, 2010 Hi Kate (and thanks to Leanne for her correspondence) Yes, this was resolved - extracts from emails that followed are below. Cheers Bill ------------------------------------------------------------------------------------------------ Hi Leanne If you change the QT boundaries [which were placed inside the active area] to SA boundaries it runs fine, with the mass error sitting at 0.1% after an hour or two from the start of the hydrograph. I left the Number of Iterations = 4 (by mistake), so try first with this commented out. I would probably still use the 0.2mm wet/dry depth (you could sensitivity test this one). For this run you can start the simulation after 2 hours as there is nothing flowing into the model until this time. FYI, QT boundaries are primarily designed for external inflows from rivers and streams, and should be located along on the perimeter of the code polygon. SA boundaries are excellent for internal flows to the model (which is the situation you have). QT boundaries tend to produce better velocity distributions across the boundary, hence why they're probably preferred for river/stream inflows, but this wouldn't be important in your model. SA boundaries seem to have much lower mass errors. I've attached the files that I modified and the SA layer (note you need to delete the QT boundaries out of the 2d_bc layer as per the 2d_bc layer attached). Hopefully this should all work well, and please get back to me if you have any queries. Cheers Bill ------------------------------------------------------------------------------------------------ Hi Bill, We've now re-run the Q100 with the files as you suggested as it all looks ok, MB error peaks at just -0.3%. We reduced the no. of iterations back to 2 but left on the 0.2mm wet/dry depth. We're currently running the Q1000 flows but hopefully it should all be ok! We noticed you'd set the IWL in the .tcf, which we left at the value you suggested. Was this value based on it being slightly lower than the level in our downstream boundary at the start of the run, or was it based on something else? Thanks for helping us to sort our model out, you've been really helpful! Cheers Leanne ------------------------------------------------------------------------------------------------ Hi Leanne Excellent and thanks for the feedback. I think I set the IWL because I started the run later into the simulation as for that event there was no inflow till around 2 hours or so. The IWL should always be set to be the same as the water level at your HT boundary at the Start Time, otherwise a "wall" of water can exist at the HT boundary when the simulation starts. Having said that, from memory I don't think there were any wet cells in the model at the start which means all cells were higher than the IWL (and downstream boundary), so in this case setting the IWL is somewhat irrelevant. You may want to check your downstream boundary values as if they are lower than the 2D cell elevations then there is something not quite right. Usually the downstream boundary levels would be above the cell elevations, hence the need for an IWL. I hope it all goes well. Let us know if you have any further queries. Cheers Bill Quote Share this post Link to post Share on other sites