Bruce Campbell 2 Posted April 17, 2020 Report Share Posted April 17, 2020 Hi -- I'm getting different water-level results with GMS 10.4.7 files when I switch from a PEST calibrated binary version to native text files. Here's the workflow: 1) Build and calibrate the models using PEST in GMS, 2) Get good results and acceptable calibration, using pilot points for aquifer horizontal hydraulic conductivities (HK) and specific yield (Sy) (for the surficial aquifer), 3) Interpolate the calibrated pilot point HK and Sy data to a 2-D grid created from the 3-D grid (same cell size and alignment) with the same interpolation methods as PEST used, 4) Replace the parameters (pilot points and drain conductances) in a new set of GMS files with calibrated values, 5) Used the Kriging interpolation method in both the PEST calibration process and when the resulting pilot point set is interpolated to the 2D grid in the native text files, 6) Write the files out as native text, do a forward run, and import results into GMS, 7) Compare key observations from the native text file results to the original binary GMS files. Here's a plot of one of the wells using outputs from the original calibrated binary GMS files. Here's the same well using outputs from the model run using the native text files. I've gone through the process of replacing the PEST variables in GMS with the final calibrated parameters twice and get the same results. One thing I have noticed is the dataset info differs between the calibrated pilot points and the version that I interpolated to the 2D grid. Here's a comparison of the dataset info for one of the pilot point datasets: Orig Cal PP Interp to 2D Grid Min 1.072 1.072 Max 904.99 734.94 Range 903.91 733.87 Mean 129.78 48.05 Median 50.132 20.75 SD 212.28 77.46 Why would these be so different? Both were done using the Kriging interpolation method. Any ideas would be appreciated! Quote Link to post Share on other sites
Michal 0 Posted April 21, 2020 Report Share Posted April 21, 2020 Hi Bruce, it is hard to tell. I suggest you to compare the HK and Sy arrays of the HDF5 and native MF version of your model. See if they are different. They should be exactly same. You could read the HDF file directly using Python for instance, see the discussion here. Good luck Michal Quote Link to post Share on other sites
Recommended Posts
Join the conversation
You can post now and register later. If you have an account, sign in now to post with your account.