This is a modern Fortran (i.e., 2003/2008) well-test simulator. It performs numerical Laplace-Hankel inversion, implementing the the main unconfined approaches in use today. The program is free software (MIT license), which can be used, modified, or redistributed for any purpose, given the license is left intact.
The publications by the author of this software (Kristopher Kuhlman) can be found at: http://kris.kuhlmans.net
This simulator is only a command-line utility, which reads a text input file, computes a solution given the inputs, and writes a simple text output file formatted for plotting using available software obtained elsewhere (e.g., MS-Excel, python matplotlib, or gnuplot). The simulator is accurate and relatively fast, using OpenMP to execute in parallel on a multi-processor Linux or Mac computer (a recent Intel compiler is needed to create parallel executables for MS-Windows).
The input parameters are explained in input-explanation.txt
-
Mishra & Neuman (2010,2011) : Unsaturated/saturated flow to a partially penetrating well. http://dx.doi.org/10.1029/2009WR008899 http://dx.doi.org/10.1029/2010WR010177 NB: The Mishra & Neuman solutions given in the WRR papers are numerically somewhat ill-behaved. his simulator implements them in three different ways.
1a. One approach to solve M/N follows the Malama (2014) http://dx.doi.org/10.1002/2013WR014909 simplified formulation - replacing the no-flow boundary condition at the land surface with a "finiteness" boundary condition. This solution has only been derived so far for the fully penetrating no wellbore storage case.
1b. A second approach to solve M/N discritizes the vadose zone using finite differences (in Laplace-Hankel space). This approach works and can be used as a "check" on the algebra and mathematics in the closed-form Laplace-space approaches. This allows partial penetration, but no wellbore storage for now.
1c. The third approach to solving M/N implements the solution listed in their paper close to the notation used in that paper. Because this approach used to fail for some combinations of parameters, it is now computed in quad precision. The J and Y Bessel functions of complex argument and fractional order needed for this solution at quad-precision accuracy using arb (http://fredrikj.net/arb/) - an arbitrary precision special function library written in C. This is a significant new dependency, since it requires the flint, mpfr, and gmp libraries. Many thanks to Fredrik Johansson who helped out using his library.
-
Malama (2011) : Alternative linearization of the moving water table boundary condition. Basically an improvement on Neuman (1974). http://dx.doi.org/10.1016/j.jhydrol.2010.11.007
-
Moench (2001,1995) : The hybrid water table boundary condition of Moench (1995), but including the multiple delayed yield (α) coefficients, as used in the large Cape Cod, Massachusetts pumping test in USGS Water Supply Paper 1629. http://dx.doi.org/10.1111/j.1745-6584.1995.tb00293.x http://pubs.usgs.gov/pp/pp1629/pdf/pp1629ver2.pdf
-
Neuman (1974,1972) : The standard moving water table solution used by most hydrologists for well-test interpretation. http://dx.doi.org/10.1029/WR008i004p01031 http://dx.doi.org/10.1029/WR010i002p00303
-
Hantush (1961) : The confined solution which includes the effects of partial penetration, but using a three-layer approach of Malama (2011), rather than the typical finite cosine transform.
-
Theis (1935) : The confined fully penetrating solution, which all other solutions build upon.
The simulator is distributed as a collection of Fortran source files and a makefile. On Linux/Unix/Mac platforms this is trivial to turn into a command-line program, by simply going to the source directory and running:
make [driver]
The driver argument is implied. An unoptimized debugging version (produces some warnings that can be safely ignored) can be compiled via:
make debug_driver
If you want to build the ARB capabilities (only needed for the Mishra-Neuman solution), specify
make ARB=1
The default is to not build with ARB.
On MS-Windows, you will need the mingw compilation environment (OpenMP doesn't work under mingw, though - so single thread only) or the Intel Fortran compiler (which works and provides OpenMP as well). I have previously compiled it (Feb 2015) with the free mingw toolchain, and can either provide assistance setting this up, or provide you with a binary (I don't have access to a Windows computer now). There are a few settings in the makefile that must be changed to get it to compile using mingw (see comments there)
The respository includes three unconfined pumping datasets for benchmarking and conducting a "beauty pageant" between the different unconfined models:
-
Moench et al., 2001 (Cape Cod, Massachusets): a 320-gpm 72-hour pumping test in a thick glacial outwash aquifer with many shallow piezometers and screened observation wells (significan partial penetration effects) the thinned PEST dataset is available from the authors electronically, and in a report (USGS Professional Paper 1629 v2). No recovery data :(
-
Wenzel, 1942 (Grand Island, Nebraska): a 540-gpm 48-hour pumping test available from an old report (USGS Water Supply Paper 887). This test also has significant partial penetration effects. I have entered this data into spreadsheet form. 36 hours of recovery data available. A large number of observation piezometers and a few screened wells (82 + pumping well) were monitored. The spreadsheets are available on Google docs at: https://docs.google.com/spreadsheet/ccc?key=0AlJMuEYu7Z-5dGJfdzBibk4zNDB4UG9DN1FpQ0FnX1E&usp=sharing
-
Bevan, 2002 (Borden, Ontario); a 12-gpm 168-hour (week long) test was obtained electronically from the authors, published in Michael Bevan's MS thesis. The aquifer was relatively thin, compared to the Cape Cod and Grand Island tests. Moisture content data and geophysical data were collected in the vadose zone before, during, and after testing. Five days of recovery data were collected.
These data are available in the git repository (i.e., see borden, cape_cod, and grand-island directories of source tree).