NAME
Modflowp - Parameter-estimation version of the modular model
ABSTRACT
This program is a new version of the U.S. Geological Survey modular,
three-dimensional, finite-difference, ground-water flow model
(MODFLOW), which, with the new Parameter-Estimation Package, can be
used to estimate parameters by nonlinear regression. The new
version of MODFLOW is called MODFLOWP (pronounced MOD-FLOW-P) and
functions nearly identically to MODFLOW when the Parameter-
Estimation Package is not used. Parameters used to compute the
following MODFLOW model inputs can be estimated: layer
transmissivity, storage, coefficient of storage, hydraulic
conductivity, and specific yield; vertical leakance; horizontal and
vertical anisotropy; hydraulic conductance of the River, Streamflow-
Routing, General-Head Boundary, and Drain Packages; areal recharge;
maximum evapotranspiration; pumpage; and the hydraulic head at
constant-head boundaries. Nearly any spatial variation in
parameters can be defined by the user. Data used to estimate
parameters can include existing independent estimates of parameter
values, observed hydraulic heads or temporal changes in hydraulic
heads, and observed gains and losses along head-dependent boundaries
(such as streams). Model output includes statistics for analyzing
the parameter estimates and the model; these statistics can be used
to quantify the reliability of the resulting model, to suggest
changes in model construction, and to compare results of models
constructed in different ways.
METHOD
Parameters are estimated by minimizing a weighted least-squares
objective function by the modified Gauss-Newton method or by a
conjugate-direction method.
HISTORY
Version 3.4 98/05/18 - (A) Integrated the Hydrauluc Flow Barrier
Package so that barriers can be defined when estimating PID=T
parameters, and so that parameters used to calculate the
hydraulic characteristic of the barrier can be estimated. (B)
PAR1AP: ADMX set. Not used presently in MODFLOWP, but used in
UCODE and may be needed to test different GN convergence options
for MODFLOWP. (C) SSEN1V: Removed GOTO 50 statement so that all
flow observations are considered for each timestep. This allows
multiple flow observations within a timestep.
Version 3.3 (A) Correction to SSEN1M for convertible layers. (B)
Added subroutine TSTAT, which produces the correct statistic for
calculating 95-percent linear confidence intervals, given the
degrees of freedom. TSTAT is called from SEN1OT. This addition
eliminates one of two functions served by the preprocessing
program BCINT. The confidence intervals on the parameter
estimates printed by MODFLOWP now use the correct critical value.
(C) Modified RESANP.F to create subroutine RESANP, which is
called from the end of MODFLOWP.F. The files of data for the
plots are generated by MODFLOWP instead of requiring that a
separate program be executed (PROGRAMMING NOT COMPLETE). (D)
Modified PAR1AP to correct floating point precision computations.
(E) PAR1AP and MAIN: Moved printing to the screen of everything
except the sum of squared, weighted residuals from MAIN to
PAR1AP. The following additional information now is printed to
the screen: DMX, JMAX, PID(JMAX). (F) SEN1OT and MAIN: Moved
printing to the screen of the sum of squared, weighted residuals
from MAIN ro SEN1OT. (G) PAR1OT: Included the standard error of
regression in the printing. Wrote data for multiple prior to
file for RESANP. Converted printing from log base e to log base
10. (H) SSEN1O: In the runs test, there is now a check for if the
number of runs equals the expected number of runs. (I) SPCG1P:
Calculated JMIN and changed WRITE statements using format 525.
This corrects a situation that produced a negative indice when
the solver converged in fewer than ITER1 iterations.
Version 3.2 97/10/09 - Corrected problems with the implementation of
the ADV Package in version 3.1 which caused erroneous regression
results (added call to SADV1O to routine SSEN1O and corrected IF
statement in SADV1L so that x-position sensitivities are
calculated correctly). Clarified some print statements in
routine PAR1OT. Added default file name (files96) for the NAME
FILE. The use of variables KSTP and KPER discussed in Hill
(1992, p.155) is now used only with IPAR=0, when sensitivity
arrays are calculated. When IPAR=-1 or 1, KSTP and KPER are used
as described for MODFLOW. This correction eliminates an
incompatibility with MODPATH. Eliminate attempts to caclulate
the Run's statistic and the correlation between ordered weighted
residuals and normal order statistics (eq. 38 of text) when
divide by errors would result.
Version 3.1 97/07/24 - ADVective-transport observation (ADV1)
Package added to distribution. This package is documented in
Anderman and Hill (1997).
Version 3.0 97/05/25 - Code updated to include MODFLOW-96
capabilities, including CHD1, GFD1, HFB1, IBS1, RES1, and TLK1
packages. Note, the MODFLOWP parameter estimation and
sensitivity analysis capabilities are not compatible with these
additional packages. MODFLOW-96 is documented in Harbaugh and
McDonald (1996a and b). This version of MODFLOWP is compatible
with data sets from all older versions of MODFLOW and MODFLOWP.
Corrected calculation of sensitivities for PID=ANI. The error
occurred in versions distributed from July 1994 to June 1997.
PCG2 changed so final interaction information for four iterations
are printed when MUTPCG=1. Corrected various initialization
problems. Changed dimensioning of some arrays back to the
original MODFLOW dimensioning to coordinate with MT3D and other
postprocessing programs.
Version 2.15 96/12/15 - In FLW1RP, apply INT in IF statement using
QCLS In OBS.F, now use NRES when assigning values to the D array.
In SEN1AL, reinstated IOUHDS and IOUFLWS in the call and print
statements. In SSEN1P, KT=0 and KB=0 moved out of if sequences.
This caused problems on Silicon Graphics machines. In YCINT,
changed nobs to nh before the Do 36 loop. In SEN1FN, when
writing format 565, the indices for ISTRM were changed to 4 and 5
from 1 and 2. This affects the printout of the information
listed in the beginning of the output for PID=KST parameters, but
does not affect calculated results. Some subroutines were
restructured by Evan Anderman which make the inclusion of new
observations more straightforward. Subroutines SSEN1F and SSEN1H
were replaced by HED1RP and FLW1RP; parts of SEN1AL were replace
by HED1AL, FLW1AL, and OBS1AL; parts of SSEN1O were replaced with
SHED1O, SFLW1O, SPR11O, SPR21O and SPR310. The new subroutines
are distributed in the file OBS.F. Added label to the composite
scaled sensitivities. In SSEN1G, checks for exceeding NSM or NSN
are now made as the data are read. Repeated information printed
when some parameters are defined was eliminated (SSEN1G).
Version 2.14 96/10/01 - Distribution includes a PostScript file of
the updated input instructions. Directory name and default file
names changed to be DOS compatible. Bug fixes; for example,
corrected SSEN1U so that the counter for multilayer wells is
updated correctly when wells are dry and for transient problems;
correction in SEN1FN related to vertical leakance parameters
distributed as log-transformed values; and changes in statements
calculating weights for flows. Updates to utility program YCINT
to accommodate full weight matrix for flows. When calculating
Beale's measure, simulation NOW will continue even if the solver
does not converge. Checks global budgets for excessive errors
(>5%), which would invalidate the calculated Beale's measure.
Version 2.10 96/04/23 - Modified MAIN to accommodate nonlinear
confidence interval subroutines that are being developed and to
accommodate computation of regression statistics when convergence
is by the SOSC criterion (DATA SET 13). Change in SEN1RP so that
Beale's measure will be calculated correctly when there are
unestimated parameter values (as indicated by negative group
numbers in DATA SET 9). Corrected indices specification for
double precision array in SEN1AL that would only affect
calculations with adjoint states. Initialized array IT1 in
pcg2ap to correct printing of 1's to indicate the first internal
pcg iterations of transient problems. Full weight matrices for
flow observations added by Steen Christensen of the University of
Aarhus, Denmark. In DATA SET 5, negative values of IUH and IUF
now can be used to eliminate printing. In the line preceding
DATA SET 2A, negative values of IUP can be used to eliminate
printing. Corrections made in subroutine SSEN1H that allow
correct representation of transient head observations in
multilayer wells.
Version 2.8 96/03/01 - Correction which applies when cells are
rewet: Update array CVWD in SEN1FN. Correction which applies
when any RCH parameter is not the last parameter and NLLI1>4 on
line 3 of the Parameter-Estimation Package input file: change in
SSEN1R. Correction which applies when laycon=3 and contributions
to vertical leakance are calculated using layer thicknesses
specified in the first multiplication arrays: Inclusion of
laycon=3 in if statement in sen1fn. Broke up some if statements
that were causing problems with some pc compilers. Corrected
write statement so that when more than 99 head-dependent boundary
cells are used to define a parameter they are written correctly.
Added statements to main so that information is printed to unit *
after the calculation of heads or sensitivities and at the end of
each parameter-estimation iteration. On most systems, unit * is
the screen. This printing can be eliminated by setting new
variable IOSTAR to 1. Changes were made to the table of
information from all parameter-estimation iterations, and changes
were made so that MODFLOWP would perform correctly when the
Parameter-Estimation Package was turned off.
Version 2.2 95/09/14 - Added error messages to help users adapt to
the new version. Added a file named VERSION to the distribution
to make it easier for users to identify what version they had.
Revised internally used revision numbers for each distributed
file. Includes nine new capabilities, including use of
convertible layers for steady-state systems. Output contains
more explanations to guide users. Added damping factor to PCG2
package.
Version 1.11 94/01/14
Version 93/11/01 - Corrected calculation of sensitivities for
PID=ANIV.
Version 92/09/01 - Dimension statements modified to achieve
compatibility with the Lahey compiler.
Version 92/03/01 - Original release.
DATA REQUIREMENTS
Input files from MODFLOW Packages modified for compatibility with
MODFLOWP, and an input file for the Parameter-Estimation Package
that defines the estimated parameters and the observations used in
the regression.
OUTPUT OPTIONS
Tabular summaries of descriptive statistics and data.
Postprocessors (bealep, bcint, resanp, and ycint) are available in
Hill (1994) and included in distribution.
SYSTEM REQUIREMENTS
MODFLOWP is written in Fortran 77 with the following extensions:
use of variable names longer than 6 characters and greater than 19
continuation lines. Generally, the program is easily installed on
most computer systems. The code has been used on UNIX-based
computers and DOS-based 386 or greater computers having a math
coprocessor and 4 mb of memory.
DOCUMENTATION
Hill, M.C., 1992, A computer program (MODFLOWP) for estimating
parameters of a transient, three-dimensional, ground-water flow
model using nonlinear regression: U.S. Geological Survey Open-
File Report 91-484, 358 p.
Hill, M.C., 1994, Five computer programs for testing weighted
residuals and calculating linear confidence and prediction
intervals on results from the ground-water parameter-estimation
program MODFLOWP: U.S. Geological Survey Open-File Report
93-481, 81 p.
RELATED DOCUMENTATION
Harbaugh, A.W., and McDonald, M.G., 1996a, User's documentation for
MODFLOW-96, an update to the U.S. Geological Survey modular
finite-difference ground-water flow model: U.S. Geological Survey
Open-File Report 96-485, 56 p.
Harbaugh, A.W., and McDonald, M.G., 1996b, Programmer's
documentation for MODFLOW-96, an update to the U.S. Geological
Survey modular finite-difference ground-water flow model: U.S.
Geological Survey Open-File Report 96-486, 220 p.
Harbaugh, A.W., 1990, A simple contouring program for gridded data:
U.S. Geological Survey Open-File Report 90-144, 37 p.
McDonald, M.G., and Harbaugh, A.W., 1988, A modular three-
dimensional finite-difference ground-water flow model: U.S.
Geological Survey Techniques of Water-Resources Investigations,
book 6, chap. A1, 586 p.
The ADVective-transport observation (ADV1) Package is documented in:
Anderman, E.R., and Hill, M.C., 1997, ADVective-transport
observation (ADV) Package, a computer program for adding
advective-transport observations of steady-state flow fields to
the three-dimensional ground-water flow parameter-estimation
model MODFLOWP: U S. Geological Survey Open-File Report 97-14, 67
p.
Version 2 of Preconditioned Conjugate Gradient Package is documented
in:
Hill, M.C., 1990, Preconditioned conjugate-gradient 2 (PCG2), a
computer program for solving ground-water flow equations: U.S.
Geological Survey Water-Resources Investigations Report 90-4048,
43 p.
The Stream Package is documented in:
Prudic, D.E., 1989, Documentation of a computer program to simulate
stream-aquifer relations using a modular, finite-difference,
ground-water flow model: U.S. Geological Survey Open-File Report
88-729, 113 p.
The Interbed-Storage and Time-Variant Specified-Head Packages are
documented in:
Leake, S.A., and Prudic, D.E., 1988, Documentation of a computer
program to simulate aquifer-system compaction using the modular
finite-difference ground-water flow model: U.S. Geological
Survey Open-File Report 88-482, 80 p.
The General Finite Difference Flow Package is documented in:
Harbaugh, A.W., 1992, A generalized finite-difference formulation
for the U.S. Geological Survey modular three-dimensional finite-
difference ground-water flow model: U.S. Geological Survey Open-
File Report 91-494, 60 p.
The Version 2 of the Block-Centered Flow Package is documented in:
McDonald, M.G., Harbaugh, A.W., Orr, B.R., and Ackerman, D.J., 1992,
A method of converting no-flow cells to variable-head cells for
the U.S. Geological Survey modular finite-difference ground-water
flow model: U.S. Geological Survey Open-File Report 91-536, 99
p.
The BCF3 Package is documented in three pieces. It builds on two
previous versions of the Block-Centered Flow (BCF) Package.
Documentation for the BCF1 Package describes the fundamental
function of all BCF Packages. This documentation is contained in
the basic model documentation (McDonald and Harbaugh, 1988). BCF2
documentation describes the addition of the capability to convert
dry cells to wet:
McDonald, M.G., Harbaugh, A.W., Orr, B.R., and Ackerman, D.J., 1992,
A method of converting no-flow cells to variable-head cells for
the U.S. Geological Survey modular finite-difference ground-water
flow model: U.S. Geological Survey Open-File Report 91-536, 99
p.
BCF3 documentation describes the addition of alternate interblock
transmissivities. The BCF3 code includes the capabilities of BCF1
and BCF2:
Goode, D.J., and Appel, C.E., 1992, Finite-difference interblock
transmissivity for unconfined aquifers and for aquifers having
smoothly varying transmissivity: U.S. Geological Survey Water-
Resources Investigations Report 92-4124, 79 p.
The HFB1 Package is documented in:
Hsieh, P.A., and Freckleton, J.R., 1993, Documentation of a computer
program to simulate horizontal-flow barriers using the U.S.
Geological Survey modular three-dimensional finite-difference
ground-water flow model: U.S. Geological Survey Open-File Report
92-477, 32 p.
The Transient-Leakage Package (TLK1) is documented in:
Leake, S.A., Leahy, P.P., and Navoy, A.S., 1994, Documentation of a
computer program to simulate transient leakage from confining
units using the modular finite-difference ground-water flow
model: U.S. Geological Survey Open-File Report 94-59, 70 p.
The DE45 Package is documented in:
Harbaugh, A.W., 1995, Direct solution package based on alternating
diagonal ordering for the U.S. Geological Survey modular finite-
difference ground-water flow model: U.S. Geological Survey Open-
File Report 95-288, 46 p.
The RES1 Package is documented in:
Fenske, J.P., Leake, S.A., and Prudic, D.E., 1996, Documentation of
a computer program (RES1) to simulate leakage from reservoirs
using the modular finite-difference ground-water flow model
(MODFLOW): U.S. Geological Survey Open-File Report 96-364, 51 p.
REFERENCES
Anderman, E.R., Hill, M.C., and Poeter, E.P., 1994, Two-dimensional
advective transport in nonlinear regression--sensitivities and
uncertainty of plume-front observations: in Warner, J. and
others, eds., 1994 Ground Water Conference, Fort Collins, Colo.,
Proceedings, p. 55-62. (Describes a useful technique for using
inverse methods to evaluate how well you might expect available
and anticipated data to estimate a defined set of parameters.)
Anderman, E.R., Hill, M.C., Poeter, E.P., 1996, Two-dimensional
advective transport in ground-water flow parameter estimation:
Ground Water, v. 34, no. 6, p. 1001-1009.
Anderman, E.R, Hill, M.C., and Poeter, E.P., 1996, Presentation and
evaluation of a new multi-stage parameter-estimation method using
advective-transport observations: in Calibration and Reliability
in Groundwater modeling, Proceedings of the 1996 Model CARE
Conference, Golden, Colo., September 1996, International
Association of Hydrological Sciences Publication 237, p. 179-188.
Barlebo, H.C., Hill, M.C., and Rosbjerg, Dan, 1996, Identification
of groundwater parameters at Columbus, Mississippi, using three-
dimensional inverse flow and transport model: in Calibration and
Reliability in Groundwater modeling, Proceedings of the 1996
Model CARE Conference, Golden, Colo., September 1996,
International Association of Hydrological Sciences Publication
237, p. 189-198. (This is not an application of MODFLOWP, but it
does provide an example of using nonlinear regression in an
interesting situation.)
Cooley, R.L., and Naff, R.L., 1990, Regression modeling of ground-
water flow: U.S. Geological Survey Techniques of Water-Resources
Investigations, book 3, chap. B4, 232 p.
D'Agnese, F.A., Faunt, C.C., Hill, M.C., and Turner, A.K., 1996,
Death Valley regional ground-water flow model calibration using
optimal parameter estimation methods and geoscientific
information systems: in Calibration and Reliability in
Groundwater modeling, Proceedings of the 1996 Model CARE
Conference, Golden, Colo., September 1996, International
Association of Hydrological Sciences Publication 237, p. 41-52.
Giacinto, J.F., 1994, An application of MODFLOWP to a Superfund case
study: in Warner, J. and others, eds., 1994 Ground Water
Conference, Fort Collins, Colo., Proceedings, p. 103-110.
Hill, M.C., 1990, Preconditioned conjugate-gradient 2 (PCG2), a
computer program for solving ground-water flow equations: U.S.
Geological Survey Water-Resources Investigations Report 90-4048,
43 p.
McDonald, M.G., Harbaugh, A.W., Orr, B.R., and Ackerman, D.J., 1991,
A method of converting no-flow cells to variable-head cells for
the U.S. Geological Survey modular finite-difference ground-water
flow model: U.S. Geological Survey Open-File Report 91-536, 99 p.
Poeter, E.P., and Hill, M.C., 1996, Unrealistic parameter values in
inverse modeling, A problem of benefit for model calibration: in
Calibration and Reliability in Groundwater modeling, Proceedings
of the 1996 Model CARE Conference, Golden, Colo., September 1996,
International Association of Hydrological Sciences Publication
237, p. 277-285.
Poeter, E.P., and Hill, M.C., 1997, Inverse models, A necessary next
step in groundwater modeling: Ground Water, v. 34, no. 2, p.
250-260.
Poeter, E.P., and McKenna, S.A., 1994, Geostatistical simulation and
inverse flow modeling to reduce uncertainty associated with flow
and transport predictions: in Warner, J. and others, eds., 1994
Ground Water Conference, Fort Collins, Colo., Proceedings, p.
47-54.
Starn, J.J., 1994, Field application of nonlinear regression to
estimate the parameters of a two-dimensional ground-water flow
model, Calvert City, Kentucky: in Warner, J. and others, eds.,
1994 Ground Water Conference, Fort Collins, Colo., Proceedings,
p. 575-582.
Starn, J.J., Arihood, L.D., and Rose, M.F., 1995, Geohydrology and
simulation of ground-water flow in the aquifer system near
Calvert City, Kentucky: U.S. Geological Survey Water-Resources
Investigations Report 94-4239, 52 p.
Tiedeman, C., and Gorelick, S.M., 1993, Analysis of uncertainty in
optimal groundwater contaminant capture design: Water Resources
Research, v. 29, no. 7, p. 2139-2153.
Yager, R.M, 1993, Simulated three-dimensional ground-water flow in
the Lockport Group, a fractured dolomite aquifer near Niagara
Falls, New York: U.S. Geological Survey Water-Resources
Investigations Report 92-4189, 43 p.
Yager, R.M., and Hill, M.C., 1991, Comparison of hypotheses used to
construct simulations of transient three-dimensional ground-water
flow by nonlinear regression, in EOS, 1992 Fall Meeting Abstract
Supplement: American Geophysical Union, p. 210.
TRAINING
Parameter Estimation for the Modular Ground-Water Flow Model
(GW3089TC), USGS National Training Center, usually offered alternate
years. Preliminary six-month correspondence course starts in August
of even numbered years; one week of classes at the USGS National
Training Center is usually the first week of the following March.
CONTACTS
Operation and Distribution:
U.S. Geological Survey
Hydrologic Analysis Software Support Program
437 National Center
Reston, VA 20192
h2osoft@usgs.gov
Official versions of U.S. Geological Survey water-resources analysis
software are available for electronic retrieval via the World Wide
Web (WWW) at:
http://water.usgs.gov/software/
and via anonymous File Transfer Protocol (FTP) from:
water.usgs.gov (path: /pub/software).
The WWW page and anonymous FTP directory from which the MODFLOWP
software can be retrieved are, respectively:
http://water.usgs.gov/software/Modflowp.html
--and--
/pub/software/ground_water/Modflowp
SEE ALSO
contour(1) - A contouring program for gridded data
moc(1) - Two-dimensional method-of-characteristics ground-
water flow and transport model
mmsp - Modular Model Statistical Processor
modflow(1) - Modular three-dimensional finite-difference
ground-water flow model
modpath(1) - Particle-tracking postprocessor program for the
modular three-dimensional finite-difference
ground-water flow model
sutra(1) - Saturated and (or) unsaturated, constant or
variable-density fluid flow, and solute or energy
transport (2-dimensional finite-element code)