Inverse Modeling with
MODFLOWP
In conventional or “forward modeling,”
model parameters (such as aquifer
properties) are specified and water levels
and flow quantities are computed (fig. 4).
For most aquifer systems, however, more
information is available from field data on
water levels, flows, and advective transport
or ground-water age than on input parameters. Typically, input parameters are
adjusted during model calibration using a
trial-and-error process. This calibration
process can yield acceptable agreement
between computed model results and field
data but is time consuming, may not
produce parameter values that result in the
best fit of field data, and does not result in
quantitative estimates of uncertainty in
model results and estimated parameter
values.
Inverse modeling is a more formal
approach to model calibration that includes
automatic parameter adjustment in order to
match field data. Program MODFLOWP
(pronounced MOD•FLOW•P) is the USGS
version of MODFLOW that includes
automatic parameter estimation (Hill,
1992). MODFLOWP uses a weighted
least-squares objective function as a
measure of how well model results agree
with field measurements. Weights are used
to reflect reliability of individual
measurements. Parameters that can be
estimated by MODFLOWP include
transmissivity; hydraulic conductivity;
storage coefficient; vertical leakance;
vertical and horizontal anisotropy;
hydraulic conductance between aquifer
systems and rivers, drains and other
features; areal recharge; maximum evapotranspiration;
pumping; and water levels at
constant-head boundaries. Measured or
externally estimated information on
parameters also can be included. Parameter
values that minimize the objective function
(fig. 5) are calculated by MODFLOWP
using the modified Gauss-Newton method
or the conjugate-direction method. The
resulting parameter values are “best-fit” in
that they provide the closest match between
measured and simulated heads and flows,
as measured by the objective function. The
task of the modeler is to evaluate how well
these calculated values represent the
physical system being simulated.
Model output includes estimates of
parameters and statistics relating to the
parameter estimates. The statistics can be
used to quantify the reliability of the
resulting model, suggest changes in model
construction, and compare results of
models constructed in different ways. Postprocessors
can be used to calculate
confidence intervals on predicted heads and
flows to depict prediction uncertainty.
To effectively use MODFLOWP, an
understanding of principles of groundwater
flow and basic statistics is needed.
Inverse Modeling withMODFLOWPIn conventional or “forward modeling,”model parameters (such as aquiferproperties) are specified and water levelsand flow quantities are computed (fig. 4).For most aquifer systems, however, moreinformation is available from field data onwater levels, flows, and advective transportor ground-water age than on input parameters. Typically, input parameters areadjusted during model calibration using atrial-and-error process. This calibrationprocess can yield acceptable agreementbetween computed model results and fielddata but is time consuming, may notproduce parameter values that result in thebest fit of field data, and does not result inquantitative estimates of uncertainty inmodel results and estimated parametervalues.Inverse modeling is a more formalapproach to model calibration that includesautomatic parameter adjustment in order tomatch field data. Program MODFLOWP(pronounced MOD•FLOW•P) is the USGSversion of MODFLOW that includesautomatic parameter estimation (Hill,1992). MODFLOWP uses a weightedleast-squares objective function as ameasure of how well model results agreewith field measurements. Weights are usedto reflect reliability of individualmeasurements. Parameters that can beestimated by MODFLOWP includetransmissivity; hydraulic conductivity;storage coefficient; vertical leakance;vertical and horizontal anisotropy;hydraulic conductance between aquifersystems and rivers, drains and otherfeatures; areal recharge; maximum evapotranspiration;pumping; and water levels atconstant-head boundaries. Measured orexternally estimated information onparameters also can be included. Parametervalues that minimize the objective function(fig. 5) are calculated by MODFLOWPusing the modified Gauss-Newton methodor the conjugate-direction method. Theresulting parameter values are “best-fit” inthat they provide the closest match betweenmeasured and simulated heads and flows,as measured by the objective function. Thetask of the modeler is to evaluate how wellthese calculated values represent thephysical system being simulated.Model output includes estimates ofparameters and statistics relating to theparameter estimates. The statistics can beused to quantify the reliability of theresulting model, suggest changes in modelconstruction, and compare results ofmodels constructed in different ways. Postprocessorscan be used to calculateconfidence intervals on predicted heads andflows to depict prediction uncertainty.To effectively use MODFLOWP, anunderstanding of principles of groundwaterflow and basic statistics is needed.
การแปล กรุณารอสักครู่..
