Improving Ground-Water Flow Model Calibration With the Advective

Document Sample
Improving Ground-Water Flow Model Calibration With the Advective Powered By Docstoc
					 Improving Ground-Water Flow Model Calibration With the
 Advective-Transport Observation (ADV) Package to MODFLOWP
     The Advective-Transport Observa-           The ADV Package                                        ment, as shown by the blue line in figure
 tion (ADV) Package (Anderman and                                                                      1. The differences in the x-, y-, and
 Hill, 1997) allows advective-transport             Observations of the advective compo-
                                                                                                       z-components between the simulated and
 observations of steady-state flow fields to   nent of contaminant transport can provide
                                                                                                       observed advective-front locations are
 be used in conjunction with the hy-           important information for the calibration
                                                                                                       then used in the nonlinear regression in
 draulic head and flow observations origi-     of ground-water flow models. The addi-
                                                                                                       conjunction with the differences between
 nally included in the three-dimensional       tion of the Advective-Transport Observa-
                                                                                                       the simulated and observed heads and
 ground-water flow parameter-estimation        tion (ADV) Package (Anderman and Hill,
                                                                                                       flows to estimate optimal model parame-
 model MODFLOWP (Hill, 1992). In the           1997) to MODFLOWP (Hill, 1992) al-
                                                                                                       ter values.
                                               lows the path and speed of advective
 ADV Package, advective-transport paths                                                                    If ground-water age dates are used, the
                                               travel to be used in conjunction with for-
 and times are represented using particle                                                              ADV Package can be used to test hypoth-
                                               mal parameter-estimation methods, which
 tracking.     This allows advective-                                                                  esized recharge areas. In this case, a
                                               are powerful tools for model calibration
 transport observations to be included in                                                              particle is tracked backwards from the
                                               (Hill, 1998; Poeter and Hill, 1997).
 the regression with minimal increase in                                                               observed age-date location and the
 computational effort. Features of the          Applying the ADV Package                               advective-transport observation consists
 package include:                                                                                      of the hypothesized recharge area; regres-
                                                   Advective-transport observations gen-
                                                                                                       sion is used to determine whether the
     E A particle-tracking routine that du-    erally cannot be measured directly, but
                                                                                                       resulting flowpath is possible with reason-
 plicates the semi-analytical method of        can be inferred from a number of data
                                                                                                       able parameter values.
 MODPATH (Pollock, 1994).                      types, such as contaminant concentrations,
                                                                                                           It is important to include the uncer-
                                               tracer tests, or, in some circumstances, age
    E Comparison of the x-, y-, and z-         datings. If contaminant concentrations are
                                                                                                       tainty in approximated advective-
 movements of the simulated and ob-                                                                    transport observations in the analysis.
                                               used, advective-transport observations
 served advective front at defined times.                                                              The weighted regression procedure used
                                               can be obtained by contouring measured
 Thus, the direction and path of travel as                                                             in MODFLOWP allows this uncertainty to
                                               concentrations and using the distribution
 well as the overall travel distance is in-                                                            be included explicitly.
                                               along the plume front to determine the
 cluded in the calibration process.            likely advective-front location, as shown               Improving Model Calibration
    E Calculation of sensitivities of the      by the triangle in figure 1. In some situa-             With the ADV Package
 particle movement to the parameters us-       tions, an analytical solution can help ap-
                                               proximate an advective-front location.                      Application of the ADV Package to
 ing the exact sensitivity-equation ap-                                                                the Otis Air Force Base sewage-discharge
                                                   A simulated equivalent to the observa-
 proach. Composite scaled sensitivities                                                                plume (Anderman and others, 1996) illus-
                                               tion is generated by tracking a particle
 provide users with a quantitative mea-                                                                trates the utility of the package.
                                               through the grid from the source location
 sure of the information provided by the                                                                   Parameter sensitivities indicate the re-
                                               for the known time of ground-water move-
 observations.                                                                                         lation between changes in parameter val-
    E Particle tracking in a forward or                                         Particle Entry Point
                                                                                                       ues (e.g. hydraulic conductivity and
                                                    OBSERVATION                                        recharge rate) and simulated values of
 backward direction can be used to evalu-           WELLS                   SOUR

 ate transport of contaminants or                                                                      head, flow, and advective transport within
                                                                                                       the model. They also provide a measure
 recharges areas.
                                               CONTAMINANT                                             of the information about the system pa-
     E Inclusion of effects such as retar-     PLUME CONTOURED                          LAKE 1

                                               FROM MEASURED
 dation by adjustment of the effective-        CONCENTRATIONS                                           Observed Advective-
 porosity value.                                                                                        Transport Location             Simulated
     Effective porosity is not included as a                                                                                           Path
 parameter to be estimated in version 1.0               LAKE 2
                                                                                                                          ∆z      ∆y
 of the ADV package.
     This publication outlines the major                                                                             ∆x        Defining the
 features of the ADV Package and gives                                                                                         discrepancy
 information on obtaining the program
 and documentation.
                                               Figure 1. Conceptual representation of advective-transport observations.

U.S. Department of the Interior                                                                                           USGS Fact Sheet FS-059-98
U.S. Geological Survey                                                                                                                    June 1998
      Composite Scaled Sensitivity   40                                                 be judged, in part, by     that more accurately simulated subsurface
                                             Head Observations                          evaluating the rea-        flow rate and direction, as indicated by the
                                                    Flow Observation                    sonableness of the         blue box in fig. 4.
                                                    Advective-Transport Observations    parameter estimates.
                                                    All Observations                    Historically, unrea-       How to Obtain ADV Package
     20                                                                                 sonable estimated pa-      Program, Documentation, and
                                                                                        rameter values have        Additional Information
                                                                                        been perceived as a
                                                                                        drawback of inverse             The ADV Package is included with
                                                                                        modeling; in fact,         MODFLOWP version 3.0 and later at no
                                                                                        they provide useful        cost from the USGS at World Wide Web
                 ic              e              n x             e ux           d
               ul ity          rg             er lu           ag Fl          on e                                  site The
             ra tiv
           d c              cha          r t h ry F        w e              P c         information         that
         Hy ndu          Re            No nda           Se arg            et n                                     site includes links to download program
                                                            h           m ct a          causes the modeler to
                                        u                sc          hu du                                         source code, test data sets, and compiled
                                      Bo               Di          As o n               scrutinize the model
                                                                                        and, perhaps, explore      versions for some computers. Also, the
Figure 2. Composite scaled sensitivities for the Otis Air Force Base                    alternative concep-        program is available by anonymous ftp
model are one measure of the information provided by different types                    tual models (Poeter        from the Internet site in di-
of data.                                                                                and Hill, 1997), re-       rectory pub/software/ground_water/.
                                                                                        sulting in a better cal-        Reports documenting USGS model
rameters present in the different data types,                                                                      programs can be purchased from
and indicate how well the parameters are ibrated model. In the Otis Air Force Base                                    U.S. Geological Survey
likely to be estimated from the available model, investigation of an unreasonable                                     Branch of Information Services
data. In the Otis Air Force Base model, value of estimated recharge revealed that                                     Box 25286
sensitivities for the advective-transport ob- the model is sensitive to a) the season in                              Denver, CO 80225-0286
servations are smaller than for the other which observations are obtained, and b)
observations (fig. 2). The additional infor- the heads at the downstream boundary                                      —E.R. Anderman and M.C. Hill
mation provided by the advective-transport conditions. This led to reevaluation of the
observations, however, reduces parameter data collection effort, concluding that
correlation (fig. 3), which measures if a probable errors in the stream elevations
coordinated change in parameter values caused significant model error.
does not change the model solution, enough                            Advective-transport observations re-
that the parameters can be estimated                              flect ground-water velocities and patterns
uniquely.                                                         over a long period of time and provide                    Particle Entry Point
    Ground-water flow model validity can more complete information about the                                                                       SOURCE
                                                                  ground-water flow system than flow mea-
     1.0                                                          surements taken at a discrete point in time.
 Parameter Correlation

                                                                  This results in different estimated parame-
   Absolute Value of

                                                                                                                        Location simulated
                                                                  ter values when different observation                  without advective-
                                                                  types are used and has a significant impact       transport observations
                                                                  on the probable accuracy of the transport
                                                                  predictions of the model. This is clearly
                                                                                                                   Plume-front location
                                                                  demonstrated by the Otis Air Force Base          approximated from
                   Only         and Flo
          Heads         Heads            Heads nsport             model. The plume movement simulated              concentration data
                                             tive Tra
                                     Advec                        by a model calibrated with only hydraulic-
Figure 3. Range of parameter correlation for                      head and flow observations significantly
                                                                                                                                                      Location simulated
the Otis Air Force Base model with different                      underpredicted the observed plume move-                                             with advective-
types of data. For correlation greater than 0.95, ment, as indicated by the yellow box in                                                             transport
parameters may not be estimated uniquely and                                                                                                          observations
                                                                  fig. 4. Including the advective transport in
predictions may be substantially in error. For                    the model calibration resulted in a model        Figure 4. Improved simulated plume move-
the data set with heads only, all values equal 1.                                                                  ment using advective-transport observations.

       References                                            in ground-water flow parameter estimation: Poeter, E.P. and Hill, M.C., 1997, Inverse mod-
                                                             Ground Water, v. 34(6), p. 1001-1009.        els: a necessary next step in groundwater
   Anderman, E.R. and Hill, M.C., 1997,
                                                          Hill, M.C., 1998, Methods and guidelines for    modeling, Ground Water: v. 35(2), p. 250-
     Advective-transport observation (ADV)
                                                             effective model calibration: U.S. Geologi-   260.
     package, a computer program for adding
     advective-transport observations of steady-             cal Survey Water-Resources Investigations Pollock, D.W., 1994, User's guide for MOD-
     state flow fields to the three-dimensional              Report 98-4005.                              PATH/MODPATH-PLOT, Version 3, A par-
     ground-water flow parameter-estimation               Hill, M. C., 1992, A computer program           ticle tracking post-processing package for
     model MODFLOWP: U.S. Geological Sur-                    (MODFLOWP) for estimating parameters         MODFLOW, the U.S. Geological Survey
     vey Open-File Report 97-14, 67 p.                       of a transient, three-dimensional, ground-   finite-difference ground-water flow model:
                                                             water flow model using nonlinear regres-     U.S. Geological Survey Open-File Report
   Anderman, E.R., Hill, M.C., and Poeter, E.P.,
                                                             sion: U.S. Geological Survey Open-File       94-464, 247 p.
     1996, Two-dimensional advective transport
                                                             Report 91-484, 358 p.