QUALITY CONTROL--Reporting and documentation of aquifer modeling studies

                         GEOLOGICAL SURVEY
                      RESTON, VIRGINIA  22092

                                               February 5, 1975


Subject:  QUALITY CONTROL--Reporting and documentation of aquifer
                           modeling studies

Ground Water Branch memo No. 75.08 described the status of digital
modeling research in WRD, summarized types of problems being
considered, and indicated the kinds of numerical approaches being
developed.  This memo provides guidelines for reporting and
documentation of modeling studies, either digital or analog.

Reporting of aquifer modeling studies
In most ground-water studies modeling analysis enhances
understanding of the ground-water flow system.  For any modeling
analysis the project report should identify for the reader the
modeling assumptions and the resultant constraints on use of the
model.  Therefore, we recommend that reports concerning model
analyses include the following information.

     1.  Define the location of the modeled area if it does
         not coincide with the project area.

     2.  Describe the hydrologic character of the model
         boundaries and how these are represented in the model.
         In many cases the model boundaries are placed where the
         aquifer terminates against nearly impermeable rocks or
         is intersected by a perennial stream show head variation
         in time and space is known.  In other cases, the aquifer
         may be so extensive relative to the area of interest
         that the model simulation can be truncated at some
         arbitrary distance; but if so, the author should show
         that selection of the arbitrary boundary location has
         not materially affected the model calibration or model
         predictions.  Internal boundaries such as streams or
         lakes should be identified and their counterpart
         representation should be defined in the report.

     3.  Justify the use of two-dimensional models.  In some
         cases pumping wells tap one or more zones of a multi-
         layered sequence of extensive sand and clay layers.
         Although three-dimensional analog models have been
         developed, a strictly three-dimensional digital model
         has not been developed.  In lieu thereof multi-aquifer
         systems are handled by a quasi three-dimensional digital
         model that uses a two-dimensional area model for each
         aquifer with interconnection of aquifers through one-
         dimensional leakage terms.  Because a strictly three-
         dimensional digital model has not been developed is not,
         of itself, adequate justification to use a two-
         dimensional model.  The burden of justification rests
         with the modeler.

     4.  Describe the aquifer system properties that are
         modeled.  Inferences are made from field data as to the
         areal variation of hydraulic properties of aquifers and
         confining beds.  The spatial variation in heads
         throughout the aquifer system at some initial time and
         the subsequent hydrologic stresses (see item 5) are
         simulated and the model-derived responses are compared
         to field observations.  If the degree of comparison is
         unacceptable, the modeler often will change the modeled
         values for one or more of the physical parameters and/or
         characterization of the physical stresses to obtain a
         better fit.  The outcome of this trial-and-error process
         essentially is an improved conceptualization of the
         aquifer system and the imposed stresses.  The final
         synthesis should be described in the report.  If the
         system differs appreciably from the conceptual model
         developed from the pre-model analysis, the modeler is
         obliged to provide evidence that modifications made to
         the model to achieve an "acceptable" fit between
         computed and observed stresses are hydrologically
         reasonable because "near best" fits between computed and
         observed responses can be obtained, for example, using
         physically absurd aquifer parameter values.

     5.  Describe the stresses and initial conditions
         modeled.  Stresses, in the form of pumpage,
         evapotranspiration from ground water, recharge from
         infiltration, river-level changes, leakage from other
         aquifers, etc. are imposed on the aquifer system.  The
         relations between observed and modeled stresses should
         be described.  For example, it might be desirable to
         provide a representative sample of actual pumping
         histories and the corresponding modeled pumping
         histories, although such data would not be necessary for
         every pumped well.

     6.  Describe the discretization network.  The numerical
         techniques most used to solve the equation describing
         ground-water flow requires that the region of interest
         be divided into subregions.  If awkward to describe in
         the text, an illustration showing the spatial
         discretization should be given.  If a digital model is
         used, the manner in which the computations are marched
         through time also should be described.

     7.  Describe the mathematical techniques used.  In most
         cases it will be desirable to briefly describe the
         mathematical equations used to approximate the flow
         problem and their method of solution.  If an analog
         model was used, the analogy between the various
         components of the hydrologic system and the electrical
         analog system should be described.  If a digital model
         was used, and the computer program used is already
         published, avoid the temptation of including a listing
         of the program unless your version is substantially
         different from the published version.  You should
         however specify any significant modifications of the
         published program.

     8.  Present the criteria used for model calibration and
         the limitations of the model's representativeness of the
         real system.  Give a representative sample of the actual
         comparisons used for calibration.  If changes in head at
         selected times were used for comparison, show these.
         Describe the data used to develop the observed heads and
         changes, and at least indicate locations of
         representative data points on maps showing observed
         conditions.  This provides a basis for the reader to
         evaluate the control for observed level conditions.

In summary, a report discussing calibration of a model should
include a description of the aquifer system, the stresses imposed
on the system, and the responses of the system to the imposed
stresses.  Real aquifer systems are too complex to be modeled
exactly.  The hydrologist develops a conceptual model that
simplifies the system to make it manageable and yet retains those
features essential to the objectives of the model study.  The
assumptions made in developing the conceptual model of the aquifer
system should be identified and justified and their implications
on model predictions specified.  The basis for model calibration
and validation should be given.

Documentation of aquifer modeling analysis
Carefully document the trial-and-error calibration of the model--
even though that material is not included in the published report.
It is good modeling practice to keep a "log" for all model runs.
We recommend that the information documented include the
assumptions made for the run, some rationale on the selection of
these conditions, and interpretations of the run results.
Remember that changes in the modeled parameters correspond to
changes in conceptualization of the physical system.  Each model
run should be logged when the results are evaluated, not months
later.  This documentation would (a) aid the modeler in keeping
track of the essentially adaptive learning process of trial-and-
error calibration, (b) be helpful during the project and report
review to discussions on the technical aspects of model
calibration, and (c) provide a useful permanent record for future
analysts of the geohydrology of the area.

                                   (s) Charles A. Appel
                                   for Gerald Meyer
                                   Chief, Ground Water Branch

WRD Distribution:  A, B, S, FO, PO