Integration of Subsurface Data Management and GIS to Facilitate Model Calibration

0
- By:

Courtesy of EarthSoft, Inc

Abstract

The development of numerical codes, as well as pre- and post-processors to support these codes, has made the discipline of groundwater modelling increasingly more automated and efficient. The value of such modelling is reliant on the site-specific data used to build and calibrate the models. The integration of EarthSoft's EQuISÒ Geology and any of several popular modelling environments results in a data management and analysis system that is more complete than the modelling system alone. Water level data are easily exported to calibration files for the appropriate modelling system. Coupled with EQuIS and other analysis applications, ArcView GIS can be used as an easy, graphical environment with which to facilitate the groundwater model calibrating process.

Introduction

Since the introduction of PLASM by Prickett and Lonnquist in 1971 to today's most advanced multi-species reactive transport models, the science of groundwater modelling has witnessed the advent of many different tools for studying the mechanics and behavior of groundwater flow systems. These groundwater models are the most effective means of predicting water levels in an aquifer based on existing or proposed parameters, assessing the effects of various pumping or injection well schemes, or estimating the areal extent of contamination where a chemical has been introduced into a hydrogeologically active environment (Anderson & Woessner, 1992). Ultimately, these tools have 'significantly expanded the nation's ability to understand and manage its water resources.' (Friedman et al.1984 quoted in Anderson & Woessner, 1992).

In addition to numerical modelling codes, several pre- and post-processors-or modelling environments-have been developed to make the use of these codes much easier and more cost-effective. These applications include the Department of Defense Groundwater Modelling System (GMS), Groundwater Vistas, Visual MODFLOW, and Argus ONE, and others. The use of GMS, for example, has resulted in considerable savings in modelling and simulation by cleanup specialists in a case study at the Schofield Army Barracks in Hawaii. The U.S. Army Corps of Engineers Waterways Experiment Station (WES) and U.S. Army Environmental Center implemented GMS to evaluate site characterization uncertainty and present modelling results. The efforts led to the acceptance by regulators of a lower-cost remedial design for cleanup; savings for this application alone were estimated at $7.5M to $10.0M. (Holland & Richards, 1998). However, just as these modelling environments have facilitated more expedient model setup and execution, the next logical step in streamlining the process from data collection to results and prediction is the integration of data management.

Data Management

When MODFLOW was first introduced it had no graphical user interface (GUI). Operation consisted of manually creating input files with a text editor and using command line prompts to perform analysis. The advent of pre- and post-processing modelling applications greatly simplified this task by providing a graphical environment which allowed modellers to easily establish model boundaries and conditions, delineate zones of constant head or no flow, and automatically create the specific files required by the numerical code. Thus, the tedious issue of manually creating input files had been resolved. However, the issue of what data (i.e. aquifer properties such as hydraulic conductivity and porosity; head values) to include in the model, and how to get that data into the modelling environment remained.

In many modelling situations, site-specific data may be non-existent, or scanty at best. Due to various factors such as budgetary constraints or land use restrictions it may be infeasible to obtain field values of parameters necessary for constructing a numerical model. In the absence of such data, estimates of aquifer properties are frequently taken from tables. However, the proper and accurate application of a groundwater model requires extensive field data for model construction and calibration. When the scenario to be investigated is of sufficient importance, the effort and expense of data collection may be well worth the cost (Anderson & Woessner, 1992).

In many modelling situations, site-specific data may be non-existent, or scanty at best. Due to various factors such as budgetary constraints or land use restrictions it may be infeasible to obtain field values of parameters necessary for constructing a numerical model. In the absence of such data, estimates of aquifer properties are frequently taken from tables. However, the proper and accurate application of a groundwater model requires extensive field data for model construction and calibration. When the scenario to be investigated is of sufficient importance, the effort and expense of data collection may be well worth the cost (Anderson & Woessner, 1992).

While data collection is a critical step in a site characterization effort, collection alone is not sufficient. Far too often, data painstakingly obtained from field investigation-particularly geologic data-are filed in a cabinet down the hall, stacked in boxes in a dusty storage facility, or even off-site in a nearly-forgotten repository. When the effort required to find and obtain data from a previous study is comparable to collecting the data in the first place, it is just as good as having no data to begin with!

The value of an electronic data management system such as EQuIS Geology is that the field data is readily available and easy to review, report, and utilize for groundwater model construction and calibration. Even projects long completed and archived can be accessed in a fraction of the time that would be required for hardcopy reports or logs. Because data is housed in an open, non-proprietary Microsoft Access database and can be quickly exported to the pre-processor for integration into the flow model, there is no longer any need store site-specific information such as hydraulic conductivity, porosity, or layer elevations on paper or even in various electronic files.

Data Usability

One of the scenarios frequently encountered among environmental data managers today is the data hostage situation. This problem may result not only from the use of a proprietary database that prevents 'back-door' access to data, but also from the process of storing data in a particular visualization or analysis application. The open systems architecture, upon which EQuIS Geology is based, provides the opportunity for a data manager to access data directly, outside of the user interface. This philosophy is rapidly gaining wide acceptance as users are able to go directly into the database to build custom queries and write need-specific applications for reporting and formatting data. Whereas the EQuIS Geology user interface provides an extensive suite of reporting tools and interfaces for sending data to many different visualization and analysis applications, open systems allow the development of custom interfaces without being bound by the cost and time requirements of the developer as is often the case with closed, proprietary systems. Another benefit of the open systems architecture is that the user is not locked into a specific visualization or analysis application. They may choose between any of several popular tools for creating borehole logs, groundwater models, solid models, or performing other types of analysis. This flexibility provides the opportunity to switch to a higher-level application if the currently used application is not adequate without migrating data to a new system. Many proprietary systems provide their own visualization and analysis tools and if these do not prove adequate, the user is at mercy of the developer or is required to migrate their data to a more suitable system.

The dangers of storing data in a specific visualization or analysis tool are illustrated by the hypothetical case of a project manager who has successfully used a groundwater modelling environment to produce the results needed by his client. However, when the client then needs borehole logs, cross-sections, or solid models in addition to the groundwater modelling, the manager is left in a quandary. Heretofore, the solution has been the costly investment not only in a product and the time required to learn the procedures necessary to produce the desired results, but also in understanding file formats and getting the appropriate data into the new application. EQuIS Geology greatly simplifies this task by facilitating the creating of borehole logs, cross-sections, fence diagrams, reports, contours, groundwater models, solid models, and more all very quickly and easily…without having to understand the intricacies of specific file formats. This mechanism permits more time to be devoted to science and analysis rather than the overhead of a specific piece of software.

Model Construction

Once site data have been entered into EQuIS Geology, either manually using simple, intuitive screens or imported using an electronic data deliverable (EDD), the data may be exported for use in constructing groundwater flow models. A simple example might include using hydraulic conductivity values associated with various geologic samples and stratigraphy or other defined geologic units from which to derive layer boundaries. Figure 1 shows the export screen for using hydraulic conductivity values in a GMS model. A query returns the samples found within a given depth range (between 15 and 30 feet in this example). The samples to be included are then selected, along with the parameters of interest and the data are saved as a GMS Scatter Point file which can then be imported, interpolated to the model grid, and used in the MODFLOW BCF Package for a given layer. Bottom elevations are dealt with similarly, only exporting Borehole Data from EQuIS Geology rather than Sample Data. In addition to aquifer characteristics, analytical chemistry data may also be exported for inclusion reactive transport or other types of analytical modelling.

Model Calibration

An essential part of any modelling task is model calibration. In order to have confidence in the model, it is necessary to demonstrate that the calculated heads are reasonably close to the observed heads. This is generally accomplished by iteratively finding the aquifer properties, boundary conditions, and hydrologic stresses that result in calculated heads within an acceptable range of error. Although automatic parameter estimation software can be used, the manual trial-and-error technique has been in use longer and in many cases is the preferred method (Anderson & Woessner, 1992).

In addition to head observations, flux may also be used in model calibration. The combination of calibration with both head and flux can significantly reduce the issue of non-unique solutions, a problem often encountered when only head is used. However, water table elevations are the most common type of field data used in MODFLOW simulations (Jones et al., 1998)

The same problem exists for model calibration as with model construction: effectively and efficiently using field data. Instead of having to build calibration coverages or targets manually in a text editor, or even using GUI tools provided in a modelling environment to enter the information, water table data stored in EQuIS Geology can be used to create these calibration files very easily and quickly. Figure 2 shows how observed field data may be used to create model calibration files.

Any given monitoring well may have many months or even years of water level measurement data. So that the model calibration is based on water levels within the appropriate time frame, a date range may be selected. Because the selected range may have multiple measurements, simple statistics may be used to calculate the calibration value. The exported target value may be the most recent, minimum, maximum, or mean of the set of measured values.

Additionally, the statistically-calculated head value may be weighted by the standard deviation (x), the reciprocal of the standard deviation (1/x), or the reciprocal of the standard deviation squared (1/x2). A linear transformation may be applied to both the head value and the weight; this feature might be used in a case where data were stored in metres, but the modelling was being conducted in feet.

The lower section of Figure 2 represents options that apply to the GMS observation coverage. The name, elevation, interpolation method, interval, and confidence may be defined for the coverage to be created. Similar files are created for model calibration in Groundwater Vistas and Visual MODFLOW.

GIS Integration

According to Liebert et al., 'Database and GIS software are natural partners for the storage and use of the digital files needed for a MODFLOW ground-water model.' They describe work done to integrate the results of numerical flow and transport modelling with the GIS environment. However, working within this environment is now more functional as certain analytical procedures can be performed within the ArcView GIS. Instead of acting solely as a graphical display and presentation tool for groundwater modelling results, the same calibration files described previously for the various modelling environments can be produced directly within the ArcView GIS. Figure 3 shows the EQuIS Geology menu that is part of the EQuIS ArcView Interface.

To preserve familiarity, the exact screens utilized in EQuIS Geology to export data for building models are invoked from within ArcView GIS. This allows the user to graphically and dynamically select the wells of interest and use those wells for much of the same functionality as found in the EQuIS Geology application.

It should be noted that the ArcView interface is designed to function as a 'casual' user interface, where junior level staff can produce results quickly and easily. For this reason, the functionality of the interface is limited in some respects.

Summary

Mathematical models have done much to allow groundwater scientists to better understand and manage water resources. However, assumptions made to simplify a model may jeopardize the accuracy of the model results. Where field data can be-and is-collected, the EQuIS Geology subsurface data management system and EQuIS ArcView Interface provide tools whereby these data can not only be utilized more effectively and efficiently for groundwater modelling studies, but also for other types of environmental investigative applications.

Future Work

Future work includes making the same type of export functionality available for the Groundwater for Windows and Argus ONE modelling environments. Additionally, supporting transient calibration targets will be added.

References

  • Anderson, M. P. & Woessner, W. W. (1992) Applied Groundwater Modelling. Academic Press, San Diego, California, USA.
  • Holland, J. P. & Richards, D. R. (1998) Overview of Department of Defense Subsurface Modelling Tools. In: MODFLOW 98 Proceedings (ed. by E. Poeter, C. Zheng, & M. Hill), 167-174.
  • Jones, N. L., Edris, E. V., & Poeter, E. P. (1998) Utilizing GIS Objects for Flux Calibration. In: MODFLOW 98 Proceedings (ed. by E. Poeter, C. Zheng, & M. Hill), 657-664.
  • Liebert, T. M., Jaumann, P. J., & Wright J. (1998) Direct Access and Use of MODFLOW Output Files in a GIS Application. In: MODFLOW 98 Proceedings (ed. by E. Poeter, C. Zheng, & M. Hill), 637-640.

Customer comments

No comments were found for Integration of Subsurface Data Management and GIS to Facilitate Model Calibration. Be the first to comment!