The Applications of GIS for Environmental Engineering



A Look at Remediation of Acid Mine Drainage



by Shane Duckworth




Table of Contents

Introduction
What is GIS?
Case Study
GIS Analysis Discussion and Conclusions
References
Glossary
Thanks!


1.0 INTRODUCTION
This brief report examines the contribution that Geographic Information Systems (GIS) can make to the field of environmental remediation of contaminated sites. The theory and structure of a typical GIS application is given, then a case study is used to illustrate the practical aspects of applying GIS to an envrionmental remediation problem.

2.0 WHAT IS GIS?
GIS stands for Geographical Information Systems, and is a term used to describe a class of software applications that display a variety of information spatially. GIS had its beginnings about 25 years ago, when the idea of using transparent map overlays was adapted to computers (1). The best way to describe a typical GIS program is to place a spatial viewer on top of a database. The database can contain any information that can be displayed and minipulated spatially. Growing in complexity, today's typical GIS programs contain digitizing packages, database management systems, statistics packages, and data processing packages. These additions allow the user(s) to input, access, minipulate, transform, and display data over geographic base maps. For example, streamflow data can be assigned to individual streams in a watershed, and correlated with such things as rainfall intensities, streambed slopes, tributary areas, etc.

It is really no secret--and should come as no surprise--that the largest effort put into GIS is data collection. This could amount to approximately 80% of the cost of implementation of GIS for any given enterprise.

GIS packages process and display information in one of two ways--using either a vector or raster representation. Vector representation takes the form of points, lines, and polygons (hence areal representation), and can link a vast number of attributes to each element. For example, if a user were to click on one of the lots shown in Figure 1, information about the survey boundaries, homeowner/landowner, street address, and assessment information may come up. Raster representation on the other hand is composed of a grid of cells that have unique (x,y,z) values; the z value can represent a variety of information from elevations to bulk resistivity. Each representation has its strengths and hence tend to be used within specific fields. For example, the vector representation is probably better suited to a municipal engineering firm that designs subdivisions because it is more concerned with attributed entities than conducting analyses with stored data. Raster representation may be better used by a franchise company that is looking to site a new store because it allows for greater statistical flexibility with less processing than the vector representation.

3.0 CASE STUDY
Having set out an initial description of what GIS is, a case study is examined here which will give insight into how GIS may be used in environmental engineering applications. The GIS program used to analyze the following case study is Idrisi, a raster-based GIS.

The Belt - Sand Coulee Coal Fields lay southeast of Great Falls, Montana, comprising a drainage area of approximately 600 square km (4), see Figure 2. The coal fields form the western portion of the Great Falls - Lewiston Coal Field. Coal was discovered there over a century ago, and mining began in 1880. Production began in earnest to fuel the trains using the newly-completed rail line through Sand Coulee, Montana. Production was completely finished at a commercial scale by World War II, although it is reported that some small-scale mining continued for domestic use. The mining operations were conducted largely by Cottonwood Coal Company, a subsidiary of Great Northern Railroad.

Acid Mine Drainage (AMD) was noted in the area as far back as the 1920s; mining records noted that the "water is practically acid" (3). However, the impetus to provide a solution to the AMD leaching out of the mines was provided by a Master's thesis by George Morris McArthur from Montana State University in 1970. Under the Surface Mining Control and Reclamation Act of 1977, the Montana Department of Environmental Quality's (DEQ) Abandoned Mine Reclamation Bureau has identified all the coal-related reclamation problems in the state, including those in the Belt - Sand Coulee area, and begun remediation work. The study area is shown in Figure 3. The red hatched areas are the mine sites of concern. The oxidization process of sulfur-rich rock to acid is shown in Appendix 1.

The general geology of the Belt-Sand Coulee area is shown in Figure 4 and Figure 4a. The coal stratum comprises the top of the Morrison Formation, and all mining works were extended underground into this geologic layer. The bedrock dips generally from 3 to 6 degrees to the north, with local variations up to 15 degrees (4). It has been suggested that the general groundwater hydrology scheme is that of water infiltration in and through the Kootenai Formation, and bounded by the relatively impermeable coal seam at the Morrison - Kootenai contact. Water flow through the coal is impeded, so the groundwater travels along this contact until it can daylight at a spring or enter the coal mine tunnels and shafts (4).

Once the groundwater enters the mine workings, it reacts with oxygen and the sulfur-rich pyrite to form sulfuric acid. Where the mine tunnels daylight, the acidic groundwater is discharged into streams which ultimately drain into the Missouri River near Great Falls.

A number of remedial and preventative methods of dealing with the AMD were suggested in McArthur's thesis. Among them were mine flooding, mine grouting/sealing, surface water diversion, in-stream neutralization, stream flow impoundment and regulation, and cropping to prevent water infiltration.

The State of Montana obviously would like to control the costs associated with the AMD remediation. DEQ has attempted to implement one method of AMD impoundment at mine portals, but the barriers were breached in floods due to uneven discharge out of the portals (3). The preferred solution to this problem is to prevent any water infiltration into the mines by planting crops at the ground surface that will take up any soil pore water. The problem is, the land overlying the mines is largely privately owned, and typically the fields lay fallow every other year. Out of frustration, one DEQ official remarked that it may be cheaper for the State of Montana to buy the required private lands outright then to operate water treatment in perpetuity. It is this statement that provided the impetus for this project.

4.0 GIS ANALYSIS
This project aims to answer the simple question: would the State be better financially advised to purchase the land over and near the mines that would contribute to groundwater infiltration into the mines, or to operate in-stream interception and neutralization in perpetuity?

The United States Geological Survey (USGS) in conjunction with DEQ has implemented a GIS database to aid in monitoring the Belt - Sand Coulee area for AMD and provide an analysis tool for examining remediation alternatives. The project is in its infancy stages, but much information has been fed into the database already, all of which has been provided at no cost for this project. Because the database is just being built, no metadata has been compiled at this point (2). The helpfulness of DEQ and Montana State Library (Natural Resource Information System) is gratefully acknowledged.

Uncertainty has been left out of my analysis due to the lack of metadata. This is commensurate with the preliminary nature of this analysis.

4.1 Input Data
The input data for this project were the following: a digital elevation model (DEM, compiled from 10 USGS quadrangle DEMs); a landuse map for the 10 quadrangles; bedrock geology for the eastern 8 of 10 quadrangles; roads, railroads, stream, and standing water bodies, all as vector overlays; and water sampling locations with an accompanying water quality and streamflow database. Soils information was included but not used in this analysis.

4.2 Targetted Output Data
Two questions were posed in light of the input data: how much money would the State be required to spend on each of the proposed remedial schemes: land buy-up and in-stream neutralization?

4.3 Methodology
Since all data provided by DEQ were in Arc/Info format, Arc/View was employed to convert the vector files into shapefiles (with an associated database), or into ASCII grids for import into Idrisi, the GIS program used to conduct the analyses. A QBasic program was written to process the ASCII grid files into a format recognized by Idrisi. Once all the shapefiles and grid files were imported into Idrisi, the landuse, geology, and DEM files were made into raster images.

The second step was to determine a land costbase map from the landuse map. Using real estate data from the Great Falls area, a cost was associated with each landuse type. Areas that were either not useable crop land (such as lakes, rivers, and wetlands) or not purchasable (such as cemetaries), a boolean image was created and then "multiplied" with the costbase map to produce an image of land purchase prices for areas that could be purchased. This costbase map is shown in Figure 5. It should be noted that one category of landuse, "Conservation Reserve Land" is land that DEQ pays farmers to plant what the State wants; in this area, DEQ has noticed a drop in groundwater discharge (3). This land is called "State Land" on the base cost and final land purchase map.

Since the mine areas were vector images of the underground mine workings, the on-screen digitizer in Idrisi was used to draw a polygon mine boundary around each mine site, enclosing all known underground mine workings. This is one area of gross uncertainty, as it was stated that there were most probably many undocumented shafts and tunnels (2). These polygons were rendered into a raster image that provided the basis for the term "mine areas" used in this report.

The DEM was then filtered twice using a low mean pass filter (3x3) to smooth out anomolies appearing down one quadrangle boundary line. This was necessary to render more representative aspect data when Idrisi calculated the aspect of each slope.

The next step caused the most problems using Idrisi. The objective at this point was to determine first the net tributary area for surface water runoff that would be caught by the mine areas--"net" meaning the tributary area that drained directly into streams was subtracted from the mine surface drainage area (it was assumed that if surface water made it to the streams, it would not enter the underground mine workings). Second, the tributary subsurface groundwater drainage area was to be determined using the assumption that the coal seam dipped due north at 5 degrees without variation (this assumption is recognized as not being valid because of local variation, jointing and discontiuities in the seam, and preferential flow paths; however to accomodate these variables would require an analysis that goes far beyond the scope of this project).

Using Idrisi's "WATRSHED" routine, the surface water runoff tributary area was calculated for the mine areas. Then, the streams lying outside the mine areas were digitized and fed into the WATRSHED routine; the resulting area was subtracted from the mine area tributary area to define the net surface water runoff tributary area. The resulting area was reclassified into 256 categories, for distance away from the mine areas (it was assumed that the further away from the mine site, the less chance a water drop would have of travelling to the mine areas).

Using an aspect image for the coal seam (all zeros = due north), WATRSHED calculated the tributary groundwater area shown in Figure 6. Clearly, the resulting image is unacceptable; there is a bias in the routine used by WATRSHED. To overcome this to some extent, the analysis was re-run and the tributary area recalculated by varying the horizon over which Idrisi looks at possible drainage paths. This new analysis is shown in Figure 7--this is theoretically correct if the flow is laminar and encounters no obstructions. Bad as these assumptions are, this flow regime was used; further analysis might make use of a friction surface to create a tributary area that fans outward upslope from the mine areas at about 30 degrees, with decreasing weighting with distance from the zero degree flow regime. Finally, tributary areas that fell outside of the top of the Morrison Formation from the geology map were excluded. This process is shown in Figure 8.

Last, assuming that groundwater infiltration does not occur only in a downward fashion, but may also travel along joints and fissures in the Kootenai Formation, radial distances from the mine areas were calculated, then weighted on a suitability scale of 0 to 255, with 255 representing areas that should be purchased. Suitability ranged from 30 m radially from the mine site to 0 at 501 m.

The relative weights of each factor were assigned as below:

These weights were assumed only from intuitive judgement, not rational.

The factor images were then placed on the suitability scale from 0 to 255, factored, then combined using Idrisi's "Image Calculator". The resulting image (Figure 9) shows the areas that must be purchased to ensure proper prevention of surface water infiltration, with decreasing importance placed on areas further away from the mine site. The green areas denote required land purchases; the colored bands designate the "buffer zones" around the mine sites that are subject to judgement on importance. To classify the areas the State should buy, an "appropriate" percentage of the total subjective land area must be specified, then the cost can be calculated. For the purposes of this analysis, 90% coverage was selected.

A boolean land purchase map was created from the image shown in Figure 9, excluding 10% of the subjective area. This image was multiplied with the costbase image to produce the final cost map shown in Figure 10. To calculate the required land costs, the area of each land type was tabulated by Idrisi, then multiplied by the respective purchase price per acre.

4.4 Cost Comparisons
The cost of remedial action by the State of Montana was calculated for two methods: land purchase for selective cropping to prevent surface water infiltration, and in-stream neutralization of the AMD. Costs for land purchase were calculated using the method defined in the previous section; costs for in-stream neutralization were calculated by applying net present value analysis to the method described in McArthur's thesis. Inflation was calculated from the Consumer, Farmer, and Producers Price Indices. The costs calculated by McArthur were modified to reflect the water quality data as collected recently by the State and USGS. It is recognized that this method is dubious at best (as technology has changed the likelihood of applying this method over another type of interception and neutralization is low), but serves the purpose of this preliminary analysis. The final costs are given below; the land purchase price is a lump-sum amount amortized over a 20-year period and does not take into account any profit realized from the sale of cash crops that may be planted by the State, and the neutralization costs are given likewise annually for a 20-year period. Cost calculations are shown in Appendix 2.

Cost for Land Purchase: US$1,475,043 per year

Cost for in-stream neutralization: US$43,012 per year

5.0 DISCUSSION AND CONCLUSIONS
The above analysis leads to several conclusions.

First, the lack of certainty leaves questions regarding the veracity of the data and assumptions used to reach any conclusions. This is why certainty becomes such an important issue in any kind of spatial analysis from which decisions will be made. The degree to which attention is paid to uncertainty in the original data and carried through the analyses should be commensurate with the importance of decisions likely to be made with the analyses.

Second, knowlege of the algorithms used to process the data is essential. The image produced in Figure 6 gives a very clear picture of the bias that may be built into a particular routine and may render the routine useless for analyses in specific conditions.

Third, the use of GIS can be very powerful in environmental engineering applications, especially at an enterprise-wide level. Despite the cost of building a useable database appearing infeasible, most of the data required to build a good GIS database is collected in the course of the environmental engineer's daily activities. If every site investigation made by a particular company were to be entered into that company's database (locations of boreholes, drill logs, water levels, contaminant concentrations, topography, landuse history, etc.), each employee would have instant access to a wealth of information on-line and in graphic format at a few keystrokes instead of being buried in company archives which are generally located off-site and expensive to retrieve (if it can be remembered that the information even exists!).

Because of this, it is highly recommended that environmental engineering firms employ GIS (if it is consistent with their mandate) to help in data analysis and decision making, particularly with large projects. However, caution is advised in its use, attention needing to be paid to the biases and limitations that a particular GIS program may contain.


REFERENCES

(1) Burrough, Peter A. and Rachael A. McDonnell, 1998. Principles of Geographical Information Systems. Oxford University Press. Oxford, NY.

(2) Highness, David. Montana State Library, Natural Resource Information System. Personal communication, October - November, 1998.

(3) Koerth, John. Montana Department of Environmental Quality, Abandoned Mine Reclamation Bureau. Personal communication, November 10, 1998.

(4) McArthur, George Morris, 1970. Acid Mine Waste Pollution Abatement, Sand Coulee Creek, Montana. Master's Thesis at Montana State University, Bozeman, MT.

(5) Noble, Cassandra and John Koerth, 1996. Montana...Bringing the Land Back to Life. Montana Department of Environmental Quality. Helena, MT.


GLOSSARY

Please use the BACK button on your browser to return to the same place you were in the text!

DEM
Digital Elevation Model, a raster grid file that contains elevations (one per cell) as the z-attribute value

Metadata
"Data about data"--information about the data set that includes source(s), method(s) of measurement, and errors and/or uncertainties.

Shapefiles
A set of export files from Arc/View that can be read by other GIS programs. Arc/View exports the vector files and a database file that contains all vector attributes as the shapefile set.

Uncertainty
Statistical uncertainty that is linked with every data point. For example, elevation data may come from differential GPS readings that could have vertical errors of, say, 1.6 m associated with them.

Vector Overlay
Although Idrisi is a raster-based GIS, it allows vectors to be placed on its maps for dispay purposes. The roads and mine areas in Figure 3 are examples of this.


THANKS!

I'd like to give many thanks to the following people who helped immensely in the preparation of this report: