Prediction of acid mine drainage has been a primary regulatory and industrial planning need for decades, yet there still is no satisfactory method. The most widely accepted method is acid base accounting (ABA) which only estimates whether there will or will not be a problem. It does not estimate ultimate acid loading, treatment costs nor duration of acid mine drainage.
However, the principles behind acid base accounting are reasonable: that there is a relationship between the acid generation and neutralization potentials of a given rock mass. This paper presents a spreadsheet which uses ABA data in a dynamic fashion to predict acid generation, loading, concentration and duration. The spreadsheet uses conventional variables plus three new ones: sulfur flux (%Sf), net deliverable neutralization (NDN) and net deliverable acidity (NDA).
The spreadsheet and preliminary validation is presented though it is recognized that a great deal of work is needed before this becomes a reliable prediction tool.
Acid Base Accounting.
Acid Base Accounting (ABA) was developed in the early 1970's by researchers at West Virginia University to identify and classify geologic strata encountered during mining (West Virginia University, 1971). A history of Acid Base Account is provided by Skousen et al. (1990).
Since its development, ABA has been used extensively in the United States and several other countries for premining coal overburden analysis. Its popularity largely stems from its simplicity. However, it has been subject to criticism since it does not account for the different rates of acid and alkali-generating reactions in rock. Modifications to ABA have been recommended (Smith and Brady, 1990; and diPretoro, 1986). Other methods have developed which accelerate or otherwise control the oxidation and leaching process in rock samples. One such procedure, that of Renton et al. (1988), was employed in this study to simulate the rate of weathering of acid-producing rock samples alone and in combination with alkaline amendments. The results are compared with traditional ABA parameters
Recent Studies.
diPretoro and Rauch (11988) found poor correlations (reported R squared = 0. 16) between a volume-weighted acid base net neutralization potential (NP) and net drainage alkalinity near thirty mine sites in West Virginia. Erickson and Hedin (1988) showed similar poor correlation between maximum potential acidity (MPA), NP, net NP from ABA and net alkalinity from drainage water. Both papers related that factors other than overburden characteristics were involved in predicting post-mining water quality.
diPretoro and Rauch (1988) found that sites which had greater than 3% calcium carbonate equivalent (NP) in overburden produced alkaline drainages while at 1 % or less acidic drainage resulted. Erickson and Hedin's results indicate that 2% calcium carbonate or less produced acidic drainage while 8% or more produced alkaline drainage. (in this later study there were no sampling points between 2% and 8%).
O'Hagan and Caruccio (1986) found that the addition of calcium carbonate at 5% by weight to a coal refuse containing 1% S produced alkaline drainage. In Minnesota, Lapakko (1988) found that 3% calcium carbonate neutralized an overburden material with 1.17%S.
Hedin and Erickson (1988) compared water quality from rocks weathered in humidity cells to ABA values. Cumulative sulfate from humidity cells was strongly correlated with total sulfur (R squared = 0.69), while cumulative acidity/alkalinity was correlated with net NP (R squared = 0.37). They also showed sulfate from humidity cells was significantly correlated to sulfate from drainage water (R squared = 0.17), but the correlation was not strong enough to predict post-mining drainage quality.
Bradham and Caruccio (1991) conducted several overburden analytical tests on pyritic wastes from Canada. They found water quality resulting from column leachings, ABA projections, and soxhlets correctly predicted eight out of ten sites where drainage was monitored from refuse piles, with weathering cells predicting ten out of ten results.
There have been several modifications in using ABA in predicting drainage quality. The Pennsylvania Department of Environmental Resources (PaDER) (Smith and Brady, 1990) developed a spreadsheet which calculates mass-weighted maximum potential acidity (MPA), NIP, and net NIP. The spreadsheet also summarizes the overburden analysis in terms of the ratio of NP/MPA and the percent sandstone. The spreadsheet of ABA data can be compared to significant thresholds or numerical limits for NP and %S and other factors can be changed to estimate the impact on drainage quality. For example, Brady and Hornberger (1989) suggested threshold values of NP equal to or greater than 3% and %S less than 0.5 as guidelines for delineating alkaline-producing strata.
In the development of its spreadsheet, PaDER (Cravotta et al., 1990) reviewed the calculation of NP in ABA. In current ABA usage, 3.125 g calcium carbonate equivalent is required to neutralize acidity resulting from oxidation of 1 g S. Cravotta et al. (1990) argue that this ratio should double to 6.25:1. Volume-weighted maximum acidities are subtracted from NP giving a positive or negative net NP for the mined area. A negative, or deficient, net NP is interpreted to indicate the amount of calcite that must be added to equalize the deficiency and prevent AMD formation.
Other alkaline materials have higher NP's than calcite. Quicklime, kiln dust and hydrated lime all have higher activities than calcite, though it is not clear that the kinetics of pyrite oxidation favor readily soluble sources of alkalinity.
Brady et al. (1990) conducted a study of 12 sites where ABA data were available. They computed net NP based on both 3.125% and 6.25% to 1 % S. Alkaline addition on the sites was conducted to abate potential AMD problems. When using 6.25%, the sign of the net NP (+ or -) matched the sign of the overall net alkalinity of water at 11 of 12 sites.
The results of their study concluded that NP and traditional estimates of MPA (e.g. 3.125% to 1% S) were not equivalent and that overburden NP must be twice the MPA to produce alkaline mine drainage. They also concluded that mining practices (such as alkaline addition, selective handling, and concurrent reclamation) enhanced the effect of alkaline addition on reducing acidity. Lastly, they concluded additional studies are needed to determine the rates, application and placement of alkaline material during mining.
Brady and Hornberger (1990), after summarizing the work on AMD prediction by ABA made the following conclusions in a recent PaDER Mining and Reclamation Manual. First, NP from ABA shows the strongest relationship with actual post-mining water quality. This relationship is only qualitative (e.g. acid vs. non-acid), and NP must significantly exceed MPA in order to produce alkaline water. If NP and MPA are similar, AMD will most likely result. Sites with less than 0.5% S will not be significant AMD producers, except where little or no NP exists. High sandstone composition in the overburden (greater than 65%) will almost always result in acid drainage.
Factors Which Induce Error in Acid Base Accounting.
The foregoing discussion makes it clear that interpretations of ABA are diverse. Given the policy and economic implications of ABA it is considered useful to better understand the basis for ABA predictions and, where acid problems are identified, to generate cost-effective solutions.
Errors in conventional application of ABA result from variance in total S content (Rymer et al. 1991), and perhaps more significantly, non-homogeneous placement of spoils. For example Schueck (1990) reported AMD generation from a surface mine site in Pennsylvania resulted largely from buried refuse and pit cleanings within an otherwise neutral to alkaline spoil matrix as identified by ABA.
Acid neutralization in spoil dumps-a paradigm '. Obviously, some spoils will be composed entirely of acid forming rocks. Others such as refuse tend to have little NP at all. But in cases where AMD forms despite significant alkalinity in the overburden, it appears to originate from localized sites within the backfill. While finding the path of least resistance to the downstream side of the dump, the acidity is influenced only the alkalinity directly in its path. Once this is overcome, AMD flows freely to the nearest stream while the remaining alkalinity persists as a spectator to the process. This is to be expected since dissolution of calcite is controlled by pH and the partial pressure of carbon dioxide. Where pore water gas is confined, and exposed to mineral acidity, its pH will remain around 6.2 the-buffering point of bicarbonate and carbonic acid. In the absence of mineral acidity, its pH will reflect bicarbonate saturation - 8.3. In either case, additional calcite will dissolve only upon addition of acidity and outgassing of carbon dioxide. So, unless contacted directly by acidity, most of the spoil calcite will simply remain in solid form. So, the presence of alkalinity in the clump does not ensure that it will be a factor in neutralizing acidity. To be an efficient process, the acid-forming and alkaline rock must be thoroughly mixed.
This largely becomes a materials handling issue. Where there is insufficient alkalinity available it would be necessary to add it to the rock. Otherwise, if one relies on random spoil dumping the system would need an overwhelming supply of alkaline rock. This probably accounts for the above reported field observations that twice or more NP is required for each unit of MPA.
The above introductory remarks are a necessary background for development and application of the AMD/TIME spreadsheet. For example, without good mixing of acid producing and alkaline rock, neither ABA nor AMD/TIME will correctly estimate the outcome. Without reliable estimates of %pyrite sulfur and neutralization potential, neither ABA nor AMD/TIME will correctly estimate the outcome.
In developing AMD/TIME the following assumptions were made:
|
sandstone |
100% |
|
shale |
50% |
|
refuse with fines |
20% |
AMD/TIME uses conventional variables plus the following:
|
%Sf/yr |
Percentage of remaining pyritic sulfur oxidized and leached per year. |
|
FLOW |
Annual rainfall (inches) X 102970 X surface area (acres) X net infiltration (%)-yielding liters per year. |
|
NDN |
Net deliverable neutralization potential. This is the proportion of NP that is exposed to acid water and is able to react with it. |
|
NDA |
Net deliverable acidity. This is the proportion of MPA that oxidizes. |
AMD/TIME operates on the Quatro Pro spreadsheet developed by Borland International, Inc. Quatro Pro is similar to Lotus 1,2,3 and, except for the graphics would probably work equally well. The spreadsheet only uses several hundred KRAM so it will work on nearly all IBM compatible desktop computers. Naturally, machine power and higher order Intel chips will make it work more quickly.
AMD/TIME was developed for simplicity, not elegance. It uses simple, empirical rather than deterministic variables. Table 1 shows the working end of AMD/TIME indicating where data is input through the variable block. The user only needs to enter the following data:
target NP/MPA ratio
years of mining
acid rock production (tons of rock produced in mining)
surface area (acres)
%Sf/yr
%S pyritic (from ABA)
%NP natural (from ABA)
%NP added
%NDN
%NDA
cost of alkaline amendment $
amendment NP (%)
cost of water treatment chemical ($/T)
life of mine coal production (T)
AMD/TIME will then estimate acid loads, concentrations, alkalinity pools for the next several hundred years. AMD/TIME automatically estimates the chemical cost of water treatment for the life of amd production. It also automatically estimates the required amount of alkaline amendment needed to reach a target NP/MPA ratio. If you enter that amount at the "%NP added" block the spreadsheet will estimate the cumulative cost of amendment. Costs in current dollars are given in absolute amounts and in dollars per raw ton of coal.
Table 2. shows the complete AMD/TIME spreadsheet extended to 70 years. In the scenario given roughly one half of the required alkaline amendment was added to the rock so the costs reflect combined amendment and chemical treatment costs.
As configured AMD/TIME is an acidity model. It can also be simplified to a sulfate model. This was used to compare various variable combinations to 11 year old 400 ton test pile data at Island Creek's Upshur Complex (Table 3). Two net infiltration values and three %Sf and %NDA rates were tested a factorial arrangement. %Sf was calculated for each pile between each two week sampling period. In each pile, %Sf was very slow during the first six months, then accelerated rapidly to a maximum within about 10 months. Three estimates of %Sf were evaluated in this study: 1) low-%Sf integrated over one year, 2) medium-%Sf integrated over the last 7 months and 3) high-%Sf integrated over the last 5 months.
The column on the left of the table indicates observed sulfate concentrations at the end of year one and at the end of year 11. The best fit for each pile and variable combination was chosen and is indicated by the shading.
The best fits were either of the two variable combinations:
|
PILES: |
1,3 |
2,4,5,10 |
|
NET INFILTRATION % |
50 |
75 |
|
%Sf |
low |
high |
|
NDA(%) |
100 |
50 |
It was surprising that only two scenarios captured the best fits for all of the test piles. Additionally, piles 1 and 3 were primarily sandstone while the other piles were mainly shale. It is logical that high NDA fits better with the sandstones given its greater porosity. Why net infiltration appeared higher on the shale than on the sandstone is a mystery unless this actually estimates residence time of water. This analysis is far from definitive. It is just the beginning of what will be a rigorous process of identifying NDA, NDN, %Sf and net infiltrations for various rock types.
AMD/TIME was applied to a site in northern West Virginia five years after initial mining and where water quality and flow data were available. Figure 1 shows acid load and acid concentrations as estimated and as observed for the site. For this case the key variables were set as suggested by the above sandstone piles (1,3):
|
NET INFILTRATION % |
50 |
|
%Sf/yr |
6% |
|
NDA(%) |
100 |
This didn't work very well, estimated acid concentrations and loads were about 4 times higher than observed in the field. Eventually a good fit (the one shown in figure 1) was found with the following settings:
|
NET INFILTRATION % |
80 |
|
%Sf/yr |
4% |
|
NDA(%) |
50 |
Figure 1 shows a shortcoming with AMD/TIME as presently configured: alkalinity is presumed to be consumed with 100% efficiency until it is exhausted, the acid load curve then leaps from zero to a high number in the following year. In reality this is, doubtless, a more gradual process though the net result is likely to be the same.
Extension of the curves in figure 1 to the point of acid exhaustion indicates that at year 113 acid loading will be 10 tons per year (figure 2.). Long before that, however, it will be negligible or easily treated in a passive system. AMD/TIME was used to estimate the required amount of alkaline amendment arid the effects on acid generation. Figure 3 shows that AMD/TIME slightly underestimated the required amount (probably by a roundoff error) such that at year 115 the added limestone was exhausted and a small (19 T/yr) peak in acid load appeared for a few years.
Treatment costs are automatically estimated. by AMD/TIME. Figure 4 shows the cumulative costs of limestone amendment versus 113 years of water treatment using hydrated lime. With alkaline amendment all costs were incurred during mining.
AMD/TIME is extremely sensitive to application of the wrong key variables. Figure 5, 6 and 7 show the effects of holding all variables constant while varying %Sf from 1 to 8. It is likely that many refuse piles resemble figure 5 (without the alkalinity) while most spoils are closer to figure 6. An annual sulfur flux such as the 8% shown in figure 7 is unlikely to occur outside the laboratory. Figure 8 shows a very unhappy situation where, due to poor mixing of alkaline materials, NDN :s only 75% while NDA is 100%. This is otherwise the same scenario as figure 6. Things get much worse much more quickly, however.
Like all predictive tools, AMD/TIME is only as good as the variables which make it run. Since it uses empirical variables it is useful to compare predictions within set boundary conditions to field observations.
The spreadsheet is not proprietary. A copy of the codes is attached in the appendix. Feel free to copy it and use it. I only ask two favors:
Let me know either way. My phone number is (304) 293 2867.
Bradham, W.F., and FT Caruccio. 1991. A comparative study of tailings analysis using acid/base accounting, cells, columns and soxhlets. pp. 157-174. In: Proceedings of the Second International Conference on the Abatement of Acidic Drainage. Montreal, P.Q.
Brady, K., and R.J. Hornberger. 1989. Mine drainage prediction and overburden analysis in Pennsylvania. In: Proceedings of the 1989 West Virginia Surface Mine Drainage Task Force Symposium. West Virginia Mining and Reclamation Association, Charleston, WV.
Brady, K., and R.J. Hornberger. 1990. A manual for premining prediction of coal mine drainage quality. Pennsylvania Department of Environmental Resources, Harrisburg, PA.
Brady, K., M.W. Smith, R.L. Beam and C.A. Cravotta. 1990. Effectiveness of the use of alkaline materials at surface coal mines in preventing or abating acid mine drainage: Part 2. Mine site case studies. pp. 227-241. In: Proceedings of the 1990 Mining and Reclamation Conference, West Virginia University, Morgantown, WV.
Cravotta, C.A., K. Brady, M.W. Smith, and R.L. Beam. 1990. Effectiveness of the addition of alkaline materials at surface coal mines in preventing or abating acid mine drainage: Part 1. Geochernical considerations. pp. 221-225. In: Proceedings of the 1990 Mining and Reclamation Conference. West Virginia University, Morgantown, WV.
diPretoro, R.S. 1986. Premining prediction of acid drainage potential from surface coal mines in northern West Virginia. M.S. Thesis, Geology Department, West Virginia University. 217 pp.
diPretoro, R.S. and H.W. Rauch. 1988. Use of acid-base accounts in premining prediction of acid drainage potential: a new approach from northern West Virginia. pp. 2-10. In: Mine Drainage and Surface Mine Reclamation. USDI Bureau of Mines Information Circular 9183.
Erickson, P.M., and R.S. Hedin. 1988. Evaluation of overburden analytical methods as means to predict post-mining coal mine drainage quality. pp. 11-19. In: Mine Drainage and Surface Mine Reclamation. USDI Bureau of Mines Information Circular 9183.
Hedin, R.S., and P.M. Erickson. 1988. Relationships between the initial geochemistry and leachate chemistry of weathering overburden samples. pp. 21-28. In: Mine Drainage and Surface Mine Reclamation. USDI Bureau of Mines Information Circular 9183.
Lapakko, K. 1988. Prediction of acid mine drainage from Duluth Complex mining wastes in northeastern Minnesota. pp. 180-190. In: Mine Drainage and Surface Mine Reclamation. USDI Bureau of Mines Information Circular 9183.
O'Hagan, M., and F.T. Caruccio. 1986. The effect of admixed limestone on rates of pyrite oxidation in low, medium, and high sulfur rocks. In: 1986 National Symposium on Mining, Hydrology, Sedimentology, and Reclamation. University of Kentucky, Lexington, KY.
Renton, JJ, Rymer, T.E. and Stiller, A.H. 1988 A laboratory procedure to evaluate the acid producing potential of coal associated rocks. Mining Science and Technology, 7 (1988) 227-235.
Rich, D.H. and Hutchison, K.R. 1990 Neutralization and stabilization of combined refuse using lime kiln dust at High Power Mountain. In: Proceedings of the 1990 Mining and Reclamation Conference. 2 vols. West Virginia University, Morgantown, WV.
Rymer, T.E., Renton, J.J., and Ziemkiewicz, P.F. 1991, Isolation of critical predictive acid producing parameters from variable field data using advanced computer technology. In: Proceedings of the Second International Conference on the Abatement of Acidic Drainage, Montreal, P.O.
Shueck, J. 1990 Using a magnetometer for investigating underground coal mine fires, burning coal refuse banks and for locating AMD source areas in surface mines. In: Proceedings of the 1990 Mining and Reclamation Conference and Exhibition. 2 vols. West Virginia University, Morgantown, WV.
Skousen, J., R.M. Smith, and J. Sencindiver. 1990. Development of the Acid-Base Account. Green Lands 20(l): 32-37.
Smith, M.W., and K. Brady. 1990. Evaluation of acid base accounting data using computer spreadsheets. pp. 213-219. In: , Proceedings of the 1990 Mining and Reclamation Conference. West Virginia University, Morgantown, WV.
West Virginia University. 1971. Mine spoil potentials for water quality and controlled erosion. 14010 EJE 12/71. Contract with, West Virginia University by the U.S. Environmental Protection Agency, Washington D.C.
