Although data generated from predictive techniques give a base for planning preventive, corrective and remedial measures, so far they have been found to have restricted application as universal predictive tool. S herlock et al ( 1 995 ) considered that predictive techniques must be applied on a site-specific basis and take into account the mineralogy of the waste material. For example, minerals present, percent sulphides and their distribution within the rock mass and along joints and other discontinuities and the likely durability of the waste rock are likely to effect the predictive test results (MEND, 1 99 1 ; Orava and Swider, 1 996). Moreover, predictive techniques alone do not account for the build-up of metal salts that may occur after disposal but prior to final reclamation. The tests also fail to incorporate an assessment of the coating of sulphide phases with unreactive coatings that may occur n aturally (Pratt et aI, 1 996) .
Most of these tests also force oxidation or neutralisation reactions that may never occur in the real situation. Acid generation (and neutralisation) are time dependent phenomenon, and until someone develops a test that takes into account time dependency, there will never be an "exact" predictor. For example a test that i ndicates the presence of neutralisers, inherent or added does not mean that the net result will be no acid. That depends on the reactivity and kinetics . If the acid is generated faster than it can be neutrali sed, the net result will be an acid effluent regardless of how much neutraliser is available.
Short-term static tests, which are conducted to determine the acid generating or acid consuming potential, will usually provide only an indirect assessment of the net acid generating potential. A longer-term kinetic test, which allows reactions to occur, will provide a more comprehensive assessment. However, even a kinetic test may not predict the net acid generating potential accuratel y because the ongoing sulphide oxidation will continue to produce acid and acid consumption by the alkalinity present in the system may be dominant only in the beginning. Conversely, the alkalinity in the system could overcome and exhaust the supply of sulphide bearing roc k present. It is therefore c lear that both the static and kinetic tests must be designed to suit the mineralogy of the waste rock being tested. B oth acid generatin g and acid consuming reaction rates in the
oxidation of pyrite and reactions with carbonate and silicate minerals must be considered for predictive acid generation tests.
The issue of appropriate NP/AP ratios is a key area of debate among the regulatory agencies and the mining industry. To date there is no comprehensive compilation of case histories of mine sites with significant ABA data, NP/AP ratios and AMD problems. The NP/ AP ratio is the most significant of the variables which regulatory agencies are attempting to use as a prescriptive measure. Ferguson and Morin ( 1 99 1 ) and Cravotta et al . ( 1 990) suggested that NP/ AP criterion separating potentially acid generating and non-acid generating samples could be about 2: 1 . However, in the data base presented by them, no sample with NP/AP > 1 produced acidic leachate in 1 66 laboratory leaching tests. There is also no field evidence of NP/AP> 1 producing AMD. The NP/ AP ratio may be considered as a safety factor with a higher safety factor probably required for mines in wet climates where carbonate minerals may be preferentially leached from the mine waste relative to the oxidation of the contained sulphide minerals.
The first interpretative use published for ABA was an estimate of NNP>S kg CaC03 fl producing alkaline conditions (Sobek et aI . , 1 978). This screening criteria was selected based on soil quality and plant growth media considerations, not mine drainage prediction. ABA later began to be applied to coal mine drainage prediction, beginning in the late 1 970's. At this time it became apparent that ABA interpretation depended on whether the end use was mine drainage prediction or mine spoil/growth media suitability. ABA is still used for both purposes today and two sets of interpretative frameworks have developed.
Several researchers have suggested that the standard ABA procedures may in fact substantially underestimate the neutrali ing material requirement of the potentially acid generating materials (Cravotta et aI . , 1990; Brady and Cravotta, 1 992; Brady et aI . , 1 994). It has been suggested that the currently used value of 3. 1 25 g CaC03 equivalent to neutralise acidity from oxidation of 1 g S should in fact be 6.25 g CaC03 to assure a neutral AMD (Cravotta et aI ., 1 990; Brady et aI ., 1 994). Field studies conducted by Brady et al . ( 1 994) found that the alkaline material requirement calculated from ABA
analysis using sulphide-S to CaC03 ratio of 3 . 1 25 was inadequate for controlling AMD. Only when the ratio was doubled to 6.25, was there an overall net alkalinity of water at 1 1 of 1 2 coal mine sites studied. Perry and Brady ( 1 995) showed that material with an NP>2 1 generall y produced alkaline drainage whereas NP< l O produced acidic drainage. The APP on the other hand, is considered adequate in predicting AMD only in materials that contained insignificant amount of carbonates « 1 % CaC03), in which case a relationship between total sulphide-S and acidity could be defined (Perry and Brady, 1 995). Other studies have shown that factors other than mine waste characteristics may be involved in the generation of AMD (DiPretoro and Rauch, 1 98 8 ; Erikson and Hedin, 1 988). They found that there was poor correlation among APP, NP and NNP from ABA and net alkalinity from drainage water. O'Hagan and Caruccio ( 1 986) found that addition of CaC03 at 5% by weight to a coal refuge containing 1 % S produced alkaline drainage whereas Lapakko ( 1 98 8 ) indicated that CaC03 > 3 % was needed to neutralise an overburden material with 1 . 1 7 % S .
The discussions above indicate that there are still discrepancies in the use o f ABA as screening tool in predicting AMD. It is also evident from their results that lime requirement assessed from ABA analysis did not always produce neutral drainage from waste rock dumpsites. There was a need for further studies on the rate, application and placement of alkaline materials in mine waste and mine sites with potential to generate AMD. There appears to be no universal set of threshold numbers for defining cut-offs on ABA. Instead, the data tend to group themselves in ranges (Brady et al., 1 990; Cravotta et al., 1 990; DiPretoro and Rauch, 1 988). However, ABA testing procedures are finnly entrenched in the mining industry and it is likely to stay in use because it is familiar to industry, consultants and regulators, and it is low cost and has rapid turnaround time.
As with the static tests, kinetic tests are also subject to queries about their accuracy in predicting real situation AMD conditions. Kinetic tests (humidity cells, columns and soxhlets) not only produce a unique leachate but also modify the sample and a significant variation in the accuracy of the results was observed (Bradham and Caruccio, 1 990) . No data is yet available from weathering tests which were run long enough to see the sulphate generation rate begin to taper off. Most weathering test
results are assessed as "acid or not acid" producing. In other words, there is not anything particularly kinetic about the data analysis. While static and kinetic tests and the associated models are far from perfect in their capacity to predict the generation or migration of AMD, they do allow for a more systematic approach to understanding the potential problems.