Mineralogical constraints on the determination of neutralization potential and prediction of acid mine drainage

 Acid-base accounting tests, commonly used as a screening tool in acid mine drainage (AMD) predictions, have limitations in (1) measuring with confidence the amount of neutralizers present in samples and (2) affording an interpretation of what the test results mean in terms of predicting the occurrence of acid mine drainage. Aside from the analytical difficulties inherent to the conventional methods, a potential source of error in neutralization potential (NP) measurements is the contribution from the dissolution of non-carbonate minerals. Non-carbonate alkalinity measured during static tests may or may not be available to neutralize acidity produced in the field. In order to assess the value-added of extending the NP with the knowledge of mineralogical composition and evaluate potential sources of errors in NP measurements, a suite of samples were examined and characterized in terms of their mineralogical and chemical compositions. The results indicate that although the acid-base accounting tests work well for simple compositions, the tests may result in overestimation or underestimation of NP values for field samples. Mineralogical constraint diagrams relating NP determinations to Ca, Mg and CO₂ concentrations were developed with the purpose to serve as supplementary guides to conventional static tests in identifying possible NP contributions from non-carbonate minerals and checking the quality of the chemical testing results. Mineralogical NP makes it possible to interpret the meaning of NP results and to assess the behaviour of samples over time by predicting the onset of AMD and calculating NP values for individual size fractions. Received: 1 June 1998 · Accepted: 6 October 1998     

Field assessment of acid mine drainage contamination in surface and ground water

Both sulfate and conductivity are useful indicators of acid mine drainage (AMD) contamination. Unlike pH, they are both extremely sensitive to AMD even where large dilutions have occurred. The advantage of using sulfate to trace AMD is that unlike other ions it is not removed to any great extent by sorption or precipitation processes, being unaffected by fluctuations in pH. These two parameters are also closely associated as would be expected, as conductivity is especially sensitive to sulfate ions. Therefore, as sulfate analysis is difficult in the field, conductivity can be used to predict sulfate concentration in both AMD and contaminated surface waters using regression analysis. Most accurate predictions are achieved by using equations given for specific conductivity ranges or AMD sources. There is also potential to use conductivity to predict approximate concentrations of key metals when the pH of the water is within their respective solubility ranges.     

Renovation of a failed constructed wetland treating acid mine drainage

 Acid mine drainage (AMD) from abandoned underground mines significantly impairs water quality in the Jones Branch watershed in McCreary Co., Kentucky, USA. A 1022-m² surface-flow wetland was constructed in 1989 to reduce the AMD effects, however, the system failed after six months due to insufficient utilization of the treatment area, inadequate alkalinity production and metal overloading. In an attempt to improve treatment efficiencies, a renovation project was designed incorporating two anoxic limestone drains (ALDs) and a series of anaerobic subsurface drains that promote vertical flow of mine water through a successive alkalinity producing system (SAPS) of limestone beds overlain by organic compost. Analytical results from the 19-month post-renovation period are very encouraging. Mean iron concentrations have decreased from 787 to 39 mg l^–1, pH increased from 3.38 to 6.46 and acidity has been reduced from 2244 to 199 mg l^–1 (CaCO₃ equivalent). Mass removal rates averaged 98% for Al, 95% for Fe, 94% for acidity, 55% for sulfate and 49% for Mn during the study period. The results indicate that increased alkalinity production from limestone dissolution and longer residence time have contributed to sufficient buffering and metal retention. The combination of ALDs and SAPS technologies used in the renovation and the sequence in which they were implemented within the wetland system proved to be an adequate and very promising design for the treatment of this and other sources of high metal load AMD. Received: 29 June 1998 · Accepted: 15 September 1998     

Environmental impact and remediation of acid mine drainage: a management problem

 Work carried out at the abandoned copper (Cu) and sulphur (S) mine at Avoca (south east Ireland) has shown acid mine drainage (AMD) to be a multi-factor pollutant. It affects aquatic ecosystems by a number of direct and indirect pathways. Major impact areas are rivers, lakes, estuaries and coastal waters, although AMD affects different aquatic ecosystems in different ways. Due to its complexity, the impact of AMD is difficult to quantify and predict, especially in riverine systems. Pollutional effects of AMD are complex but can be categorized as (a) metal toxicity, (b) sedimentation processes, (c) acidity, and (d) salinization. Remediation of such impacts requires a systems management approach which is outlined. A number of working procedures which have been developed to characterise AMD sites, to produce surface water quality management plans, and to remediate mine sites and AMD are all discussed. Received: 16 January 1996 · Accepted: 5 March 1996     

Sampling riverine sediments impacted by acid mine drainage: problems and solutions

 Sampling acid mine drainage (AMD) or natural acid rock drainage (ARD)-impacted sediments is complex, requiring appropriate field sampling techniques to ensure representative samples that are both repeatable and reproducible. The important factors affecting sampling of riverine sediments are examined. These include sample site location, field observations, representative sampling, sample collection techniques, and sample preservation. A recommended sampling and processing protocol is presented for AMD- and ARD-impacted riverine sediments, which includes sediment sampling, Fe hydroxide floc sampling, chemical analysis, interstitial (pore) water collection, sediment elutriates, sediment fractionation, and physical analysis. The importance of bioassay testing is discussed, as is quality assurance and assessment approaches to define sediment quality criteria. Received: 18 September 1995 · Accepted: 23 October 1995     

Hydrogeochemical characteristics of acid mine drainage and water pollution at Makum Coalfield,India

Acid mine drainage (AMD) is one of the severe environmental problems that coal mines are facing. Generation of AMD in the northeastern part of India due to the coal mining activities has long been reported. However detailed geochemical characterization of AMD and its impact on water quality of various creeks, river and groundwater in the area has never been reported. Coal and coal measure rocks in the study area show finely disseminated pyrite crystals. Secondary solid phases, resulted due to oxidation of pyrite, occur on the surface of coal, and are mainly consisting of hydrated sulphate complexes of Fe and Mg (copiapite group of minerals). The direct mine discharges are highly acidic (up to pH 2.3) to alkaline (up to pH 7.6) in nature with high concentration of SO₄²⁻. Acidic discharges are highly enriched with Fe, Al, Mn, Ni, Pb and Cd, while Cr, Cu, Zn and Co are below their maximum permissible limit in most mine discharges. Creeks that carrying the direct mine discharges are highly contaminated; whereas major rivers are not much impacted by AMD. Ground water close to the collieries and AMD affected creeks are highly contaminated by Mn, Fe and Pb. Through geochemical modeling, it is inferred that jarosite is stable at pH less than 2.5, schwertmannite at pH less than 4.5, ferrihydrite above 5.8 and goethite is stable over wide range of pH, from highly acidic to alkaline condition.     

Hydrogeochemical characteristics of acid mine drainage in the vicinity of an abandoned mine, Daduk Creek, Korea

Acid mine drainage discharged from the abandoned Daduk mine towards the Daduk creek has a pH of 3.3, and concentrations of Al, Mn, Fe, Zn and SO₄ of 18, 41, 45, 38 and 1940 mg/L, respectively. In particular, As concentration in acid mine drainage is 1000 μg/L. Removing order of metal ions normalized by SO₄ concentration downstream from discharge point is Fe > As > Al > Cu > Zn > Mn > Cd > Pb. In the Daduk creek, Fe and As are the most rapidly depleted downstream from acid mine drainage because As adsorbs, coprecipitates and forms compounds with ferric oxyhydroxide. From the results of geochemical modeling using the Phreeq C program, goethite (FeOOH) is oversaturated, and schwertmannite (Fe₈O₈(OH)_4.5(SO₄)_1.75) is the most stable solid phase at low pH in the Daduk creek. Yellowish red (orange ochre) precipitates that occurred in the study area are probably composed of goethite or schwertmannite.     

Heavy metal geochemistry of the acid mine drainage discharged from the Hejiacun uranium mine in central Hunan,China

The acid mine drainage (AMD) discharged from the Hejiacun uranium mine in central Hunan (China) was sampled and analyzed using ICP-MS techniques. The analyzing results show that the AMD is characterized by the major ions Fe_Total, Mn, Al and Si, and is concentrated with heavy metals and metalloids including Cd, Co, Ni, Zn, U, Cu, Pb, Tl, V, Cr, Se, As and Sb. During the AMD flowing downstream, the dissolved heavy metals were removed from the AMD waters through adsorption onto and co-precipitation with metal-oxhydroxides coated on the streambed. Among these metals, Cd, Co, Ni, Zn, U, Cu, Pb and Tl are negatively correlated to pH values, and positively correlated to major ions Fe, Al, Si, Mn, Mg, Ca and K. The metals/metalloids V, Cr, Se, As and Sb are conservative in the AMD solution, and negatively-correlated to major ions Na, Ca and Mg. Due to the above different behaviors of these chemical elements, the pH-negatively related metals (PM) and the conservative metals (CM) are identified; the PM metals include Cd, Co, Ni, Zn, U, Cu, Pb and Tl, and the CM metals V, Cr, Se, As and Sb. Based on understanding the geochemistry of PM and CM metals in the AMD waters, a new equation: EXT = (Acidity + PM)/pH + CM × pH, is proposed to estimate and evaluate extent of heavy-metal pollution (EXT) of AMD. The evaluation results show that the AMD and surface waters of the mine area have high EXT values, and they could be the potential source of heavy-metal contamination of the surrounding environment. Therefore, it is suggested that both the AMD and surface waters should be treated before they are drained out of the mine district, for which the traditional dilution and neutralization methods can be applied to remove the PM metals from the AMD waters, and new techniques through reducing the pH value of the downstream AMD waters should be developed for removal of the CM metals.     

Acid mine drainage and acid rock drainage processes in the environment of Herrerías Mine (Iberian Pyrite Belt, Huelva-Spain) and impact on the Andevalo Dam

The present work describes the process of acid water discharge into the Andévalo Dam (Iberian Pyrite Belt, Huelva-Spain) starting from the interpretation of rainfall data and chemical analyses regarding pH, conductivity, metal and sulphate content in water, from a time series corresponding to the sampling of two confluent channels that discharge water into the referred dam. Statistical data treatment allows us to conclude the existence of acid mine drainage processes in the Chorrito Stream, which are translated into very low pH values and high sulphate and metal concentrations in the water coming from Herrerías Mine. On the other hand, the Higuereta Stream shows, for the same parameters, much lower values that can be interpreted as the channel response to acid rock drainage processes in its drainage basin induced by the rocky outcrops of the Iberian Pyrite Belt.     

Hydrochemical characterization of acute acid mine drainage at Iron Duke mine, Mazowe, Zimbabwe

 Acid mine drainage (AMD) with a minimum pH of 0.52 was recorded at Iron Duke mine near Mazowe, Zimbabwe during an investigation of the environmental geochemistry of mine waters in the Greenstone Belts of Zimbabwe. Hydrochemical data for waters emanating from the Iron Duke waste-rock pile indicate their super-saturation with respect to Fe and SO₄ ^2–. Extremely high dissolved concentrations of Al, Zn, Cu, Co, Ni, V, Cr, Cd and As also prevail. Substantial losses of metals from solution occur within 400 m of the AMD source through the precipitation of crystalline sulphates, principally melanterite. Further downstream, hydrous oxide precipitation forms the dominant mechanism of metal attenuation in waters characteristically under-saturated with respect to Fe sulphates. Speciation and saturation index data generated using the equilibrium model WATEQ4F, suggest that such codes have broad utility for generic prediction of the mineralogical contraints on metal mobility in acute AMD systems. Major discrepancies between modelled and empirical hydrochemistries are, however, evident for super-saturated waters in which the kinetics of Fe precipitation are slow, and in which total ionic strengths markedly exceed their theoretical maximum. Received: 28 August 1998 · Accepted: 7 December 1998     

Prediction of acid mine drainage generation potential in selected mines in the Ashanti Metallogenic Belt using static geochemical methods

Acid mine/rock drainage (AMD/ARD) is the biggest environmental threat facing the mining industry. This study investigates AMD/ARD possibilities in three mines in the Ashanti Belt, using acid base accounting (ABA) and net acid generation pH (NAGpH) tests. Twenty-eight samples of rock units and mine spoil from these mines were collected for ABA and NAGpH tests. Two tailing dumps at Prestea and Nsuta were confirmed by both methods as acid generating with NAGpH of 4.5 and 4.6 and neutralization potential ratio values of 4.38 and 4.60, respectively. Six other samples are classified as potentially acid generating using a variety of established classification criteria. The rest of the samples either exhibited very low sulphur and carbonate content or had excess carbonate over sulphur. Consistency between results from ABA and NAGpH tests validates these tests as adequate tools for preliminary evaluation of AMD/ARD possibilities in any mining project in the Ashanti Belt.     

The potential use of serpentinite in the passive treatment of acid mine drainage: batch experiments

The efficiency of serpentinite as an alternative alkalinity generating material for the passive treatment of acid mine drainage (AMD) was assessed in the laboratory. Three series of batch experiments were designed for the passive treatment of a low pH (1.6) AMD synthetic solution containing 2,500 ppm Fe²⁺, 6,600 ppm SO₄^2–, 10.5 ppm Al, 15 ppm Ni, and traces of Cr, Mn and Cu. The influencing factors studied were: the effect of water/rock ratio, residence time, type of the alkalinity generating material (dolomite, magnesite, marble, serpentinite), and nature of the system (open vs. closed cells). The variations in solution chemistry observed in the open cells indicate that a lower water/rock ratio (0.33 ml/g) was the most efficient for metals removal. The optimal residence time in open cells was 24 h to reach the higher pH values. In the closed cells laboratory setup, synthetic AMD was placed in contact with the various alkaline materials for three different contact times (24, 48, 72 h). The optimal pH was reached after 48 h and did not change appreciably for longer contact time, and the best results for metal removals were obtained with marble and serpentinite. Single treatment efficiency was compared with a successive treatment approach. The most promising results were obtained with a five step treatment: (1) pre-treatment in a closed cell using serpentinite, (2) aeration and settling, (3) treatment in an open cell using marble, (4) final aeration and settling, and (5) filtration with a coarse silica sand. With this configuration, the final pH was 6.5 and pronounced metals depletion was achieved (100% for Al, 99.95% for Fe, 85.7% for Ni).     

Sulfate reducing bioreactor dependence on organic substrates for remediation of coal-generated acid mine drainage: Field experiments

Field experiments were conducted over a 460-day period to assess the efficiency of different mixtures of organic substrates to remediate coalmine-generated acid mine drainage (AMD). Five pilot-scale, flow-through bioreactors containing mixtures of herbaceous and woody organic substrates along with one control reactor containing only limestone were constructed at the Tab-Simco site and exposed to AMD in situ. Tab-Simco is an abandoned coal mine near Carbondale, Illinois that produces AMD with pH ∼2.5 and notably high average concentrations of SO₄ (5050 mg/L), Fe (950 mg/L), Al (200 mg/L), and Mn (44 mg/L). Results showed that the sequestration of SO₄ and metals was achieved in all reactors; however, the presence and type of organic carbon matrix impacted the overall system dynamics and the AMD remediation efficiency. All organic substrate-based reactors established communities of sulfate reducing microorganisms that contributed to enhanced removal of SO₄, Fe, and trace metals (i.e., Cu, Cd, Zn, Ni) via microbially-mediated reduction followed by precipitation of insoluble sulfides. Additional mechanisms of contaminant removal were active in all reactors and included Al- and Fe-rich phase precipitation and contaminant surface sorption on available organic and inorganic substrates. The organic substrate-based reactors removed more SO₄, Fe, and Al than the limestone-only control reactor, which achieved an average removal of ∼19 mol% SO₄, ∼49 mol% Fe, 36 mol% Al, and 2 mol% Mn. In the organic substrate-based reactors, increasing herbaceous content correlated with increased removal efficiency of SO₄ (26–35 mol%), Fe (36–62 mol%), Al (78–83 mol%), Mn (2–6 mol%), Ni (64–81 mol%), Zn (88–95 mol%), Cu (72–85 mol%), and Cd (90–92 mol%), while the diversity of the intrinsic microbial community remained relatively unchanged. The extrapolation of these results to the full-scale Tab-Simco treatment system indicated that, over the course of a 460-day period, the predominantly herbaceous bioreactors could remove up to 92,500 kg SO₄, 30,000 kg Fe, 8,950 kg Al, and 167 kg Mn, which represents a 18.3 wt%, 36.8 wt%, 4.1 wt% and 82.3 wt% increase in SO₄, Fe, Al, and Mn, respectively, removal efficiency compared to the predominantly ligneous bioreactors.The results imply that anaerobic organic substrate bioreactors are promising technologies for remediation of coal-generated AMD and that increasing herbaceous content in the organic substrate matrix can enhance contaminant sequestration. However, in order to improve the remediation capacity, future designs must optimize not only the organic carbon substrate but also include a pretreatment phase in which the bulk of dissolved Fe/Al-species are removed from the influent AMD prior to entering the bioreactor because of 1) seasonal variations in temperature and redox gradients could induce dissolution of the previously formed redox sensitive compounds, and 2) microbially-mediated sulfate reduction activity may be inhibited by the excessive precipitation of Al- and Fe-rich phases.     

On the neutralization of acid rock drainage by carbonate and silicate minerals

The net result of acid-generating and-neutralizing reactions within mining wastes is termed acid rock drainage (ARD). The oxidation of sulfide minerals is the major contributor to acid generation. Dissolution and alteration of various minerals can contribute to the neutralization of acid. Definitions of alkalinity, acidity, and buffer capacity are reviewed, and a detailed discussion of the dissolution and neutralizing capacity of carbonate and silicate minerals related to equilibium conditions, dissolution mechanism, and kinetics is provided. Factors that determine neutralization rate by carbonate and silicate minerals include: pH, P_{CO 2}, equilibrium conditions, temperature, mineral composition and structure, redox conditions, and the presence of foreign ions. Similar factors affect sulfide oxidation. Comparison of rates shows sulfides react fastest, followed by carbonates and silicates. The differences in the reaction mechanisms and kinetics of neutralization have important implications in the prediction, control, and regulation of ARD. Current static and kinetic prediction methods upon which mine permitting, ARD control, and mine closure plans are based do not consider sample mineralogy or the kinetics of the acid-generating and-neutralizing reactions. Erroneous test interpretations and predictions can result. The importance of considering mineralogy for site-specific interpretation is highlighted. Uncertainty in prediction leads to difficulties for the mine operator in developing satisfactory and cost-effective control and remediation measures. Thus, the application of regulations and guidelines for waste management planning need to beflexible.     

A laboratory study of covers made of low-sulphide tailings to prevent acid mine drainage

Covers with capillary barrier effects (CCBE) are considered to be one of the most effective ways to control acid mine drainage (AMD) production from mine wastes. The use of low-sulphide tailings in CCBE has been proposed recently for cases where other types of material may be unavailable near the mining site. This paper presents leaching column test results showing that CCBEs with a moisture-retaining layer made of slightly reactive tailings, with three different sulphide contents, can effectively limit the production of AMD from the acid-generating tailings placed underneath. With these layered covers, the leachate pH was maintained near neutrality throughout the testing period. When compared to uncovered tailings, the efficiency of the cover systems for reducing the amount of contaminants in the percolated water was determined to be greater than 99% for zinc, copper and iron. This study shows that the use of low-sulphide tailings can improve the ability of a CCBE to limit gas diffusion by consuming a fraction of the migrating oxygen.     

Characterization of Kolar gold field mine tailings for cyanide and acid drainage

The gold mining process at Kolar gold field (KGF) mines has generated about 32 million tons of tailings. Gold was extracted from the mined ores using cyanidation technique that involved dissolution of gold in the ore by water soluble alkali metal cyanides (example, sodium cyanide or potassium cyanide). Of the several dumps that received the mine tailings only the Kennedy’s Line dump was active prior to closure of the KGF mines in the year 2000. The Kennedy’s Line dump received sulfide bearing tailings in slurry form that comprised of spent ore and process water bearing soluble alkali metal cyanide. Depending on the pH of the tailing slurry, the free cyanides may exist as aqueous hydrogen cyanide that can escape to the atmosphere as hydrogen cyanide gas or occur as soluble cyanide (CN⁻) ions that can be leached by infiltrating water to the sub-surface environment. Additionally, the presence of pyrite minerals in the Kennedy’s Line dump makes them susceptible to acid drainage. This study examines the potential of gold tailings of Kennedy’s Line dump to release cyanide ions (CN⁻) and acid drainage to the sub-surface environment by performing physico-chemical and leaching tests with tailing samples collected from various depths of the dump, sub-surface soil samples beneath the dump and groundwater samples from vicinity of Kennedy’s Line dump. The chemical mechanisms responsible for the ambient cyanide and pH levels of the tailing dump, sub-surface soil samples and groundwater are also inferred from the laboratory results.     

The status of the passive treatment systems for acid mine drainage in South Korea

This study was performed to investigate the operating status, evaluate the problems, and discuss possible improvement methods of passive treatment systems for acid mine drainage (AMD) in South Korea. Thirty-five passive treatment systems in 29 mines have been constructed from 1996 to 2002 using successive alkalinity producing systems (SAPS) as the main treatment process. We investigated 29 systems (two for metal mines), 19 of which revealed various problems. Overflows of drainage from SAPS, wetland, or oxidation ponds were caused by the flow rate exceeding the capacities of the facilities or by the reduced permeability of the organic substance layer. Leakages occurred at various parts of the systems. In some cases, clogged and broken pipes at the mouths of the mine adits made the whole system unusable. Some systems showed very low efficiencies without apparent leakage or overflow. Even though the systems showed fairly good efficiencies in metal removal ratios (mainly iron) and pH control; sulfate removal rates were very poor except in three systems, which may indicate very poor sulfate reductions with sulfate reducing bacteria (SRB) as a means.     

Experimental study of the stability of metals associated with iron oxyhydroxides precipitated in acid mine drainage

Iron oxyhydroxide precipitates associated with acid mine drainage (AMD) from the Stearns Coal Zone in southeastern Kentucky were analyzed for their metal (Al, Cu, Pb, Mn, Ni, and Zn) content. The most concentrated metals within these sediments are nickel (27–32×10³μmol/kg), manganese (16–29×10³μmol/kg), and aluminum (13–22×10³μmol/kg) as determined by HCl-HNO₃ digestion. Metal concentrations associated with the organic fraction as determined by H₂O₂ digestion were generally far lower, with the exception of aluminum. "Batch" experiments (at initial pH=2.0) were used to analyze the stability of these metals associated with a contaminated soil. Aluminum was the most mobile of the metals, presumably the result of the formation of aluminum-sulfate aqueous complexes. The solubilization rates for nickel and iron were very similar, suggesting that nickel, unlike the other metals, coprecipitated with iron in these sulfatic oxyhydroxides. Received: 9 October 1997 · Accepted: 15 December 1997     

Interaction of acid mine drainage with waters and sediments at the Corona stream, Lousal mine (Iberian Pyrite Belt, Southern Portugal)

This study investigates the geochemical characteristics of the acid mine drainage discharged from the abandoned mine adits and tailing piles in the vicinity of the Lousal mine and evaluates the extent of pollution on water and on the stream sediments of the Corona stream. Atmospheric precipitation interacting with sulphide minerals in exposed tailings produces runoff water with pH values as low as 1.9–2.9 and high concentrations of (9,249–20,700 mg l⁻¹), Fe (959–4,830 mg l⁻¹) and Al (136–624 mg l⁻¹). The acidic effluents and mixed stream water carry elevated Cu, Pb, Zn, Cd and As concentrations that exceed the water quality standards. However, the severity of contamination generally decreases 4 km downstream of the source due to mixing with fresh waters, which causes the dilution of dissolved toxic metals and neutralization of acidity. Some natural attenuation of the contaminants also occurs due to the general reduced solubility of most trace metals, which may be removed from solution, by either co-precipitation or adsorption to the iron and aluminium precipitates.     

Fuzzy modelling of acid mine drainage environments using geochemical,ecological and mineralogical indicators

Fuzzy logic was applied to model acid mine drainage (AMD) and to obtain a classification index of the environmental impact in a contaminated riverine system. The data set used to develop this fuzzy model (a fuzzy classifier) concerns an abandoned mine in Northern Portugal—Valdarcas mining site. Here, distinctive drainage environments (spatial patterns) can be observed based on the AMD formed in the sulphide-rich waste-dumps. Such environments were established, as the effluent flows through the mining area, using several kinds of indicators. These are physical–chemical, ecological and mineralogical parameters, being expressed in a quantitative or qualitative basis. The fuzzy classifier proposed in this paper is a min–max fuzzy inference system, representing the spatial behaviour of those indicators, using the AMD environments as patterns. As they represent different levels (classes) of contamination, the fuzzy classifier can be used as a tool, allowing a more reasonable approach, compared with classical models, to characterize the environmental impact caused by AMD. In a general way it can be applied to other sites where sulphide-rich waste-dumps are promoting the pollution of superficial water through the generation of AMD. Electronic supplementary material  The online version of this article (doi:) contains supplementary material, which is available to authorized users.