Samples were taken during May 2013 along pre-established calcite saturation index transects (Mayes et al., 2008, 2018; Riley and Mayes, 2015) within Dene Burn and Howden Burn (Figure 1) for further analyses in the laboratory. In these streams, calcite saturation is enhanced by four alkaline discharges within the former Consett Iron and Steelworks. The site was operational from the middle of the nineteen century until decommissioning in the early 1908s (Figure 1). Waste up to 45 m depth, including slag, flue dusts, ashes and construction and demolition rubble were accumulated after working closure in an area of 2.9 km² (Harber and Forth, 2001). These materials are now stored in landfill, and the leachates emerging from them are alkaline ([OH^–] = 10–130 mg.L^–1; [CO₃^2–] = 10–110 mg.L^–1; [HCO₃^–] = 110 mg.L^–1) because of the bulk chemical composition meteoric waters develop after inter-acting with these materials in the subsurface (Mayes et al., 2008, 2018; Riley and Mayes, 2015).

These landfills can be a source of pollution to surface and ground waters (Mayes et al., 2008) (Figure 1). The production of iron results in the emission of sulfur dioxide gas, ore and coke dust particles (Harber and Forth, 2001). Other potential contaminants such as arsenic, chromium, boron and copper are present in iron ore (Harber and Forth, 2001). Coke making provides inorganic compounds, for example zinc, fluoride, sulfates, phosphates and organic compounds such as fuels and oils (Harber and Forth, 2001). Iron and steel flue dusts contain metals including lead, cadmium, halides, zinc, chromium, arsenic, copper, nickel, and alkali metals (Harber and Forth, 2001). Fluoride and zinc are the main pollutants of steel-making wastewaters (Harber and Forth, 2001).

Half of Dene Burn flow comes from a subterranean drainage network beneath the Grove Heaps which consist of blast furnace bottom and steel slag (Mayes et al., 2008), whereas the Howden burn drains an area previously occupied by the working blast furnaces, power station and steel plant (Mayes et al., 2008) (Figure 1). The hydrochemical facies of the leachates are seen to vary notably between the landfills resulting in the streams Howden Burn and Dene Burn (Mayes et al., 2008) (Figure 1). Dene Burn leachate source is dominated by Ca, Mg and HCO₃ characteristic of Coal Measures drainage in the region and Ca-OH type leachate from the dissolution of portlandite in the steel slag mounds (Mayes et al., 2008). Dene Burn source leachate also shows elevated Na and Cl concentrations from de-icing salt runoff (Mayes et al., 2008). Howden Burn leachate sources differ considerably from Dene Burn drainage system and are mixed waters rich in K, Na, Ca, and sulfate (Mayes et al., 2008) (Figure 1). Sulfate concentrations are exceptionally elevated [2,500 mg.L^–1] typical of a drainage from the former power station and coke works area of the workings (Figure 1). Within these locations, coke works spoil, ashes, and iron slags, which have a high sulfur content, were co-deposited with steel and blast furnace slag (Harber and Forth, 2001). The extremely important K content of Howden Burn leachate sources probably derive from highly soluble potassium oxides associated with flue dusts and ashes deposited with slag (Mayes et al., 2008).

Four samples reported here (HU-MR-/3CON1, HU-MR-/3CON2, HU-MR-/3CON4, HU-MR-/3CON5) are from Dene Burn (Figure 1) and further two samples (HU-MR-/3CON6, HU-MR-/3CON7) originate from Howden Burn (Figure 1).

During our sampling, on-site measurements of major physico-chemical parameters (pH, electrical conductivity and water temperature) were performed using a Myron L Ultrameter^® calibrated on each sample day with pH 4, 7, and 10 buffer solutions and a 1,413 μS conductivity standard, to confirm the system had not changed since previous sampling. Sample alkalinity was also obtained in the field using a two-stage titration against 1.6 N H₂SO₄ with phenolphthalein (to pH 8.3) and bromocresol green-methyl red indicators (to pH 4.6) to facilitate calculation of the constituents of sample alkalinity [i.e., hydroxyl, carbonate and bicarbonate alkalinity using the United States Geological Survey (USGS) Alkalinity calculator (U.S. Geological Survey [USGS], 2014)]. For each sample, three polypropylene bottles were filled, one of which was acidified with trace analysis grade concentrated nitric acid (for total cation and trace element analysis), one of which was filtered (with 0.45-μm cellulose nitrate filters) prior to acidification (for dissolved cation and trace element analyses) and the other left untreated (for anion analysis).

Hydrochemistry at each of the sampling sites on the day of sampling is summarized in Table 1, and is consistent with previous (Mayes et al., 2008) and more recent (Hull et al., 2014; Riley and Mayes, 2015; Hobson et al., 2018; Mayes et al., 2018) monitoring of these systems. Some calcium is sourced from background seepage of karst groundwater into both streams, but the majority is derived from steel slag leachate in the form of Ca(OH)₂. Although Ca²⁺_(aq) is elevated throughout the system, it is dramatically enhanced in the relatively proximal Howden Burn (161–239 mg.L^–1) even compared to the more proximal Dene Burn (11 to 14 mg.L^–1). CO₂ is supplied to the system from the air, and in-gasses vigorously due to the high pH of the stream waters, and carbonate alkalinity is high where pH is high (7–18 mg.L^–1 at Dene Burn and 103–109 mg.L^–1 at Howden Burn). Total alkalinity shows the same enhancement at proximal sites (55–59 mg.L^–1 in Dene Burn and 264–332 mg.L^–1 in Howden Burn). Given this water composition, it is unsurprising that calcite saturation index is very high (+0.04 to +2.47), and mineral precipitation is high enough to be lethal to local microbenthos by smothering (Hull et al., 2014).

Precipitation of carbonate material occurs immediately when the water contacts air, and waters seeping from the landfill package precipitate calcite rafts in shallow pools and slow moving water on emergence (Figure 2). The meniscus layer mineralization is a direct consequence of CO₂ in-gassing, but extends for only a few meters downstream. Precipitation is greatest in the environment of waterfalls (Figure 2). Fine micrite sediments fill the upper pools within the Howden Burn waterfall-pool system (Figure 2). Dark patches on the waterfall surface indicates the presence of substantial diatomaceous biofilms, and a less common yellow discoloration reflects the occurrence of cyanobacteria.

There is a significant hydrochemical gradient between the Dene Burn and Howden Burn streams (Table 1). Parameters such as pH, conductivity, alkalinity, CSI and R ([pH = 11.5–11.6]; [conductivity = 2160–3190 μS]; [alkalinity = 260–330mg.L^–1]; [CSI = 2.3–2.5]; [R = 4.10–10 – 5.10–10m.s−1]) are dramatically more elevated in Howden Burn than in Dene Burn ([pH = 9–10.7]; [conductivity = 425–450 μS]; [alkalinity = 55–66 mg.L^–1]; [CSI = 0.04–1]; [R = 9.10–13 – 3.10–11m.s−1]) (Table 1).

Location C4 in Dene Burn describes the lowest CSI (0.04), then followed by location C5 (0.25), C2 (0.67), and C1 respectively (0.96) (Table 1). In Howden Burn, the values are almost doubled at the C6 (2.3) and C7 (2.5) sites (Table 1) which are correlated with higher concentrations of carbonate (100–140 mg.L^–1) and calcium (160–240 mg.L^–1) (Table 1).

R is the lowest at location C5 (9.10–13m.s−1), then increases quite progressively at locations C2 (1.10–11m.s−1), C1 (3,1.10–11m.s−1), and C4 (3,4.10–11m.s−1) (Table 1). In Howden burn, R faces a major augmentation (C6: R = 4.10–10m.s-1; C7: R = 5.10–10m.s−1) (Table 1).

Calcite saturation index (CSI), ionic strength and activity of Ca²⁺ and CO₃^2– ions were obtained by using the PHREEQC Interactive software. The calcite growth rate (R) was calculated following the formula established by Wolthers et al. (2012):

R = I - 0.004 ⁢ p ⁢ H - 10.71 ⁢ r a ⁢ q - 0.35 ⁢ ( S - 1 ) 2

where I is the ionic strength, r_aq = {Ca²⁺}/{CO₃^2–} activity ratio, S is the saturation ratio (Ω^1/2) and R is expressed in m s^–1. Growth rate differs from saturation index in that it predicts how quickly crystals will actually assemble, rather than how thermodynamically “favored” that assembly is. Consequently, this value is intrinsically much more closely linked to crystal assembly controls than saturation is. This parameter was indeed used previously for understanding the precipitation of terrestrial carbonates (Kandianis et al., 2008).

X-ray powder diffraction (XRD) data were collected from ground sample (HU-MR-/3CON6) mounted in stainless steel sample holders. Data collection range 5 ≤ 2θ/° ≤ 80 with a step size 0.0262606° and a counting time of 304.725 s per step. Raw data were examined using the program PANAlytical HighScore (Plus) which is an implementation of the PDF2 database. This sample was also analyzed by Fourier transform infrared (FTIR) spectroscopy using an Agilent 4300 Handheld Portable FTIR and the peaks were analyzed and identified by the MicroLab FTIR Agilent Software.

Thin-sections (6 × 3 cm) were made for each sample, impregnated with blue-epoxy and were studied with a GXM-MXP Series L3230 optical microscope at the School of Environmental Sciences (University of Hull, England, United Kingdom) to determine the mineralogy, texture and the fabric of the studied carbonates. To identify the micro-texture and the micro-scale fabrics, a Scanning Electron Microscope (Zeiss EVO 60) attached to an EDS detector (Oxford Instruments INCA System350 with Silicon Drift Detector) was also used on the same samples for the characterization of their growth and diagenetic processes. The terminology used for petrographic descriptions follow the nomenclature from Pedley (1992), Jones et al. (2005), and Flügel (2013). The definitions of Jones (2017) for single crystal and polycrystal are also employed in this study. Single crystal corresponds to a solid object with only one grain or crystal and hence, no grain boundaries in which an orderly three-dimensional arrangement of the atoms, ions, or molecules is repeated throughout the entire volume. A polycrystal is an aggregate of several crystals or grains where the boundary between the grain is the grain boundary across which the orientation of the crystal changes and the point at which three boundaries meet is called the triple junction.

A key element of the petrographic part of this study is the definition of “shrubby.” We use the definition of Chafetz and Folk (1984): “three-dimensional, arborescent structures composed of CaCO₃ that expand away from the substrate.” This differs from dendrites, for which we use the definition of Jones et al. (2005): “100–200 nm thick calcite fibers that form 3D lattice-like domains and in each dendrite domain, fibers have three structurally equal orientations.”

XRD data show that HU-MR-/3CON6 is highly crystalline and corresponds to pure calcite (Figure 3). However, FTIR measurements indicate a material which is not calcite via 3 significant peaks (λ = 600 cm^–1; λ = 850 cm^–1; λ = 1100 cm^–1) (Figure 4). These absorptions are consistent with the presence of silica (SiO stretching in silica network defects, symmetric stretching of Si-O-Si bonds and silicon–oxygen covalent bonds vibrations respectively; Al-Oweini and El-Rassy, 2009). The absence of crystalline silica scattering in the XRD patterns indicate that this material must be amorphous, and is present in most samples at >1% concentrations. Consistent with field observations, SEM investigation indicated the presence of a large number of diatom frustules on the sample surface and within the shallow subsurface. The presence of substantial concentrations of these within the limestones explains the persistent presence of amorphous silica in these samples, indicating that the presence of diatomaceous biofilms at these sites during precipitation is persistent.

All samples are light brown limestone with clear sedimentary structures corresponding to brown-dark laminations except the HU-MR-/3CON6 sample which is light beige exhibiting a circular ring structure with beige brown laminations (Figure 5). In the latter, the beige layers have a fibrous texture (Figure 5). This specimen developed as a stalactite growing on the surface of an annual grass and hanging from the roof at the mouth of the drainage tunnel emerging near the top of Howden Burn. The ∼8 mm of growth of this specimen have therefore occurred within a few months before sampling, and the rate of accumulation at this site is likely >>1 cm yr^–1.

Eight types of microfacies have been identified on the basis of fabric, mineralogy and texture (Figure 6). We present a small number of replicates (6), which limits our ability to draw widely applicable conclusions about the distribution of specific features within the sediment. However, they present a wide range of microfabrics, some of which are rarely observed growing in natural environments today. We therefore limit our analysis to identification of these rare features, and to providing a framework for future investigations.

Shrubs occur exclusively at the most proximal location, C7, and is the only microfacies identified at that site (Figure 15A). Dendrite is the most common microfacies represented at location C6, but clotted micrite, sparry carbonate crust and cluster-shaped carbonate are present as well (Figure 15A). Microbial rim and sparry carbonate crust are the only microfacies described within location C1 (Figure 15A). Cluster-shaped carbonate is abundant at location C2 surrounded by some blocky calcite and sparry-carbonate crust (Figure 15A). Location C5 displays the most diverse microfacies, with clotted micrite, microbial rim, blocky calcite and multi-shaped calcite all contributing equal proportions of the sediment (Figure 15A). Multi-shaped calcite is a unique microfacies present only at location C4, and even these it is not abundant (Figure 15A).

In proximal Howden Burn, locations C6 and C7 show deposition principally of shrubby calcite (C7), and dendrite (C6) (pH = 11.5; CSI = 2.5; R = 5.10^–10m.s−1) (Figure 15A). These sites are characterized by very high pH (11.5–11.6), calcium concentration (161–240 mg.L^–1) and carbonate concentration (104–140 mg.L^–1) (Table 1). Consequently, calcite saturation (CSI = 2.3–2.5) and calcite growth rate (R = 4.4.10^–10–5.4.10^–10m.s−1) are also very elevated (Figure 15A). As the solution moves away from the most proximal sites, calcium and alkalinity are consumed steadily reducing CSI. However, this near-linear change in saturation does not well reflect the disappearance of shrub and dendrite microfacies, which do not occur beyond C6. This better reflects R, which is 100–1000 times lower in more distal Dene Burn sites than in Howden Burn (10^–12–3.10^–13m.s^–1) (Figure 15A). In locations C1 and C2, despite medium CSI (1–1.3) and high pH (10.3–10.7), R is low (1.10^–11–3.10^–11 m.s^–1) (Figure 15A), and this relatively weak kinetic forcing corresponds to formation of microbial rim (C1), sparry-carbonate crust (C1) and cluster shaped carbonate (C2). The most distal sites exhibiting significant carbonate deposits (C4 and C5) exhibit the lowest CSI (0.04–0.25), R (3.10^–11–1.10^–12) and pH (9–10.3) and are dominated by multi-shaped calcite (C4-C5), clotted micrite (C5), and blocky calcite (C5).

The juxtaposition of the hydrochemistry and distribution of microfacies shows that shrubby and dendritic crystals develop with high kinetic forcing, blocky calcite at low kinetic forcing and classic “microbialite” rim and clotted micrite forming at moderate-low forcing. We therefore conclude that formation of these microfacies in this system exhibits a kinetic rather than thermodynamic control. This is highly comparable with findings from materials chemistry summarized by Meldrum and Cölfen (2008). Kinetic control of crystallization arises from enhanced nucleation onto kinetically favored crystal faces and step-edges Ostwald’s step rule. Skeletal biomineral morphologies will self-assemble in such a setting due to Berg effect, where supersaturation over a flat face is low at the center and high near the edges, causing first “hopper like” then true single-crystalline ordered dendrites, disordered polycrystalline side branches, and finally disordered polycrystalline dendrites and dense branching morphologies as the kinetic forcing becomes more extreme. Sunagawa (1999) shows that it is a general rule that growth rate anisotropy will determine the forms of polyhedral crystal, as seen in the development of snow crystals (Libbrecht, 2012), where solid hexagonal plates crystallize at low supersaturation, whereas increasing the supersaturation leads to dendritic growth and flower-like morphologies with six petals. Beck and Andreassen (2012) report similar kinetic control on vaterite formation, where low levels of supersaturation lead to the growth of hexagonal, plate-like crystals while increasing kinetic forcing force causes a shift of the particle growth mechanisms toward dendritic growth.

Where kinetic forcing is reduced (lower R), diffusive controls on crystal formation re-assert themselves and non-dendritic fabrics occur (blocky calcite which displays a polyhedral shape in a single crystal). This is as would be expected from fundamental crystal assembly controls at relatively low rates of supersaturations (Meldrum and Cölfen, 2008). Curiously, we find the minimum of biofilm and microbe presence within these fabrics, indicating a process of formation dominated by inorganic mineralization, as defined by Dupraz et al. (2009). Classic “microbialite” fabrics (microbial rim and clotted micrite) occur within the middle and low saturation regime. Diatom and cyanobacterial biofilms are preserved directly as thin brown bands recognized on the thin-section material (Figure 6), and indirectly within the biologically influenced microfacies where the carbonate precipitation is indirectly modified by a precursor organomineral (Dupraz et al., 2009). It is also important to note that the diatoms reported generally develop in environmental conditions of freshwaters and represent biotopes of low to (at most) moderate electrolyte content (Sterrenburg, 1995; Van de Vijver et al., 2013). There were found living epilithically in several smaller rivers and brooks with an almost slightly alkaline pH (7.5) and low specific conductance levels (75–100 μS.cm–1) (Sterrenburg, 1995; Van de Vijver et al., 2013). In our study, it is therefore surprising to find this abundance of diatoms in such very salty and alkaline conditions.

Cluster-shaped carbonate was developed among bacterial filaments with an organic film forming sticky thin black layer around crystals (Figures 11A–D) which could correspond to EPS. These organic components associated with the micritic matrix of this facies along with its chemical characteristics (high alkalinity, medium pH and SI) suggest that they come from biologically influenced mineralization as well. The distribution of bio-influenced facies is not ordered with respect to the single-crystal inorganic precipitates, but rather both types of precipitation occur apparently randomly.

Kinetic forcing of dendrite growth can be enhanced by “poisoning” of some step-edges by additive molecules, particularly polymerized, polar organic molecules (Meldrum and Cölfen, 2008). We noted that biofilms were common even in proximal sites within Howden and Dene Burns during sampling, and both diatom frustules and organic carbon is pervasive through the majority of samples. Consequently, it is possible that organic additives are promoting the formation of extreme skeletal crystals in these systems at relatively low R. Whether skeletal crystals would form spontaneously in the absence of biofilm EPS was not possible to test during this study. However, we were able to demonstrate that within shrubby fabrics, no indication of microbial biofilm presence (i.e., classic microbialite features) could be identified, even when diatoms were touching the carbonate: it seems that if kinetic forcing is permitted by local hydrochemistry, all the carbonate product generated at that site will comprise skeletal composite crystals. Consequently, the absence of microbial textures does not appear to be a criterion for assuming no microbial influence on precipitation.

The co-existence of different fabrics at a same location (e.g., at C1, C2, C5, and C6 sites) indicates that the rather “static” analysis we present above is oversimplified, and that some spatial or temporal structure in the environment is present to cause physical association of the mineral products. This will be at least in part due to a variable chemistry caused by weather fluctuations (rainy, dry, low or high temperatures according to season) (Riley and Mayes, 2015) (Figure 15A). Mayes et al. (2008) showed that the hydrochemical system in Howden Burn and Dene Burn is relatively stable with a very elevated pH (pH = 11–12) over the 36-year data series but when heavy rainfall occurs, the river does adapt very rapidly. Spate flow changes the hydraulics of the flow altering gas exchange, (Gomes et al., 2017), reduces carbonate and calcium concentration and pH, and therefore suppressed CSI and R (Riley and Mayes, 2015). As reduced R will suppress kinetically forced fabrics and promote diffusion-forced crystal formation, this process alone will explain co-occurrence at any given site. For example, in the case of C6, short phases of reduced R and CSI may produce brief hiatuses in growth, forcing creation of concentric rings of dendritic fabrics (Figure 6). In addition to this abiotic forcing, it is likely that the distribution of biofilms is not spatially smooth, and patchy bio-influence will result in patchy occurrence of bio-influenced microfabrics. The apparently random co-occurrence of diffusion-forced and bio-influenced fabrics in Dene Burn likely reflects this spatial complexity within the sites microbiology.

Although we acknowledge the relatively small sample size in this case and the need for further investigation of such systems, given the hydrogeochemical and facies variation at the site a depositional model can be forwarded. This model is based on the key interpretation that lateral distribution of microfacies at the site relates to geochemical forcing (Figures 15A,B). Shrub and dendrite microfacies disappear with distance from the source, where CSI and R are excessively high (Figure 15B). In the most distal part of the river, microbial rim are quite continuous, occurring in most sites (Figure 15B). Cluster-shaped carbonate is the only facies represented upstream (C2) and sparry carbonate crust co-appear in C1 and C2 locations (Figure 15B). Blocky calcite and multi-shaped calcite are present in the downstream part of Dene Burn (C4 and C5) (Figure 15B). Clotted micrite is exclusively represented in both proximal and distal zones of the river (C1, C5, C6) (Figure 15B).

Although this is the first systematic analysis of the sedimentology of an Anthropocene limestone, the Consett site is not unique, and similar hyperalkaline sites occur in most post-industrial landscapes. These extremely alkaline environments have been found for instance in man-made river bed after repository building measures in Austria (Boch et al., 2015), slag-fill aquifers in the Lake Calumet region of Chicago (Roadcap et al., 2005, 2006), and Coatham Marsh steel slag leachate discharges in United Kingdom (Mayes et al., 2008). Examples such as non-ferrous metal industry waste disposal sites in Slovakia (Pristas et al., 2015), and technosol contaminated by former mining activity in Slovak Republic (Šimonovièová et al., 2017) can be cited as well. Understanding their operation in terms of mass deposition is a key part of fully constraining their behavior, and thus being able to manage them successfully. In particular, our finding that precipitation in proximal parts of the system are not diffusion-controlled is likely to result in deviation of element distribution coefficients from empirical values, and higher incorporation of trace metals and metalloids is anticipated in these sites. We recommend that further sedimentological investigation of these “accidental laboratories” is likely to yield further insights, of use both to geoscientists and environmental scientists working with very different scientific motivations.

A striking feature of recent saline alkaline lakes in rift tectonic settings (such as the East African Rift System, EARS) is the strong dissimilarity between alkaline metal concentration (calcium and magnesium) and the alkalinity of the waters (measured as carbonate and bicarbonate contents). During the evolution of these saline lakes, they tend to become Ca and Mg-poor systems due to the combined effects of the evaporative concentration and the relative mineral solubility which triggers the sequential precipitation of minerals from carbonates to silicates to complex salts (Eugster and Hardie, 1978; Deocampo and Renaut, 2016).

Many recent EARS lakes (e.g., Lake Magadi or Lake Bogoria; Jones et al., 1977; Renaut et al., 2002) have been considered tectonically and hydrologically similar to the ancient Pre-Salt lakes. However, they do not show clear evidence of substantial accumulation of spherulitic carbonate sediments, questioning whether EARS usual Ca and Mg-poor chemistries are encouraging spherulitic carbonate precipitation (Rogerson et al., 2017).

Furthermore, the record of anthropogenically mediated carbonate deposits such as those described here or in an asbestos open pit pond in Yukon, Canada (Power et al., 2011) is increasingly demonstrating that spherulitic carbonate growth can easily arise from waters anomalously enriched in calcium and magnesium, apart from elevated alkalinities and pH. In these sites, the alkalinity engine is thought to be related to the progressive atmospheric CO₂ ingassing (Rogerson et al., 2017) or the dissolution of the carbonate bedrock (Power et al., 2011) rather than increasing evaporation as occurs in alkaline lakes (Eugster and Hardie, 1978).

In addition, both natural alkaline, saline lakes and industrial hyperalkaline ponds are known to precipitate primary aragonite and/or low-Mg calcite minerals. This is mainly driven by the Mg/Ca ratios and the supersaturation state of the precipitating waters (De Choudens-Sánchez and González, 2009; Wang et al., 2012). An interesting observation from the Pre-Salt Aptian carbonate record is the repeated occurrence of Mg-rich and Sr-rich calcites (Saller et al., 2016) which are thought to be primary fabrics due to the lack of an original aragonite precursor. Also, shrubs described in the Kwanza upper case B were interpreted as originally precipitated as Mg-rich calcites and lately transformed to low-Mg calcites (Saller et al., 2016). These features strongly suggest that the Pre-Salt lake waters were likely much richer in Mg and sustained higher supersaturation states than those recorded in both the recent EARS and man-made alkaline counterparts. Thus, a major future challenge is to constrain the hydrological conditions enabling natural alkaline, saline waterbodies to chemically behave, at least in part, as anthropogenic spherulite-producing sites (Mercedes-Martín et al., 2019).