The mine-water signal: Nuasjarvi

Question 1: when

Saltiest under the ice, exported at the melt

The risk is not random, it has a seasonal shape that repeats. The lake bed is saltiest through the ice-covered months, when the deep layer accumulates the dense water with no turnover to mix it. The band below is the 25–75% range of bottom-water conductivity by calendar month, with the median through it.

Concentration is only half the story. What the river downstream actually carries is load, concentration times flow. Pairing the lake-outflow discharge with the sulphate shows the second driver: the largest tonnage leaves not under the ice but at the spring melt, when snowmelt flow flushes the stored winter load downstream. Illustrative downstream export: monthly mean outflow discharge x monthly median lake sulphate. Concentration is grab-sampled; treat the seasonal shape, not the exact tonnage.^[2,1]

Sixteen years of monthly profiles confirm the cycle repeats, with the peak height tracking how hard the mine is discharging.

Question 2: is it the mine

The mine, the road, or the rock

Conductivity says something is dissolved; it cannot say what. The ion ratios can. Road salt and a saline aquifer raise sodium and chloride; a sulphide mine raises sulphate and metals. Plotting sulphate against chloride separates them cleanly: the discharge-bay samples climb the sulphate axis while chloride stays put. This is the ore's salinity, not winter road maintenance.

The black shale under the lake leaches sulphate and metals naturally too, which is the honest confounder. What points to the mine is the combination and the magnitude, concentrated at the discharge bay and absent from the reference basins, not a lake-wide drift the bedrock could explain.

Is the catchment just naturally metal-rich?

The strongest version of the confounder is that Kainuu is black-shale country, so maybe the whole catchment's soil is metalliferous and the lake would run enriched with or without a mine. GTK's background geochemistry answers it directly. Across 365 till samples in the Talvivaara-Nuasjarvi catchment, the natural soil is ordinary Finnish till: nickel, cobalt, chromium, zinc, copper and vanadium all sit within about 10% of the national median, not multiples above it.^[18]

The till settles the solid phase; the water has its own answer. Reading SYKE's open record for the region's natural streams, away from the mine, gives the dissolved starting point in the same units as the lake. Across 343 Kainuu river stations, natural running water carries a median 3.23 mg/l of sulphate and 3.1 mS/m of conductivity^[19]. Two things follow. The stream conductivity, 3.1 mS/m, lands right on the lake's own pre-mine baseline of 3.15, an independent confirmation that the natural background here really is that low. And against a natural 3.23 mg/l, the discharge bay's 220 mg/l of sulphate is about 68× the regional stream water: the mine is not adding to a naturally briny system, it is the dominant dissolved-sulphate source in its own catchment.

Dissolved nickel is the honest exception. Across the 45 stream stations that measure it, the median is 3 µg/l but the spread is wide (1 to 12), because some Kainuu streams drain naturally mineralized black-shale terrain and run elevated in nickel with no mine at all^[19]. So the dissolved nickel background is not zero, which is exactly why the attribution leans on the bay-versus-reference magnitude and the pipe-era before-after step, not on a single concentration.

The verification gap

What the discharge adds, and why the headline hides it

Two readings of the same lake disagree, and the disagreement is the point. Compare the discharge bay with the rest of the lake today and everything is enriched: sulphate ×22.4, calcium and manganese several-fold, and the metals the ore predicts are not at background at all, nickel runs about ×13.1 the open lake, with cobalt and arsenic clearly raised. In absolute terms the bay nickel, about 17 µg/l, is roughly four times the EU environmental quality standard of 4 µg/l, the regulator's line, though that standard is set on the bioavailable fraction at a defined compliance point, not the deep bay^[4]. Read that as a screening signal, not a formal compliance finding.

Now the reading most monitoring reports lead with: the lake-wide surface record against its own pre-mine baseline. Here only the salt moved, sulphate, calcium, magnesium and manganese rose modestly, while nickel, zinc, cobalt and arsenic look flat or lower. Taken alone it reads as a clean bill of health.

Both readings are right

Both are correct. The dense plume sinks and is retained in the deep bay, so the open lake's surface barely registers it and the lake-wide median, dominated by clean reference stations, stays near background. The metals are not absent; they are concentrated where a coarse, surface, lake-wide program does not look. The treatment removes most of the load the raw ore would leach, but "controlled" is not "gone," and the gap between the reassuring average and the enriched bay is exactly the kind of reported-versus-real distance this site exists to measure.

The mechanism is chemical, and it is the reason the load is salt rather than metal. The discharge is neutralised, and the bay measures about pH 7^[1], so at near-neutral pH much of the nickel, cobalt and zinc leaves the dissolved phase, by precipitation, adsorption onto iron and manganese solids, and settling, while sulphate stays conservative and manganese largely soluble, both passing downstream. Iron follows the metals, not the salt: at the bay it sits near background (about ×1.1), already precipitated, and it is the solid that scavenges the nickel and cobalt out of solution. pH is the master variable: the same ore draining acid, with no neutralisation, would carry the metals too. So the residual signal here is salinity by design, and the metal control depends on holding the pH, which is why the chemistry, not just the concentration, is what to watch.

The black-shale marker metals make the same case from the other direction. Uranium and molybdenum are the trace elements a critic would expect a black-shale mine to leach, and in the lake they do not appear: uranium sits at its pre-mine background and molybdenum near its detection limit^[1]. Note where they are measured, though. The operator's compliance stations in the discharge bay log the acute-toxicity metals, nickel, cobalt, zinc, copper, arsenic, iron, but not uranium, molybdenum or manganese; those are read only at the routine lake network. So the very elements that fingerprint the ore are absent from the monitoring at the point of discharge, and the reassuring reading for them comes from the wider lake, not the plume.

Not a harmless residue

And the salinity is not a harmless residue. A peer-reviewed paleolimnological study of this lake found the discharge deepened the stratification, starved the near-bottom water of oxygen and degraded the benthic community by the pipe, even as the open-water communities changed little^[5]. Auditable is not the same as harmless: the open record measures both the metals held back and the salt's own ecological cost.

Where to look next

It also says where to look next. The metals that leave the water are in the solids, so the treatment sludge and the bay sediment, not the discharge, are the thing here that would repay an actual sample, for the residual cobalt, scandium and rare earths as much as for the uranium liability. The Finnish mining page carries that sampling read.

The divergence is recent and clean. Before the pipe, the discharge bay and the rest of the lake tracked each other. After November 2015 the bay's bottom water pulls away, the reference basins do not, and the gap is the mine's increment net of any regional or climate trend.

Question 3: does the lake confirm the report

The operator's numbers, against the water

Everything above reads the lake. The other half of a reported-versus-real check is the report itself. Terrafame discharges to Nuasjarvi under a sulphate quota: the Vaasa Administrative Court set the pipe limit at 15,000 t/yr in 2016, with a 2,000 t monthly cap in the open-water season, and the operative permit quota for sulphate in discharged water is now 15,780 t/yr^[7,12]. Its own disclosures put the discharge under that line every year ● measured: 12,424 t in 2024 and 10,088 t in 2025, and as low as 3,434 t in 2018, when bioleaching ran closed-cycle for months and little water left the site^[6,10,12,11]. The check below runs three routes against that report: the tonnage against its permit quota, its year-to-year shape against the measured downstream flux, and its implied mechanism against the spatial plume read above.

Three views of the same discharge now agree. The operator reports a permitted, sub-quota tonnage. The lake shows it as a sharp local plume, sulphate at the bay runs about ×22.4 the rest of the lake, while lake-wide the same sulphate is only about 1.5× the pre-mine baseline, diluted toward background away from the pipe. And the reported nickel discharge is tiny beside the sulphate, about 73 kg in 2018 against thousands of tonnes of sulphate^[6], the reported-side echo of the mechanism this page reads from the water: the salt passes, the metals are held. The operator's books, the bay chemistry and the lake-wide average are consistent readings of one managed discharge ◐ inferred.

Can open monitoring confirm the report year by year? Partly. Index each series to its own mean and the measured downstream sulphate flux at the outflow and the reported discharge fall to their shared minimum in 2018-2019: the closed-cycle dip is visible in the lake ◐ inferred. But the year-to-year correspondence is loose. The outflow integrates the whole Oulujoki catchment's natural sulphate and swings with flow, so the mine's smaller moves sit below that noise.

The annual series is coarse, and it hid the reported side's real shape. Terrafame's water-emissions reports tabulate the pipe load month by month, and rolled to quarters the sub-annual structure appears● measured: an open-water rhythm that lifts discharge through summer and autumn, and the 2018-2019 closed-cycle collapse to near zero, the same event the lake corroborated above. This is the reported driver at the resolution a severe test needs, and it fills the years the annual tables left as a bracket.

The cleaner way to read this is as a negative control. The annual outlet load fails as an estimator of the mine's contribution, because the discharge is embedded in a much larger, hydrologically variable, background-dominated catchment flux: the black schist of the Talvivaara area leaches sulphate on its own, and only a small part of the Nuasjarvi catchment drains the mine^[13]. Sparse grab sampling and flow-driven variance make a single annual load a poor attribution tool in any case^[14]. That failure is itself the result: it rules out a misleading discharged-versus-exported mass balance and forces verification onto triangulation, the reported load against the quota, the local bay and depth enrichment, the lake-wide dilution, and a before-after contrast against reference basins ◐ inferred. Open data verifies the big swing and the spatial plume; the fine-grained tonnage still rests on the operator's reporting.

What would close the attribution gap: the sulphur and oxygen isotopes of sulphate (δ34S and δ18O) fingerprint the source, separating mine sulphate from black-shale weathering and other inputs in a way a bulk load number cannot^[15]. That is the measured upgrade this read would need to move from corroboration to apportionment.

When the driver swings, the lake shouts

The upstream lakes, and a crisis read from open data

The subtlety above is a feature of a stable, well-run operation. Terrafame's discharge barely varies, so the lake barely moves, and the honest check is a pre-registered forward test. The opposite case is instructive. When the driver swings hard, the same open data reads the mine unambiguously. Through the 2010-2014 Talvivaara crisis, before the Nuasjarvi pipe, the mine's sulphate load and the November 2012 gypsum-pond leak discharged into the small near-field lakes to the south. Their sulphate ran from a few mg/l before mining to hundreds and thousands during the crisis, then fell back after the 2015 rerouting sent effluent north. The whole arc sits in the open record ● measured.

This is the method's reach and its limit on one mine. Where the operation is stable, the signal is subtle and the honest move is a severe forward test, the Nuasjarvi outflow above. Where the driver swung, the signal is overwhelming and open water data alone reconstructs the operational history. Logged as a backtest, a retrospective worked example, never scored in any tally ◐ inferred.

The method generalises

The same read on other ore types

The bridge from one mine to a method is the ore type. Each ore leaches a predictable signature, and the same open monitoring reads it on the receptor water of any Finnish mine. A black-shale and a gold mine and a chromite mine should look nothing alike, and they do not. Each tile is the receptor-water conductivity over time with the ore's signature element; Terrafame's figures here are receptor-wide and so far milder than the discharge bay shown above.

What moves what

The dependency edges, with their standing

Everything above reads as prose; this section is the same analysis as machine-checkable objects. A claim on this site states what is, and carries a measured, inferred or projected tier. A dependency edge is the layer above: it states what moves what, through which mechanism, with which sign and lag, and it carries a pre-registered falsifier. An edge starts as conjectured, earns corroborated only by surviving an out-of-sample test against named null models, and is marked refuted when it fails. Nothing may be cited as established while it sits under test. These are the three edges this page has earned so far, exactly as they sit in the registry, including the two where the honest verdict is that no severe test currently passes.

Method and limits

Conductivity does the detection because it integrates the whole dissolved load and is measured most often; the ion suite does the attribution; depth, season and discharge turn a number into a mechanism. Three honest limits. Conductivity is non-specific, so it needs the ions to tell a mine plume from road salt or a saline aquifer. The black shale under this lake leaches sulphate and metals on its own, so separating geogenic background from mining-induced load is the real scientific work and the credibility risk. And most stations are grab-sampled every few weeks, so this is seasonal and event risk-windowing, not a day-ahead forecast, except at the continuously logged stations a discharge permit forces. The load figure is illustrative, lake-outflow discharge times a grab-sampled concentration, so read its seasonal shape rather than its exact tonnage. The same method runs on any mine with a downstream monitoring record; Finland is where the open data is deepest, which is why it is the reference build.

How the numbers are built, so the reading can be checked. The discharge-bay set is five stations at and beside the pipe mouth, Jormaslahti 3 and Nuasjärvi 47 to 50; every other Nuasjärvi station is the lake reference. Each sampling visit is reduced to a surface value, the shallowest sample, and a bottom value, the deepest, and the medians are taken unweighted across visits and stations, not area- or volume-weighted. The windows are fixed and differ by panel: the pre-mine baseline runs to 2015 (conductivity back to 1962, each ion and metal from when it enters the record); recent conductivity is 2016 onward and the lake-wide chemistry 2023 onward; the bay-versus-reference comparison and pH use 2018 onward for wider coverage of the sparse metals; the ion fingerprint uses post-pipe samples from late 2015, and the upstream-downstream control is plotted from 2000 so the split after the pipe shows. Values are used as reported by VESLA with no below-detection substitution, so at the clean reference stations a few low-level metals sit near their detection limit and the bay-to-reference ratios should be read as indicative, not exact. Every figure is reproducible from the open VESLA and Hydrology APIs with the filters above. The reported-discharge figures are Terrafame's own, from its annual and sustainability reports; they are a site total that is almost entirely the Nuasjarvi pipe, and 2021 and 2023 are not separately tabulated. The downstream-flux series is a targeted outflow load (lake-outflow discharge times lake sulphate, summed by year) that carries the whole catchment's sulphate, so it is read as shape against the reported series, not as the mine's own tonnage.