Chromium and nickel in African reservoir sediments trace geology, not industry

Sediments at the bottom of Nandoni Reservoir, a drinking water supply impoundment on the Luvuvhu River in Limpopo Province, South Africa, carry chromium and nickel concentrations high enough to trip consensus sediment quality benchmarks at more than half of the sampled sites. Yet the same sediments show almost no enrichment of the elements that normally betray industrial pollution, and their concentrations rise and fall in lockstep with aluminium, the backbone of the natural detrital mineral matrix. A new open access study published in Environmental Geochemistry and Health argues that this combination is not a paradox but a lesson: contamination indices, ecological screening benchmarks and human exposure models answer three different questions, and collapsing them into a single verdict can send water managers chasing pollution sources that geology supplies for free.

Reservoirs are natural sediment traps. Dam construction interrupts the downstream transport of riverborne particles, creating quiet depositional zones where fine grains and the potentially toxic elements bound to them accumulate over years. These deposits record what the catchment delivers, but they can also become secondary sources when water levels drop, exposed margins erode or dredged sediment is resuspended. The conventional assessment toolkit combines three analytically distinct components. Contamination indices, including the contamination factor, the geo accumulation index, the pollution load index and the potential ecological risk index, measure how far measured concentrations depart from an assumed geochemical background. Sediment quality guidelines, such as the consensus threshold and probable effect concentrations derived by matching paired datasets of sediment chemistry and biological response, estimate the probability of adverse effects. Exposure models estimate risk under specified assumptions about contact, uptake and toxicity. Treating these outputs as interchangeable produces two opposite errors: a sediment whose composition is set by parent rock may be flagged for source control that cannot reduce a concentration the geology supplies, or a geogenic element may be dismissed as harmless even though it exceeds an effects based benchmark and reaches a nearby receptor.

The research team, led by Tatenda Dalu with colleagues from several South African institutions, sampled surface sediment from zero to five centimetres depth at 21 sites across the reservoir in October 2023, at the end of the dry season and before the main summer inflow. Sites were allocated by stratified purposive design across riverine, transitional and lacustrine zones, spanning water depths from 1.8 to 22.3 metres and both shorelines. An Ekman grab was chosen because its lightweight design closes with minimal bow wave, preserving the flocculent surface layer that carries most of the trace element load. At each site, three grab deployments were combined into a single composite sample. Twenty six elements were quantified by inductively coupled plasma mass spectrometry following US EPA Method 6020B at an accredited laboratory, with certified reference material recoveries for the priority elements between 81.3 and 111.4 percent and duplicate agreement within six percent. The digestion recovers pseudo total rather than bioavailable concentrations, a constraint that shapes every downstream comparison.

The source partitioning results point strongly toward matrix control. Spearman rank correlations with aluminium reached 0.98 for lead and copper, 0.95 for nickel and zinc, 0.94 for arsenic and cobalt, 0.90 for cadmium and 0.87 for chromium, with coefficients of variation of the element to aluminium ratios at 0.36 or below for all four priority metals. In plain terms, chromium, nickel, copper and lead maintained nearly constant proportions to the detrital aluminosilicate fraction across the whole basin. Aluminium normalised enrichment factors relative to average shale placed lead, zinc, cadmium, arsenic and mercury at or below crustal proportions, with geo accumulation index values for lead, zinc, arsenic and cadmium below zero at every site. Chromium ranged from 2.2 to 6.1 and copper from 2.1 to 3.9, values as compatible with mafic parent material as with moderate anthropogenic addition. The authors are explicit that these diagnostics describe matrix association and cannot by themselves establish source dominance; a diffuse anthropogenic input transported as fine particles could remain correlated with aluminium. What the data show is the absence of a large, independently varying anthropogenic signal, not proof that no anthropogenic input exists.

Ecological screening told a different story. Chromium exceeded the consensus threshold effect concentration at all 21 sites and the probable effect concentration at 14 sites, while nickel exceeded the threshold at 20 sites and the probable effect concentration at 13. Copper exceeded the threshold at 15 sites but the probable effect concentration at only two, and arsenic, cadmium, lead, zinc and quantified mercury stayed below their thresholds everywhere. The mean probable effect concentration quotient ranged from 0.18 to 0.85, with a mean of 0.52, exceeding 0.5 at 11 of 21 sites. Water depth was positively associated with aluminium, chromium, nickel, copper, lead and the quotient itself, consistent with a depositional gradient in which deeper, lower energy settings accumulate finer, matrix rich particles. The authors stress that guideline exceedance is neither evidence of contamination nor of demonstrated toxicity, because the consensus benchmarks are indifferent to element provenance and cannot resolve local speciation, grain size, organic carbon or sulphide binding. What it establishes is a screening level trigger for site specific verification.

Human health screening added a third, receptor dependent interpretation. For an adult maintenance worker assumed to handle dredged sediment during intake maintenance, with an ingestion rate of 100 milligrams per day over a 35 year career, the deterministic hazard index ranged from 0.073 to 0.335 and stayed below one at every site. Ingestion accounted for 86.2 percent of the summed index. For a child assumed to have incidental contact with exposed shoreline sediment during low water periods, with an ingestion rate of 200 milligrams per day against a body mass of only 15 kilograms, the hazard index ranged from 0.360 to 1.665 and exceeded one at nine of 21 sites, with ingestion contributing 93.7 percent. Cobalt, chromium, manganese and vanadium were the largest contributors to the child index. The contrast between receptors is a property of the exposure model, particularly the ratio of ingestion rate to body mass, rather than a measure of differences in sediment toxicity. Summed incremental lifetime cancer risk fell within the commonly used 10 to the minus 6 to 10 to the minus 4 risk management range, with chromium and nickel contributing roughly 95 percent of the total in both scenarios.

Chromium speciation emerged as the single largest source of uncertainty in the health screening. Because the analytical method determines total chromium without resolving oxidation state, the primary calculation conservatively treated all measured chromium as hexavalent chromium, the form that maximises both the hazard quotient and cancer risk. A parallel sensitivity calculation treating all chromium as trivalent, using an oral reference dose of 1.5 milligrams per kilogram per day and applying no slope factor, reduced the worker hazard index to 0.048 to 0.251 and the child hazard index to 0.253 to 1.301, cutting the number of sites with a child hazard index above one from nine to three. Cancer risk estimates fell by approximately 47 percent. Reduced, organic rich profundal sediment derived from mafic parent material would be expected to favour the trivalent form, so the true value probably lies nearer the lower bracket, but the authors caution that this cannot be asserted without speciation data. The two calculations bracket the plausible range rather than compete.

The most striking demonstration of methodological fragility came from the background sensitivity analysis. When average shale was replaced with a lower pre industrial reference set, the pollution load index inflated by a factor of 4.04 at every site without exception. The authors show analytically that this inflation factor is the geometric mean of the element wise background ratios and is therefore identical everywhere and completely independent of the measured concentrations; the apparent fourfold increase carries no information about Nandoni sediment at all. The potential ecological risk index rose from a mean of 27.6, classifying all sites as low risk, to a mean of 237.5, classifying eight sites as moderate and eight as high risk. A decomposition revealed that nickel, copper and chromium produced 98.7 percent of the change, precisely the three elements whose assumed pre industrial backgrounds of five to six milligrams per kilogram are most discordant with the mafic and ultramafic influenced geology of the Soutpansberg catchment, where parent material alone would supply background concentrations of several tens of milligrams per kilogram.

The study has clear limits. Sampling occurred on a single occasion at the end of one dry season, with one composite per site, so seasonal variability and within site heterogeneity remain uncharacterised, and very shallow or obstructed marginal areas are under represented. Grain size, total organic carbon, acid volatile sulphide, pore water chemistry, mineralogy and iron and manganese oxide content were not measured, precluding multivariable modelling, and no locally derived geochemical baseline exists for the catchment, leaving both reference sets as external comparators. The authors recommend four follow up measurements: a local geochemical baseline built from catchment soils, pre impoundment or deep core sediment and screened reference sites; chromium speciation together with operationally defined bioavailability; routine grain size and organic carbon measurement; and biological validation through paired benthic macroinvertebrate sampling and whole sediment toxicity testing across the observed quotient range. Their central conclusion stands regardless: geogenic elevation does not render a potential hazard irrelevant, and generic background indices should not be used alone to trigger source control decisions in geologically heterogeneous catchments, particularly where a defensible local baseline is unavailable.

Subject of Research: Lithogenic chromium and nickel drive sediment screening exceedances in a South African water supply reservoir

Article Title: Predominantly lithogenic chromium and nickel elevate sediment guideline quotients and exposure screening in a subtropical African water supply reservoir

Article References: Dalu, T., Dondofema, F., Keates, C., & Munyai, L. F. (2026). Predominantly lithogenic chromium and nickel elevate sediment guideline quotients and exposure screening in a subtropical African water supply reservoir. Environmental Geochemistry and Health, 48(16), Article 665. https://doi.org/10.1007/s10653-026-03558-1

Image Credits: AI Generated

DOI: 10.1007/s10653-026-03558-1

Keywords: Nandoni Reservoir, sediment quality guidelines, chromium, nickel, aluminium normalisation, enrichment factor, pollution load index, probable effect concentration quotient, human health risk assessment, hazard index, geochemical background, South Africa