Doyle, Szojda, Onwukwe and Anguilano (Brunel University of London and the Silesian University of Technology) report in Discover Sustainability (2026, Article in Press, open access, CC BY) the first comparative screen of phytomining potential on legacy coal colliery spoil heaps — one at the former Barnsley Main Colliery in South Yorkshire, UK, the other at the former KWK Dębieńsko spoil complex at Czerwionka-Leszczyny in Upper Silesia, Poland. The premise is attractive: thousands of hectares of coal spoil in Europe alone, spontaneously colonised by vegetation that is already tolerating the substrate, with Zn and Cu as candidate recoverables. The work was funded under the EU’s MSCA-RISE staff-exchange scheme (project CSTO2NE, Grant Agreement 101086302) — one small sign of where Horizon-era resource-recovery money is flowing.
What was actually measured
Naturally established plants were sampled on the spoil, above-ground material was washed with deionised water, oven-dried at 65 °C for 24 h, ground, and analysed by energy-dispersive XRF (Rigaku NEX DE; 60 kV, 12 W, 100 s count) — a bulk, non-destructive technique that quantifies Ti, Mn, Ni, Cu, Zn, Al and Fe without digestion. Barnsley was sampled repeatedly from June 2023 to July 2024 across a small set of dominant taxa (Jacobaea vulgaris, Leucanthemum vulgare, Agrostis stolonifera, Zea mays); Dębieńsko was sampled once, in April 2024, across a far more diverse assemblage of twenty taxa from Dryopteris filix-mas to Pinus sylvestris. Crucially, neither spoil substrate nor roots were sampled.
The numbers, held to the threshold
At Barnsley the maxima were 903 mg Zn kg⁻¹ in J. vulgaris (November 2023) and 497 mg Zn kg⁻¹ in L. vulgare (March 2024); copper peaked at 97 mg kg⁻¹. At Dębieńsko the single-pass screen found 314 mg Zn kg⁻¹ in Pimpinella anisum, 72 mg Mn kg⁻¹ in Reynoutria japonica, 2,343 mg Fe kg⁻¹ in Lactuca sativa, and 17,465 mg Al kg⁻¹ with 181 mg Ti kg⁻¹ in Tussilago farfara. For context against the current screening thresholds (van der Ent et al. 2013; see our zinc and copper hubs): Zn hyperaccumulation starts at 3,000 mg kg⁻¹ in shoot dry matter — the best value here is 30% of the bar, and the comparison benchmark, Noccaea caerulescens, routinely exceeds it severalfold. Copper’s threshold of 300 mg kg⁻¹ was never approached. The authors are explicit that these are accumulation results, not hyperaccumulation: with no substrate or root data, no bioaccumulation or translocation factor can be computed, so plant-to-soil enrichment is simply unknown. One internal blemish to note: the running text cites L. vulgare at 113 mg Zn kg⁻¹ in November 2023, while Table 2 prints 133 — the table governs.
The mechanism lesson: EDXRF cannot see where the metal sits
The most instructive values in the paper are not the Zn figures but the aluminium and iron. Even after careful deionised-water washing, J. vulgaris from March 2024 carried 38,529 mg Al kg⁻¹ and 11,960 mg Fe kg⁻¹, and L. vulgare 21,646 and 2,322. Because EDXRF measures the total elemental content of the analysed surface, it cannot distinguish metal that has been taken up, translocated and sequestered by the plant from fine spoil particles adhering to or embedded in the foliage — and on heterogeneous mine spoil, particulate contamination is the default hypothesis for lithogenic elements at these levels. The Zn and Cu numbers therefore sit under the same cloud: they are upper bounds on physiological accumulation, not measurements of it. This is precisely the failure mode that inflated many historical copper hyperaccumulation records — our copper hub notes how dust-contaminated samples had to be critically reassessed — and it is why pairing foliar screens with substrate chemistry and a washing validation is non-negotiable before any agromining claim.
What it means for phytomining practice
The honest read is a screening result with a clear operational split. Barnsley-type sites — few dominant, metal-tolerant colonists and the higher maxima — are the better candidates for targeted cultivation: choose the best accumulator, manage biomass, and pursue metal yield per hectare, where concentration × harvestable biomass, not concentration alone, decides economics. Dębieńsko-type sites, with diverse, moderately accumulating communities, fit integrated phytostabilisation with gradual extraction rather than cropping. The real opportunity the paper documents is methodological: EDXRF is cheap and fast enough to triage hundreds of spoil-heap populations before anyone commits to agronomy, and the authors’ own prescription — same species at both sites, paired substrate and root sampling, seasonal replication, biomass measurement — is a modest, fundable next step. The genuine risk is the mirror image: cheap screens published without those pairings read as stronger evidence than they are, and can steer cultivation investment toward phantoms.
Limits a reader should keep attached
This is an Article in Press, shared early and subject to further edits. The two sites share no sampled species, and Dębieńsko has a single sampling date, so the site contrast is descriptive, not controlled — the climate comparison (Table 4) is context, not mechanism. The paper’s own conclusion that cross-site accumulation “confirm[s] bioavailability” overreaches its limitations sections, which state bioavailability cannot be established without substrate measurements; the cautious framing is the defensible one. A handful of Dębieńsko’s “spontaneous” records — Lactuca sativa, Pimpinella anisum, Daucus carota — are garden plants the paper does not explain. And despite the candidate-pool tag, this is not a nickel story: Ni stayed low (max 75 mg kg⁻¹), consistent with coal spoil’s Zn–Cu rather than ultramafic-style Ni–Co geochemistry. Phytomining on coal spoil, on today’s evidence, is a biomass-and-metal-price calculation awaiting paired soil–plant data — not a new hyperaccumulator play.
Source: open-access full text (CC BY) read in full via the Brunel University Research Archive (BURA) repository PDF of the Article in Press; every load-bearing number verified verbatim by three independent vendor reviewers (Claude, GPT, Kimi), each re-retrieving the primary source independently, with all three flagging the same two paper-internal issues (the 113/133 typo and the conclusion’s bioavailability overstatement); retrieval and cross-check notes per /methodology/.