Aminatun, Idrus, Rakhmawati, Atun, Restele and Kwarta — Universitas Negeri Yogyakarta and Universitas Gadjah Mada with PT Bumi Suksesindo — publish today in the Journal of Water and Land Development (16 October 2026, open access, CC BY-NC-ND) an on-site gold-phytomining experiment run in demonstration plots on the heap-leach spent ore of the Tujuh Bukit gold mine, East Java. It is, to my knowledge, the first field trial of gold phytomining that combines rhizobacterial inoculation with compost amendment at a working mine — and it is a useful lesson in how a concentration ratio can be made to say much more than the underlying measurements.
The setup is genuinely field-scale, and short
The spent ore itself is the interesting character here. The composite sample contained 0.33 ppm Au alongside 1,425.99 ppm As, 422.41 ppm Pb, 321.31 ppm Cu, 10.28 ppm Ag, 4.76 ppm Mo and 0.005 ppm Hg — a classic high-sulphidation pathfinder suite. Into thirty 1 × 1 m plots went three plants each of the Chinese brake fern Pteris vittata, a documented arsenic hyperaccumulator already colonizing the heap-leach area; the ferns were harvested just 1.5 months after planting. Ten treatments crossed three bacterial isolates from their own earlier work (Bacillus cereus RTKS, Pseudomonas aeruginosa RTKP1, Stenotrophomonas geniculata RTKP2) with compost at 1:1, 1:2 or none.
The headline numbers — and the control they ride on
The best treatment, B. cereus RTKS plus compost at 1:1 (PBK1), produced the paper’s headline: post-treatment ore Au of 0.220 ppm, shoot Au of 2.793 ppm, and a shoot BCF of 19.11, root BCF of 14.73 and TF of 2.98 — all the highest in the table. But look at the untreated control before applauding. Control shoots held 2.535 ppm Au — about 10% below PBK1 — while control roots held 1.820 ppm, comfortably above PBK1’s 1.116 ppm. The ANOVA tells the same story from another angle: only shoot BCF differed significantly among treatments (p = 0.008); TF (p = 0.554) and root BCF (p = 0.390) did not. The claimed translocation advantage is statistically unsupported; what differs is the denominator.
That denominator matters because the BCF here is computed against post-treatment ore concentration. Any treatment that leaves less Au in the ore mechanically inflates its own BCF, and ore Au at these grades is brutally heterogeneous: PBK1’s three replicate ore values spanned 0.04–0.41 ppm (SD 0.19 against a mean of 0.22), and its shoots spanned 0.86–6.62 ppm (SD 3.31 against a mean of 2.79 — a coefficient of variation near 118%). Averaging per-plot ratios against that 0.04-ppm outlier is how you get 19.11; dividing the table means instead gives 12.7. The paper itself concedes the substrate problem in a single sentence — non-uniform Au distribution, citing a kriging study — but the mass-balance problem is worse than a caveat: the control’s post-treatment ore Au (0.640 ppm) is roughly twice the initial composite (0.33 ppm), and eight of ten treatments likewise “gained” ore gold they were never given. No ore mass, bulk density or plant biomass is reported anywhere, so no balance can be closed and no grams of recovered gold can be stated — the paper never claims any, and neither should anyone citing it.
What the fern is actually harvesting: arsenic
Here is the non-obvious implication. The same Table 1 that reports the gold shows P. vittata shoots carrying 1,047.51–1,940.32 ppm As across treatments — against 1,426 ppm in the ore — a consistent, documented expression of the fern’s arsenic hyperaccumulation physiology (the authors attribute its tolerance to the fern’s arsenic transporter). At this particular site, the honest reading of the experiment is not “gold phytomining” but arsenic phytoextraction with a gold footnote: the fern removed meaningful As from an As-rich waste pile while accumulating gold at ~2.8 mg kg⁻¹, roughly three orders of magnitude below the
1,000 mg kg⁻¹ shoot criterion the community uses for metals like Ni (the paper’s “BCF > 1 = hyperaccumulator” gloss, borrowed from a Pb classification scheme, should not be read as an established gold threshold). For the site’s owner the remediation value is real either way — see our Pteris vittata species page and the /arsenic hub for the fern’s track record on exactly this job.
The finding worth keeping: growth and accumulation are decoupled
Strip the BCF framing away and there is a sound, deployable result inside this paper. Treatments without compost grew best (control height fell 5.0 cm; PBK1 fell 9.32 cm with frond loss, plausibly from competition between inoculated bacteria and the compost’s native microbiota), yet the compost-plus-B. cereus combination moved the one metric that reached significance. B. cereus siderophore production is a plausible mobilization mechanism, and the paper’s own conclusion — that bacteria, compost and plant performance “do not always have synergistic effects” — is the kind of negative, dose-dependent interaction that field phytomining programmes keep rediscovering the hard way. The agronomy is real; the gold arithmetic on top of it is not yet.
Risks before this goes anywhere
Three cautions for anyone tempted to cite the 19.11. First, the plants were collected from the contaminated heap-leach area itself and no baseline tissue analysis was done, so part of the shoot metal burden may predate the experiment — the authors note this for As and it applies to Au as well. Second, the study carries a structural disclosure tension: a co-author is a PT Bumi Suksesindo employee, all ore preparation ran in the company’s laboratory, the company is thanked as project partner — and the paper declares no conflict of interest. That doesn’t invalidate the data, but independent replication matters more than usual here. Third, at ~2.8 ppm in shoot tissue with no biomass yield, no mass balance and a 1.5-month cycle, no economic recovery figure can be honestly derived; gold at this scale in biomass is a detection problem, not a product.
Provenance: full open-access text read (publisher PDF, all tables extracted; CC BY-NC-ND). Every load-bearing number independently re-retrieved from the primary source by three separate reviewers on a shared claim template plus one independent source re-download, with zero cross-vendor discrepancies on any digit; paper-internal issues surfaced and reported above rather than smoothed over (denominator-driven BCF, irreproducible mean-of-ratios, the control’s ore Au exceeding the baseline, n=3 variance, absent biomass/mass balance, dry-vs-fresh weight basis unstated); methods and limits at /methodology/.