The Augusta–Priolo–Melilli petrochemical district in eastern Sicily (37°13′49″N, 15°13′10″E) is one of Italy’s most contaminated industrial corridors — a designated Site of National Interest where refineries have left soils laced with Pb, Cd, Hg and As alongside PAHs and dioxins, under a Mediterranean climate with only 500–600 mm of rain a year. Into that setting, Di Agosto, Armaro and colleagues project an integrated agrivoltaic-phytoremediation system: solar panels mounted 4–4.5 m above crops that clean the soil while the land also generates electricity (Di Agosto et al. 2026, Environmental Science and Pollution Research, CC BY). The study is a framework plus simulation — PVsyst energy modelling and a techno-economic assessment, calibrated with literature-derived crop responses, not a field trial — and that distinction matters for how far its numbers travel.
Three tolerant workhorses, three panel pairings
The species shortlist is telling for anyone who follows this site: Cannabis sativa (hemp, C3), Chrysopogon zizanioides (vetiver, C4) and Arundo donax (giant reed, C3). These are high-biomass, contamination-tolerant crops — excluder and stabilizer physiologies chosen for biomass productivity, not hyperaccumulator shoots. The assumed baseline yields are 75 t ha⁻¹ yr⁻¹ for vetiver, 35 for giant reed and 20 for hemp. Each crop is matched to a panel architecture that fits its light requirement: giant reed under tilted monofacial modules (highest electricity, 568,660 kWh yr⁻¹ per simulated hectare), hemp under semi-transparent modules (285,638 kWh yr⁻¹), and the shade-sensitive C4 vetiver under vertically mounted bifacial modules spaced 10 m apart, whose more even light distribution limits its interception penalty (297,987 kWh yr⁻¹; 1,008 panels of 440 Wp in 11 rows).
Contamination and shade draw on the same biomass account
The paper’s central quantitative result is that the two stresses act on the same variable — biomass, which is also what phytoextraction throughput is made of. From literature pot studies the authors assume stress penalties of up to −50% for hemp at 1,500 mg kg⁻¹ Ni (the threshold that also bites hardest in the /metals/nickel/ context), −40% for giant reed above 600 mg kg⁻¹ Cr, and −50% for vetiver at critical Zn, Pb and Cd levels. Panel shade is budgeted the same way: PAR reductions of 25–30% (monofacial), 20–25% (semi-transparent) and 10–15% (vertical bifacial) are “conservatively translated” into biomass losses of 20–25%, 25–30% and 15–20% respectively — note the deliberate swap, the C4 grass under the most even light loses least. In the scenario arithmetic this lands vetiver at 37.5 t ha⁻¹ yr⁻¹ and a net €3,681 ha⁻¹ yr⁻¹ under severe contamination, versus 60.0 t ha⁻¹ and €6,730 under shading; hemp sinks to 10.0 t and €995 when Ni stress is worst. Non-obvious consequence: on a contaminated site, every percentage point of panel shade is a percentage point of metal removal foregone — the light budget is the remediation budget, and the vetiver–vertical-bifacial pairing wins precisely because it spends least of it.
The ash clause — liability in the model, feedstock in agromining
Here is the piece a phytomining reader should sit up for. The harvested biomass is valorised through gasification (≈60% efficiency, syngas at 5–6 MJ Nm⁻³, ≈30% electrical conversion, sold at an average European ≈€0.215 kWh⁻¹), giving reported revenues of €8,762 ha⁻¹ yr⁻¹ for unstressed vetiver down to €995 for stressed hemp. But the model then charges €100–200 per tonne for disposal of contaminated ash, at an ash yield of 5–8% of dry mass — enough, with harvest and transport costs of €20–30, €5–15 and €10–15 per tonne plus €100–150 per tonne for outsourced pruning, to cut net revenues by 30–50%. Read that against the chemistry: gasification drives off carbon and leaves the non-combustible fraction behind, so by the paper’s own ash-yield range the potentially toxic elements are concentrated roughly 12–20-fold into 5–8% of the original mass. The study prices that concentrate exclusively as a landfill liability and never asks what it contains — there is no line item, anywhere in the analysis, for recovering Ni, Zn, Cd or Pb from the ash. That is precisely the inversion agromining proposes: the same thermochemical step that makes biomass an energy feedstock makes it a metal-enriched one, and at 12–20× concentration the ash of a remediation crop is a low-grade ore whose value the model sets to zero. For a site whose contamination is nickel-bearing industrial soil, the difference between “€100–200 per tonne to bury” and “recoverable metal product” is the difference between phytoremediation as a subsidised cleanup and phytomining as a revenue line — this study, like most of the remediation economics literature, does not bridge it.
Honest bounds
The numbers are scenario assumptions, not measurements: stress responses are extrapolated from pot studies (Testa et al. 2023 for hemp; Barbosa et al. 2015 for giant reed; Ng et al. 2020 for vetiver), and the shade response comes from a C3/C4 crop meta-analysis that remains an AgriRxiv preprint (Laub et al. 2021), not peer-reviewed literature. No hyperaccumulator entered the candidate pool, so the framework optimises energy-and-biomass economics while shoot concentrations — and therefore metal-removal rates per hectare — stay capped by excluder-type physiology. The authors themselves flag giant reed’s invasiveness as a deployment constraint and call explicitly for field validation. And a careful reader should note the paper’s gross-versus-net revenue labels do not reconcile cleanly with its own Table 7 — the scenario figures quoted above are the conclusion’s “net revenue” values, which three independent checks confirmed verbatim.
Source: Di Agosto, Armaro, Di Novo, Salmeri, Randazzo & Sergi (2026), Environmental Science and Pollution Research, DOI 10.1007/s11356-026-38149-1 (CC BY, published online 2026-09-02). Every load-bearing number was verified verbatim against the full text by three independent reviewers; one disagreement (the panel-row count and the A. donax invasiveness caveat, both confirmed against the retrieved full text) is noted above. See /methodology/ for how we source and check analysis pieces.