Soil Health, Not Single Species, Rewrites Metal Risk Maps in Tianjin

For decades, the standard way to judge whether a polluted soil is dangerous has been to ask how many individual species can tolerate the contaminant in question. Toxicologists compile laboratory thresholds for earthworms, plants, and other soil-dwelling organisms, fit a cumulative curve to the values, and read off a hazardous concentration that protects some chosen percentage of species. This tool, the species sensitivity distribution, or SSD, underpins environmental quality criteria worldwide. But a new study in iScience argues that the approach may systematically overstate the risk that metals pose to real soils, because it treats every species as an isolated test subject rather than as part of a functioning ecosystem.

Researchers led by Qiuyun Xu and Yajuan Shi of the Research Center for Eco-Environmental Sciences developed an alternative framework they call the soil health sensitivity distribution, or SHSD. Instead of asking how many species are harmed by a given concentration of lead, copper, or cadmium, the SHSD asks how many soil functions are impaired. Soil health, in this framing, is the sustained capacity of an ecosystem to function as a vital living system supporting plants, animals, and humans. The team built standardized soil microcosms containing wheat (Triticum aestivum), the compost earthworm Eisenia fetida, and the indigenous microbial community of Tianjin field soil, then exposed them to graded concentrations of the three metals for 30 days.

The choice of organisms was not arbitrary. The researchers compiled functional trait profiles for candidate plant and invertebrate species and used principal coordinate analysis together with Bray-Curtis and Gower dissimilarity metrics to identify the taxa closest to the centroid of each group’s trait space. Wheat and Eisenia fetida emerged as the most functionally representative choices, meaning their trait profiles best stand in for the broader community of soil plants and animals. Nine soil health indicators were then retained to represent complementary functional domains: microbial biomass, plant biomass, earthworm biomass change, nitrate content, the activities of four enzymes (FDA hydrolase, beta-glucosidase, urease, and neutral phosphatase), and glomalin content, a glycoprotein linked to soil structural stability.

The responses of these indicators to metal exposure were strikingly complex. Biomass measures generally declined with increasing metal concentrations, but the enzyme activities behaved in nonlinear, sometimes non-monotonic ways. Under cadmium stress, beta-glucosidase and neutral phosphatase showed U-shaped responses, while urease followed an inverted U. For copper, several enzymes rose at intermediate concentrations before falling at high doses. The team used principal component analysis to integrate these non-monotonic patterns, identifying concentration ranges at which combined carbon, nitrogen, and phosphorus cycling shifted adversely, and benchmark dose modeling to derive indicator-specific effect thresholds. For cadmium, the most sensitive indicator was glomalin content, with a threshold of just 0.89 milligrams per kilogram, whereas microbial biomass tolerated up to 53.44 milligrams per kilogram.

These indicator thresholds were then assembled into cumulative sensitivity distributions, from which hazardous concentrations were read. For lead, the model-averaged SHSD placed the concentration affecting 5 percent of soil health indicators at 113.89 milligrams per kilogram; for copper, the HC5 was 34.00 milligrams per kilogram; and for cadmium it was just 2.59 milligrams per kilogram. Coupling these distributions with spatially resolved exposure data from 86 topsoil samples collected across Tianjin in June 2022, the researchers produced probabilistic risk estimates for the entire municipality. Copper posed the highest regional ecological risk, with a mean estimate of 9.07 percent, followed by lead at 1.13 percent and cadmium at 0.52 percent. Spatially, copper risks peaked in the urban Core Area at 14.88 percent, while lead and cadmium risks were highest in Jinghai District.

When the same exposure data were run through conventional species sensitivity distributions, the risk estimates were dramatically higher: 22.42 percent for copper, 3.57 percent for lead, and 3.33 percent for cadmium. The SHSD-based risks were lower by 84.38 percent for cadmium, 61.90 percent for lead, and 59.55 percent for copper, a difference the authors report as statistically significant. The ranking of the three metals was broadly consistent between the two frameworks, and the spatial hotspots largely overlapped, suggesting that exposure patterns, shaped by historical industrial emissions, traffic-related deposition, and long-term fertilizer inputs, drive much of the geographic structure in both assessments.

Why the large gap? One obvious suspect is soil chemistry. SSD toxicity values are typically drawn from studies conducted in wildly different soils, where pH, organic matter, and cation exchange capacity strongly modulate metal bioavailability. When the researchers normalized the literature toxicity data to the physicochemical conditions of their microcosm soil using species-specific equations, part of the discrepancy vanished. For lead, normalization reduced the SSD-SHSD deviation by nearly 100 percent at low concentrations but the benefit shrank nonlinearly as concentrations rose, disappearing above roughly 3,223 milligrams per kilogram where direct toxicity dominates. For copper, normalization cut the deviation by 29.73 percent at the regional mean concentration. For cadmium, whose greater solubility and weaker binding to soil particles make it less sensitive to soil context, the effect was smaller and even reversed slightly at intermediate concentrations.

Normalization alone could not close the gap, however, and the residual difference points to something more intriguing: the integrated microcosm appears genuinely less sensitive than the sum of its parts. Using a partitioning framework originally proposed by Loreau and Hector, the team constructed expected SHSDs from additive combinations of microbial-only, earthworm-microbe, and plant-microbe assemblages. The observed SHSD from the full three-component system shifted consistently to the right, toward higher tolerated concentrations, for all three metals. The authors propose several candidate mechanisms. Earthworm bioturbation may promote stable metal-organic complexes and microbial colonization; plant root exudates may enrich carbon substrates and stimulate metal-immobilizing microbes; and microbial metabolism can mediate metal reduction, precipitation, and complexation. Functional redundancy adds another layer, as enhanced enzyme activities under moderate metal exposure suggest microbial consortia preserved nutrient cycling even when plant productivity declined. Feedback loops, such as glomalin-driven aggregate formation increasing the sorptive capacity of soil colloids, may further decouple organismal exposure from total contaminant load.

The authors are careful to frame these results as a demonstration rather than a replacement. The SHSD thresholds are microcosm-based functional tolerance values derived from a defined soil matrix, fixed exposure duration, and selected representative taxa; they are not yet universal regulatory criteria. Important determinants of bioavailability and ecosystem complexity, including pH, soil organic carbon, and microbial diversity, were not included as direct response indicators, and no field measurements of functional impairment were available to independently validate the thresholds. Sensitivity analyses also revealed that the choice of indicator weighting matters, particularly for cadmium: excluding glomalin content raised the cadmium HC5 from 2.59 to 6.81 milligrams per kilogram and collapsed the regional risk estimate to 0.01 percent, showing that a single highly sensitive structural endpoint can dominate the lower tail of the distribution.

Even with those caveats, the regulatory implications are considerable. The authors envision SHSD-derived thresholds combined with spatial exposure maps to support risk-based zoning and hotspot prioritization, joint probability estimates to rank remediation urgency among regions and contaminants, and monitoring programs built around indicators directly tied to ecosystem functioning, from microbial biomass to enzyme activities and nutrient-cycling metrics. Most compellingly, SHSD could slot into a tiered assessment strategy in which conventional SSDs provide screening-level species protection and SHSD supplies higher-tier information on how real soil systems respond under defined conditions. If validated across contrasting soils, pollution histories, and biological assemblages, the approach could shift soil risk assessment from a ledger of how many species die toward a measure of whether the living system itself keeps working, a question that matters far more for the farms, cities, and coastal wetlands sitting on contaminated ground.

Subject of Research: Soil health-based versus species-based sensitivity distributions for assessing ecological risks of metal contamination in soils

Article Title: Comparing soil-health- and species-based sensitivity distributions for metal risk assessment in Tianjin soils

Article References: Xu, Q., Shi, Y., Xiong, X., Qian, L., Fang, L., Zhou, X., & Shao, X. (2026). Comparing soil-health- and species-based sensitivity distributions for metal risk assessment in Tianjin soils. iScience, 29(10), Article 117644. https://doi.org/10.1016/j.isci.2026.117644

Image Credits: AI Generated

DOI: 10.1016/j.isci.2026.117644

Keywords: soil health, ecological risk assessment, species sensitivity distribution, heavy metals, lead, copper, cadmium, soil microcosm, Tianjin, soil enzymes, glomalin, environmental toxicology