Herbicide leakage into seawater impacts primary productivity and zooplankton globally

The study investigates how widespread herbicide contamination in coastal waters affects marine primary productivity and higher trophic levels. Herbicides, many of which are photosystem II inhibitors, frequently enter the ocean via runoff and are detected globally at concentrations capable of inhibiting phytoplankton photosynthesis. Sensitive diatom taxa (e.g., Phaeodactylum tricornutum and Chaetoceros spp.) can be strongly inhibited at low concentrations, raising concerns about impacts on oceanic carbon fixation and potential coastal "desertification." The authors aim to quantify global coastal herbicide pollution, establish concentration–response relationships at the phytoplankton community level for multiple herbicides via toxicity equivalence, and link herbicide usage on land to marine residue patterns to assess and predict ecological impacts on primary production and zooplankton communities.

Previous work has documented extensive herbicide use worldwide with substantial fractions transported to marine systems. Many herbicides are photosynthetic inhibitors, implicated in effects such as coral bleaching in the Great Barrier Reef. Toxicity studies demonstrate significant impacts on algal photosynthetic physiology, nutrient uptake, and carbon sequestration gene expression, with sensitivity differences among phytoplankton species. Most risk assessments have focused on single herbicides and limited regions, despite evidence of additive toxicity among compounds sharing modes of action. The importance of phytoplankton community structure and size spectra for determining primary production and energy transfer is well established; smaller pico- and nano-phytoplankton often channel production through the microbial loop, while larger nano- and micro-phytoplankton feed more efficiently into classical food chains. A gap remains in understanding in situ, community-level impacts of realistic multi-herbicide mixtures on primary productivity and subsequent trophic effects in coastal oceans.

Global compilation and analysis: Using Web of Science, the authors compiled 568 publications (1990–2022) and extracted herbicide types, concentrations, dates, and locations for 661 stations across 15 bays/gulfs grouped into seven sea areas. Where needed, data were digitized from figures. Temporal trends were evaluated by one-way ANOVA across three periods (1990–2000, 2001–2011, 2012–2022). Toxicity equivalence (TEQ) framework: Twelve triazine herbicides (most prevalent class) were assessed for short-term toxicity to the model diatom Phaeodactylum tricornutum Pt-1. Cultures were exposed across 0.1 nmol L−1 to 32 µmol L−1; concentration–response curves (logistic/Weibull; R²=0.989–0.999) yielded ECx values and EC50s (4.3±0.3 to 849.1±21.7 nmol L−1). Relative toxicities were used with the concentration addition (CA) model to convert measured in situ concentrations of each triazine to an atrazine-equivalent concentration (TEQ). Site-level multi-herbicide burdens were expressed as atrazine TEQs for comparability across stations. Microcosm experiments (natural seawater): A controlled 30-day microcosm used herbicide-free coastal seawater (Shilaoren Bay, Qingdao), with atrazine added at 0, 0.5, 5, and 50 nmol L−1 (triplicate 80 L bottles per treatment). Subsampling occurred over days 0–30. Primary productivity proxies (chlorophyll a) were measured fluorometrically, including size-fractionated Chl a (<2 µm pico-, 2–20 µm nano-, 20–200 µm micro-). Differences from controls were assessed with two-sided t-tests (p≤0.05). Community composition and particle size: On day 21, phytoplankton and micro-zooplankton communities were profiled via 18S rRNA gene amplicon sequencing (primers 18S-82F/Ek-516R; Illumina MiSeq 2×300). Reads were processed (QIIME/FLASH/USEARCH/UCLUST) to 97% OTUs, rarefied (30,000 reads/sample), and analyzed for alpha/beta diversity (Kruskal–Wallis with FDR correction; PCoA). BLAST-based taxonomic assignments were linked to literature-derived cell size information to interpret size-structured shifts. Micro-zooplankton were also morphologically identified and counted (Sedgwick-Rafter chamber) for validation. Growth and grazing (dilution method): On day 4, dilution-series incubations (100–20% ISW diluted with 0.2 µm-filtered seawater) quantified intrinsic growth rate (µ), net growth rate (NGR), and micro-zooplankton grazing rate (g) for pico-, nano-, and micro-phytoplankton fractions from each atrazine treatment. Rates were derived from linear regressions of ln(Pt/P0)/t vs dilution (Landry-Hassett approach). In situ coastal surveys: The Bohai and Yellow Seas (adjacent to high-risk agricultural regions) were surveyed in autumn 2017 and spring 2018 at 64 stations for 22 herbicides (10 triazines, 6 phenylureas, 6 amides) using SPE-LC-MS/MS (with deuterated internal standards). LOQs were 0.004–0.016 nmol L−1 for triazines, 0.007–0.023 ng L−1 phenylureas, 0.005–0.01 nmol L−1 amides. Spatial and seasonal patterns were analyzed. Global agricultural risk mapping: Using PEST-CHEMGRIDSv1 (crop-specific application rates for 52 widely used herbicides at 5′ resolution), FAOSTAT, and USGS data, predicted environmental concentrations (PECs) were estimated via the Environmental Potential Risk Indicator approach. Risk quotients (RQ=PEC/PNEC) employed PNECs from LC50s of earthworms (soil) and fish (water) with safety factors (÷1000). Grid-cell risk points (RP=log ΣRQ_i) were categorized: negligible (≤0), low (0–1), medium (1–3), high (>3). Correlations between land RP and adjacent coastal residues were examined. Modeling inhibition at global stations: The TEQ-based atrazine concentrations for each station were mapped to an experimentally fitted community-level atrazine–Chl a inhibition logistic curve to estimate percent inhibition of primary productivity at 661 stations.

- Global coastal contamination: Among 32 herbicides in 5 categories, 12 triazines comprised 37.5% of types and were detected at 95% of 661 stations. High-concentration regions included the Gulf of Mexico, East Asia, and Vilaine Bay, with median triazine concentrations of 3.84, 2.28, and 1.64 nmol L−1, and maxima ~13.7, 12.1, and 11.7 nmol L−1, respectively. Overall, across the global dataset, summed triazines had median, third quartile, and maximum concentrations of 0.18, 1.27, and 29.50 nmol L−1 (95% CI reported as 1.06, 1.47 for quartiles). Temporal analyses showed significant increases in triazines in several regions (e.g., East Asia, US East Coast, South Africa). - Toxicity equivalence: Atrazine-equivalent concentrations (TEQs) for triazine mixtures across stations ranged 0–47.58 nmol L−1, with median 0.54 nmol L−1, third quartile 5.09 nmol L−1, and maximum 47.58 nmol L−1. High-TEQ areas included Camps Bay/Maputo Bay (median 9.47 nmol L−1), Yellow/Bohai Seas (median 8.74 nmol L−1), and Gulf of Mexico (median 7.31 nmol L−1); maxima reached 47.58, 23.16, and 22.91 nmol L−1, respectively. - Community-level dose–response: The fitted atrazine–Chl a inhibition curve indicated 5%, 10%, and 25% inhibition at 5.1, 11.9, and 35.2 nmol L−1 atrazine, respectively. Applying TEQs, primary productivity was inhibited >5% at 25% (167/661) of stations and >10% at 10% (67/661) of stations. Median inhibition was highest in South African bays (11.59%), the Yellow/Bohai Seas (7.72%), and the Gulf of Mexico (6.60%); station-level maxima near estuaries reached 48.97%, 17.86%, and 12.23%. - Microcosm outcomes at realistic doses: Atrazine (0.5, 5, 50 nmol L−1) reduced total Chl a by 9.1%, 8.8%, and 18.8% after 2 days; by day 30, reductions were 16.9% (5 nM) and 24.5% (50 nM). Nano-phytoplankton contributions declined markedly (e.g., from 75.3% in controls to 44.9%, 23.7%, 16.7% at 0.5, 5, 50 nM by day 30), while micro-phytoplankton increased (from 21.7% to 52.7%, 66.7%, 74.1%). - Community shifts: Initially, Bacillariophyta (77.4±5.87%) dominated; under 0.5, 5, 50 nM atrazine, Bacillariophyta fell to 15.5±5.21%, 10.7±3.35%, and 0.7±0.23%, while Dinophyceae rose from 17.8±2.16% to 47.5±6.87%, 53.6±10.06%, and 79.2±11.17%. Chaetoceros declined from 73.0% to 5.2%, 2.9%, and 0.3% (p≤0.0007). The dominant species shifted from Chaetoceros tenuissimus to Gyrodinium jinhaense. - Growth and grazing: Intrinsic growth rates decreased under atrazine, especially for nano- (−36.5% at 5 nM; −52.7% at 50 nM) and pico-fractions (−14.9%; −71.9%); micro decreased by −18.5% and −32.3%. Zooplankton grazing rates declined under medium/high doses across size classes (e.g., micro −18.1% and −37.7% at 5 and 50 nM on day 4). - Zooplankton secondary effects: Micro-zooplankton shifted from copepod larvae-dominated (∼97%) to ciliate-dominated at high atrazine (ciliates to 40%; copepod larvae down to ∼21%). Morphological counts corroborated sequencing: copepod larvae abundances fell by an order of magnitude at medium/high doses, while ciliates (Tintinnopsis, Euplotes) increased. Network analyses linked copepod declines to sharp reductions in nano-diatoms (e.g., Chaetoceros), not direct herbicide toxicity. - Land–sea linkage and risk: 65.02% (~23.63 million km²) of global agricultural land showed herbicide risk (RP>0), with 16.84% (~6.12 million km²) at high risk (RP>3), concentrated in Latin America (29% of its ag land), Europe (25%), Asia (22%), and North America (11%). Adjacent seas exhibited correspondingly high residues; in the Bohai/Yellow Seas, detection rates for 10 triazines, 4 phenylureas, and 3 amides were 100%, with category maxima up to 12.07, 6.97, and 2.79 nmol L−1, exceeding EU water safety standards (2.5 nM) in many locations. Concentrations were higher near estuaries and in spring, reflecting agricultural schedules. - Global carbon implications: A 5% reduction in coastal primary productivity could decrease annual phytoplankton carbon fixation by 3.75–8.75×10^9 tons, comparable to Amazon rainforest sequestration, with further losses expected from inefficient energy transfer via a strengthened microbial loop.

This work demonstrates that realistic, mixed-herbicide exposures in coastal waters are sufficiently high and widespread to suppress phytoplankton primary productivity at community scales, especially in regions adjacent to high-use agricultural areas. By converting multi-component triazine mixtures to atrazine-equivalent concentrations and applying an empirically derived community-level dose–response, the authors quantify inhibition levels across 661 stations—showing nontrivial declines (>5–10%) at a substantial fraction of sites. Microcosm experiments reveal mechanistic underpinnings: herbicides selectively inhibit abundant, sensitive diatoms (e.g., Chaetoceros), drive succession toward more tolerant Dinophyceae, and shift size structure from nano- toward larger micro-phytoplankton. These compositional changes, together with direct reductions in intrinsic growth rates (particularly for nano- and pico-fractions), reduce total primary productivity and slow its growth dynamics. The altered phytoplankton size spectrum and community structure propagate to higher trophic levels: micro-zooplankton communities transition from larger copepod larvae to smaller ciliates as nano-diatoms decline, decoupling classical grazing pathways and preferentially routing energy into the less efficient microbial loop. This prolongs transfer through the food web, compounding productivity losses with decreased transmission efficiency to higher consumers. The documented correlation between terrestrial herbicide application risk and adjacent marine residues supports a land–sea management nexus: intensifying agricultural herbicide use is likely to exacerbate coastal ecosystem impacts without mitigation. Overall, the findings directly address the research question by quantifying the magnitude, mechanisms, and downstream ecological consequences of widespread herbicide contamination on coastal primary productivity.

The study provides a global assessment of triazine herbicide contamination in coastal bays and gulfs, introduces a toxicity-equivalence framework to aggregate mixed herbicide burdens, and empirically links these burdens to community-level primary productivity inhibition. It reveals consistent community and size-structure shifts in phytoplankton and secondary restructuring of micro-zooplankton, together reducing productivity and energy transfer efficiency. Land-based herbicide risks strongly predict adjacent marine residues, highlighting opportunities for targeted monitoring and mitigation. Future directions include: expanding mixture assessments beyond triazines to other herbicide classes and co-occurring agrochemicals; conducting long-term, in situ ecosystem-scale measurements to validate and refine model predictions; integrating nutrient enrichment effects to resolve trade-offs between fertilization-driven productivity gains and herbicide-induced suppression; improving species- and size-resolved response functions to reduce uncertainty in toxicity equivalence; and developing land–sea management strategies to reduce non-point-source herbicide loads while maintaining agricultural productivity.

- The toxicity equivalence approach assumes concentration addition and uses Phaeodactylum tricornutum as a representative diatom; interspecific variability in sensitivity may introduce bias. - Community metabarcoding (18S rRNA) can be biased by differential rDNA copy numbers, read length limits, and growth rate differences across taxa; although morphological validation was used, residual biases are possible. - Microcosm conditions may overestimate or underestimate in situ effects due to simplified hydrodynamics, light, and nutrient regimes; ocean mixing and self-purification can reduce local exposure, yet widespread detections offshore indicate persistent exposure. - The global inhibition estimates primarily consider triazines; other herbicide classes and pesticide mixtures present in seawater were not fully incorporated, likely leading to conservative (under-)estimates. - Risk mapping relies on PECs constrained by available usage data and generic PNECs with safety factors; uncertainties in application rates, degradation, and transport can affect RP classifications and land–sea correlations.