1. The study uses the Concentration Addition (CA) model to calculate cumulative risk (RQsum) for complex mixtures of 132 chemicals. This model assumes similar modes of action. Given the diverse chemical classes detected (herbicides, pharmaceuticals, corrosion inhibitors), how do you justify this assumption? Have you considered the Independent Action (IA) model or conducted a sensitivity analysis to determine if the qualitative findings (e.g., >70%>70% of sites at risk) are robust to the choice of mixture model?
2. The paper uses a metric of nucleotide divergence (π) as a measure of “genetic diversity.” However, eDNA metabarcoding data are prone to PCR and sequencing errors. How did you control for these errors to ensure that the observed decline in ππ along the chemical gradient reflects true genetic erosion and not, for instance, a reduction in the abundance of rare taxa or an artifact of differential PCR efficiency in contaminated samples? Is it possible that the loss of “sequence diversity” is simply a proxy for a reduction in the number of unique haplotypes within the dominant species?
3. The structural equation model (SEM) posits a causal pathway where chemical stressors “mediate” the effect on beta diversity through sequence diversity. Given the single spatial survey and the inherent correlation between land use, BOD, and chemical concentrations, how do you rule out the possibility that an unmeasured co-stressor (e.g., physical habitat degradation, flow alteration) is the primary driver of both the chemical stressor gradients and the observed biodiversity losses? The model’s path coefficients do not establish causation.
4. The 5% Hazard Concentration (HC5) threshold is presented as a key metric for algae sensitivity. However, the reported HC5 values (0.51 – 0.62) are unitless. What are the actual units of the RQalgae gradient used for this quantile regression? Without units, this HC5 value is uninterpretable and cannot be used for any regulatory or comparative purpose.
5. The study is based on a single cross-sectional survey in September. How do you account for the high temporal variability in both chemical concentrations (e.g., seasonal pesticide pulses) and eDNA signals (e.g., life-history events, decay rates)? A single snapshot may not represent the long-term exposure that drives biodiversity effects, potentially confounding the observed relationships.
