Study Design (est-study-design)

ES&T reviewers are demanding about whether a design can actually support its environmental claim. The failure modes are predictable: unrealistic conditions, missing controls, no replication, an unclosed mass balance, or QA/QC bolted on after the fact. Design to pre-empt them. Execution and reporting of results live in est-data-analysis.

When to trigger

• Planning a lab/mesocosm/field study, sampling campaign, or modeling experiment
• Choosing concentrations, matrices, controls, replicates, and endpoints
• Setting up the QA/QC and mass/energy-balance plan before generating data
• A reviewer questioned environmental relevance, controls, or replication

Design principles ES&T expects

1. Environmental relevance. Use concentrations, matrices, pH/ionic strength, light, temperature,

and timescales representative of the target system — not only idealized lab spikes. Justify any accelerated/exaggerated conditions.

1. Controls that isolate the mechanism. Include the controls that rule out abiotic loss,

sorption, volatilization, photolysis, blanks, and matrix effects — whatever could mimic your effect.

1. Replication & randomization. Biological/experimental replicates (not just technical);

randomize/rotate where position or batch could confound; power your design for the effect size.

1. Mass / energy balance. Where the design implies one, plan to account for inputs, products,

sorbed/volatilized fractions, and losses — unexplained gaps are a top rejection reason.

1. QA/QC by design. Pre-plan blanks (method/field), spikes/recoveries, CRMs, LOD/LOQ,

calibration, surrogate/internal standards, and duplicates (see est-data-analysis).

1. Dose–response / kinetics. For toxicity or reaction studies, design enough points to fit

curves/rate constants, not just a single dose or time point.

1. Models. State assumptions, domain, boundary/initial conditions, calibration vs validation data,

and sensitivity/uncertainty analysis up front.

Controls a reviewer expects, by claim type

The fastest way to lose an environmental-relevance argument is to omit the control that rules out a competing process. Match the control set to what you are claiming:

Worked micro-example (illustrative — designing the PFAS biotransformation study)

To support "precursor X biotransforms to PFHxA in river water," the design (illustrative) builds in the defenses before any sample is run:

• Environmental relevance: spike at ~50 ng/L (illustrative — near observed field levels), in

filtered river water at ambient pH and temperature, not a buffered idealized matrix.

• Controls: an autoclaved (abiotic) control to separate biotransformation from sorption/hydrolysis;

a no-spike blank; a sorption check on the vessel walls.

• Replication & power: triplicate microcosms per timepoint (biological replicates), randomized

incubator position; enough timepoints (e.g., 0, 1, 3, 7, 14 d) to fit a first-order rate constant.

• Mass balance: measure precursor, intermediates, and terminal acid plus a sorbed-fraction

extraction, targeting ≥80% closure (illustrative) and reporting the gap.

• QA/QC by design: per-batch spikes/recoveries, field and method blanks, per-analyte LOQ, and

surrogate standards pre-specified — not improvised after the run.

The design choice that pre-empts the top rejection: the abiotic control plus the mass balance together make the biotransformation claim falsifiable, which is exactly what the analytical reviewer checks.

Anti-patterns

• Lab spikes orders of magnitude above environmental levels presented as relevant
• No abiotic/sorption/photolysis control to isolate the claimed process
• n = 1 or technical replicates passed off as independent replication
• A transformation/treatment study with no attempt at a mass balance
• QA/QC improvised after data collection; no blanks or recoveries planned
• A model with undisclosed assumptions and no validation or sensitivity analysis

Output format

【Design】lab / mesocosm / field / modeling + endpoints
【Environmental relevance】conditions match target system? [Y/N + justification]
【Controls】which confounders ruled out
【Replication/power】n, randomization, effect size
【Mass/energy balance】planned? how closed?
【QA/QC plan】blanks / spikes / CRM / LOD-LOQ / calibration
【Next】est-data-analysis

Supplementary resources

• ../../resources/external_tools.md — instruments, fate/transport models, QA/QC backbone
• ../../resources/official-source-map.md — ES&T scope and rigor expectations