MESSAI.iO

Reproducibility · ISMET EU 2026

Five-parameter minimum checklist

Reporting these five experimental parameters reduces inter-study power-density variance by more than 40% (487 → 278 mW/m², p < 0.05) across 63 validation papers from the MESSAI corpus.

Score your own paper, browse the 289-paper corpus distribution, or use the comprehensive 18-parameter view for a deeper audit.

No DOI handy?

Score interactively

Already know what your paper reports? Toggle the criteria below to see your score immediately — no lookup needed.

Jump to the interactive scorer

Interactive scorer

Toggle each criterion to see how your reporting compares against the ISMET 2026 minimum. Each criterion links to the literature it’s grounded in.

Electrode Spacing

Cell GeometryCriticalcm

Distance between anode and cathode electrodes (cm). Controls ohmic resistance and mass-transport boundary layer thickness; the single largest determinant of power density variance across papers in the abstract's correlation analysis (r = 0.37).

Typical range: 1 – 20 cm

1 reference
  • Liu & Logan 2004 (Environ Sci Technol 38:4040)
Current status: Missing

Electrode Surface Area

Cell GeometryCriticalcm²

Projected geometric area of the anode and cathode (cm²). Required to convert raw current to areal current density. Without it, any power-density claim is a unit-less number — cross-study comparison becomes impossible.

Typical range: 1 – 1,000 cm²

1 reference
  • Logan 2008 (Microbial Fuel Cells, Wiley)
Current status: Missing

External Resistance

CircuitCriticalΩ

Load resistance across the circuit (Ω). Sets the operating point on the polarization curve. Maximum power is reported at R_ext = R_internal; papers omitting R_ext typically also omit polarization sweeps, making the reported power point uncomparable.

Typical range: 10 – 10,000 Ω

1 reference
  • Logan et al. 2006 (Environ Sci Technol 40:5181)
Current status: Missing

Measurement Method

MethodologyCritical

How power / current were measured: data-logger vs handheld multimeter, sample interval, voltage-sweep protocol if a polarization curve was run. 8% of inter-study variance in the abstract's correlation analysis. Without it, a 26 mW/m² instantaneous reading is indistinguishable from a 26 mW/m² 24-hour average — a 4× real-world performance gap masquerading as agreement.

1 reference
  • Logan et al. 2019 (Nat Rev Microbiol 17:307)
Current status: Missing

Unit Normalization Basis

MethodologyCritical

Whether power is normalized to electrode area (mW/m²), reactor volume (W/m³), or anode mass (W/kg). The abstract's headline finding: power density spans five orders of magnitude (CoV 1,285%) largely because authors silently change normalization basis. Stating it explicitly is the single highest-leverage reporting improvement (most extreme rows in the corpus collapse onto a ~10× range when re-normalized).

1 reference
  • Schröder 2007 (Phys Chem Chem Phys 9:2619)
Current status: Missing

ISMET 5-param minimum

0%

Not yet comparable

Fewer than 60% of criteria reported. The reported power / current numbers are not safely cross-comparable against the corpus — context is missing.

Reported
0
Omitted
0
Missing
5
Weighted score
0/5

Validation cohort

43% IQR reduction

Across 63 papers in the MESSAI corpus, those reporting all five minimum criteria had an inter-quartile range of 278 mW/m² on power density — vs 487 mW/m² for the rest.

Mann-Whitney U test, p < 0.05. Cohort defined by retrospective audit of the corpus against the checklist criteria; not a randomized trial.

Without checklist (n=46)487 mW/m²
With 5-param checklist (n=17)278 mW/m²

Axis: power-density inter-quartile range across each cohort, mW/m². Source: docs/abstracts/messai-eu-ismet-2026.pdf Figure 2C.

Corpus baseline

23.1% average completeness

Across 289 papers manually scored from the 10,824-paper indexed corpus, average reporting completeness is 23.1% — nowhere near the level needed for quantitative cross-study comparison.

MFC papers report on average 8.7 parameters; MEC papers report 3.6. MEC has the largest gap to close.

The five-parameter minimum is calibrated to be the lowest-friction subset that materially closes this gap.

SystemPapersElectrode
Specs		Operating
Conds		Electrical
Meas		BiologicalData
Reporting
MFC3,84772%58%45%35%32%
MES4,84268%55%42%32%28%
BES1,09665%48%38%28%25%
MEC96670%52%40%30%28%
MDC7355%38%28%20%15%

Heat scale: red < 20% · orange 20-35% · amber 35-50% · green 50-70% · deep green ≥ 70%. Source: abstract Fig 2A.

Coverage by experiment type

Reproducibility isn’t one number

Each study type needs a different minimum-reporting set, so we score each paper against its kind of study — a polarization curve, a continuous-flow treatment study, and a CO&sub2;-electrosynthesis run are judged on different criteria. Lower coverage = harder to reproduce.

Coverage:<15% critical15–35% partial>35% better

The dominant comparability gap

86.1% of papers reporting a power/current-density value never state which area it’s normalized to.

Areal vs volumetric — and which area (anode / cathode / membrane / projected) — differ by orders of magnitude; an unstated basis is the single biggest driver of the ~1,285% spread in reported power density.

When a basis IS stated: electrode (generic) 4% · anode area 1.9% · projected/geometric 0.7% · cathode area 0.6% · membrane area 0.3% · unspecified 92.4% · n = 1,613

Mean reporting coverage by study type, with the density normalization basis and the worst-reported criterion for each type.
Study typePapersMean coverageBy axisDensity basis statedWorst-reported criterion
Bioelectrochemical system (general)1,0375%cond 9%outc 3%meth 0%equi 0%0%Electrode spacing (0.1%)
Modeling / simulation4796.1%cond 6%outc 5%meth 7%equi 8%25.7%Sensitivity / uncertainty propagation (0.4%)
Electrode / membrane materials4,04711.1%cond 20%outc 8%meth 3%equi 2%10.6%How electrode area was determined (0.7%)
Sediment / plant MFC22812.9%cond 19%outc 15%meth 5%equi 2%23.5%Reading convention + sampling interval (0.9%)
Startup / enrichment dynamics1,55713.3%cond 23%outc 10%meth 3%equi 2%19%Reading convention + sampling interval (0.7%)
Wastewater treatment / resource recovery1,77514.3%cond 20%outc 17%meth 4%equi 3%28.2%Pumps + flow control (0.9%)
Microbial desalination cell10514.4%cond 21%outc 17%meth 7%equi 1%38.9%Potentiostat / electrochemical workstation (0%)
Bioelectrochemical sensor12815%cond 21%outc 17%meth 11%equi 4%16.7%Potentiostat / electrochemical workstation (0%)
MFC — power generation2,05815.5%cond 19%outc 25%meth 4%equi 2%11.9%Reading convention + sampling interval (0.8%)
MEC — hydrogen / gases37818%cond 21%outc 28%meth 7%equi 7%33.3%Reading convention + sampling interval (0.3%)
Fundamental / EET kinetics27121.5%cond 30%outc 21%meth 14%equi 9%29.5%Data acquisition system (3%)
Long-term operation / durability52221.9%cond 29%outc 32%meth 6%equi 5%27.7%Reading convention + sampling interval (1.3%)
Microbial electrosynthesis16127.1%cond 37%outc 30%meth 15%equi 15%36.1%Product selectivity / distribution (0.6%)
Electrochemical characterization46231.7%cond 41%outc 38%meth 22%equi 13%32.1%Reading convention + sampling interval (3.5%)

Baseline comparability floor (% of all papers reporting)

  • 1.6% Electrode spacing
  • 1.8% Buffer type + concentration
  • 3.2% Electrolyte conductivity / ionic strength
  • 3.7% Unit normalization basis
  • 5.1% Replication (n) + variance reporting
  • 5.5% Electrode surface area
  • 7% External resistance / load
  • 15.4% Operating temperature
  • 17.8% Anolyte / catholyte pH
  • 47.2% Substrate / carbon source

Local committed sample (papers with DOI + >=1 extracted param). Re-run against a full DB export for corpus-scale numbers. Method: scripts/analysis/experiment_type_coverage.py · generated 2026-07-22.

Methodology & equipment

What you measured with, not just what you measured

A polarization curve swept at 1 mV/s and one stepped through a resistor box to steady state give different peak powers for the same reactor. These decisions never appear as an extracted number — they live in your methods section — so they are the reporting gap the corpus can’t see. Pick your study type for the protocol and instrument choices it implies.

MFC — power generation

A microbial fuel cell reporting power or current density, typically from a polarization curve. Reproducibility hinges on geometry, load, and exactly how power was sampled and normalized.

Methodology

Polarization / power-curve protocol

critical

The same cell yields substantially different peak power under a fast sweep versus a resistor box held to steady state at each point — the fast sweep inflates power because the biofilm never re-equilibrates. The protocol is not a detail; it sets the number.

Available options — report which one you used

  • Variable external resistance, steady state per point — Resistor stepped across a decade range, each point held until voltage is stable (typically ≥20–30 min, or one full fed-batch cycle). The reference method; slowest and most defensible.
  • Linear sweep voltammetry (LSV) — Potentiostat sweeps the cell at a stated scan rate (commonly 0.1–1 mV/s). Fast, but scan rate must be reported — faster sweeps systematically overstate peak power.
  • Galvanostatic stepped discharge — Current stepped and voltage recorded at each hold. Good control of the operating point; report step size and hold time.
  • Single-cycle potentiodynamic sweep vs reference — Half-cell resolved; separates anode and cathode contributions but needs a reference electrode.

Logan et al. 2006 (Environ Sci Technol 40:5181) · Zhao et al. 2009 (Electroanalysis 21:1449)

Reading convention + sampling interval

critical

An instantaneous peak, a cycle maximum and a 24-hour average are three different numbers from one reactor. Without the convention and the logging interval, a 26 mW/m² instantaneous reading is indistinguishable from a 26 mW/m² daily mean — a fourfold real-world gap masquerading as agreement.

Available options — report which one you used

  • Instantaneous peak — Maximum observed reading. Highest and least reproducible; must be labelled as such.
  • Steady-state / plateau value — Value after the reading has stabilised, with the stability criterion stated (e.g. <5% drift over 1 h).
  • Cycle maximum (fed-batch) — Peak within a feeding cycle, averaged over a stated number of cycles. Report cycle count and cycle length.
  • Time-averaged over the run — Mean across continuous operation. The most conservative and the most useful for energy claims.

How the density was computed

critical

The arithmetic behind the density: which area or volume went in the denominator, and whether it was projected, real (BET) or electrochemically active area. Stating the basis without stating which area was used leaves the number ambiguous.

Available options — report which one you used

  • Projected (geometric) anode area — The most common convention and the recommended default for cross-study comparison.
  • Projected cathode area — Legitimate when the cathode limits; must be labelled, since it typically differs from anode area.
  • Membrane / separator area — Used for stacked and flow systems; often the smallest of the three, inflating the density.
  • Real surface area (BET) or electrochemically active area (EASA) — Physically meaningful for materials comparison but orders of magnitude larger than projected area — never comparable to a projected-area figure.
  • Total reactor / net liquid volume — Volumetric basis (W/m³). Specify total vs net liquid volume — they differ by the headspace and the electrode volume.

How electrode area was determined

high

The denominator of every density. Projected area, BET area and electrochemically active area on one brush anode can span two orders of magnitude — the method must travel with the number.

Available options — report which one you used

  • Projected / geometric from dimensions — Measured with calipers; state whether one face or both faces were counted — a factor-of-two error that is endemic in the literature.
  • Manufacturer specific surface area × mass — Common for felts and brushes; cite the datasheet value, since it is a supplier claim rather than a measurement.
  • BET measured on the actual electrode — Directly measured total area.
  • Electrochemically active area from double-layer capacitance — The area actually participating; the most defensible basis for a catalytic comparison.

COD / BOD / TOC determination

critical

Removal percentages hinge on the assay. Closed-reflux dichromate COD, a Hach kit and a TOC analyser return different numbers on the same sample, and filtered versus total COD differ by the biomass.

Available options — report which one you used

  • Closed-reflux colorimetric COD (SM 5220D) — The reference method. State whether COD was total or filtered (0.45 µm) — the difference is the suspended biomass.
  • Commercial digestion vial kit (Hach/Merck) — Practical and widely used; cite the range and the interference check for chloride.
  • BOD5 (SM 5210B) — Biological demand over 5 days with a stated seed and dilution; not interchangeable with COD.
  • TOC analyser (combustion or UV-persulfate) — Carbon rather than oxygen demand; the cleanest for defined substrates but not comparable to COD without a conversion factor.

Statistical treatment

critical

What test, on how many independent reactors, with what significance threshold. Distinguishing biological replicates from technical replicates is the difference between a result and an artefact of measuring one reactor three times.

Available options — report which one you used

  • Independent reactor replicates (biological) — Separate reactors inoculated and run independently. The only basis for a claim about the system.
  • Technical replicates on one reactor — Repeat measurements of a single unit. Valid for instrument precision; not a basis for a performance claim — label it as such.
  • Hypothesis test with stated α — t-test or ANOVA with post-hoc correction, reporting n, the statistic and the exact p.
  • Effect size / confidence interval — More informative than a p-value alone for cross-study comparison, and directly usable in meta-analysis.

Equipment & tools

Potentiostat / electrochemical workstation

critical

Instrument class bounds what a measurement can be. A resistor box and a voltmeter can produce a power curve but never a CV, an EIS spectrum or a poised-potential experiment; channel count decides whether replicate reactors were run simultaneously or sequentially.

Available options — report which one you used

  • Multi-channel potentiostat — e.g. BioLogic VMP/VSP, Gamry Interface, Metrohm Autolab, Ivium-n-Stat, Admiral Squidstat. Required for simultaneous replicates and for any half-cell work.
  • Single-channel potentiostat — Full electrochemical capability, one reactor at a time — note whether replicates were therefore run sequentially.
  • Data logger + fixed/variable resistor — Whole-cell voltage only. Perfectly adequate for power curves and long-term monitoring; cannot poise a potential or run CV/EIS.
  • Handheld multimeter, manual reads — Lowest-resolution option. Report the reading cadence, since manual sampling cannot capture fed-batch transients.

Data acquisition system

high

Sampling interval and input impedance decide what the record can show. A logger sampling every 30 min cannot resolve a fed-batch peak, and a low-input-impedance meter loads the cell and depresses the voltage it is measuring.

Available options — report which one you used

  • Multi-channel data logger — e.g. Keithley 2700, Agilent/Keysight 34970A, Campbell Scientific. Report channel count and interval.
  • DAQ card + custom software — e.g. NI USB-6000 series with LabVIEW/Python. Report resolution and whether the acquisition code is available.
  • Microcontroller logger (Arduino / Raspberry Pi) — Increasingly common and entirely acceptable — report ADC resolution and input impedance, which are the real limits.
  • Potentiostat's own acquisition — Highest fidelity; note the channel count constraint on replicates.

Temperature + pH control hardware

high

Whether the condition was controlled or merely observed. An incubator holding 30 ± 0.5 °C and an unheated room 'at about 25 °C' are different experiments, and the second explains a lot of seasonal variance in the literature.

Available options — report which one you used

  • Incubator / climate chamber — Active control; report setpoint and tolerance.
  • Thermostatted water bath or jacket — Direct reactor-side control; the tightest option for kinetic work.
  • Temperature-controlled room — Ambient control; report the actual measured range, not the nominal setpoint.
  • Uncontrolled ambient — Acceptable, but the observed range must be reported — it is often the largest uncontrolled variable in the study.
  • pH controller / pH-stat with dosing — Active pH hold; report setpoint, titrant and dosing volume, since the titrant adds ions that change conductivity.

Reactor hardware specification

high

Body material, port layout, gasket and sealing decide oxygen ingress, dead volume and whether a cell can be rebuilt from the paper. Acrylic and glass differ in oxygen permeability, which shows up directly in coulombic efficiency.

Available options — report which one you used

  • Machined acrylic / PMMA block — The most common custom build; note oxygen permeability relative to glass.
  • Glass (H-cell, bottle) — Low oxygen permeability and easy to sterilise; limited geometry flexibility.
  • 3D-printed body — Report material and print parameters — and share the CAD or STL, which makes the reactor genuinely reproducible.
  • Commercial reactor platform — Cite model and supplier; the cleanest reproducibility case since geometry is fixed.
  • Field / in-situ deployment — For sediment and benthic systems; report the physical mounting, deployment depth and site coordinates.

Comprehensive 18-parameter view

Importance-weighted scoring across reactor geometry, materials, substrate, biology, operating conditions, and performance metrics. Use this for a full reproducibility audit beyond the five-parameter minimum.

Cell Geometry

Electrode Spacing

Cell GeometryCriticalcm

Distance between anode and cathode electrodes (cm). Controls ohmic resistance and mass-transport boundary layer thickness; the single largest determinant of power density variance across papers in the abstract's correlation analysis (r = 0.37).

Typical range: 1 – 20 cm

1 reference
  • Liu & Logan 2004 (Environ Sci Technol 38:4040)
Current status: Missing

Electrode Surface Area

Cell GeometryCriticalcm²

Projected geometric area of the anode and cathode (cm²). Required to convert raw current to areal current density. Without it, any power-density claim is a unit-less number — cross-study comparison becomes impossible.

Typical range: 1 – 1,000 cm²

1 reference
  • Logan 2008 (Microbial Fuel Cells, Wiley)
Current status: Missing

Reactor Volume

Cell GeometryHighL

Total working volume of the bioelectrochemical system. Needed for volumetric power density calculation if that's the reported normalization basis.

Typical range: 0.01 – 1,000 L

1 reference
  • Logan et al. 2006
Current status: Missing

Circuit

External Resistance

CircuitCriticalΩ

Load resistance across the circuit (Ω). Sets the operating point on the polarization curve. Maximum power is reported at R_ext = R_internal; papers omitting R_ext typically also omit polarization sweeps, making the reported power point uncomparable.

Typical range: 10 – 10,000 Ω

1 reference
  • Logan et al. 2006 (Environ Sci Technol 40:5181)
Current status: Missing

Methodology

Measurement Method

MethodologyCritical

How power / current were measured: data-logger vs handheld multimeter, sample interval, voltage-sweep protocol if a polarization curve was run. 8% of inter-study variance in the abstract's correlation analysis. Without it, a 26 mW/m² instantaneous reading is indistinguishable from a 26 mW/m² 24-hour average — a 4× real-world performance gap masquerading as agreement.

1 reference
  • Logan et al. 2019 (Nat Rev Microbiol 17:307)
Current status: Missing

Unit Normalization Basis

MethodologyCritical

Whether power is normalized to electrode area (mW/m²), reactor volume (W/m³), or anode mass (W/kg). The abstract's headline finding: power density spans five orders of magnitude (CoV 1,285%) largely because authors silently change normalization basis. Stating it explicitly is the single highest-leverage reporting improvement (most extreme rows in the corpus collapse onto a ~10× range when re-normalized).

1 reference
  • Schröder 2007 (Phys Chem Chem Phys 9:2619)
Current status: Missing

Operating Conditions

Operating Temperature

Operating ConditionsHigh°C

Ambient or controlled-bath temperature. Drives the Arrhenius factor on biofilm kinetics and ohmic resistance.

Typical range: 15 – 35 °C

1 reference
  • Liu et al. 2005
Current status: Missing

Anolyte / Catholyte pH

Operating ConditionsHigh–

pH of each chamber. Sets Nernst potential offset + biofilm health envelope.

Typical range: 6.5 – 8.5

1 reference
  • Torres et al. 2008
Current status: Missing

Hydraulic Retention Time

Operating ConditionsHighh

For continuous-flow systems, the time a unit volume spends in the reactor (h). Identified in the abstract's analysis (8% impact) as the second most consequential underreported parameter after electrode spacing.

Typical range: 1 – 72 h

1 reference
  • Logan 2008
Current status: Missing

Materials

Anode Material

MaterialsHigh

Type and composition of the anode (carbon cloth, carbon felt, graphite brush, MXene-coated foam, etc.). Drives biocompatibility and electron-transfer kinetics.

1 reference
  • Wei et al. 2011
Current status: Missing

Cathode Material

MaterialsHigh

Type and composition of the cathode (Pt/C, MnO₂, air-cathode, biocathode). Dominates the oxygen-reduction overpotential.

1 reference
  • Rismani-Yazdi et al. 2008
Current status: Missing

Membrane / Separator

MaterialsHigh

Type of ion-exchange membrane or separator (Nafion 117, CEM, AEM, J-cloth, none). Sets ion transport resistance and crossover behaviour.

1 reference
  • Rozendal et al. 2006
Current status: Missing

Substrate

Substrate Type

SubstrateHigh

Carbon source: acetate, glucose, lactate, synthetic wastewater, real wastewater. Determines maximum theoretical electron yield and biofilm community.

1 reference
  • Logan 2009
Current status: Missing

Substrate Concentration

SubstrateHighg/L

Initial concentration of the organic substrate (mg/L or g/L COD). Sets the Monod saturation regime.

Typical range: 0.1 – 10 g/L

1 reference
  • Liu et al. 2004
Current status: Missing

Electrolyte Composition

SubstrateMedium

Buffer composition + ionic strength of anolyte / catholyte. Sets solution conductivity which appears in the ohmic resistance.

1 reference
  • Lovley 2006
Current status: Missing

Biological

Inoculum Source

BiologicalMedium

Source + type of microbial inoculum (anaerobic digester sludge, soil, defined pure culture). Drives biofilm community structure.

1 reference
  • Logan & Regan 2006
Current status: Missing

Startup Period

BiologicalMediumweeks

Time required for system startup + biofilm establishment (weeks). Without it, reported peak performance can be a transient overshoot vs steady state.

Typical range: 1 – 8 weeks

1 reference
  • Liu & Logan 2004
Current status: Missing

Performance

Coulombic Efficiency

PerformanceMedium%

Fraction of electrons recovered as current vs. those theoretically available from substrate oxidation. Closes the energy-balance loop alongside power density.

Typical range: 10 – 90 %

1 reference
  • Logan et al. 2006
Current status: Missing

Comprehensive (18-param)

0%

Not yet comparable

Fewer than 60% of criteria reported. The reported power / current numbers are not safely cross-comparable against the corpus — context is missing.

Reported
0
Omitted
0
Missing
18
Weighted score
0/55

Citing this tool in your research

Frons, S. (2026). MESSAI: An open-source platform for cross-study
comparison and reproducibility in microbial electrochemical systems.
EU-ISMET 2026. https://messai.io/tools/reproducibility