Suitability and optimization of FAO’s small-scale aquaponics systems for joint production of lettuce (Lactuca sativa) and fish (Carassius auratus)

Metadata

Cite key: perezurrestarazuSuitabilityOptimizationFAO2019
Item type: Journal Article
Authors: L. Pérez-Urrestarazu, J. Lobillo-Eguíbar, R. Fernández-Cañero, V.M. Fernández-Cabanás
Affiliation: Area of Agro-Forestry Engineering, Universidad de Sevilla, ETSIA, Seville, Spain (Pérez-Urrestarazu); Departamento de Ciencias Agroforestales, Universidad de Sevilla, ETSIA, Seville, Spain (Lobillo-Eguíbar, Fernández-Cañero, Fernández-Cabanás)
Journal: Aquacultural Engineering 85 (2019) 129-137
Date: 2019 (received 6 Feb 2019, accepted 16 Apr 2019, online 18 Apr 2019)
Date added: [not reported] — no zotero-export.csv “Date Added” value distinguishable for this single-entry item
DOI: 10.1016/j.aquaeng.2019.04.001
Funding: “This study was financed by the University of Seville (Spain). Ayudas de innovación y mejora docente (II Plan Propio de Docencia. Convocatoria 2013-2017; Ref. 1.10).”
URL: https://doi.org/10.1016/j.aquaeng.2019.04.001
PDF: Pérez-Urrestarazu et al. - 2019 - Suitability and optimization of FAO's small-scale .pdf

Opinion

A genuinely useful head-to-head of three FAO/Somerville-manual hydroponic subsystem designs (NFT, floating raft, vertical felt living wall) against a hydroponic control, run twice with different lettuce cultivars and fish sizes — exactly the kind of “which sub-system should I build” comparison the field needs more of. The honest caveat is scale: each design is a single, unreplicated physical unit per test, so the striking productivity gap (NFT beating floating raft 5-25x, and vertical felt losing to everything by 1-2 orders of magnitude) rests on n=1 systems with plants as the only pseudo-replicate, not on independently replicated aquaponic installations. The fish side is almost an afterthought by comparison — goldfish grew and survived well in all three designs with no significant design effect reported, so this paper’s real contribution is a hydroponic-engineering result wearing an aquaponics label. The vertical felt system’s failure is well diagnosed (uneven water/nutrient delivery through the pockets, light limitation from the wall’s own geometry, algae fouling) and is a useful negative result for anyone tempted to bolt a living wall onto a RAS. Table 6’s SGR and FCR unit labels are internally inconsistent with their own stated formulas — verified by recomputation, not just asserted — which anyone reusing these numbers should know before citing them at face value.

Abstract

Aquaponics is a developing technique that combines the simultaneous production of plants (hydroponics) and fish (aquaculture). With it, the use of resources (i.e., water, nutrients, land) is reduced whilst at the same time minimising residues’ discharge to the environment. Among its benefits, it allows the production of healthy vegetables and fish in reduced spaces by means of small-scale systems. In this work, three of them based on FAO models with different hydroponic subsystems (nutrient film technique -NFT-, floating raft, and vertical felt) are tested to produce lettuce (Lactuca sativa) and goldfish (Carassius auratus). Water parameters as well as the growth of plants and fishes were monitored in two different production cycles. The hydroponic subsystem that outperformed the best was the NFT, both in terms of crop production and water consumption. All systems showed similar results in fish production. Further research is needed to corroborate the outputs obtained when using other combinations of plants and fishes. Small-scale aquaponic systems are particularly interesting for self-production and even more so in urban environments with reduced available space.

Summary

The authors built three small-scale aquaponic systems (SAS1-SAS3) at the University of Seville, each modified from FAO’s Somerville et al. (2014) small-scale aquaponics manual and differing only in hydroponic subsystem: SAS1 used nutrient film technique (NFT, five 3 m PVC pipes), SAS2 used a floating raft (two 0.48 m3 raft tanks), and SAS3 used a vertical felt living wall (a 2.4 x 2 m sloped felt structure with expanded-clay pockets). Each system had its own single 1 m3 fish tank stocked with goldfish (Carassius auratus) and was compared against a non-aquaponic hydroponic/perlite control (Ctrl, Hoagland solution). The comparison was run twice as sequential seasonal tests — a 44-day spring/summer 2016 cycle growing Lollo rosso and Lollo bionda lettuce, and a 54-day autumn/winter 2016-17 cycle growing romaine lettuce with larger fish — with water pH/EC/nitrate, plant morphometrics and yield, and fish growth/survival/FCR/SGR monitored throughout. NFT (SAS1) substantially outproduced both other aquaponic designs in both tests (e.g. 3.38 vs 0.98 vs 0.24 kg/m2 in test 2) and matched or exceeded the hydroponic control in the second test, while the vertical felt wall (SAS3) performed worst in both tests, attributed to uneven water/nutrient delivery to the felt pockets, light limitation from the wall’s geometry, and algae fouling that increased water loss. NFT also used the least water per kg of lettuce produced. In contrast, fish production (survival >90% in all systems, similar weight gain and FCR ranges) showed no clear effect of hydroponic subsystem design — the paper’s central finding is that hydroponic engineering choice, not the aquaponic principle itself, drove the crop-yield differences observed.

Experiment data

Location: 65 m2 greenhouse, School of Agricultural Engineering (ETSIA), University of Seville, Seville, Spain (37.3518 N, -5.9368 W)

Design: Three small-scale aquaponic system (SAS) designs modified from FAO/Somerville et al. (2014) guidelines, differing only in hydroponic subsystem (SAS1=NFT, SAS2=floating raft, SAS3=vertical felt living wall), each a single unreplicated physical unit (one 1 m3 IBC fish tank + one hydroponic subsystem), compared against a non-aquaponic Ctrl (perlite pots + Hoagland solution); run as two sequential seasonal test cycles with different lettuce cultivars and larger fish in test 2. Statistical comparisons (ANOVA/Kruskal-Wallis, Tukey/Games-Howell) use individual plants within each system as the analysis unit — see Extraction notes on the quasi-experiment classification.

Replicates / n: 1 physical AP system per treatment per test (unreplicated at system level); 20 plants/treatment in test 1 (10 Lollo rosso + 10 Lollo bionda pooled), 24 plants/treatment in test 2 (romaine)

Duration: Test 1: 9 May-22 Jun 2016 (44 d); Test 2: 8 Nov 2016-11 Jan 2017 (54 d)

Organisms: Goldfish (Carassius auratus) / Lettuce (Lactuca sativa) (Lollo rosso + Lollo bionda, test 1; romaine, test 2)

Statistics: One-way ANOVA (SPSS 24); Shapiro-Wilk normality test; Kruskal-Wallis when non-normal; HSD Tukey and Games-Howell post-hoc tests; significance at P<0.05

Lettuce productivity (kg/m2): Test 1: SAS1(NFT) 0.49, SAS2(raft) 0.09, SAS3(felt) 0.02, Ctrl 1.24. Test 2: SAS1 3.38, SAS2 0.98, SAS3 0.24, Ctrl 2.84. NFT was the only aquaponic design to match/exceed Ctrl (test 2).

Water consumption (DWC, L/day): Test 1: SAS1 21.6, SAS2 15.1, SAS3 23.3. Test 2: SAS1 13.3, SAS2 15.4, SAS3 32.2. NFT (SAS1) used least water in test 2; floating raft (SAS2) least in test 1.

Fish production: Survival 92-98% (test 1), 100-104% (test 2, >100% from in-tank reproduction in SAS1/SAS2); Feed Conversion Rate (FCR) 2.88-7.65 (test1), 2.95-3.29 (test2); no significant design effect reported on fish growth/survival.

Hydroponic subsystem effect on lettuce yield

This paper: NFT (SAS1) produced 0.49 kg/m2 in test 1 and 3.38 kg/m2 in test 2, outperforming both the floating raft (SAS2: 0.09, 0.98 kg/m2) and vertical felt wall (SAS3: 0.02, 0.24 kg/m2) in both tests by a wide margin, and matching/exceeding the hydroponic control (Ctrl: 1.24, 2.84 kg/m2) in test 2. The authors attribute NFT’s advantage partly to higher water temperature in the pipes under direct sun favouring nitrification and nutrient uptake, and partly to good root-zone aeration. Floating raft’s weaker performance is attributed to deficient root-zone aeration (dissolved oxygen adequate for fish but probably suboptimal for plant roots). Vertical felt’s poor performance (worst in both tests) is attributed to insufficient/uneven water and nutrient delivery to the felt pockets despite continuous wetting, reduced light influx from the wall’s own geometry, and algae proliferation on the felt causing competition, emitter blockage and extra water loss.

Compared with:

todo Lennard and Leonard 2006 — 21-day gravel-bed/floating-raft/NFT comparison with Murray cod and Green Oak lettuce; gravel bed best, NFT least productive, opposite ranking to this paper’s NFT result; also found no hydroponic-subsystem effect on fish growth, corroborating this paper’s fish-side finding. (p.135, secondary)
See graberAquaponicSystemsNutrient2009 — cited for the general observation that AP systems often show low K/Fe/Ca, motivating this paper’s foliar K2SO4 and chelated-iron supplementation. (p.130)
See nozziNutrientManagementAquaponics2018 — floating-raft lettuce with varied nutrient supplementation, 4-6.13 kg/m2 vs. 5.65 kg/m2 hydroponic control; used as a yield benchmark for this paper’s much lower floating-raft (SAS2) results. (p.133, secondary)
See delaideLettuceLactucaSativa2016 — NFT aquaponic (tilapia) vs. hydroponic lettuce, no significant yield difference (80.55/35.72 vs. 98.17/39.64 g/plant across two trials); also appears in Table 7 as an NFT yield benchmark (35-81 g/plant). (p.133, secondary)
todo Maucieri et al. 2018 — micro floating-raft AP system, Lactuca sativa cv. Bionda Ricciolina di Trieste, 0.4 kg/m2; used as a direct productivity benchmark for this paper’s floating-raft (SAS2) result. (p.133, secondary)
todo Seawright et al. 1998 — romaine lettuce cv. Jericho in floating rafts, 2 kg/m2; used as a benchmark exceeding this paper’s SAS2 romaine result (0.98 kg/m2). (p.133, secondary)
todo Johnson et al. 2017 — grow-bed Lollo lettuce (0.258 kg/m2, significantly below a 0.47 kg/m2 pure-hydroponic control) and grow-bed romaine (0.53 vs. 0.49 kg/m2, ns); cited alongside this paper’s own cultivar comparison (green/Lollo vs. red cultivars). (p.133, secondary)
todo Geisenhoff et al. 2016 — NFT lettuce cv. Verônica with intensive tilapia, 86-96 g/plant, 2.27 kg/m2; Table 7 NFT benchmark. (p.133, secondary)
See jordanYieldLettuceGrown2018 — NFT lettuce cv. Pira verde on different substrates, 104-200 g/plant, 2-4 kg/m2; Table 7 NFT benchmark. (p.133, secondary)

Fish production and design effect

This paper: All three SAS designs produced broadly similar goldfish outcomes: survival 92-98% (test 1) and 100-104% (test 2, with reproduction pushing SAS1/SAS2 above 100%); specific growth rate 0.19-0.56 (test 1) and 0.37-0.41 (test 2, tighter range as fish were larger/older); FCR 2.88-7.65 (test 1, SAS2’s 7.65 the outlier) and 2.95-3.29 (test 2). The paper explicitly frames this as the inverse of the crop-side result: “unlike crop production results, in terms of fish production, SAS3 worked slightly better than the other SAS,” attributed to more frequent water replenishment diluting metabolites in that system, rather than any structural advantage of the vertical-felt design for fish.

Compared with:

todo Sarkar and Upadhyay 2012 — goldfish under varied photoperiod, FCR 1.5-4.18, SGR 1.02-2.85; this paper’s FCR range (2.88-7.65) partly overlaps, its SGR values (0.19-0.56, unscaled per the WARN-MINOR unit note) read far lower on a %/day basis. (p.134, secondary)
todo Bandyopadhyay et al. 2005 — goldfish fed varied local agro-produce diets, FCR 1.94-2.61, SGR 1.56-2.13; cited as a comparison point for both FCR and SGR. (p.134, secondary)
todo Shete et al. 2013 — goldfish + spinach aquaponics with varied water-circulation periods, FCR 4.5-4.8, SGR 1.30-1.44 (larger starting fish); FCR range overlaps this paper’s. (p.134, secondary)
todo Souto et al. 2013 — goldfish protein:energy ratio study, FCR 3.74-13.06, SGR 0.15-0.45; the only cited study with SGR values in the same low range as this paper’s (0.19-0.56), suggesting the unit/scale may be shared rather than this paper being an outlier. (p.134, secondary)

Water consumption and footprint

This paper: Daily water consumption (DWC) varied by design and test: NFT (SAS1) was most efficient in test 2 (13.3 L/day, a 38% reduction from test 1’s 21.6 L/day, attributed to seasonal temperature differences), while floating raft (SAS2) was most efficient in test 1 (15.1 L/day). Vertical felt (SAS3) was consistently worst (23.3, 32.2 L/day), attributed to direct evaporation from the exposed felt surface and, in test 2, algae-driven emitter blockage and spillage. All water-replenishment rates (0.76-2.48% of system volume per day) fell in the lower half of the 0.5-10% range the authors cite as typical for aquaponics. The paper’s own (author-computed, not this extraction’s) water-footprint-per-kg-of-produce figures make the design gap stark: as low as 41 L/kg (SAS1, romaine, test 2) versus 1950-25,200 L/kg for SAS2/SAS3 in test 1.

Compared with:

todo Mekonnen and Hoekstra 2011 — conventional lettuce water footprint ~237 L/kg; used as the paper’s baseline comparator for its own SAS/kg-produce figures. (p.135, secondary)
todo Barbosa et al. 2015 — lettuce water footprint ~250 L/kg (land-based), ~20 L/kg (hydroponic); cited alongside Mekonnen and Hoekstra as a second baseline. (p.135, secondary)
See delaideEffectWastewaterPikeperch2019 is a different, later Delaide paper already in this vault; the paper actually cited here for a 50/50 crop/fish water-need split and a 100-250 L/kg range is todo Delaide et al. 2017 (PAFF Box small-scale aquaponic system, J. Aquac. Eng. Fish. Res. 78) — NOT yet in the vault, kept separate to avoid conflating the two Delaide papers. (p.135, secondary)
todo Silva et al. 2018 — dynamic root floating technique reduced electric power consumption 11.4% in aquaponics; cited as an example energy-optimisation strategy, not a water-footprint comparator. (p.136, secondary)

Linked claims

Citations to chase

Extraction notes

TYPE CLASSIFICATION JUDGMENT CALL: recorded as quasi-experiment, not experiment. The paper has defined treatments (3 SAS hydroponic-subsystem designs vs. a hydroponic control) and formal statistical testing (one-way ANOVA/Kruskal-Wallis, Tukey/Games-Howell post-hoc, P<0.05), which superficially fits experiment. However, per SCHEMA.md Part 2 decision rule 2, each SAS treatment is exactly ONE physical system (one IBC fish tank + one hydroponic subsystem) per test cycle — there is no independent system-level replicate for any treatment anywhere in the paper, and “randomised” is never used for treatment-to-system assignment. The reported significance tests are computed using individual plants within that single system as pseudo-replicates, not independent replicate systems. This is a stricter case than pantanellaAquaponicsHydroponicsProduction2012 (3 independent replicate systems per treatment per crop, judged experiment despite also lacking the word “randomised”): here there is no system-level replication at all, so quasi-experiment was judged the better fit. Flagging this explicitly since it is a judgment call that another reader might resolve differently.

Trial structure: Six trials.csv rows (3 SAS designs x 2 sequential test cycles), each paired with that test’s own hydroponic control (T1-T3 share the test-1 Ctrl; T4-T6 share the test-2 Ctrl, a different physical/numerical control, not a repeat of T1-T3’s). Test 1 grew a 50/50 pooled mix of Lollo rosso and Lollo bionda per SAS (20 plants); test 2 grew pure romaine (24 plants) with larger, older fish. Confirmed via recomputation that Table 4/5’s per-plant averages (FWP) equal TFW divided by plant count for every SAS in both tests, so the pooled-cultivar test-1 figures are genuine per-plant averages across both cultivars combined, not a single-cultivar subset.

WARN-CHECK NO3-N vs. NO3- (ion) basis, all six trials, Table 1 (p.132). Table 1 explicitly labels this parameter “NO3- (ppm)” — the nitrate ion, not nitrate-nitrogen — and the paper never states or performs a conversion to NO3-N. SCHEMA.md’s NO3-N column specifically wants nitrate-nitrogen, and NO3 vs. NO3-N differ by a factor of 4.43 (SCHEMA.md’s own worked example of this exact ambiguity). The Methods only names the instrument (“an RQflex 10 plus (MERK, Darmstadt, Germany)”), and Merck RQflex nitrate test strips are sold in both NO3- and NO3-N calibrations, so the basis cannot be confirmed from the paper. Recorded the paper’s own printed value unconverted in every trial’s NO3-N cell, with the basis explicitly flagged as NO3- (ion) rather than NO3-N. Affects all six trials’ NO3-N cells. Added to REVIEW.md by the batch merge step.

WARN-MINOR FCR unit label, all six trials (Table 6, Eq. 3, p.131-132). Table 6 and the text label FCR as ”(%)” (“The FCR was between 3 and 4%, except for the case of SAS2 in test 1 (7.6%)”), but recomputing FCR = FI/(TWf-TW0) from the paper’s own stated total feed intake (742 g test 1, 3247 g test 2, identical across all three SAS within each test) against Table 6’s weight-gain figures reproduces every one of Table 6’s six FCR values exactly as a dimensionless ratio (e.g. 742/191=3.885≈3.88; 742/97=7.649≈7.65), not a percentage. The ”%” label is a units mislabel, not a distinct measured value; the recorded FCR cells use Table 6’s own printed ratio (matching SCHEMA.md’s “ratio, dimensionless” FCR definition) unchanged. Recomputation is verification evidence only, per SCHEMA.md’s “recomputing to check is not derivation” allowance.

WARN-MINOR SGR unit label, all six trials (Table 6, Eq. 1, p.131-134). Table 6 labels SGR as “(g·day-1)” and gives Eq. 1 as SGR=(ln AWFt - ln AWFt-1)/Δt with no x100 factor shown, which is dimensionally a fraction/day, not a weight/day quantity, and does not match SCHEMA.md’s %/day SGR unit (which implies a x100 factor the stated formula omits). Recorded Table 6’s printed values unchanged (test1: 0.36/0.19/0.56; test2: 0.37/0.41/0.40, matching the text’s stated ranges exactly) without rescaling, since SCHEMA.md’s no-derivation rule prohibits correcting a paper’s own reported number even when its stated unit looks inconsistent with its own formula. Flagged so the mismatch is visible before anyone cites these as %/day figures.

Fish reproduction note: T4/T5 (test 2, SAS1/SAS2) show survival rate >100% (104%, 102%) because the Results text states fish “even reproduced in test 2 (SAS1 and SAS2)”; SVR is defined as final/initial fish count, so in-tank reproduction pushed the ratio above 100%. Not an error — recorded as stated, with the explanation in each affected trial’s Experimental Remarks.

Fish size initial/final (all trials): Table 6 gives per-SAS AWF0/AWFf values (used in the cells); Methods 2.2 separately states single rounded averages across all three SAS combined (30 g / ~35 g test 1; 53.3 g / ~74 g test 2). Not a contradiction (different aggregation level, both explicitly stated as such) but noted for anyone reconciling the two figures.

[not reported]/NR fields, grouped:

Fish: feed N/P/K composition beyond crude protein (24.4%); per-fish weight gain (AWF0/AWFf both given but their difference never itself stated — NOT DERIVED); Fish Category beyond the paper’s own “ornamental fishes” framing.
Water: Water type/Water classification (no categorical label used by the paper); TAN/NH4-N, NO2-N (not measured — only pH, EC, nitrate and temperature were monitored); Dissolved Oxygen (equipment described, no trial-mean value given); pHOptimal (no stated pH target/setpoint for the AP loop, only the observed range); single combined system water volume (only fish-tank volume of 1 m3 is stated as one figure; downstream component volumes given individually, never summed by the paper — NOT DERIVED); Water temperature as a single combined mean (only separate Tmax/Tmin trial means given, recorded as a range).
Plant: Days Plant after transplant (fish-trial duration and transplant offset both given but their difference never itself stated — NOT DERIVED); Plants/m2 (plant counts and component dimensions given individually but no single system growing-area figure stated — NOT DERIVED; kg/m2 productivity used directly in AP/HYD instead); SPAD, Tissue nitrate AP/HYD (not measured at all); Plant Category (no categorical term applied by the paper beyond specific cultivar names).
FUE AP/HYD, WUE, Remineralization, pH Buffers, Climate control, Artificial Lighting, Average room Temperature: none addressed by the paper in these specific terms.

No plant tissue analyte data (plant.csv is header-only): this paper reports only plant morphometrics (leaf count, height, shoot diameter, fresh/dry weight, dry-matter content) that map directly onto existing trials.csv columns. No biochemistry (pigments, phenolics), mineral (leaf tissue elemental panel), microbiology, or proximate-composition data was measured or reported anywhere in the paper, so plant_measurements.csv’s four categories have nothing to receive — out/perezurrestarazuSuitabilityOptimizationFAO2019.plant.csv contains only the header row.

NO COLUMN items (kept in each trial row’s Experimental Remarks, not discarded): Table 1 EC initial/final snapshot values (distinct from the trial-mean EC used in the EC cell); total water consumption (TWC, L) per test; the paper’s own author-computed water-footprint-per-kg-produce figures (Discussion, p.135, a secondary calculation, not a trial-mean measurement); productivity-per-kg-fish-food figures (Results, p.132); Table 2/3 shoot diameter (a third plant morphometric alongside height/leaf count, no dedicated schema column); per-system construction cost (~1000 EUR each, Discussion p.136); Table 7’s literature-comparison figures (secondary, other studies’ data).

Tags judgment call: Tagged Meta/Fish/Goldfish (reusing the existing vault facet from abbeyBasilOcimumBasilicum2022) since Carassius auratus is the sole aquaculture species studied and reared throughout. Tagged Meta/Plant/Lettuce (reusing the existing widely-used vault facet) even though three distinct cultivars (Lollo rosso, Lollo bionda, romaine) were used, consistent with how other lettuce-cultivar papers in the vault (e.g. pantanellaAquaponicsHydroponicsProduction2012) are tagged at the species level, not per cultivar. Meta/Type/Quasi-experiment reused from goddekComparisonLactucaSativa2018/delaideLettuceLactucaSativa2016/graberAquaponicSystemsNutrient2009/lennardComparisonPlantGrowth2019/nozziNutrientManagementAquaponics2018, all already using this exact facet.

New wikilink targets introduced: L. Pérez-Urrestarazu and V.M. Fernández-Cabanás reused exactly as already spelled in fernandezcabanasComparativeAnalysisHorizontal2020 (per Zotero export’s own spelling, “Cabanás” with acute accent, not Crossref’s “Cabañas” with tilde — CLAUDE.md treats the Zotero export as authoritative over Crossref for author spelling when both are present, and the existing vault note already established this exact spelling as the wikilink target). J. Lobillo-Eguíbar and R. Fernández-Cañero are new to the vault (no existing notes found). Goldfish (Carassius auratus), Lettuce (Lactuca sativa), Feed Conversion Rate (FCR) reused in their existing canonical forms. Four new claim-note wikilinks introduced (see Linked claims) — checked against the vault, no existing fragmented variants found.

PDF quality: Clean text layer throughout (9 pages, standard two-column Elsevier typesetting), fully extractable via pdftotext/pdfplumber. One extraction pitfall worth recording: pdftotext -layout badly mis-columns Table 6 (SGR/AGR/FCR/SVR rows appear transposed/interleaved under wrong headers) because of the table’s stacked two-line column headers; the values were recovered correctly by re-extracting with pdfplumber’s reading-order text mode (no -layout) and cross-verified against every narrative figure quoted in the Results/Discussion text (DWC reduction percentages, FCR exception for SAS2, SGR ranges, SVR reproduction note) — every cross-check matched exactly, giving high confidence in the final Table 6 values used here. Table 1 extracted cleanly on the first attempt via the same non-layout method.

Severity tally: 0 BLOCK, 0 MATERIAL, 1 CHECK (NO3-N vs. NO3- ion basis, recurring across all 6 trials but a single systemic issue), 2 MINOR (FCR unit label, SGR unit label) → quality: ok (0 BLOCK, ≤2 MATERIAL).

Source: Pérez-Urrestarazu et al. - 2019 - Suitability and optimization of FAO's small-scale .pdf

Data Tables

Structured data extracted from this paper into the vault's trials.csv / plant_measurements.csv datasets. Fields the paper didn't report are omitted. Download the full datasets (measurements).

Trial Parameters

perezurrestarazuSuitabilityOptimizationFAO2019-T1

Fish

Field	Value
Fish	Goldfish (Carassius auratus)
Fish Category	Ornamental fish (goldfish are explicitly described as ‘ornamental fishes’ with a higher market value than other AP species, p.130)
Initial Stock density	1.6 kg/m3 (test 1, stated as a target/achieved stocking density, Methods 2.2)
FCR	3.88
SGR	0.36
Protein	24.4
% of body weight	1-2 (ramped from 1% of total fish weight at test start to 2% by test end, incremented every fortnight, Methods 2.3)
Fish size initial	30.0
Fish size final	34.8
Feed routine	Twice daily (Prodac International S.r.l., Cittadella, Italy, pond-stick feed)
Feed regime	1% of total fish weight at start, incremented fortnightly to 2% of total weight (Methods 2.3)
Total Feed (kg)	0.742
Fish biomass created (kg)	0.191
Fish survival rate	96
Fish trial duration (days)	44

Water

Field	Value
Water recycle	41.7-83.3 (converted from 2500-5000 L/h, Eheim Compact+5000 pump, Methods 2.1)
Water volume in the system	1000 (fish tank only, 1 m3 IBC container, x3 SAS identical; downstream component volumes given individually, not summed by the paper — see Experimental Remarks)
Daily Water exchange rate	1.66
Aq pH	7.9 ± 0.2
EC	0.329 ± 0.055
Water temperature	Tmin 23.0 ± 3.6 / Tmax 27.0 ± 2.5 (degC, Table 1; no single combined trial mean reported, see Experimental Remarks)
NO3-N	38.6 ± 21.3 (NO3- as ion per paper’s own Table 1 label, NOT confirmed as NO3-N — see WARN-CHECK in Experimental Remarks)

Plant

Field	Value
Plant	Lettuce (Lactuca sativa L.)
Details	Lollo rosso (LR) + Lollo bionda (LB), 10 plants each (20 total); pre-cultivated in seedbed trays 21 days; transplanted 2 weeks after fish stocking; test 1 (9 May-22 Jun 2016, 44 d)
Plant height	23.7 ± 3.5
Leaf count	17.2 ± 1.1
Plant fresh weight	58.8 ± 6.3
Plant dry matter	13.0 ± 2.2

System & Setup

Field	Value
System type	Nutrient film technique (NFT)
Media Details	5 x 3 m PVC pipes, 0.11 m diameter, spaced 0.11 m, 1% slope; 12 holes/pipe (0.05 m dia.) spaced 0.25 m; mechanical filter + biofilter (2 x 0.21 m3 cylindrical tanks, 1000 standard + 600 handmade PVC bioballs); Eheim Compact+5000 pump, 80% water to fish tank / 20% to NFT pipes (Methods 2.1)
Biological system already in use	N (Each 4-week fishless nitrifying-bacteria cycling period preceded BOTH tests (Biodigest Pro inoculant + artificial ammonia dosing, Methods 2.3) — system was newly cycled at the start of each test, not carried over from a prior operating state)
Air supplement	Y (Eheim Air pump 400 (400 L/h) for fish-tank DO; BOYU ACQ-903 air compressor for supplementary fish/raft-tank aeration (Methods 2.1))
Iron supplemented	Y (0.1 L chelated iron solution (1%, EDDHA Sequestrene 138 Fe) added directly to water, Fridays every fortnight (SAS1/SAS3), Methods 2.3)
Nutrient supplemented	Y (K2SO4 (1.5%) foliarly applied twice weekly (Mondays and Fridays) via manual sprayer to alleviate observed K/Fe/Ca insufficiency symptoms (Methods 2.3))
Equipment	Eheim Air pump 400 (Germany, 400 L/h) for fish-tank aeration; Eheim Compact+5000 pump (2500-5000 L/h adjustable) for biofilter-to-fish-tank/hydroponic recirculation; BOYU ACQ-903 air compressor (Raoping Guangdong, China) for supplementary fish/raft tank aeration; TFA (Germany) maximum-minimum thermometer; Crimson Instruments GLP 22 pH-meter; Crimson Instruments EC-Meter BASIC 30; Merck RQflex 10 plus reflectometer (nitrate); SPSS 24 statistical package
Control Parameters	Water pH, EC and nitrate measured periodically with pH-meter/EC-meter/RQflex reflectometer; Tmax/Tmin logged daily via max-min thermometer; daily water consumption (DWC) measured daily to derive total water consumption (TWC) and water replenishment rate (WR); fish counted/weighed every fortnight (test1) or every 3 weeks (test2); feed ration set at 1% of total fish weight at test start, incremented fortnightly up to 2% of total weight; foliar K2SO4 (1.5%) applied twice weekly and chelated iron (EDDHA Sequestrene 138 Fe, 1%) applied fortnightly to address nutrient insufficiency; plant leaf count/height/shoot diameter recorded weekly
Combination	Goldfish (Carassius auratus, ornamental fish) and lettuce (Lactuca sativa: Lollo rosso + Lollo bionda cultivars in test 1, Romaine cultivar in test 2) compared across three small-scale aquaponic system designs modified from FAO/Somerville et al. (2014) guidelines — SAS1 (Nutrient Film Technique), SAS2 (floating raft), SAS3 (vertical felt living wall) — each run as a single unreplicated physical unit, against a non-aquaponic hydroponic/substrate control (Ctrl: perlite pots + Hoagland nutrient solution), across two sequential seasonal test cycles (spring/summer 2016 and autumn/winter 2016-17).

Site

Field	Value
Region	Europe
Country	Spain
Lat	37.3518
Long	-5.9368

Results & Statistics

Field	Value
Measured Unit	kg/m2 (yield/productivity); g (TFW/FWP/DWP); % (DMC, dry matter content); cm (height, diameter); count (leaves); mg/L=ppm (NO3-); dS/m converted from uS/cm (EC); degC (temperature)
Statistic Details	One-way ANOVA (SPSS 24); Shapiro-Wilk normality test; nonparametric Kruskal-Wallis ANOVA when data not normally distributed; HSD Tukey post-hoc test; Games-Howell post-hoc test when variances unequal; significance at P<0.05 (Methods 2.4)
Statistically analysed	Y
Replicates (n)	20 plants/treatment (10 Lollo rosso + 10 Lollo bionda, pooled; 1 physical AP system per treatment, not independently replicated — see Extraction notes / TYPE CLASSIFICATION note)
AP	0.49
HYD	1.24

Experimental Remarks: TRIAL DEFINITION: T1 = SAS1 (NFT), test 1 of 2, vs. Ctrl (hydroponic, perlite pots + Hoagland solution) of the SAME test cycle, recorded in the HYD-labelled cells (AP=0.49, HYD=1.24 kg/m2). Six aquaponic trial rows total (3 SAS designs x 2 sequential test cycles), each paired with that test cycle’s own hydroponic control (Ctrl), recorded in the HYD-labelled cells — Ctrl was NOT itself replicated or repeated identically between tests (different lettuce cultivar, different plant count, different productivity result each time), so T1-T3 share one Ctrl (test 1) and T4-T6 share a different Ctrl (test 2). Each SAS was built and operated as a SINGLE physical unit per test (one IBC fish tank + one hydroponic subsystem per treatment) — there is no independent system-level replicate of any SAS within a test; the individual lettuce plants (20/treatment in test 1 pooling 10 Lollo rosso + 10 Lollo bionda, 24/treatment in test 2, all Romaine) served as the pseudo-replicates for the reported ANOVA/Kruskal-Wallis/Tukey/Games-Howell comparisons across the four treatments (SAS1/SAS2/SAS3/Ctrl) within each test. | TYPE CLASSIFICATION JUDGMENT CALL: recorded as quasi-experiment. The paper has defined treatments (3 SAS designs vs. hydroponic control) and formal statistical testing (ANOVA/Kruskal-Wallis, Tukey/Games-Howell post-hoc, P<0.05), which would suggest ‘experiment’ under SCHEMA.md’s simple test. However, per SCHEMA.md Part 2 decision rule 2 (‘Randomised treatments with replication? Yes -> experiment. Treatments without randomisation or true replication -> quasi-experiment’), each SAS treatment consists of exactly ONE physical system (one IBC fish tank + one hydroponic subsystem) per test cycle — there is no independent system-level replicate for any treatment, and the word ‘randomised’ is never used for treatment-to-system assignment. The reported significance tests are computed on individual plants within that single system (pseudo-replication), not on independent replicate systems. This is a stricter case than pantanellaAquaponicsHydroponicsProduction2012 (which had 3 independent replicate physical systems per treatment per crop and was judged ‘experiment’ despite similarly lacking the word ‘randomised’); here there is no system-level replication at all, so ‘quasi-experiment’ was judged the better fit. | WARN-CHECK NO3-N vs NO3- (ion) basis. Table 1 reports this parameter explicitly as ‘NO3- (ppm)’ (nitrate ion, not nitrate-nitrogen); the paper never states or converts to NO3-N. SCHEMA.md’s NO3-N column specifically wants nitrate-nitrogen (mg/L), and NO3 vs NO3-N differ by a factor of 4.43 (SCHEMA.md’s own worked example of this exact ambiguity). Converting would require assuming the paper’s ‘NO3-’ is pure NO3 ion and dividing by 4.43 — an assumption not confirmable from the Methods (only ‘an RQflex 10 plus (MERK, Darmstadt, Germany)’ reflectometer is named, and Merck RQflex nitrate strips are sold in both NO3- and NO3-N calibrations). Recorded the paper’s own printed value UNCONVERTED, with basis explicitly named as NO3- (ion), not NO3-N — per SCHEMA.md CHECK convention (‘cell takes the value whose basis is clearest, with the basis named’). Flagged for REVIEW.md (batch merge step) so downstream use can decide whether to apply the x1/4.43 conversion. | WARN-MINOR FCR unit label. Table 6 and Eq. 3 label FCR as ‘FCR (%)’ and the text states ‘The FCR was between 3 and 4%, except for the case of SAS2 in test 1 (7.6%)’ (p.132), implying a percentage. Recomputing from the paper’s own stated inputs (Eq. 3: FCR=FI/(TWf-TW0); total feed intake FI=742 g in test 1 and 3247 g in test 2, identical across all three SAS within each test, per p.131-132) against Table 6’s total-weight-gain values reproduces Table 6’s FCR figures exactly as a DIMENSIONLESS RATIO, not a percentage: test1 SAS1 742/191=3.885 (~3.88 recorded), SAS2 742/97=7.649 (~7.65), SAS3 742/258=2.876 (~2.88); test2 (FI=3247) SAS1 3247/1040=3.122 (~3.12), SAS2 3247/1101=2.949 (~2.95), SAS3 3247/986=3.294 (~3.29). All six recomputed values match Table 6 digit-for-digit, confirming Table 6’s numbers are the ratio FI/gain, not that ratio x100. The paper’s own ’%’ label and text (‘3 and 4%’, ‘7.6%’) are therefore a units mislabel, not a distinct measured value. Recorded the ratio exactly as printed in Table 6 (matches SCHEMA.md’s FCR definition, ‘ratio, dimensionless’); no cell value is affected, only the unit label. Recomputation shown here as verification evidence only, per SCHEMA.md’s ‘recomputing to check is not derivation’ allowance — the recorded FCR value is Table 6’s own printed number, not a value newly computed by this extraction. | WARN-MINOR SGR unit label. Eq. 1 and Table 6 label SGR as ‘SGR (g.day-1)’, but the formula given (SGR = (lnAWFt - lnAWFt-1)/Delta t, no x100 factor shown) produces a dimensionless-per-day quantity, not a weight/day quantity, and does not match SCHEMA.md’s expected SGR unit of %/day (which would require a x100 factor the paper’s stated formula omits). Recorded Table 6’s printed values exactly as given (test1: SAS1 0.36, SAS2 0.19, SAS3 0.56; test2: SAS1 0.37, SAS2 0.41, SAS3 0.40 — matching the text’s ‘SGR values ranged from 0.19 (SAS2) to 0.56 (SAS3) in test 1 and were between 0.37 and 0.41 in the second’, p.134) without rescaling by 100, since SCHEMA.md’s no-derivation rule prohibits correcting/rescaling a paper’s own reported number even when its unit label looks inconsistent with the formula shown. Flagged so the mismatch between the paper’s stated unit and the SCHEMA.md column unit (%/day) is visible to anyone using this figure downstream. | Fish size initial/final: Table 6 gives per-SAS-specific AWF0/AWFf (e.g. test1 AWF0 30.0/30.6/30.2 g, AWFf 34.8/33.1/38.0 g for SAS1/SAS2/SAS3); Methods 2.2 separately gives single rounded overall figures (‘average weight of 30 g per fish’ stocked, ‘finishing with an average of 35 g’ for test1; ‘initial average weight of 53.3 g per fish’ for test2, ‘changed… to 74g’ on average). These are not in conflict — the Methods figures are explicit averages ACROSS all three SAS tanks combined (stated as such: fish ‘distributed among the three SAS tanks in order to have a similar stocking density’), while Table 6 gives the resulting per-tank values — but are noted here for anyone reconciling the two sources. Table 6’s per-SAS values used in Fish size initial/final cells as the more granular, trial-specific figures. | Water volume in the system: only the fish-tank volume (1 m3 = 1000 L IBC container) is explicitly stated as a single figure, identical across all three SAS and both tests (Methods 2.1). Downstream component volumes are given individually but never summed by the paper into one system total: SAS1/SAS3 share two 0.21 m3 (210 L) cylindrical tanks (mechanical filter + biofilter); SAS2 uses a 0.21 m3 clarifier (same as SAS1), a 0.48 m3 (360 L pre-washed expanded clay) biofilter tank, a 0.54 m3 sump tank, and two 0.48 m3 floating-raft tanks (0.96 m3 combined). Recording a summed total would be derivation (SCHEMA.md no-derivation rule); recorded fish-tank-only volume with component breakdown here. | UNIT CONVERSION ONLY: EC Table 1 gives trial mean +/- SD in uS/cm; SCHEMA.md EC column unit is dS/m (1 dS/m = 1000 uS/cm, divide by 1000): this trial’s uS/cm value 329 ± 55 -> 0.329 ± 0.055 dS/m (used in the EC cell). | UNIT CONVERSION ONLY: coordinates 37 deg 21 min 6.45 sec N, 5 deg 56 min 12.35 sec W (p.130, University of Seville, ETSIA) -> both DMS strings valid (minutes/seconds under 60) -> decimal Lat 37 + 21/60 + 6.45/3600 = 37.3518 N; Long 5 + 56/60 + 12.35/3600 = 5.9368, West is negative -> -5.9368. | Water temperature: Table 1 gives separate Tmax and Tmin trial means +/-SD (daily maximum/minimum logged via max-min thermometer), not one combined single trial mean; recorded both in the Water temperature cell as a range per SCHEMA.md’s ‘record the range… no trial mean reported’ guidance (adapted here: two distinct summary statistics rather than one). This trial: Tmin 23.0 ± 3.6 degC - Tmax 27.0 ± 2.5 degC (Table 1). | NOT DERIVED, left NR: Fish weight gain (per-fish) — AWF0 (30.0 g) and AWFf (34.8 g) are both explicitly given (Table 6) but their difference (a per-fish weight-gain figure) is never itself stated as a number in the paper; recording it would be derivation per SCHEMA.md’s explicit example of this exact temptation. Total tank-level biomass gain (Fish biomass created) IS explicitly stated (Table 6 Delta-TW) and is recorded in that column instead. Days Plant after transplant — fish trial duration (44 d) and the transplant offset (‘two weeks after’ test 1 / ‘one week after’ test 2) are both given, but their difference (harvest day counted from transplant) is never itself stated as a number; computing it would be derivation. Plants/m2 — plant counts (20 test1 / 24 test2) are stated and component dimensions are given individually (NFT pipe/hole spacing, raft tank footprint, felt wall area) but no single system growing-area figure is stated by the paper for any SAS, so density cannot be recorded without derivation; kg/m2 productivity IS explicitly stated and is used directly in the AP/HYD cells instead. Total Feed composition (N/P/K %) beyond crude protein (24.4%) — not given. Water recycle flow — recorded as a range (41.7-83.3 L/min, converted from the stated 2500-5000 L/h Eheim Compact+5000 pump spec, Methods 2.1); same pump model used across all three SAS and both tests, not restated per-SAS. | NO COLUMN items (no home in trials.csv/plant.csv): Table 1 EC initial/final snapshot values (distinct from the trial-mean+/-SD EC row used in the EC cell) — test1 SAS1 285->392, SAS2 274->357, SAS3 256->406 uS/cm; test2 SAS1 271->481, SAS2 246->455, SAS3 268->438 uS/cm. Total water consumption (TWC, L) over the whole test (Table 1): test1 SAS1 972.0, SAS2 680.0, SAS3 1048.4 L; test2 SAS1 666.9, SAS2 770.9, SAS3 1609.9 L. Author-computed water-footprint-per-kg-produce figures (Discussion, p.135, the paper’s own secondary calculation, not a trial-mean measurement): test1(LB/LR combined, SAS1) 413 L/kg, (SAS2) 1950 L/kg, (SAS3) 25200 L/kg; test2 (RL) 41 L/kg (SAS1), 196 L/kg (SAS2), 2446 L/kg (SAS3); best-case combined fish+crop footprint 72.9 L/kg (SAS1, test2). Productivity per unit of fish food (Results, p.132): test1 1.59/0.24/0.03 kg lettuce per kg fish food for SAS1/SAS2/SAS3; test2 0.36/0.05/0.01 kg lettuce per kg fish food. Plant shoot diameter (Table 2/3, cm, third morphometric measured alongside height/leaf count, no dedicated schema column): test1 SAS1 17.2+/-2.3, SAS2 8.7+/-0.8, SAS3 4.2+/-0.6, Ctrl 24.1+/-1.1; test2 SAS1 38.0+/-1.8, SAS2 23.8+/-1.4, SAS3 8.7+/-0.9, Ctrl 39.7+/-1.5. Per-system construction cost (‘each SAS cost around 1000 EUR’, Discussion p.136). Literature comparison table (Table 7, secondary figures from other studies, not this paper’s own data).

perezurrestarazuSuitabilityOptimizationFAO2019-T2

Fish

Field	Value
Fish	Goldfish (Carassius auratus)
Fish Category	Ornamental fish (goldfish are explicitly described as ‘ornamental fishes’ with a higher market value than other AP species, p.130)
Initial Stock density	1.6 kg/m3 (test 1, stated as a target/achieved stocking density, Methods 2.2)
FCR	7.65
SGR	0.19
Protein	24.4
% of body weight	1-2 (ramped from 1% of total fish weight at test start to 2% by test end, incremented every fortnight, Methods 2.3)
Fish size initial	30.6
Fish size final	33.1
Feed routine	Twice daily (Prodac International S.r.l., Cittadella, Italy, pond-stick feed)
Feed regime	1% of total fish weight at start, incremented fortnightly to 2% of total weight (Methods 2.3)
Total Feed (kg)	0.742
Fish biomass created (kg)	0.097
Fish survival rate	98
Fish trial duration (days)	44

Water

Field	Value
Water recycle	41.7-83.3 (converted from 2500-5000 L/h, Eheim Compact+5000 pump, Methods 2.1)
Water volume in the system	1000 (fish tank only, 1 m3 IBC container, x3 SAS identical; downstream component volumes given individually, not summed by the paper — see Experimental Remarks)
Daily Water exchange rate	0.76
Aq pH	8.1 ± 0.2
EC	0.309 ± 0.033
Water temperature	Tmin 21.2 ± 2.2 / Tmax 25.6 ± 2.9 (degC, Table 1; no single combined trial mean reported, see Experimental Remarks)
NO3-N	22.2 ± 8.7 (NO3- as ion per paper’s own Table 1 label, NOT confirmed as NO3-N — see WARN-CHECK in Experimental Remarks)

Plant

Field	Value
Plant	Lettuce (Lactuca sativa L.)
Details	Lollo rosso (LR) + Lollo bionda (LB), 10 plants each (20 total); pre-cultivated in seedbed trays 21 days; transplanted 2 weeks after fish stocking; test 1 (9 May-22 Jun 2016, 44 d)
Plant height	4.2 ± 0.5
Leaf count	8.7 ± 1.0
Plant fresh weight	8.7 ± 1.3
Plant dry matter	11.3 ± 0.9

System & Setup

Field	Value
System type	Floating raft
Media Details	Clarifier (0.21 m3, same as SAS1) -> 0.48 m3 tank (1 x 1.2 x 0.4 m) filled with 360 L pre-washed expanded clay as biofilter, fitted with bell syphon -> 0.54 m3 sump tank (1 x 1.2 x 0.45 m) -> Eheim Compact+5000 pump, 80% to fish tank / 20% to two 0.48 m3 raft tanks; rafts = 0.04 m XPS foam sheets (1 x 1.2 m), holes 0.06 m diameter spaced 0.25 m; BOYU ACQ-903 air compressor for raft-tank aeration (Methods 2.1)
Biological system already in use	N (Each 4-week fishless nitrifying-bacteria cycling period preceded BOTH tests (Biodigest Pro inoculant + artificial ammonia dosing, Methods 2.3) — system was newly cycled at the start of each test, not carried over from a prior operating state)
Air supplement	Y (Eheim Air pump 400 (400 L/h) for fish-tank DO; BOYU ACQ-903 air compressor for supplementary fish/raft-tank aeration (Methods 2.1))
Iron supplemented	Y (0.14 L chelated iron solution (1%, EDDHA Sequestrene 138 Fe) added directly to water, Fridays every fortnight (SAS2, higher circulating water volume), Methods 2.3)
Nutrient supplemented	Y (K2SO4 (1.5%) foliarly applied twice weekly (Mondays and Fridays) via manual sprayer to alleviate observed K/Fe/Ca insufficiency symptoms (Methods 2.3))
Equipment	Eheim Air pump 400 (Germany, 400 L/h) for fish-tank aeration; Eheim Compact+5000 pump (2500-5000 L/h adjustable) for biofilter-to-fish-tank/hydroponic recirculation; BOYU ACQ-903 air compressor (Raoping Guangdong, China) for supplementary fish/raft tank aeration; TFA (Germany) maximum-minimum thermometer; Crimson Instruments GLP 22 pH-meter; Crimson Instruments EC-Meter BASIC 30; Merck RQflex 10 plus reflectometer (nitrate); SPSS 24 statistical package
Control Parameters	Water pH, EC and nitrate measured periodically with pH-meter/EC-meter/RQflex reflectometer; Tmax/Tmin logged daily via max-min thermometer; daily water consumption (DWC) measured daily to derive total water consumption (TWC) and water replenishment rate (WR); fish counted/weighed every fortnight (test1) or every 3 weeks (test2); feed ration set at 1% of total fish weight at test start, incremented fortnightly up to 2% of total weight; foliar K2SO4 (1.5%) applied twice weekly and chelated iron (EDDHA Sequestrene 138 Fe, 1%) applied fortnightly to address nutrient insufficiency; plant leaf count/height/shoot diameter recorded weekly
Combination	Goldfish (Carassius auratus, ornamental fish) and lettuce (Lactuca sativa: Lollo rosso + Lollo bionda cultivars in test 1, Romaine cultivar in test 2) compared across three small-scale aquaponic system designs modified from FAO/Somerville et al. (2014) guidelines — SAS1 (Nutrient Film Technique), SAS2 (floating raft), SAS3 (vertical felt living wall) — each run as a single unreplicated physical unit, against a non-aquaponic hydroponic/substrate control (Ctrl: perlite pots + Hoagland nutrient solution), across two sequential seasonal test cycles (spring/summer 2016 and autumn/winter 2016-17).

Site

Field	Value
Region	Europe
Country	Spain
Lat	37.3518
Long	-5.9368

Results & Statistics

Field	Value
Measured Unit	kg/m2 (yield/productivity); g (TFW/FWP/DWP); % (DMC, dry matter content); cm (height, diameter); count (leaves); mg/L=ppm (NO3-); dS/m converted from uS/cm (EC); degC (temperature)
Statistic Details	One-way ANOVA (SPSS 24); Shapiro-Wilk normality test; nonparametric Kruskal-Wallis ANOVA when data not normally distributed; HSD Tukey post-hoc test; Games-Howell post-hoc test when variances unequal; significance at P<0.05 (Methods 2.4)
Statistically analysed	Y
Replicates (n)	20 plants/treatment (10 Lollo rosso + 10 Lollo bionda, pooled; 1 physical AP system per treatment, not independently replicated — see Extraction notes / TYPE CLASSIFICATION note)
AP	0.09
HYD	1.24

Experimental Remarks: TRIAL DEFINITION: T2 = SAS2 (floating raft), test 1 of 2, vs. Ctrl (hydroponic, perlite pots + Hoagland solution) of the SAME test cycle, recorded in the HYD-labelled cells (AP=0.09, HYD=1.24 kg/m2). Six aquaponic trial rows total (3 SAS designs x 2 sequential test cycles), each paired with that test cycle’s own hydroponic control (Ctrl), recorded in the HYD-labelled cells — Ctrl was NOT itself replicated or repeated identically between tests (different lettuce cultivar, different plant count, different productivity result each time), so T1-T3 share one Ctrl (test 1) and T4-T6 share a different Ctrl (test 2). Each SAS was built and operated as a SINGLE physical unit per test (one IBC fish tank + one hydroponic subsystem per treatment) — there is no independent system-level replicate of any SAS within a test; the individual lettuce plants (20/treatment in test 1 pooling 10 Lollo rosso + 10 Lollo bionda, 24/treatment in test 2, all Romaine) served as the pseudo-replicates for the reported ANOVA/Kruskal-Wallis/Tukey/Games-Howell comparisons across the four treatments (SAS1/SAS2/SAS3/Ctrl) within each test. | TYPE CLASSIFICATION JUDGMENT CALL: recorded as quasi-experiment. The paper has defined treatments (3 SAS designs vs. hydroponic control) and formal statistical testing (ANOVA/Kruskal-Wallis, Tukey/Games-Howell post-hoc, P<0.05), which would suggest ‘experiment’ under SCHEMA.md’s simple test. However, per SCHEMA.md Part 2 decision rule 2 (‘Randomised treatments with replication? Yes -> experiment. Treatments without randomisation or true replication -> quasi-experiment’), each SAS treatment consists of exactly ONE physical system (one IBC fish tank + one hydroponic subsystem) per test cycle — there is no independent system-level replicate for any treatment, and the word ‘randomised’ is never used for treatment-to-system assignment. The reported significance tests are computed on individual plants within that single system (pseudo-replication), not on independent replicate systems. This is a stricter case than pantanellaAquaponicsHydroponicsProduction2012 (which had 3 independent replicate physical systems per treatment per crop and was judged ‘experiment’ despite similarly lacking the word ‘randomised’); here there is no system-level replication at all, so ‘quasi-experiment’ was judged the better fit. | WARN-CHECK NO3-N vs NO3- (ion) basis. Table 1 reports this parameter explicitly as ‘NO3- (ppm)’ (nitrate ion, not nitrate-nitrogen); the paper never states or converts to NO3-N. SCHEMA.md’s NO3-N column specifically wants nitrate-nitrogen (mg/L), and NO3 vs NO3-N differ by a factor of 4.43 (SCHEMA.md’s own worked example of this exact ambiguity). Converting would require assuming the paper’s ‘NO3-’ is pure NO3 ion and dividing by 4.43 — an assumption not confirmable from the Methods (only ‘an RQflex 10 plus (MERK, Darmstadt, Germany)’ reflectometer is named, and Merck RQflex nitrate strips are sold in both NO3- and NO3-N calibrations). Recorded the paper’s own printed value UNCONVERTED, with basis explicitly named as NO3- (ion), not NO3-N — per SCHEMA.md CHECK convention (‘cell takes the value whose basis is clearest, with the basis named’). Flagged for REVIEW.md (batch merge step) so downstream use can decide whether to apply the x1/4.43 conversion. | WARN-MINOR FCR unit label. Table 6 and Eq. 3 label FCR as ‘FCR (%)’ and the text states ‘The FCR was between 3 and 4%, except for the case of SAS2 in test 1 (7.6%)’ (p.132), implying a percentage. Recomputing from the paper’s own stated inputs (Eq. 3: FCR=FI/(TWf-TW0); total feed intake FI=742 g in test 1 and 3247 g in test 2, identical across all three SAS within each test, per p.131-132) against Table 6’s total-weight-gain values reproduces Table 6’s FCR figures exactly as a DIMENSIONLESS RATIO, not a percentage: test1 SAS1 742/191=3.885 (~3.88 recorded), SAS2 742/97=7.649 (~7.65), SAS3 742/258=2.876 (~2.88); test2 (FI=3247) SAS1 3247/1040=3.122 (~3.12), SAS2 3247/1101=2.949 (~2.95), SAS3 3247/986=3.294 (~3.29). All six recomputed values match Table 6 digit-for-digit, confirming Table 6’s numbers are the ratio FI/gain, not that ratio x100. The paper’s own ’%’ label and text (‘3 and 4%’, ‘7.6%’) are therefore a units mislabel, not a distinct measured value. Recorded the ratio exactly as printed in Table 6 (matches SCHEMA.md’s FCR definition, ‘ratio, dimensionless’); no cell value is affected, only the unit label. Recomputation shown here as verification evidence only, per SCHEMA.md’s ‘recomputing to check is not derivation’ allowance — the recorded FCR value is Table 6’s own printed number, not a value newly computed by this extraction. | WARN-MINOR SGR unit label. Eq. 1 and Table 6 label SGR as ‘SGR (g.day-1)’, but the formula given (SGR = (lnAWFt - lnAWFt-1)/Delta t, no x100 factor shown) produces a dimensionless-per-day quantity, not a weight/day quantity, and does not match SCHEMA.md’s expected SGR unit of %/day (which would require a x100 factor the paper’s stated formula omits). Recorded Table 6’s printed values exactly as given (test1: SAS1 0.36, SAS2 0.19, SAS3 0.56; test2: SAS1 0.37, SAS2 0.41, SAS3 0.40 — matching the text’s ‘SGR values ranged from 0.19 (SAS2) to 0.56 (SAS3) in test 1 and were between 0.37 and 0.41 in the second’, p.134) without rescaling by 100, since SCHEMA.md’s no-derivation rule prohibits correcting/rescaling a paper’s own reported number even when its unit label looks inconsistent with the formula shown. Flagged so the mismatch between the paper’s stated unit and the SCHEMA.md column unit (%/day) is visible to anyone using this figure downstream. | Fish size initial/final: Table 6 gives per-SAS-specific AWF0/AWFf (e.g. test1 AWF0 30.0/30.6/30.2 g, AWFf 34.8/33.1/38.0 g for SAS1/SAS2/SAS3); Methods 2.2 separately gives single rounded overall figures (‘average weight of 30 g per fish’ stocked, ‘finishing with an average of 35 g’ for test1; ‘initial average weight of 53.3 g per fish’ for test2, ‘changed… to 74g’ on average). These are not in conflict — the Methods figures are explicit averages ACROSS all three SAS tanks combined (stated as such: fish ‘distributed among the three SAS tanks in order to have a similar stocking density’), while Table 6 gives the resulting per-tank values — but are noted here for anyone reconciling the two sources. Table 6’s per-SAS values used in Fish size initial/final cells as the more granular, trial-specific figures. | Water volume in the system: only the fish-tank volume (1 m3 = 1000 L IBC container) is explicitly stated as a single figure, identical across all three SAS and both tests (Methods 2.1). Downstream component volumes are given individually but never summed by the paper into one system total: SAS1/SAS3 share two 0.21 m3 (210 L) cylindrical tanks (mechanical filter + biofilter); SAS2 uses a 0.21 m3 clarifier (same as SAS1), a 0.48 m3 (360 L pre-washed expanded clay) biofilter tank, a 0.54 m3 sump tank, and two 0.48 m3 floating-raft tanks (0.96 m3 combined). Recording a summed total would be derivation (SCHEMA.md no-derivation rule); recorded fish-tank-only volume with component breakdown here. | UNIT CONVERSION ONLY: EC Table 1 gives trial mean +/- SD in uS/cm; SCHEMA.md EC column unit is dS/m (1 dS/m = 1000 uS/cm, divide by 1000): this trial’s uS/cm value 309 ± 33 -> 0.309 ± 0.033 dS/m (used in the EC cell). | UNIT CONVERSION ONLY: coordinates 37 deg 21 min 6.45 sec N, 5 deg 56 min 12.35 sec W (p.130, University of Seville, ETSIA) -> both DMS strings valid (minutes/seconds under 60) -> decimal Lat 37 + 21/60 + 6.45/3600 = 37.3518 N; Long 5 + 56/60 + 12.35/3600 = 5.9368, West is negative -> -5.9368. | Water temperature: Table 1 gives separate Tmax and Tmin trial means +/-SD (daily maximum/minimum logged via max-min thermometer), not one combined single trial mean; recorded both in the Water temperature cell as a range per SCHEMA.md’s ‘record the range… no trial mean reported’ guidance (adapted here: two distinct summary statistics rather than one). This trial: Tmin 21.2 ± 2.2 degC - Tmax 25.6 ± 2.9 degC (Table 1). | NOT DERIVED, left NR: Fish weight gain (per-fish) — AWF0 (30.6 g) and AWFf (33.1 g) are both explicitly given (Table 6) but their difference (a per-fish weight-gain figure) is never itself stated as a number in the paper; recording it would be derivation per SCHEMA.md’s explicit example of this exact temptation. Total tank-level biomass gain (Fish biomass created) IS explicitly stated (Table 6 Delta-TW) and is recorded in that column instead. Days Plant after transplant — fish trial duration (44 d) and the transplant offset (‘two weeks after’ test 1 / ‘one week after’ test 2) are both given, but their difference (harvest day counted from transplant) is never itself stated as a number; computing it would be derivation. Plants/m2 — plant counts (20 test1 / 24 test2) are stated and component dimensions are given individually (NFT pipe/hole spacing, raft tank footprint, felt wall area) but no single system growing-area figure is stated by the paper for any SAS, so density cannot be recorded without derivation; kg/m2 productivity IS explicitly stated and is used directly in the AP/HYD cells instead. Total Feed composition (N/P/K %) beyond crude protein (24.4%) — not given. Water recycle flow — recorded as a range (41.7-83.3 L/min, converted from the stated 2500-5000 L/h Eheim Compact+5000 pump spec, Methods 2.1); same pump model used across all three SAS and both tests, not restated per-SAS. | NO COLUMN items (no home in trials.csv/plant.csv): Table 1 EC initial/final snapshot values (distinct from the trial-mean+/-SD EC row used in the EC cell) — test1 SAS1 285->392, SAS2 274->357, SAS3 256->406 uS/cm; test2 SAS1 271->481, SAS2 246->455, SAS3 268->438 uS/cm. Total water consumption (TWC, L) over the whole test (Table 1): test1 SAS1 972.0, SAS2 680.0, SAS3 1048.4 L; test2 SAS1 666.9, SAS2 770.9, SAS3 1609.9 L. Author-computed water-footprint-per-kg-produce figures (Discussion, p.135, the paper’s own secondary calculation, not a trial-mean measurement): test1(LB/LR combined, SAS1) 413 L/kg, (SAS2) 1950 L/kg, (SAS3) 25200 L/kg; test2 (RL) 41 L/kg (SAS1), 196 L/kg (SAS2), 2446 L/kg (SAS3); best-case combined fish+crop footprint 72.9 L/kg (SAS1, test2). Productivity per unit of fish food (Results, p.132): test1 1.59/0.24/0.03 kg lettuce per kg fish food for SAS1/SAS2/SAS3; test2 0.36/0.05/0.01 kg lettuce per kg fish food. Plant shoot diameter (Table 2/3, cm, third morphometric measured alongside height/leaf count, no dedicated schema column): test1 SAS1 17.2+/-2.3, SAS2 8.7+/-0.8, SAS3 4.2+/-0.6, Ctrl 24.1+/-1.1; test2 SAS1 38.0+/-1.8, SAS2 23.8+/-1.4, SAS3 8.7+/-0.9, Ctrl 39.7+/-1.5. Per-system construction cost (‘each SAS cost around 1000 EUR’, Discussion p.136). Literature comparison table (Table 7, secondary figures from other studies, not this paper’s own data).

perezurrestarazuSuitabilityOptimizationFAO2019-T3

Fish

Field	Value
Fish	Goldfish (Carassius auratus)
Fish Category	Ornamental fish (goldfish are explicitly described as ‘ornamental fishes’ with a higher market value than other AP species, p.130)
Initial Stock density	1.6 kg/m3 (test 1, stated as a target/achieved stocking density, Methods 2.2)
FCR	2.88
SGR	0.56
Protein	24.4
% of body weight	1-2 (ramped from 1% of total fish weight at test start to 2% by test end, incremented every fortnight, Methods 2.3)
Fish size initial	30.2
Fish size final	38.0
Feed routine	Twice daily (Prodac International S.r.l., Cittadella, Italy, pond-stick feed)
Feed regime	1% of total fish weight at start, incremented fortnightly to 2% of total weight (Methods 2.3)
Total Feed (kg)	0.742
Fish biomass created (kg)	0.258
Fish survival rate	92
Fish trial duration (days)	44

Water

Field	Value
Water recycle	41.7-83.3 (converted from 2500-5000 L/h, Eheim Compact+5000 pump, Methods 2.1)
Water volume in the system	1000 (fish tank only, 1 m3 IBC container, x3 SAS identical; downstream component volumes given individually, not summed by the paper — see Experimental Remarks)
Daily Water exchange rate	1.79
Aq pH	8.1 ± 0.1
EC	0.320 ± 0.057
Water temperature	Tmin 21.0 ± 2.8 / Tmax 25.6 ± 3.1 (degC, Table 1; no single combined trial mean reported, see Experimental Remarks)
NO3-N	26.8 ± 10.7 (NO3- as ion per paper’s own Table 1 label, NOT confirmed as NO3-N — see WARN-CHECK in Experimental Remarks)

Plant

Field	Value
Plant	Lettuce (Lactuca sativa L.)
Details	Lollo rosso (LR) + Lollo bionda (LB), 10 plants each (20 total); pre-cultivated in seedbed trays 21 days; transplanted 2 weeks after fish stocking; test 1 (9 May-22 Jun 2016, 44 d)
Plant height	3.4 ± 0.3
Leaf count	4.2 ± 0.7
Plant fresh weight	1.1 ± 0.2
Plant dry matter	4.5 ± 1.5

System & Setup

Field	Value
System type	Vertical felt living wall
Media Details	Same fish tank/mechanical filter/biofilter as SAS1; hydroponic component = galvanised steel structure 2.4 m high x 2 m wide at 20 deg from vertical, 4 felt modules (0.75 x 1 m, 2 upper + 2 lower); each module = outer porous aeration layer + inner geotextile distribution layer sewn into a 0.2 x 0.25 m grid of pockets filled with expanded clay; max capacity 20 plants/m2, not fully planted (Methods 2.1)
Biological system already in use	N (Each 4-week fishless nitrifying-bacteria cycling period preceded BOTH tests (Biodigest Pro inoculant + artificial ammonia dosing, Methods 2.3) — system was newly cycled at the start of each test, not carried over from a prior operating state)
Air supplement	Y (Eheim Air pump 400 (400 L/h) for fish-tank DO; BOYU ACQ-903 air compressor for supplementary fish/raft-tank aeration (Methods 2.1))
Iron supplemented	Y (0.1 L chelated iron solution (1%, EDDHA Sequestrene 138 Fe) added directly to water, Fridays every fortnight (SAS1/SAS3), Methods 2.3)
Nutrient supplemented	Y (K2SO4 (1.5%) foliarly applied twice weekly (Mondays and Fridays) via manual sprayer to alleviate observed K/Fe/Ca insufficiency symptoms (Methods 2.3))
Equipment	Eheim Air pump 400 (Germany, 400 L/h) for fish-tank aeration; Eheim Compact+5000 pump (2500-5000 L/h adjustable) for biofilter-to-fish-tank/hydroponic recirculation; BOYU ACQ-903 air compressor (Raoping Guangdong, China) for supplementary fish/raft tank aeration; TFA (Germany) maximum-minimum thermometer; Crimson Instruments GLP 22 pH-meter; Crimson Instruments EC-Meter BASIC 30; Merck RQflex 10 plus reflectometer (nitrate); SPSS 24 statistical package
Control Parameters	Water pH, EC and nitrate measured periodically with pH-meter/EC-meter/RQflex reflectometer; Tmax/Tmin logged daily via max-min thermometer; daily water consumption (DWC) measured daily to derive total water consumption (TWC) and water replenishment rate (WR); fish counted/weighed every fortnight (test1) or every 3 weeks (test2); feed ration set at 1% of total fish weight at test start, incremented fortnightly up to 2% of total weight; foliar K2SO4 (1.5%) applied twice weekly and chelated iron (EDDHA Sequestrene 138 Fe, 1%) applied fortnightly to address nutrient insufficiency; plant leaf count/height/shoot diameter recorded weekly
Combination	Goldfish (Carassius auratus, ornamental fish) and lettuce (Lactuca sativa: Lollo rosso + Lollo bionda cultivars in test 1, Romaine cultivar in test 2) compared across three small-scale aquaponic system designs modified from FAO/Somerville et al. (2014) guidelines — SAS1 (Nutrient Film Technique), SAS2 (floating raft), SAS3 (vertical felt living wall) — each run as a single unreplicated physical unit, against a non-aquaponic hydroponic/substrate control (Ctrl: perlite pots + Hoagland nutrient solution), across two sequential seasonal test cycles (spring/summer 2016 and autumn/winter 2016-17).

Site

Field	Value
Region	Europe
Country	Spain
Lat	37.3518
Long	-5.9368

Results & Statistics

Field	Value
Measured Unit	kg/m2 (yield/productivity); g (TFW/FWP/DWP); % (DMC, dry matter content); cm (height, diameter); count (leaves); mg/L=ppm (NO3-); dS/m converted from uS/cm (EC); degC (temperature)
Statistic Details	One-way ANOVA (SPSS 24); Shapiro-Wilk normality test; nonparametric Kruskal-Wallis ANOVA when data not normally distributed; HSD Tukey post-hoc test; Games-Howell post-hoc test when variances unequal; significance at P<0.05 (Methods 2.4)
Statistically analysed	Y
Replicates (n)	20 plants/treatment (10 Lollo rosso + 10 Lollo bionda, pooled; 1 physical AP system per treatment, not independently replicated — see Extraction notes / TYPE CLASSIFICATION note)
AP	0.02
HYD	1.24

Experimental Remarks: TRIAL DEFINITION: T3 = SAS3 (vertical felt), test 1 of 2, vs. Ctrl (hydroponic, perlite pots + Hoagland solution) of the SAME test cycle, recorded in the HYD-labelled cells (AP=0.02, HYD=1.24 kg/m2). Six aquaponic trial rows total (3 SAS designs x 2 sequential test cycles), each paired with that test cycle’s own hydroponic control (Ctrl), recorded in the HYD-labelled cells — Ctrl was NOT itself replicated or repeated identically between tests (different lettuce cultivar, different plant count, different productivity result each time), so T1-T3 share one Ctrl (test 1) and T4-T6 share a different Ctrl (test 2). Each SAS was built and operated as a SINGLE physical unit per test (one IBC fish tank + one hydroponic subsystem per treatment) — there is no independent system-level replicate of any SAS within a test; the individual lettuce plants (20/treatment in test 1 pooling 10 Lollo rosso + 10 Lollo bionda, 24/treatment in test 2, all Romaine) served as the pseudo-replicates for the reported ANOVA/Kruskal-Wallis/Tukey/Games-Howell comparisons across the four treatments (SAS1/SAS2/SAS3/Ctrl) within each test. | TYPE CLASSIFICATION JUDGMENT CALL: recorded as quasi-experiment. The paper has defined treatments (3 SAS designs vs. hydroponic control) and formal statistical testing (ANOVA/Kruskal-Wallis, Tukey/Games-Howell post-hoc, P<0.05), which would suggest ‘experiment’ under SCHEMA.md’s simple test. However, per SCHEMA.md Part 2 decision rule 2 (‘Randomised treatments with replication? Yes -> experiment. Treatments without randomisation or true replication -> quasi-experiment’), each SAS treatment consists of exactly ONE physical system (one IBC fish tank + one hydroponic subsystem) per test cycle — there is no independent system-level replicate for any treatment, and the word ‘randomised’ is never used for treatment-to-system assignment. The reported significance tests are computed on individual plants within that single system (pseudo-replication), not on independent replicate systems. This is a stricter case than pantanellaAquaponicsHydroponicsProduction2012 (which had 3 independent replicate physical systems per treatment per crop and was judged ‘experiment’ despite similarly lacking the word ‘randomised’); here there is no system-level replication at all, so ‘quasi-experiment’ was judged the better fit. | WARN-CHECK NO3-N vs NO3- (ion) basis. Table 1 reports this parameter explicitly as ‘NO3- (ppm)’ (nitrate ion, not nitrate-nitrogen); the paper never states or converts to NO3-N. SCHEMA.md’s NO3-N column specifically wants nitrate-nitrogen (mg/L), and NO3 vs NO3-N differ by a factor of 4.43 (SCHEMA.md’s own worked example of this exact ambiguity). Converting would require assuming the paper’s ‘NO3-’ is pure NO3 ion and dividing by 4.43 — an assumption not confirmable from the Methods (only ‘an RQflex 10 plus (MERK, Darmstadt, Germany)’ reflectometer is named, and Merck RQflex nitrate strips are sold in both NO3- and NO3-N calibrations). Recorded the paper’s own printed value UNCONVERTED, with basis explicitly named as NO3- (ion), not NO3-N — per SCHEMA.md CHECK convention (‘cell takes the value whose basis is clearest, with the basis named’). Flagged for REVIEW.md (batch merge step) so downstream use can decide whether to apply the x1/4.43 conversion. | WARN-MINOR FCR unit label. Table 6 and Eq. 3 label FCR as ‘FCR (%)’ and the text states ‘The FCR was between 3 and 4%, except for the case of SAS2 in test 1 (7.6%)’ (p.132), implying a percentage. Recomputing from the paper’s own stated inputs (Eq. 3: FCR=FI/(TWf-TW0); total feed intake FI=742 g in test 1 and 3247 g in test 2, identical across all three SAS within each test, per p.131-132) against Table 6’s total-weight-gain values reproduces Table 6’s FCR figures exactly as a DIMENSIONLESS RATIO, not a percentage: test1 SAS1 742/191=3.885 (~3.88 recorded), SAS2 742/97=7.649 (~7.65), SAS3 742/258=2.876 (~2.88); test2 (FI=3247) SAS1 3247/1040=3.122 (~3.12), SAS2 3247/1101=2.949 (~2.95), SAS3 3247/986=3.294 (~3.29). All six recomputed values match Table 6 digit-for-digit, confirming Table 6’s numbers are the ratio FI/gain, not that ratio x100. The paper’s own ’%’ label and text (‘3 and 4%’, ‘7.6%’) are therefore a units mislabel, not a distinct measured value. Recorded the ratio exactly as printed in Table 6 (matches SCHEMA.md’s FCR definition, ‘ratio, dimensionless’); no cell value is affected, only the unit label. Recomputation shown here as verification evidence only, per SCHEMA.md’s ‘recomputing to check is not derivation’ allowance — the recorded FCR value is Table 6’s own printed number, not a value newly computed by this extraction. | WARN-MINOR SGR unit label. Eq. 1 and Table 6 label SGR as ‘SGR (g.day-1)’, but the formula given (SGR = (lnAWFt - lnAWFt-1)/Delta t, no x100 factor shown) produces a dimensionless-per-day quantity, not a weight/day quantity, and does not match SCHEMA.md’s expected SGR unit of %/day (which would require a x100 factor the paper’s stated formula omits). Recorded Table 6’s printed values exactly as given (test1: SAS1 0.36, SAS2 0.19, SAS3 0.56; test2: SAS1 0.37, SAS2 0.41, SAS3 0.40 — matching the text’s ‘SGR values ranged from 0.19 (SAS2) to 0.56 (SAS3) in test 1 and were between 0.37 and 0.41 in the second’, p.134) without rescaling by 100, since SCHEMA.md’s no-derivation rule prohibits correcting/rescaling a paper’s own reported number even when its unit label looks inconsistent with the formula shown. Flagged so the mismatch between the paper’s stated unit and the SCHEMA.md column unit (%/day) is visible to anyone using this figure downstream. | Fish size initial/final: Table 6 gives per-SAS-specific AWF0/AWFf (e.g. test1 AWF0 30.0/30.6/30.2 g, AWFf 34.8/33.1/38.0 g for SAS1/SAS2/SAS3); Methods 2.2 separately gives single rounded overall figures (‘average weight of 30 g per fish’ stocked, ‘finishing with an average of 35 g’ for test1; ‘initial average weight of 53.3 g per fish’ for test2, ‘changed… to 74g’ on average). These are not in conflict — the Methods figures are explicit averages ACROSS all three SAS tanks combined (stated as such: fish ‘distributed among the three SAS tanks in order to have a similar stocking density’), while Table 6 gives the resulting per-tank values — but are noted here for anyone reconciling the two sources. Table 6’s per-SAS values used in Fish size initial/final cells as the more granular, trial-specific figures. | Water volume in the system: only the fish-tank volume (1 m3 = 1000 L IBC container) is explicitly stated as a single figure, identical across all three SAS and both tests (Methods 2.1). Downstream component volumes are given individually but never summed by the paper into one system total: SAS1/SAS3 share two 0.21 m3 (210 L) cylindrical tanks (mechanical filter + biofilter); SAS2 uses a 0.21 m3 clarifier (same as SAS1), a 0.48 m3 (360 L pre-washed expanded clay) biofilter tank, a 0.54 m3 sump tank, and two 0.48 m3 floating-raft tanks (0.96 m3 combined). Recording a summed total would be derivation (SCHEMA.md no-derivation rule); recorded fish-tank-only volume with component breakdown here. | UNIT CONVERSION ONLY: EC Table 1 gives trial mean +/- SD in uS/cm; SCHEMA.md EC column unit is dS/m (1 dS/m = 1000 uS/cm, divide by 1000): this trial’s uS/cm value 320 ± 57 -> 0.320 ± 0.057 dS/m (used in the EC cell). | UNIT CONVERSION ONLY: coordinates 37 deg 21 min 6.45 sec N, 5 deg 56 min 12.35 sec W (p.130, University of Seville, ETSIA) -> both DMS strings valid (minutes/seconds under 60) -> decimal Lat 37 + 21/60 + 6.45/3600 = 37.3518 N; Long 5 + 56/60 + 12.35/3600 = 5.9368, West is negative -> -5.9368. | Water temperature: Table 1 gives separate Tmax and Tmin trial means +/-SD (daily maximum/minimum logged via max-min thermometer), not one combined single trial mean; recorded both in the Water temperature cell as a range per SCHEMA.md’s ‘record the range… no trial mean reported’ guidance (adapted here: two distinct summary statistics rather than one). This trial: Tmin 21.0 ± 2.8 degC - Tmax 25.6 ± 3.1 degC (Table 1). | NOT DERIVED, left NR: Fish weight gain (per-fish) — AWF0 (30.2 g) and AWFf (38.0 g) are both explicitly given (Table 6) but their difference (a per-fish weight-gain figure) is never itself stated as a number in the paper; recording it would be derivation per SCHEMA.md’s explicit example of this exact temptation. Total tank-level biomass gain (Fish biomass created) IS explicitly stated (Table 6 Delta-TW) and is recorded in that column instead. Days Plant after transplant — fish trial duration (44 d) and the transplant offset (‘two weeks after’ test 1 / ‘one week after’ test 2) are both given, but their difference (harvest day counted from transplant) is never itself stated as a number; computing it would be derivation. Plants/m2 — plant counts (20 test1 / 24 test2) are stated and component dimensions are given individually (NFT pipe/hole spacing, raft tank footprint, felt wall area) but no single system growing-area figure is stated by the paper for any SAS, so density cannot be recorded without derivation; kg/m2 productivity IS explicitly stated and is used directly in the AP/HYD cells instead. Total Feed composition (N/P/K %) beyond crude protein (24.4%) — not given. Water recycle flow — recorded as a range (41.7-83.3 L/min, converted from the stated 2500-5000 L/h Eheim Compact+5000 pump spec, Methods 2.1); same pump model used across all three SAS and both tests, not restated per-SAS. | NO COLUMN items (no home in trials.csv/plant.csv): Table 1 EC initial/final snapshot values (distinct from the trial-mean+/-SD EC row used in the EC cell) — test1 SAS1 285->392, SAS2 274->357, SAS3 256->406 uS/cm; test2 SAS1 271->481, SAS2 246->455, SAS3 268->438 uS/cm. Total water consumption (TWC, L) over the whole test (Table 1): test1 SAS1 972.0, SAS2 680.0, SAS3 1048.4 L; test2 SAS1 666.9, SAS2 770.9, SAS3 1609.9 L. Author-computed water-footprint-per-kg-produce figures (Discussion, p.135, the paper’s own secondary calculation, not a trial-mean measurement): test1(LB/LR combined, SAS1) 413 L/kg, (SAS2) 1950 L/kg, (SAS3) 25200 L/kg; test2 (RL) 41 L/kg (SAS1), 196 L/kg (SAS2), 2446 L/kg (SAS3); best-case combined fish+crop footprint 72.9 L/kg (SAS1, test2). Productivity per unit of fish food (Results, p.132): test1 1.59/0.24/0.03 kg lettuce per kg fish food for SAS1/SAS2/SAS3; test2 0.36/0.05/0.01 kg lettuce per kg fish food. Plant shoot diameter (Table 2/3, cm, third morphometric measured alongside height/leaf count, no dedicated schema column): test1 SAS1 17.2+/-2.3, SAS2 8.7+/-0.8, SAS3 4.2+/-0.6, Ctrl 24.1+/-1.1; test2 SAS1 38.0+/-1.8, SAS2 23.8+/-1.4, SAS3 8.7+/-0.9, Ctrl 39.7+/-1.5. Per-system construction cost (‘each SAS cost around 1000 EUR’, Discussion p.136). Literature comparison table (Table 7, secondary figures from other studies, not this paper’s own data).

perezurrestarazuSuitabilityOptimizationFAO2019-T4

Fish

Field	Value
Fish	Goldfish (Carassius auratus)
Fish Category	Ornamental fish (goldfish are explicitly described as ‘ornamental fishes’ with a higher market value than other AP species, p.130)
Initial Stock density	2.5 kg/m3 (test 2, stated as a target/achieved stocking density, Methods 2.2)
FCR	3.12
SGR	0.37
Protein	24.4
% of body weight	1-2 (ramped from 1% of total fish weight at test start to 2% by test end, incremented every fortnight, Methods 2.3)
Fish size initial	51.2
Fish size final	69.9
Feed routine	Twice daily (Prodac International S.r.l., Cittadella, Italy, pond-stick feed)
Feed regime	1% of total fish weight at start, incremented fortnightly to 2% of total weight (Methods 2.3)
Total Feed (kg)	3.247
Fish biomass created (kg)	1.040
Fish survival rate	104
Fish trial duration (days)	54

Water

Field	Value
Water recycle	41.7-83.3 (converted from 2500-5000 L/h, Eheim Compact+5000 pump, Methods 2.1)
Water volume in the system	1000 (fish tank only, 1 m3 IBC container, x3 SAS identical; downstream component volumes given individually, not summed by the paper — see Experimental Remarks)
Daily Water exchange rate	1.03
Aq pH	7.0 ± 0.6
EC	0.370 ± 0.058
Water temperature	Tmin 16.8 ± 0.8 / Tmax 23.0 ± 2.2 (degC, Table 1; no single combined trial mean reported, see Experimental Remarks)
NO3-N	62.7 ± 19.8 (NO3- as ion per paper’s own Table 1 label, NOT confirmed as NO3-N — see WARN-CHECK in Experimental Remarks)

Plant

Field	Value
Plant	Lettuce (Lactuca sativa L.)
Details	Romaine lettuce (RL), 24 plants; pre-cultivated in seedbed trays 21 days; transplanted 1 week after fish stocking; test 2 (8 Nov 2016-11 Jan 2017, 54 d)
Plant height	33.9 ± 1.1
Leaf count	30.4 ± 1.7
Plant fresh weight	337.9 (author-computed average from TFW/plant count, no SD — per-plant measurement not done in test 2, Methods 2.3)
Plant dry matter	5.6 (author-computed average, no SD, Methods 2.3)

System & Setup

Field	Value
System type	Nutrient film technique (NFT)
Media Details	Same construction as T1 (SAS1 test 1); system reused/rebuilt identically for test 2 (Methods 2.1)
Biological system already in use	N (Each 4-week fishless nitrifying-bacteria cycling period preceded BOTH tests (Biodigest Pro inoculant + artificial ammonia dosing, Methods 2.3) — system was newly cycled at the start of each test, not carried over from a prior operating state)
Air supplement	Y (Eheim Air pump 400 (400 L/h) for fish-tank DO; BOYU ACQ-903 air compressor for supplementary fish/raft-tank aeration (Methods 2.1))
Iron supplemented	Y (0.1 L chelated iron solution (1%, EDDHA Sequestrene 138 Fe) added directly to water, Fridays every fortnight (SAS1/SAS3), Methods 2.3)
Nutrient supplemented	Y (K2SO4 (1.5%) foliarly applied twice weekly (Mondays and Fridays) via manual sprayer to alleviate observed K/Fe/Ca insufficiency symptoms (Methods 2.3))
Equipment	Eheim Air pump 400 (Germany, 400 L/h) for fish-tank aeration; Eheim Compact+5000 pump (2500-5000 L/h adjustable) for biofilter-to-fish-tank/hydroponic recirculation; BOYU ACQ-903 air compressor (Raoping Guangdong, China) for supplementary fish/raft tank aeration; TFA (Germany) maximum-minimum thermometer; Crimson Instruments GLP 22 pH-meter; Crimson Instruments EC-Meter BASIC 30; Merck RQflex 10 plus reflectometer (nitrate); SPSS 24 statistical package
Control Parameters	Water pH, EC and nitrate measured periodically with pH-meter/EC-meter/RQflex reflectometer; Tmax/Tmin logged daily via max-min thermometer; daily water consumption (DWC) measured daily to derive total water consumption (TWC) and water replenishment rate (WR); fish counted/weighed every fortnight (test1) or every 3 weeks (test2); feed ration set at 1% of total fish weight at test start, incremented fortnightly up to 2% of total weight; foliar K2SO4 (1.5%) applied twice weekly and chelated iron (EDDHA Sequestrene 138 Fe, 1%) applied fortnightly to address nutrient insufficiency; plant leaf count/height/shoot diameter recorded weekly
Combination	Goldfish (Carassius auratus, ornamental fish) and lettuce (Lactuca sativa: Lollo rosso + Lollo bionda cultivars in test 1, Romaine cultivar in test 2) compared across three small-scale aquaponic system designs modified from FAO/Somerville et al. (2014) guidelines — SAS1 (Nutrient Film Technique), SAS2 (floating raft), SAS3 (vertical felt living wall) — each run as a single unreplicated physical unit, against a non-aquaponic hydroponic/substrate control (Ctrl: perlite pots + Hoagland nutrient solution), across two sequential seasonal test cycles (spring/summer 2016 and autumn/winter 2016-17).

Site

Field	Value
Region	Europe
Country	Spain
Lat	37.3518
Long	-5.9368

Results & Statistics

Field	Value
Measured Unit	kg/m2 (yield/productivity); g (TFW/FWP/DWP); % (DMC, dry matter content); cm (height, diameter); count (leaves); mg/L=ppm (NO3-); dS/m converted from uS/cm (EC); degC (temperature)
Statistic Details	One-way ANOVA (SPSS 24); Shapiro-Wilk normality test; nonparametric Kruskal-Wallis ANOVA when data not normally distributed; HSD Tukey post-hoc test; Games-Howell post-hoc test when variances unequal; significance at P<0.05 (Methods 2.4)
Statistically analysed	Y
Replicates (n)	24 plants/treatment (Romaine, pooled; 1 physical AP system per treatment, not independently replicated — see Extraction notes / TYPE CLASSIFICATION note)
AP	3.38
HYD	2.84

Experimental Remarks: TRIAL DEFINITION: T4 = SAS1 (NFT), test 2 of 2, vs. Ctrl (hydroponic, perlite pots + Hoagland solution) of the SAME test cycle, recorded in the HYD-labelled cells (AP=3.38, HYD=2.84 kg/m2). Six aquaponic trial rows total (3 SAS designs x 2 sequential test cycles), each paired with that test cycle’s own hydroponic control (Ctrl), recorded in the HYD-labelled cells — Ctrl was NOT itself replicated or repeated identically between tests (different lettuce cultivar, different plant count, different productivity result each time), so T1-T3 share one Ctrl (test 1) and T4-T6 share a different Ctrl (test 2). Each SAS was built and operated as a SINGLE physical unit per test (one IBC fish tank + one hydroponic subsystem per treatment) — there is no independent system-level replicate of any SAS within a test; the individual lettuce plants (20/treatment in test 1 pooling 10 Lollo rosso + 10 Lollo bionda, 24/treatment in test 2, all Romaine) served as the pseudo-replicates for the reported ANOVA/Kruskal-Wallis/Tukey/Games-Howell comparisons across the four treatments (SAS1/SAS2/SAS3/Ctrl) within each test. | TYPE CLASSIFICATION JUDGMENT CALL: recorded as quasi-experiment. The paper has defined treatments (3 SAS designs vs. hydroponic control) and formal statistical testing (ANOVA/Kruskal-Wallis, Tukey/Games-Howell post-hoc, P<0.05), which would suggest ‘experiment’ under SCHEMA.md’s simple test. However, per SCHEMA.md Part 2 decision rule 2 (‘Randomised treatments with replication? Yes -> experiment. Treatments without randomisation or true replication -> quasi-experiment’), each SAS treatment consists of exactly ONE physical system (one IBC fish tank + one hydroponic subsystem) per test cycle — there is no independent system-level replicate for any treatment, and the word ‘randomised’ is never used for treatment-to-system assignment. The reported significance tests are computed on individual plants within that single system (pseudo-replication), not on independent replicate systems. This is a stricter case than pantanellaAquaponicsHydroponicsProduction2012 (which had 3 independent replicate physical systems per treatment per crop and was judged ‘experiment’ despite similarly lacking the word ‘randomised’); here there is no system-level replication at all, so ‘quasi-experiment’ was judged the better fit. | WARN-CHECK NO3-N vs NO3- (ion) basis. Table 1 reports this parameter explicitly as ‘NO3- (ppm)’ (nitrate ion, not nitrate-nitrogen); the paper never states or converts to NO3-N. SCHEMA.md’s NO3-N column specifically wants nitrate-nitrogen (mg/L), and NO3 vs NO3-N differ by a factor of 4.43 (SCHEMA.md’s own worked example of this exact ambiguity). Converting would require assuming the paper’s ‘NO3-’ is pure NO3 ion and dividing by 4.43 — an assumption not confirmable from the Methods (only ‘an RQflex 10 plus (MERK, Darmstadt, Germany)’ reflectometer is named, and Merck RQflex nitrate strips are sold in both NO3- and NO3-N calibrations). Recorded the paper’s own printed value UNCONVERTED, with basis explicitly named as NO3- (ion), not NO3-N — per SCHEMA.md CHECK convention (‘cell takes the value whose basis is clearest, with the basis named’). Flagged for REVIEW.md (batch merge step) so downstream use can decide whether to apply the x1/4.43 conversion. | WARN-MINOR FCR unit label. Table 6 and Eq. 3 label FCR as ‘FCR (%)’ and the text states ‘The FCR was between 3 and 4%, except for the case of SAS2 in test 1 (7.6%)’ (p.132), implying a percentage. Recomputing from the paper’s own stated inputs (Eq. 3: FCR=FI/(TWf-TW0); total feed intake FI=742 g in test 1 and 3247 g in test 2, identical across all three SAS within each test, per p.131-132) against Table 6’s total-weight-gain values reproduces Table 6’s FCR figures exactly as a DIMENSIONLESS RATIO, not a percentage: test1 SAS1 742/191=3.885 (~3.88 recorded), SAS2 742/97=7.649 (~7.65), SAS3 742/258=2.876 (~2.88); test2 (FI=3247) SAS1 3247/1040=3.122 (~3.12), SAS2 3247/1101=2.949 (~2.95), SAS3 3247/986=3.294 (~3.29). All six recomputed values match Table 6 digit-for-digit, confirming Table 6’s numbers are the ratio FI/gain, not that ratio x100. The paper’s own ’%’ label and text (‘3 and 4%’, ‘7.6%’) are therefore a units mislabel, not a distinct measured value. Recorded the ratio exactly as printed in Table 6 (matches SCHEMA.md’s FCR definition, ‘ratio, dimensionless’); no cell value is affected, only the unit label. Recomputation shown here as verification evidence only, per SCHEMA.md’s ‘recomputing to check is not derivation’ allowance — the recorded FCR value is Table 6’s own printed number, not a value newly computed by this extraction. | WARN-MINOR SGR unit label. Eq. 1 and Table 6 label SGR as ‘SGR (g.day-1)’, but the formula given (SGR = (lnAWFt - lnAWFt-1)/Delta t, no x100 factor shown) produces a dimensionless-per-day quantity, not a weight/day quantity, and does not match SCHEMA.md’s expected SGR unit of %/day (which would require a x100 factor the paper’s stated formula omits). Recorded Table 6’s printed values exactly as given (test1: SAS1 0.36, SAS2 0.19, SAS3 0.56; test2: SAS1 0.37, SAS2 0.41, SAS3 0.40 — matching the text’s ‘SGR values ranged from 0.19 (SAS2) to 0.56 (SAS3) in test 1 and were between 0.37 and 0.41 in the second’, p.134) without rescaling by 100, since SCHEMA.md’s no-derivation rule prohibits correcting/rescaling a paper’s own reported number even when its unit label looks inconsistent with the formula shown. Flagged so the mismatch between the paper’s stated unit and the SCHEMA.md column unit (%/day) is visible to anyone using this figure downstream. | Fish size initial/final: Table 6 gives per-SAS-specific AWF0/AWFf (e.g. test1 AWF0 30.0/30.6/30.2 g, AWFf 34.8/33.1/38.0 g for SAS1/SAS2/SAS3); Methods 2.2 separately gives single rounded overall figures (‘average weight of 30 g per fish’ stocked, ‘finishing with an average of 35 g’ for test1; ‘initial average weight of 53.3 g per fish’ for test2, ‘changed… to 74g’ on average). These are not in conflict — the Methods figures are explicit averages ACROSS all three SAS tanks combined (stated as such: fish ‘distributed among the three SAS tanks in order to have a similar stocking density’), while Table 6 gives the resulting per-tank values — but are noted here for anyone reconciling the two sources. Table 6’s per-SAS values used in Fish size initial/final cells as the more granular, trial-specific figures. | Water volume in the system: only the fish-tank volume (1 m3 = 1000 L IBC container) is explicitly stated as a single figure, identical across all three SAS and both tests (Methods 2.1). Downstream component volumes are given individually but never summed by the paper into one system total: SAS1/SAS3 share two 0.21 m3 (210 L) cylindrical tanks (mechanical filter + biofilter); SAS2 uses a 0.21 m3 clarifier (same as SAS1), a 0.48 m3 (360 L pre-washed expanded clay) biofilter tank, a 0.54 m3 sump tank, and two 0.48 m3 floating-raft tanks (0.96 m3 combined). Recording a summed total would be derivation (SCHEMA.md no-derivation rule); recorded fish-tank-only volume with component breakdown here. | UNIT CONVERSION ONLY: EC Table 1 gives trial mean +/- SD in uS/cm; SCHEMA.md EC column unit is dS/m (1 dS/m = 1000 uS/cm, divide by 1000): this trial’s uS/cm value 370 ± 58 -> 0.370 ± 0.058 dS/m (used in the EC cell). | UNIT CONVERSION ONLY: coordinates 37 deg 21 min 6.45 sec N, 5 deg 56 min 12.35 sec W (p.130, University of Seville, ETSIA) -> both DMS strings valid (minutes/seconds under 60) -> decimal Lat 37 + 21/60 + 6.45/3600 = 37.3518 N; Long 5 + 56/60 + 12.35/3600 = 5.9368, West is negative -> -5.9368. | Water temperature: Table 1 gives separate Tmax and Tmin trial means +/-SD (daily maximum/minimum logged via max-min thermometer), not one combined single trial mean; recorded both in the Water temperature cell as a range per SCHEMA.md’s ‘record the range… no trial mean reported’ guidance (adapted here: two distinct summary statistics rather than one). This trial: Tmin 16.8 ± 0.8 degC - Tmax 23.0 ± 2.2 degC (Table 1). | Fish survival rate >100% (T4, T5): Table 6 gives SVR(%) = 104 (SAS1) and 102 (SAS2) for test 2, apparently exceeding 100%. This is explained, not an error: the Results text states ‘Fishes even reproduced in test 2 (SAS1 and SAS2)’ (p.132) — SVR is defined (Methods 2.3) as ‘the ratio between the initial and the final number of fishes in each SAS’, so in-tank reproduction increased the final fish count above the initial stocked count in these two systems, producing SVR>100%. SAS3 test2 (no stated reproduction) = 100% (all survived, none reproduced or died); test1 values (96/98/92%, no reproduction reported) are all <=100% as expected. | NOT DERIVED, left NR: Fish weight gain (per-fish) — AWF0 (51.2 g) and AWFf (69.9 g) are both explicitly given (Table 6) but their difference (a per-fish weight-gain figure) is never itself stated as a number in the paper; recording it would be derivation per SCHEMA.md’s explicit example of this exact temptation. Total tank-level biomass gain (Fish biomass created) IS explicitly stated (Table 6 Delta-TW) and is recorded in that column instead. Days Plant after transplant — fish trial duration (54 d) and the transplant offset (‘two weeks after’ test 1 / ‘one week after’ test 2) are both given, but their difference (harvest day counted from transplant) is never itself stated as a number; computing it would be derivation. Plants/m2 — plant counts (20 test1 / 24 test2) are stated and component dimensions are given individually (NFT pipe/hole spacing, raft tank footprint, felt wall area) but no single system growing-area figure is stated by the paper for any SAS, so density cannot be recorded without derivation; kg/m2 productivity IS explicitly stated and is used directly in the AP/HYD cells instead. Total Feed composition (N/P/K %) beyond crude protein (24.4%) — not given. Water recycle flow — recorded as a range (41.7-83.3 L/min, converted from the stated 2500-5000 L/h Eheim Compact+5000 pump spec, Methods 2.1); same pump model used across all three SAS and both tests, not restated per-SAS. | NO COLUMN items (no home in trials.csv/plant.csv): Table 1 EC initial/final snapshot values (distinct from the trial-mean+/-SD EC row used in the EC cell) — test1 SAS1 285->392, SAS2 274->357, SAS3 256->406 uS/cm; test2 SAS1 271->481, SAS2 246->455, SAS3 268->438 uS/cm. Total water consumption (TWC, L) over the whole test (Table 1): test1 SAS1 972.0, SAS2 680.0, SAS3 1048.4 L; test2 SAS1 666.9, SAS2 770.9, SAS3 1609.9 L. Author-computed water-footprint-per-kg-produce figures (Discussion, p.135, the paper’s own secondary calculation, not a trial-mean measurement): test1(LB/LR combined, SAS1) 413 L/kg, (SAS2) 1950 L/kg, (SAS3) 25200 L/kg; test2 (RL) 41 L/kg (SAS1), 196 L/kg (SAS2), 2446 L/kg (SAS3); best-case combined fish+crop footprint 72.9 L/kg (SAS1, test2). Productivity per unit of fish food (Results, p.132): test1 1.59/0.24/0.03 kg lettuce per kg fish food for SAS1/SAS2/SAS3; test2 0.36/0.05/0.01 kg lettuce per kg fish food. Plant shoot diameter (Table 2/3, cm, third morphometric measured alongside height/leaf count, no dedicated schema column): test1 SAS1 17.2+/-2.3, SAS2 8.7+/-0.8, SAS3 4.2+/-0.6, Ctrl 24.1+/-1.1; test2 SAS1 38.0+/-1.8, SAS2 23.8+/-1.4, SAS3 8.7+/-0.9, Ctrl 39.7+/-1.5. Per-system construction cost (‘each SAS cost around 1000 EUR’, Discussion p.136). Literature comparison table (Table 7, secondary figures from other studies, not this paper’s own data).

perezurrestarazuSuitabilityOptimizationFAO2019-T5

Fish

Field	Value
Fish	Goldfish (Carassius auratus)
Fish Category	Ornamental fish (goldfish are explicitly described as ‘ornamental fishes’ with a higher market value than other AP species, p.130)
Initial Stock density	2.5 kg/m3 (test 2, stated as a target/achieved stocking density, Methods 2.2)
FCR	2.95
SGR	0.41
Protein	24.4
% of body weight	1-2 (ramped from 1% of total fish weight at test start to 2% by test end, incremented every fortnight, Methods 2.3)
Fish size initial	56.4
Fish size final	79.6
Feed routine	Twice daily (Prodac International S.r.l., Cittadella, Italy, pond-stick feed)
Feed regime	1% of total fish weight at start, incremented fortnightly to 2% of total weight (Methods 2.3)
Total Feed (kg)	3.247
Fish biomass created (kg)	1.101
Fish survival rate	102
Fish trial duration (days)	54

Water

Field	Value
Water recycle	41.7-83.3 (converted from 2500-5000 L/h, Eheim Compact+5000 pump, Methods 2.1)
Water volume in the system	1000 (fish tank only, 1 m3 IBC container, x3 SAS identical; downstream component volumes given individually, not summed by the paper — see Experimental Remarks)
Daily Water exchange rate	0.77
Aq pH	7.7 ± 0.4
EC	0.366 ± 0.055
Water temperature	Tmin 17.3 ± 1.4 / Tmax 20.4 ± 3.4 (degC, Table 1; no single combined trial mean reported, see Experimental Remarks)
NO3-N	36.8 ± 12.9 (NO3- as ion per paper’s own Table 1 label, NOT confirmed as NO3-N — see WARN-CHECK in Experimental Remarks)

Plant

Field	Value
Plant	Lettuce (Lactuca sativa L.)
Details	Romaine lettuce (RL), 24 plants; pre-cultivated in seedbed trays 21 days; transplanted 1 week after fish stocking; test 2 (8 Nov 2016-11 Jan 2017, 54 d)
Plant height	27.8 ± 0.9
Leaf count	26.7 ± 1.4
Plant fresh weight	81.9 (author-computed average from TFW/plant count, no SD — per-plant measurement not done in test 2, Methods 2.3)
Plant dry matter	6.4 (author-computed average, no SD, Methods 2.3)

System & Setup

Field	Value
System type	Floating raft
Media Details	Same construction as T2 (SAS2 test 1); system reused/rebuilt identically for test 2 (Methods 2.1)
Biological system already in use	N (Each 4-week fishless nitrifying-bacteria cycling period preceded BOTH tests (Biodigest Pro inoculant + artificial ammonia dosing, Methods 2.3) — system was newly cycled at the start of each test, not carried over from a prior operating state)
Air supplement	Y (Eheim Air pump 400 (400 L/h) for fish-tank DO; BOYU ACQ-903 air compressor for supplementary fish/raft-tank aeration (Methods 2.1))
Iron supplemented	Y (0.14 L chelated iron solution (1%, EDDHA Sequestrene 138 Fe) added directly to water, Fridays every fortnight (SAS2, higher circulating water volume), Methods 2.3)
Nutrient supplemented	Y (K2SO4 (1.5%) foliarly applied twice weekly (Mondays and Fridays) via manual sprayer to alleviate observed K/Fe/Ca insufficiency symptoms (Methods 2.3))
Equipment	Eheim Air pump 400 (Germany, 400 L/h) for fish-tank aeration; Eheim Compact+5000 pump (2500-5000 L/h adjustable) for biofilter-to-fish-tank/hydroponic recirculation; BOYU ACQ-903 air compressor (Raoping Guangdong, China) for supplementary fish/raft tank aeration; TFA (Germany) maximum-minimum thermometer; Crimson Instruments GLP 22 pH-meter; Crimson Instruments EC-Meter BASIC 30; Merck RQflex 10 plus reflectometer (nitrate); SPSS 24 statistical package
Control Parameters	Water pH, EC and nitrate measured periodically with pH-meter/EC-meter/RQflex reflectometer; Tmax/Tmin logged daily via max-min thermometer; daily water consumption (DWC) measured daily to derive total water consumption (TWC) and water replenishment rate (WR); fish counted/weighed every fortnight (test1) or every 3 weeks (test2); feed ration set at 1% of total fish weight at test start, incremented fortnightly up to 2% of total weight; foliar K2SO4 (1.5%) applied twice weekly and chelated iron (EDDHA Sequestrene 138 Fe, 1%) applied fortnightly to address nutrient insufficiency; plant leaf count/height/shoot diameter recorded weekly
Combination	Goldfish (Carassius auratus, ornamental fish) and lettuce (Lactuca sativa: Lollo rosso + Lollo bionda cultivars in test 1, Romaine cultivar in test 2) compared across three small-scale aquaponic system designs modified from FAO/Somerville et al. (2014) guidelines — SAS1 (Nutrient Film Technique), SAS2 (floating raft), SAS3 (vertical felt living wall) — each run as a single unreplicated physical unit, against a non-aquaponic hydroponic/substrate control (Ctrl: perlite pots + Hoagland nutrient solution), across two sequential seasonal test cycles (spring/summer 2016 and autumn/winter 2016-17).

Site

Field	Value
Region	Europe
Country	Spain
Lat	37.3518
Long	-5.9368

Results & Statistics

Field	Value
Measured Unit	kg/m2 (yield/productivity); g (TFW/FWP/DWP); % (DMC, dry matter content); cm (height, diameter); count (leaves); mg/L=ppm (NO3-); dS/m converted from uS/cm (EC); degC (temperature)
Statistic Details	One-way ANOVA (SPSS 24); Shapiro-Wilk normality test; nonparametric Kruskal-Wallis ANOVA when data not normally distributed; HSD Tukey post-hoc test; Games-Howell post-hoc test when variances unequal; significance at P<0.05 (Methods 2.4)
Statistically analysed	Y
Replicates (n)	24 plants/treatment (Romaine, pooled; 1 physical AP system per treatment, not independently replicated — see Extraction notes / TYPE CLASSIFICATION note)
AP	0.98
HYD	2.84

Experimental Remarks: TRIAL DEFINITION: T5 = SAS2 (floating raft), test 2 of 2, vs. Ctrl (hydroponic, perlite pots + Hoagland solution) of the SAME test cycle, recorded in the HYD-labelled cells (AP=0.98, HYD=2.84 kg/m2). Six aquaponic trial rows total (3 SAS designs x 2 sequential test cycles), each paired with that test cycle’s own hydroponic control (Ctrl), recorded in the HYD-labelled cells — Ctrl was NOT itself replicated or repeated identically between tests (different lettuce cultivar, different plant count, different productivity result each time), so T1-T3 share one Ctrl (test 1) and T4-T6 share a different Ctrl (test 2). Each SAS was built and operated as a SINGLE physical unit per test (one IBC fish tank + one hydroponic subsystem per treatment) — there is no independent system-level replicate of any SAS within a test; the individual lettuce plants (20/treatment in test 1 pooling 10 Lollo rosso + 10 Lollo bionda, 24/treatment in test 2, all Romaine) served as the pseudo-replicates for the reported ANOVA/Kruskal-Wallis/Tukey/Games-Howell comparisons across the four treatments (SAS1/SAS2/SAS3/Ctrl) within each test. | TYPE CLASSIFICATION JUDGMENT CALL: recorded as quasi-experiment. The paper has defined treatments (3 SAS designs vs. hydroponic control) and formal statistical testing (ANOVA/Kruskal-Wallis, Tukey/Games-Howell post-hoc, P<0.05), which would suggest ‘experiment’ under SCHEMA.md’s simple test. However, per SCHEMA.md Part 2 decision rule 2 (‘Randomised treatments with replication? Yes -> experiment. Treatments without randomisation or true replication -> quasi-experiment’), each SAS treatment consists of exactly ONE physical system (one IBC fish tank + one hydroponic subsystem) per test cycle — there is no independent system-level replicate for any treatment, and the word ‘randomised’ is never used for treatment-to-system assignment. The reported significance tests are computed on individual plants within that single system (pseudo-replication), not on independent replicate systems. This is a stricter case than pantanellaAquaponicsHydroponicsProduction2012 (which had 3 independent replicate physical systems per treatment per crop and was judged ‘experiment’ despite similarly lacking the word ‘randomised’); here there is no system-level replication at all, so ‘quasi-experiment’ was judged the better fit. | WARN-CHECK NO3-N vs NO3- (ion) basis. Table 1 reports this parameter explicitly as ‘NO3- (ppm)’ (nitrate ion, not nitrate-nitrogen); the paper never states or converts to NO3-N. SCHEMA.md’s NO3-N column specifically wants nitrate-nitrogen (mg/L), and NO3 vs NO3-N differ by a factor of 4.43 (SCHEMA.md’s own worked example of this exact ambiguity). Converting would require assuming the paper’s ‘NO3-’ is pure NO3 ion and dividing by 4.43 — an assumption not confirmable from the Methods (only ‘an RQflex 10 plus (MERK, Darmstadt, Germany)’ reflectometer is named, and Merck RQflex nitrate strips are sold in both NO3- and NO3-N calibrations). Recorded the paper’s own printed value UNCONVERTED, with basis explicitly named as NO3- (ion), not NO3-N — per SCHEMA.md CHECK convention (‘cell takes the value whose basis is clearest, with the basis named’). Flagged for REVIEW.md (batch merge step) so downstream use can decide whether to apply the x1/4.43 conversion. | WARN-MINOR FCR unit label. Table 6 and Eq. 3 label FCR as ‘FCR (%)’ and the text states ‘The FCR was between 3 and 4%, except for the case of SAS2 in test 1 (7.6%)’ (p.132), implying a percentage. Recomputing from the paper’s own stated inputs (Eq. 3: FCR=FI/(TWf-TW0); total feed intake FI=742 g in test 1 and 3247 g in test 2, identical across all three SAS within each test, per p.131-132) against Table 6’s total-weight-gain values reproduces Table 6’s FCR figures exactly as a DIMENSIONLESS RATIO, not a percentage: test1 SAS1 742/191=3.885 (~3.88 recorded), SAS2 742/97=7.649 (~7.65), SAS3 742/258=2.876 (~2.88); test2 (FI=3247) SAS1 3247/1040=3.122 (~3.12), SAS2 3247/1101=2.949 (~2.95), SAS3 3247/986=3.294 (~3.29). All six recomputed values match Table 6 digit-for-digit, confirming Table 6’s numbers are the ratio FI/gain, not that ratio x100. The paper’s own ’%’ label and text (‘3 and 4%’, ‘7.6%’) are therefore a units mislabel, not a distinct measured value. Recorded the ratio exactly as printed in Table 6 (matches SCHEMA.md’s FCR definition, ‘ratio, dimensionless’); no cell value is affected, only the unit label. Recomputation shown here as verification evidence only, per SCHEMA.md’s ‘recomputing to check is not derivation’ allowance — the recorded FCR value is Table 6’s own printed number, not a value newly computed by this extraction. | WARN-MINOR SGR unit label. Eq. 1 and Table 6 label SGR as ‘SGR (g.day-1)’, but the formula given (SGR = (lnAWFt - lnAWFt-1)/Delta t, no x100 factor shown) produces a dimensionless-per-day quantity, not a weight/day quantity, and does not match SCHEMA.md’s expected SGR unit of %/day (which would require a x100 factor the paper’s stated formula omits). Recorded Table 6’s printed values exactly as given (test1: SAS1 0.36, SAS2 0.19, SAS3 0.56; test2: SAS1 0.37, SAS2 0.41, SAS3 0.40 — matching the text’s ‘SGR values ranged from 0.19 (SAS2) to 0.56 (SAS3) in test 1 and were between 0.37 and 0.41 in the second’, p.134) without rescaling by 100, since SCHEMA.md’s no-derivation rule prohibits correcting/rescaling a paper’s own reported number even when its unit label looks inconsistent with the formula shown. Flagged so the mismatch between the paper’s stated unit and the SCHEMA.md column unit (%/day) is visible to anyone using this figure downstream. | Fish size initial/final: Table 6 gives per-SAS-specific AWF0/AWFf (e.g. test1 AWF0 30.0/30.6/30.2 g, AWFf 34.8/33.1/38.0 g for SAS1/SAS2/SAS3); Methods 2.2 separately gives single rounded overall figures (‘average weight of 30 g per fish’ stocked, ‘finishing with an average of 35 g’ for test1; ‘initial average weight of 53.3 g per fish’ for test2, ‘changed… to 74g’ on average). These are not in conflict — the Methods figures are explicit averages ACROSS all three SAS tanks combined (stated as such: fish ‘distributed among the three SAS tanks in order to have a similar stocking density’), while Table 6 gives the resulting per-tank values — but are noted here for anyone reconciling the two sources. Table 6’s per-SAS values used in Fish size initial/final cells as the more granular, trial-specific figures. | Water volume in the system: only the fish-tank volume (1 m3 = 1000 L IBC container) is explicitly stated as a single figure, identical across all three SAS and both tests (Methods 2.1). Downstream component volumes are given individually but never summed by the paper into one system total: SAS1/SAS3 share two 0.21 m3 (210 L) cylindrical tanks (mechanical filter + biofilter); SAS2 uses a 0.21 m3 clarifier (same as SAS1), a 0.48 m3 (360 L pre-washed expanded clay) biofilter tank, a 0.54 m3 sump tank, and two 0.48 m3 floating-raft tanks (0.96 m3 combined). Recording a summed total would be derivation (SCHEMA.md no-derivation rule); recorded fish-tank-only volume with component breakdown here. | UNIT CONVERSION ONLY: EC Table 1 gives trial mean +/- SD in uS/cm; SCHEMA.md EC column unit is dS/m (1 dS/m = 1000 uS/cm, divide by 1000): this trial’s uS/cm value 366 ± 55 -> 0.366 ± 0.055 dS/m (used in the EC cell). | UNIT CONVERSION ONLY: coordinates 37 deg 21 min 6.45 sec N, 5 deg 56 min 12.35 sec W (p.130, University of Seville, ETSIA) -> both DMS strings valid (minutes/seconds under 60) -> decimal Lat 37 + 21/60 + 6.45/3600 = 37.3518 N; Long 5 + 56/60 + 12.35/3600 = 5.9368, West is negative -> -5.9368. | Water temperature: Table 1 gives separate Tmax and Tmin trial means +/-SD (daily maximum/minimum logged via max-min thermometer), not one combined single trial mean; recorded both in the Water temperature cell as a range per SCHEMA.md’s ‘record the range… no trial mean reported’ guidance (adapted here: two distinct summary statistics rather than one). This trial: Tmin 17.3 ± 1.4 degC - Tmax 20.4 ± 3.4 degC (Table 1). | Fish survival rate >100% (T4, T5): Table 6 gives SVR(%) = 104 (SAS1) and 102 (SAS2) for test 2, apparently exceeding 100%. This is explained, not an error: the Results text states ‘Fishes even reproduced in test 2 (SAS1 and SAS2)’ (p.132) — SVR is defined (Methods 2.3) as ‘the ratio between the initial and the final number of fishes in each SAS’, so in-tank reproduction increased the final fish count above the initial stocked count in these two systems, producing SVR>100%. SAS3 test2 (no stated reproduction) = 100% (all survived, none reproduced or died); test1 values (96/98/92%, no reproduction reported) are all <=100% as expected. | NOT DERIVED, left NR: Fish weight gain (per-fish) — AWF0 (56.4 g) and AWFf (79.6 g) are both explicitly given (Table 6) but their difference (a per-fish weight-gain figure) is never itself stated as a number in the paper; recording it would be derivation per SCHEMA.md’s explicit example of this exact temptation. Total tank-level biomass gain (Fish biomass created) IS explicitly stated (Table 6 Delta-TW) and is recorded in that column instead. Days Plant after transplant — fish trial duration (54 d) and the transplant offset (‘two weeks after’ test 1 / ‘one week after’ test 2) are both given, but their difference (harvest day counted from transplant) is never itself stated as a number; computing it would be derivation. Plants/m2 — plant counts (20 test1 / 24 test2) are stated and component dimensions are given individually (NFT pipe/hole spacing, raft tank footprint, felt wall area) but no single system growing-area figure is stated by the paper for any SAS, so density cannot be recorded without derivation; kg/m2 productivity IS explicitly stated and is used directly in the AP/HYD cells instead. Total Feed composition (N/P/K %) beyond crude protein (24.4%) — not given. Water recycle flow — recorded as a range (41.7-83.3 L/min, converted from the stated 2500-5000 L/h Eheim Compact+5000 pump spec, Methods 2.1); same pump model used across all three SAS and both tests, not restated per-SAS. | NO COLUMN items (no home in trials.csv/plant.csv): Table 1 EC initial/final snapshot values (distinct from the trial-mean+/-SD EC row used in the EC cell) — test1 SAS1 285->392, SAS2 274->357, SAS3 256->406 uS/cm; test2 SAS1 271->481, SAS2 246->455, SAS3 268->438 uS/cm. Total water consumption (TWC, L) over the whole test (Table 1): test1 SAS1 972.0, SAS2 680.0, SAS3 1048.4 L; test2 SAS1 666.9, SAS2 770.9, SAS3 1609.9 L. Author-computed water-footprint-per-kg-produce figures (Discussion, p.135, the paper’s own secondary calculation, not a trial-mean measurement): test1(LB/LR combined, SAS1) 413 L/kg, (SAS2) 1950 L/kg, (SAS3) 25200 L/kg; test2 (RL) 41 L/kg (SAS1), 196 L/kg (SAS2), 2446 L/kg (SAS3); best-case combined fish+crop footprint 72.9 L/kg (SAS1, test2). Productivity per unit of fish food (Results, p.132): test1 1.59/0.24/0.03 kg lettuce per kg fish food for SAS1/SAS2/SAS3; test2 0.36/0.05/0.01 kg lettuce per kg fish food. Plant shoot diameter (Table 2/3, cm, third morphometric measured alongside height/leaf count, no dedicated schema column): test1 SAS1 17.2+/-2.3, SAS2 8.7+/-0.8, SAS3 4.2+/-0.6, Ctrl 24.1+/-1.1; test2 SAS1 38.0+/-1.8, SAS2 23.8+/-1.4, SAS3 8.7+/-0.9, Ctrl 39.7+/-1.5. Per-system construction cost (‘each SAS cost around 1000 EUR’, Discussion p.136). Literature comparison table (Table 7, secondary figures from other studies, not this paper’s own data).

perezurrestarazuSuitabilityOptimizationFAO2019-T6

Fish

Field	Value
Fish	Goldfish (Carassius auratus)
Fish Category	Ornamental fish (goldfish are explicitly described as ‘ornamental fishes’ with a higher market value than other AP species, p.130)
Initial Stock density	2.5 kg/m3 (test 2, stated as a target/achieved stocking density, Methods 2.2)
FCR	3.29
SGR	0.40
Protein	24.4
% of body weight	1-2 (ramped from 1% of total fish weight at test start to 2% by test end, incremented every fortnight, Methods 2.3)
Fish size initial	52.3
Fish size final	73.3
Feed routine	Twice daily (Prodac International S.r.l., Cittadella, Italy, pond-stick feed)
Feed regime	1% of total fish weight at start, incremented fortnightly to 2% of total weight (Methods 2.3)
Total Feed (kg)	3.247
Fish biomass created (kg)	0.986
Fish survival rate	100
Fish trial duration (days)	54

Water

Field	Value
Water recycle	41.7-83.3 (converted from 2500-5000 L/h, Eheim Compact+5000 pump, Methods 2.1)
Water volume in the system	1000 (fish tank only, 1 m3 IBC container, x3 SAS identical; downstream component volumes given individually, not summed by the paper — see Experimental Remarks)
Daily Water exchange rate	2.48
Aq pH	7.6 ± 0.4
EC	0.378 ± 0.059
Water temperature	Tmin 16.3 ± 2.7 / Tmax 19.4 ± 2.0 (degC, Table 1; no single combined trial mean reported, see Experimental Remarks)
NO3-N	39.4 ± 16.3 (NO3- as ion per paper’s own Table 1 label, NOT confirmed as NO3-N — see WARN-CHECK in Experimental Remarks)

Plant

Field	Value
Plant	Lettuce (Lactuca sativa L.)
Details	Romaine lettuce (RL), 24 plants; pre-cultivated in seedbed trays 21 days; transplanted 1 week after fish stocking; test 2 (8 Nov 2016-11 Jan 2017, 54 d)
Plant height	17.0 ± 1.4
Leaf count	13.9 ± 1.8
Plant fresh weight	13.7 (author-computed average from TFW/plant count, no SD — per-plant measurement not done in test 2, Methods 2.3)
Plant dry matter	11.4 (author-computed average, no SD, Methods 2.3)

System & Setup

Field	Value
System type	Vertical felt living wall
Media Details	Same construction as T3 (SAS3 test 1); system reused/rebuilt identically for test 2 (Methods 2.1)
Biological system already in use	N (Each 4-week fishless nitrifying-bacteria cycling period preceded BOTH tests (Biodigest Pro inoculant + artificial ammonia dosing, Methods 2.3) — system was newly cycled at the start of each test, not carried over from a prior operating state)
Air supplement	Y (Eheim Air pump 400 (400 L/h) for fish-tank DO; BOYU ACQ-903 air compressor for supplementary fish/raft-tank aeration (Methods 2.1))
Iron supplemented	Y (0.1 L chelated iron solution (1%, EDDHA Sequestrene 138 Fe) added directly to water, Fridays every fortnight (SAS1/SAS3), Methods 2.3)
Nutrient supplemented	Y (K2SO4 (1.5%) foliarly applied twice weekly (Mondays and Fridays) via manual sprayer to alleviate observed K/Fe/Ca insufficiency symptoms (Methods 2.3))
Equipment	Eheim Air pump 400 (Germany, 400 L/h) for fish-tank aeration; Eheim Compact+5000 pump (2500-5000 L/h adjustable) for biofilter-to-fish-tank/hydroponic recirculation; BOYU ACQ-903 air compressor (Raoping Guangdong, China) for supplementary fish/raft tank aeration; TFA (Germany) maximum-minimum thermometer; Crimson Instruments GLP 22 pH-meter; Crimson Instruments EC-Meter BASIC 30; Merck RQflex 10 plus reflectometer (nitrate); SPSS 24 statistical package
Control Parameters	Water pH, EC and nitrate measured periodically with pH-meter/EC-meter/RQflex reflectometer; Tmax/Tmin logged daily via max-min thermometer; daily water consumption (DWC) measured daily to derive total water consumption (TWC) and water replenishment rate (WR); fish counted/weighed every fortnight (test1) or every 3 weeks (test2); feed ration set at 1% of total fish weight at test start, incremented fortnightly up to 2% of total weight; foliar K2SO4 (1.5%) applied twice weekly and chelated iron (EDDHA Sequestrene 138 Fe, 1%) applied fortnightly to address nutrient insufficiency; plant leaf count/height/shoot diameter recorded weekly
Combination	Goldfish (Carassius auratus, ornamental fish) and lettuce (Lactuca sativa: Lollo rosso + Lollo bionda cultivars in test 1, Romaine cultivar in test 2) compared across three small-scale aquaponic system designs modified from FAO/Somerville et al. (2014) guidelines — SAS1 (Nutrient Film Technique), SAS2 (floating raft), SAS3 (vertical felt living wall) — each run as a single unreplicated physical unit, against a non-aquaponic hydroponic/substrate control (Ctrl: perlite pots + Hoagland nutrient solution), across two sequential seasonal test cycles (spring/summer 2016 and autumn/winter 2016-17).

Site

Field	Value
Region	Europe
Country	Spain
Lat	37.3518
Long	-5.9368

Results & Statistics

Field	Value
Measured Unit	kg/m2 (yield/productivity); g (TFW/FWP/DWP); % (DMC, dry matter content); cm (height, diameter); count (leaves); mg/L=ppm (NO3-); dS/m converted from uS/cm (EC); degC (temperature)
Statistic Details	One-way ANOVA (SPSS 24); Shapiro-Wilk normality test; nonparametric Kruskal-Wallis ANOVA when data not normally distributed; HSD Tukey post-hoc test; Games-Howell post-hoc test when variances unequal; significance at P<0.05 (Methods 2.4)
Statistically analysed	Y
Replicates (n)	24 plants/treatment (Romaine, pooled; 1 physical AP system per treatment, not independently replicated — see Extraction notes / TYPE CLASSIFICATION note)
AP	0.24
HYD	2.84

Experimental Remarks: TRIAL DEFINITION: T6 = SAS3 (vertical felt), test 2 of 2, vs. Ctrl (hydroponic, perlite pots + Hoagland solution) of the SAME test cycle, recorded in the HYD-labelled cells (AP=0.24, HYD=2.84 kg/m2). Six aquaponic trial rows total (3 SAS designs x 2 sequential test cycles), each paired with that test cycle’s own hydroponic control (Ctrl), recorded in the HYD-labelled cells — Ctrl was NOT itself replicated or repeated identically between tests (different lettuce cultivar, different plant count, different productivity result each time), so T1-T3 share one Ctrl (test 1) and T4-T6 share a different Ctrl (test 2). Each SAS was built and operated as a SINGLE physical unit per test (one IBC fish tank + one hydroponic subsystem per treatment) — there is no independent system-level replicate of any SAS within a test; the individual lettuce plants (20/treatment in test 1 pooling 10 Lollo rosso + 10 Lollo bionda, 24/treatment in test 2, all Romaine) served as the pseudo-replicates for the reported ANOVA/Kruskal-Wallis/Tukey/Games-Howell comparisons across the four treatments (SAS1/SAS2/SAS3/Ctrl) within each test. | TYPE CLASSIFICATION JUDGMENT CALL: recorded as quasi-experiment. The paper has defined treatments (3 SAS designs vs. hydroponic control) and formal statistical testing (ANOVA/Kruskal-Wallis, Tukey/Games-Howell post-hoc, P<0.05), which would suggest ‘experiment’ under SCHEMA.md’s simple test. However, per SCHEMA.md Part 2 decision rule 2 (‘Randomised treatments with replication? Yes -> experiment. Treatments without randomisation or true replication -> quasi-experiment’), each SAS treatment consists of exactly ONE physical system (one IBC fish tank + one hydroponic subsystem) per test cycle — there is no independent system-level replicate for any treatment, and the word ‘randomised’ is never used for treatment-to-system assignment. The reported significance tests are computed on individual plants within that single system (pseudo-replication), not on independent replicate systems. This is a stricter case than pantanellaAquaponicsHydroponicsProduction2012 (which had 3 independent replicate physical systems per treatment per crop and was judged ‘experiment’ despite similarly lacking the word ‘randomised’); here there is no system-level replication at all, so ‘quasi-experiment’ was judged the better fit. | WARN-CHECK NO3-N vs NO3- (ion) basis. Table 1 reports this parameter explicitly as ‘NO3- (ppm)’ (nitrate ion, not nitrate-nitrogen); the paper never states or converts to NO3-N. SCHEMA.md’s NO3-N column specifically wants nitrate-nitrogen (mg/L), and NO3 vs NO3-N differ by a factor of 4.43 (SCHEMA.md’s own worked example of this exact ambiguity). Converting would require assuming the paper’s ‘NO3-’ is pure NO3 ion and dividing by 4.43 — an assumption not confirmable from the Methods (only ‘an RQflex 10 plus (MERK, Darmstadt, Germany)’ reflectometer is named, and Merck RQflex nitrate strips are sold in both NO3- and NO3-N calibrations). Recorded the paper’s own printed value UNCONVERTED, with basis explicitly named as NO3- (ion), not NO3-N — per SCHEMA.md CHECK convention (‘cell takes the value whose basis is clearest, with the basis named’). Flagged for REVIEW.md (batch merge step) so downstream use can decide whether to apply the x1/4.43 conversion. | WARN-MINOR FCR unit label. Table 6 and Eq. 3 label FCR as ‘FCR (%)’ and the text states ‘The FCR was between 3 and 4%, except for the case of SAS2 in test 1 (7.6%)’ (p.132), implying a percentage. Recomputing from the paper’s own stated inputs (Eq. 3: FCR=FI/(TWf-TW0); total feed intake FI=742 g in test 1 and 3247 g in test 2, identical across all three SAS within each test, per p.131-132) against Table 6’s total-weight-gain values reproduces Table 6’s FCR figures exactly as a DIMENSIONLESS RATIO, not a percentage: test1 SAS1 742/191=3.885 (~3.88 recorded), SAS2 742/97=7.649 (~7.65), SAS3 742/258=2.876 (~2.88); test2 (FI=3247) SAS1 3247/1040=3.122 (~3.12), SAS2 3247/1101=2.949 (~2.95), SAS3 3247/986=3.294 (~3.29). All six recomputed values match Table 6 digit-for-digit, confirming Table 6’s numbers are the ratio FI/gain, not that ratio x100. The paper’s own ’%’ label and text (‘3 and 4%’, ‘7.6%’) are therefore a units mislabel, not a distinct measured value. Recorded the ratio exactly as printed in Table 6 (matches SCHEMA.md’s FCR definition, ‘ratio, dimensionless’); no cell value is affected, only the unit label. Recomputation shown here as verification evidence only, per SCHEMA.md’s ‘recomputing to check is not derivation’ allowance — the recorded FCR value is Table 6’s own printed number, not a value newly computed by this extraction. | WARN-MINOR SGR unit label. Eq. 1 and Table 6 label SGR as ‘SGR (g.day-1)’, but the formula given (SGR = (lnAWFt - lnAWFt-1)/Delta t, no x100 factor shown) produces a dimensionless-per-day quantity, not a weight/day quantity, and does not match SCHEMA.md’s expected SGR unit of %/day (which would require a x100 factor the paper’s stated formula omits). Recorded Table 6’s printed values exactly as given (test1: SAS1 0.36, SAS2 0.19, SAS3 0.56; test2: SAS1 0.37, SAS2 0.41, SAS3 0.40 — matching the text’s ‘SGR values ranged from 0.19 (SAS2) to 0.56 (SAS3) in test 1 and were between 0.37 and 0.41 in the second’, p.134) without rescaling by 100, since SCHEMA.md’s no-derivation rule prohibits correcting/rescaling a paper’s own reported number even when its unit label looks inconsistent with the formula shown. Flagged so the mismatch between the paper’s stated unit and the SCHEMA.md column unit (%/day) is visible to anyone using this figure downstream. | Fish size initial/final: Table 6 gives per-SAS-specific AWF0/AWFf (e.g. test1 AWF0 30.0/30.6/30.2 g, AWFf 34.8/33.1/38.0 g for SAS1/SAS2/SAS3); Methods 2.2 separately gives single rounded overall figures (‘average weight of 30 g per fish’ stocked, ‘finishing with an average of 35 g’ for test1; ‘initial average weight of 53.3 g per fish’ for test2, ‘changed… to 74g’ on average). These are not in conflict — the Methods figures are explicit averages ACROSS all three SAS tanks combined (stated as such: fish ‘distributed among the three SAS tanks in order to have a similar stocking density’), while Table 6 gives the resulting per-tank values — but are noted here for anyone reconciling the two sources. Table 6’s per-SAS values used in Fish size initial/final cells as the more granular, trial-specific figures. | Water volume in the system: only the fish-tank volume (1 m3 = 1000 L IBC container) is explicitly stated as a single figure, identical across all three SAS and both tests (Methods 2.1). Downstream component volumes are given individually but never summed by the paper into one system total: SAS1/SAS3 share two 0.21 m3 (210 L) cylindrical tanks (mechanical filter + biofilter); SAS2 uses a 0.21 m3 clarifier (same as SAS1), a 0.48 m3 (360 L pre-washed expanded clay) biofilter tank, a 0.54 m3 sump tank, and two 0.48 m3 floating-raft tanks (0.96 m3 combined). Recording a summed total would be derivation (SCHEMA.md no-derivation rule); recorded fish-tank-only volume with component breakdown here. | UNIT CONVERSION ONLY: EC Table 1 gives trial mean +/- SD in uS/cm; SCHEMA.md EC column unit is dS/m (1 dS/m = 1000 uS/cm, divide by 1000): this trial’s uS/cm value 378 ± 59 -> 0.378 ± 0.059 dS/m (used in the EC cell). | UNIT CONVERSION ONLY: coordinates 37 deg 21 min 6.45 sec N, 5 deg 56 min 12.35 sec W (p.130, University of Seville, ETSIA) -> both DMS strings valid (minutes/seconds under 60) -> decimal Lat 37 + 21/60 + 6.45/3600 = 37.3518 N; Long 5 + 56/60 + 12.35/3600 = 5.9368, West is negative -> -5.9368. | Water temperature: Table 1 gives separate Tmax and Tmin trial means +/-SD (daily maximum/minimum logged via max-min thermometer), not one combined single trial mean; recorded both in the Water temperature cell as a range per SCHEMA.md’s ‘record the range… no trial mean reported’ guidance (adapted here: two distinct summary statistics rather than one). This trial: Tmin 16.3 ± 2.7 degC - Tmax 19.4 ± 2.0 degC (Table 1). | NOT DERIVED, left NR: Fish weight gain (per-fish) — AWF0 (52.3 g) and AWFf (73.3 g) are both explicitly given (Table 6) but their difference (a per-fish weight-gain figure) is never itself stated as a number in the paper; recording it would be derivation per SCHEMA.md’s explicit example of this exact temptation. Total tank-level biomass gain (Fish biomass created) IS explicitly stated (Table 6 Delta-TW) and is recorded in that column instead. Days Plant after transplant — fish trial duration (54 d) and the transplant offset (‘two weeks after’ test 1 / ‘one week after’ test 2) are both given, but their difference (harvest day counted from transplant) is never itself stated as a number; computing it would be derivation. Plants/m2 — plant counts (20 test1 / 24 test2) are stated and component dimensions are given individually (NFT pipe/hole spacing, raft tank footprint, felt wall area) but no single system growing-area figure is stated by the paper for any SAS, so density cannot be recorded without derivation; kg/m2 productivity IS explicitly stated and is used directly in the AP/HYD cells instead. Total Feed composition (N/P/K %) beyond crude protein (24.4%) — not given. Water recycle flow — recorded as a range (41.7-83.3 L/min, converted from the stated 2500-5000 L/h Eheim Compact+5000 pump spec, Methods 2.1); same pump model used across all three SAS and both tests, not restated per-SAS. | NO COLUMN items (no home in trials.csv/plant.csv): Table 1 EC initial/final snapshot values (distinct from the trial-mean+/-SD EC row used in the EC cell) — test1 SAS1 285->392, SAS2 274->357, SAS3 256->406 uS/cm; test2 SAS1 271->481, SAS2 246->455, SAS3 268->438 uS/cm. Total water consumption (TWC, L) over the whole test (Table 1): test1 SAS1 972.0, SAS2 680.0, SAS3 1048.4 L; test2 SAS1 666.9, SAS2 770.9, SAS3 1609.9 L. Author-computed water-footprint-per-kg-produce figures (Discussion, p.135, the paper’s own secondary calculation, not a trial-mean measurement): test1(LB/LR combined, SAS1) 413 L/kg, (SAS2) 1950 L/kg, (SAS3) 25200 L/kg; test2 (RL) 41 L/kg (SAS1), 196 L/kg (SAS2), 2446 L/kg (SAS3); best-case combined fish+crop footprint 72.9 L/kg (SAS1, test2). Productivity per unit of fish food (Results, p.132): test1 1.59/0.24/0.03 kg lettuce per kg fish food for SAS1/SAS2/SAS3; test2 0.36/0.05/0.01 kg lettuce per kg fish food. Plant shoot diameter (Table 2/3, cm, third morphometric measured alongside height/leaf count, no dedicated schema column): test1 SAS1 17.2+/-2.3, SAS2 8.7+/-0.8, SAS3 4.2+/-0.6, Ctrl 24.1+/-1.1; test2 SAS1 38.0+/-1.8, SAS2 23.8+/-1.4, SAS3 8.7+/-0.9, Ctrl 39.7+/-1.5. Per-system construction cost (‘each SAS cost around 1000 EUR’, Discussion p.136). Literature comparison table (Table 7, secondary figures from other studies, not this paper’s own data).

Quartz 5

Explorer

Suitability and optimization of FAO's small-scale aquaponics systems for joint production of lettuce (Lactuca sativa) and fish (Carassius auratus)

Suitability and optimization of FAO’s small-scale aquaponics systems for joint production of lettuce (Lactuca sativa) and fish (Carassius auratus)

Metadata

Abstract

Summary

Hydroponic subsystem effect on lettuce yield

Fish production and design effect

Water consumption and footprint

Linked claims

Citations to chase

Extraction notes

Data Tables

Trial Parameters

perezurrestarazuSuitabilityOptimizationFAO2019-T1

perezurrestarazuSuitabilityOptimizationFAO2019-T2

perezurrestarazuSuitabilityOptimizationFAO2019-T3

perezurrestarazuSuitabilityOptimizationFAO2019-T4

perezurrestarazuSuitabilityOptimizationFAO2019-T5

perezurrestarazuSuitabilityOptimizationFAO2019-T6

Graph View

Table of Contents

Backlinks