Comparisons between Aquaponic and Conventional Hydroponic Crop Yields: A Meta-Analysis

Metadata

Cite key: ayipioComparisonsAquaponicConventional2019
Item type: Journal Article
Authors: E. Ayipio, D.E. Wells, A. McQuilling, A.E. Wilson
Affiliation: Department of Horticulture, Auburn University, Auburn, AL, USA (Ayipio, Wells); CSIR-Savanna Agricultural Research Institute, Nyankpala-Tamale, Ghana (Ayipio, secondary affiliation); Department of Energy and Environment, Southern Research, Birmingham, AL, USA (McQuilling); School of Fisheries, Aquaculture, and Aquatic Sciences, Auburn University, Auburn, AL, USA (Wilson)
Journal: Sustainability 11(22) (2019) 6511
Date: 11/2019 (Received 22 October 2019; Accepted 11 November 2019; Published 19 November 2019)
Date added: [not reported]
DOI: 10.3390/su11226511
Funding: “This research received no external funding.” Acknowledgments thank Adelia Grabowsky for input on initial drafts, Adelia Grabowsky and Claudine Jenda jointly for help fine-tuning literature-search terms, Carlo Nicoletto for responding to a request for study details, and three anonymous reviewers.
URL: https://doi.org/10.3390/su11226511
PDF: sustainability-11-06511.pdf (MDPI’s DOI-style download filename — not yet renamed to the vault’s Author et al. - YEAR - Title.pdf Zotero convention; treated identically to any other paper per extraction rules)

Opinion

A genuinely quantitative meta-analysis (PRISMA-informed screening, a defined effect-size metric, random-effects pooling, subgroup analysis, and meta-regression), correctly self-labelled meta-analysis rather than a narrative review despite the MDPI section header reading “Review.” The authors are unusually candid about the analysis’s own limits: they chose an unweighted bootstrapped pooling method specifically because most of the 22 underlying studies never reported a variance measure, and they flag repeatedly (abstract, results, discussion) that the headline “no significant difference” result should be read with caution for exactly that reason. The subgroup work is the more useful output — nutrient supplementation is the one moderator whose pooled response-ratio confidence interval clearly excludes the null (2.96, CI 1.24–7.08), which lines up with the wider vault’s evidence that unsupplemented aquaculture effluent is usually deficient in plant micronutrients. Two numbers in the paper do not reconcile on close reading (see Extraction notes): the year-by-year study count in §3.1 sums to 31 against a stated total of 22 studies, and the aquatic-organism heterogeneity contribution is given as both 43.9% (§3.4) and 37% (§4.2) for what reads as the same statistic. Neither affects any cell in this vault’s CSVs, since a meta-analysis contributes no trial rows here — but a reader citing either number should check the original PDF. The paper is also unusually self-effacing about its own generalisability: with only 50 effect sizes across 22 studies and I²=100% heterogeneity, the authors are explicit that “pooled comparison is not quite feasible, although has been attempted.”

Abstract

Aquaponic is a relatively new system of farming, which has received much research attention due to its potential for sustainability. However, there is no consensus on comparability between crop yields obtained from aquaponics (AP) and conventional hydroponics (cHP). Meta-analysis was used to synthesize the literature on studies that compared crop yields of AP and cHP. Factors responsible for differences were also examined through subgroup analysis. A literature search was conducted in five databases with no time restriction in order to capture any publication on AP and cHP crop yield comparisons. The search was, however, limited to journal and conference articles published in English. Study characteristics and outcome measures of food crops were extracted. A natural log response ratio effect size measure was used to transform study outcomes. An unweighted meta-analysis was conducted through bootstrapping to calculate overall effect size and its confidence interval. Between-study heterogeneity (I²) was estimated using a random effects model. Subgroup and meta-regression were used to assess moderators, in an attempt to explain heterogeneity in the effect size. The results showed that although crop yield in AP was lower than conventional cHP, the difference was not statistically significant. However, drawing conclusions on the overall effect size must be done with caution due to the use of unweighted meta-analysis. There were statistically significant effects of aquatic organism, hydroponic system type, and nutrient supplementation used in the studies on crop yield comparisons. Nutrient supplementation, particularly, led to on average higher crop yield in AP relative to cHP. These findings are a vital information source for choosing factors to include in an AP study. These findings also synthesize the current trends in AP crop yields in comparison with cHP.

Summary

The authors systematically searched five databases (Web of Science, CAB Abstracts, Agricola, ASFA, ProQuest Dissertations and Theses Global) for studies directly comparing aquaponic (AP) and conventional hydroponic (cHP) crop yield, screened them against PRISMA-style criteria, and pooled 50 effect sizes from 22 eligible studies (2009–2018) using the natural log response ratio (lnR = ln(YAp/YHp)) as the outcome measure. Because most primary studies did not report a variance measure, they used an unweighted, bootstrapped random-effects (REML) model to estimate the overall pooled effect and ran subgroup analyses plus meta-regression to explore the very high between-study heterogeneity (I²=100%). The headline finding is that pooled AP yield was somewhat lower than cHP but not statistically significantly so (overall log-scale effect size −0.19, 95% CI −0.43 to 0.067). Six of seven tested moderators (aquatic organism, age uniformity, aquaponic coupling type, hydroponic system type, grow media, crop species) had pooled response-ratio confidence intervals that included 1 (no clear net direction); only nutrient supplementation had a confidence interval that excluded 1, with supplemented systems averaging a substantially higher AP-to-cHP yield ratio than unsupplemented ones. A funnel plot and Rosenthal’s fail-safe number (353, versus a stated inconsequential-bias threshold of 120) were used to check for publication bias. The paper’s own conclusion is cautious rather than triumphant: pooled comparison across such heterogeneous, mostly unweighted studies is only a “rough idea of the trend,” and the authors call for more consistent reporting of variance measures in future primary aquaponics-versus-hydroponics trials so a proper weighted meta-analysis becomes possible.

Experiment data

Location: Not a sited trial — a global literature synthesis. Search executed by the authors at Auburn University (Auburn, AL, USA); underlying primary studies are not geographically restricted and are not individually geolocated in this paper.

Design: PRISMA-informed literature search (5 databases, search terms "aquaponic* AND hydroponic* AND crop yield" for Agricola/CAB Abstracts and "(aquaponic* OR recirculat* aquaculture AND hydroponic*) AND TOPIC: (crop* yield OR crop* growth OR vegetable*)" for Web of Science/ASFA/ProQuest, search began 4 March 2019, no publication-date limit, English-language only, reviews and editorials excluded) → PRISMA screening (4 inclusion criteria: compares AP with conventional HP; food crop; replicated controlled trial; reports a mean and sample size — variance measures not required) → random-effects meta-analysis of the natural log response ratio (Eq. 1, lnR = ln(YAp/YHp)), REML estimator, unweighted bootstrapped estimation of overall effect and 95% CI (1000 bootstraps, OpenMEE software) → subgroup analysis (OpenMEE + R metafor) on aquatic species, AP coupling type, HP system type, grow media, crop, nutrient-supplementation status → meta-regression (R meta package, bubble function) for the two continuous moderators only (fish mean stocking density, feed crude protein content).

Replicates / n: “In total, 50 effect sizes were obtained from 22 studies” (p.3, §2.4) — this 22-study total is independently cross-checked by the paper’s own fail-safe-number arithmetic (5k+10 = 5×22+10 = 120, matching the stated “greater than 120” threshold, p.5–6). ⚠️ See Extraction notes — a year-by-year study count given later in the same section (p.4) sums to 31, not 22.

Duration: Not a single trial duration — pooled primary studies were published 2009–2018 (per p.4’s year distribution); literature search itself began 4 March 2019 with no end date stated in the main text.

Organisms: Aquatic species and crops vary across the 22 pooled primary studies and were not themselves raised or grown by these authors (see Extraction notes on why no Meta/Fish/ or Meta/Plant/ tag was applied). Most-represented fish: tilapia (K=8), carp (K=6, includes koi); most-represented crop: lettuce, Lactuca sativa (K=10), with spinach, strawberry, tomato, basil, cucumber, eggplant and babyleaf also pooled in smaller numbers.

Statistics: Random-effects model with REML estimator; heterogeneity via I² (classified 0–25% small, 30–50% moderate, ≥75% substantial, per the authors’ own bins, citing Higgins & Thompson 2002); publication-bias assessment via Rosenthal’s fail-safe number and a sample-size-based funnel plot (both via R metafor); meta-regression (R meta, REML) for continuous moderators only.

Effect size (statistics): Overall pooled effect size (log scale) = −0.19 (95% CI −0.43 to 0.067) — AP lower than cHP on average, but the CI spans zero, so not statistically significant (Figure 1, p.5).

Effect size (statistics): Response-ratio subgroup means (linear scale, back-transformed, mean [95% CI], Figure 4 p.6): Aquatic Organism 0.94 [0.66, 1.33]; Age Uniformity 0.85 [0.32, 2.24]; Aquaponic (coupling) Type 0.77 [0.27, 2.20]; Hydroponic Type 0.73 [0.37, 1.44]; Grow media 1.01 [0.51, 1.99]; Crop 1.00 [0.63, 1.59]; Nutrient Supplemented 2.96 [1.24, 7.08] — the only subgroup mean whose CI excludes 1 (parity between AP and cHP).

Heterogeneity explained by moderator (% of total heterogeneity): Hydroponic system type 11.25% (p.6); grow media 0.44% (p.6); nutrient supplementation 29.43% (p.9, “results not shown”); aquatic organism ⚠️ given as both 43.9% (p.8) and 37% (p.11) — see Extraction notes.

Publication bias: Rosenthal’s fail-safe number (FSN) = 353, against a stated “inconsequential bias” threshold of >120 (5k+10, k=22); funnel plot (Figure 3, p.6) built on sample size rather than standard error (because of the unweighted approach) and described by the authors as visibly asymmetric.

Overall AP vs. cHP yield comparison and heterogeneity

This paper: Across all 50 pooled effect sizes, AP crop yield was on average lower than cHP but the difference was not statistically significant (log-scale effect size −0.19, 95% CI −0.43 to 0.067; Figure 1). The authors immediately qualify this: between-study heterogeneity was “substantial” by their own classification (I²=100%, not shown as a numeric CI), which they attribute to how idiosyncratic individual AP systems are — “no two AP systems are identical. Small modifications lead to substantial differences” (p.10) — and they state explicitly that “pooled comparison is not quite feasible, although has been attempted in the current study” (p.10), framing the headline effect size as a rough trend indicator rather than a precise population estimate. Effect sizes also varied considerably by publication year (Figure 2): apart from 2009 and 2017, every other study year favoured cHP over AP.

Compared with:

Delaide et al. 2016 — cited (ref [24]) as one of the studies showing nutrient-supplemented AP can match or outperform cHP; already in vault as delaideLettuceLactucaSativa2016.md.
Jordan et al. 2018 — cited (ref [25]) for a coconut-fiber/crushed-stone substrate combination giving higher AP lettuce yield; already in vault as jordanYieldLettuceGrown2018.md.
Goddek and Vermeulen 2018 — cited (ref [26]) among studies supporting nutrient supplementation improving AP comparability with cHP; already in vault as goddekComparisonLactucaSativa2018.md.
todo Blidariu, Drasovean and Grozea 2013 — phosphorus level in green lettuce, conventional vs. aquaponic (ref [10]), cited among studies finding AP crop yields lower than cHP.
todo Roosta and Hamidpour 2013 — tomato mineral nutrient content in aquaponic vs. hydroponic systems (ref [11]), cited among studies finding AP crop yields lower than cHP.
todo Reyes-Flores et al. 2016 — aquaponics nutrient concentration in effluent for tomato production (ref [12]), cited among studies finding AP crop yields lower than cHP.
todo Alcarraz et al. 2018 — lettuce (Lactuca sativa) quality in aquaponic and hydroponic systems (ref [15]), cited among studies finding higher or similar AP crop yield vs. cHP.
Anderson et al. 2017 — butterhead lettuce growth/tissue elemental composition, hydroponic vs. aquaponic (ref [16]), cited among studies finding higher or similar AP crop yield vs. cHP; already in vault as andersonGrowthTissueElemental2017.md.

Nutrient supplementation

This paper: Nutrient supplementation was the moderator with the clearest effect: supplemented systems had a pooled response ratio of 2.96 (95% CI 1.24–7.08, Figure 4), the only subgroup CI excluding 1, and it accounted for 29.43% of overall heterogeneity (p.9). Figure 8B shows supplemented studies (n=25 effect sizes) clustering above the null line and unsupplemented studies (n=25) clustering below it — i.e., unsupplemented AP tended to under-yield cHP while supplemented AP tended to match or exceed it. Of the 22 pooled studies, 13 did not supplement their aquaculture effluent at all (p.4). The authors frame this as expected, since aquaculture effluent is “usually low in essential crop nutrients required for optimum plant growth” (p.2), especially micronutrients, and note that at least chelated iron supplementation appears necessary for healthy AP plant biomass (p.11), while flagging that heavy reliance on synthetic supplementation could work against AP’s own sustainability rationale (p.10).

Compared with:

todo Goddek et al. 2018 — nutrient mineralization and organic-matter reduction performance of RAS-based sludge in sequential UASB-EGSB reactors (ref [13]), cited as an alternative (anaerobic sludge digestion) route to recovering nutrients without external supplementation.
todo Buhmann, Waller, Wecker and Papenbrock 2015 — halophyte biofiltration for nutrient-rich saline water (ref [40]), cited re: chelated-iron supplementation being required for healthy AP plant biomass.
todo Bittsánszky et al. 2016 — nutrient supply of plants in aquaponic systems (ref [43]), cited to support that fish effluent is typically nutrient-poor, especially in micronutrients.
todo Goddek et al. 2015 — challenges of sustainable and commercial aquaponics (ref [44]), cited on the tension between nutrient supplementation and AP’s sustainability goals.

Hydroponic system type and grow media

This paper: Hydroponic system type accounted for 11.25% of total heterogeneity (p.6); pooled by system, deep-water culture (DWC, K=9 studies) showed the best average AP performance (mostly positive effect sizes beyond the null), media-based systems (K=6) were intermediate, and nutrient film technique (NFT, K=7) performed worst (Figure 5A). Grow media contributed comparatively little to heterogeneity (0.44%, p.6): pooled media-based AP performed poorly overall, but organic media specifically outperformed inorganic media types (Figure 5B) — light expanded clay aggregate (LECA, K=1 study) had a favourable response ratio, while perlite (K=3) performed worst among named media, and studies using no distinct grow media (DWC/NFT, grouped as “none,” K=15) were intermediate.

Compared with:

todo Sirakov et al. 2017 — comparison of two production technologies and two substrate types in an experimental aquaponic system (ref [50]), cited re: lettuce performing better in raft (DWC) than LECA media-bed technology.
todo Schmautz et al. 2016 — tomato productivity/quality across three hydroponic methods in aquaponics (ref [51]), cited as a contrasting case where HP system type had no significant influence on tomato yield.
todo Roosta and Afsharipoor 2012 — cultivation-media effects on strawberry growth in hydroponic vs. aquaponic systems (ref [14]), cited re: organic/inorganic media-ratio effects on AP crop yield.

Crop species

This paper: Lettuce (K=10 studies) generally showed similar or better AP performance relative to cHP than other crops (Figure 6), which the authors attribute to most lettuce trials using DWC and/or nutrient-supplemented effluent. Tomato (n=5 effect sizes), eggplant (n=1) and spinach (n=1) showed no clear AP/cHP difference, though the authors caution these are inconclusive given so few effect sizes per crop; the babyleaf result (n=13) is flagged as coming from a single study and therefore potentially subject to within-study bias rather than a genuine crop effect. Crop species overall was a comparatively weak moderator (pooled response ratio 1.00, 95% CI 0.63–1.59, i.e. essentially no net effect once pooled across crops), and the authors attribute crop-level differences to nutrient demand — fruiting crops like tomato and cucumber need more nutrients than most AP systems (especially unsupplemented ones) can supply.

Compared with:

Buzby and Lin 2014 — evaluating aquaponic crops in a freshwater flow-through fish culture system (ref [27]/[52]), cited re: nutrient uptake being strongly influenced by crop species; already in vault as buzbyScalingAquaponicSystems2014.md.

Aquatic organism / fish species and age composition

This paper: Tilapia (K=8) and carp (K=6, all carp types including koi pooled together) were the most-studied fish; rainbow trout and Pangasius each appeared in two studies, with crayfish and shrimp also represented. The paper reports no significant net difference between aquatic organisms overall (pooled response ratio 0.94, 95% CI 0.66–1.33), but flags substantial subgroup-level heterogeneity (I²>80%) and states aquatic organism type explains a large share of overall heterogeneity — ⚠️ given as both 43.9% (p.8) and 37% (p.11), see Extraction notes. Effluent from tilapia, carp and Pangasius was associated with poorer relative AP-vs-cHP yield (Figure 7A); results for perch, crayfish, shrimp and catfish are described as inconclusive due to small sample sizes, and the Pangasius result is flagged as coming from a single study. Fish age composition also mattered: homogeneous-age fish populations (K=12) were associated with lower relative AP crop yield than heterogeneous-age populations (K=9), which the authors suggest is confounded by heterogeneous-age studies disproportionately using higher-effect-size taxa (crayfish, shrimp, catfish).

Compared with:

todo Knaus and Palm 2017a — effects of fish biology on ebb-and-flow aquaponically cultured herbs in northern Germany (ref [32], Aquaculture 466), cited re: fish species influencing crop species choice and yield.
todo Pinho, Lemos de Mello, Fitzsimmons and Emerenciano 2018 — integrated pacu/red-tilapia production with scallion and parsley garnish (ref [33]), cited as a contrasting case finding no significant fish-species effect on crop yield.
todo Knaus and Palm 2017b — effects of fish species choice on vegetables in aquaponics, spring–summer conditions, northern Germany (ref [53], Aquaculture 473), cited for the subgroup finding that tilapia and carp have similar effects on AP/cHP yield comparison.

Aquaponic coupling type (recirculating vs. decoupled)

This paper: Both coupling types (recirculating/single-loop, K=20, and decoupled/multi-loop, K=1) showed lower relative AP crop yield than cHP (Figure 8A; pooled aquaponic-type response ratio 0.77, 95% CI 0.27–2.20). One study did not report coupling type. The authors caution strongly against generalizing the decoupled result, since it comes from a single study, and argue that coupling type alone does not determine yield outcomes — what matters is whether growth conditions (pH, nutrient concentration) are actively adjusted for the HP side, citing a case where a decoupled system underperformed cHP specifically because conditions were not adjusted.

Compared with:

todo Pickens 2015 — PhD thesis integrating RAS effluent with greenhouse cucumber and tomato production (ref [56]), cited for lower cucumber yield in a decoupled AP system vs. cHP fertilizer, attributed to unadjusted growth conditions.
Goddek et al. 2016 — system-dynamics design approach for decoupled aquaponic systems (ref [54]), cited for the definitions of recirculating/single-loop vs. decoupled/multi-loop coupling; already in vault as goddekNavigatingDecoupledAquaponic2016.md.

Stocking density and feed crude protein content (meta-regression)

This paper: Neither fish mean stocking density (range 0.5–~53 kg·m⁻³ across studies; 8 of 22 studies did not report it; most common values 7 kg·m⁻³ and 6.4 kg·m⁻³, K=3 each) nor feed crude protein content (range 30–48% across most studies, with one outlier study’s feed CP estimated at 84% from its feed description) showed a significant relationship with effect size in meta-regression (Figure 9), and removing the identified outlier did not change the conclusion (per the paper’s Supplementary Materials, not read as part of this extraction). The authors conclude that, within the range of practice represented in the pooled literature, increasing stocking density or feed protein content did not translate into higher AP crop yield relative to cHP.

Compared with: (no external comparisons cited in this subsection beyond the paper’s own meta-regression)

Publication bias

This paper: Publication bias was assessed via a funnel plot (built on sample size rather than standard error, because of the unweighted analytic approach) and Rosenthal’s fail-safe number. The observed FSN of 353 exceeds the stated inconsequential-bias threshold of >120 (5k+10, k=22 studies), which the authors cite as evidence against publication bias being a major concern — while simultaneously noting the funnel plot itself appeared visibly asymmetric, and that the unweighted method and reliance on a formula from Brown and Sutton (cited as ref [49], though printed in-text as “Bown, M.J.; Sutton, A.J.” — see Extraction notes) both limit how much weight to place on this reassurance.

Compared with: (methodological reference only — no additional primary aquaponics studies)

Linked claims

Citations to chase

Extraction notes

Type classification (meta-analysis): Straightforward call. The authors pool effect sizes (natural log response ratio) across 22 independently-published primary studies using a formal random-effects model, run subgroup analysis and meta-regression on moderators, and collect no new empirical data of their own — this is the textbook meta-analysis case in SCHEMA.md, not systematic-review (there is no evidence a full PRISMA flow diagram/count reconciliation was produced — see the K=22 vs. K=31 flag below) and not narrative-review (a quantitative pooled effect size with a stated CI is present, which narrative reviews by definition lack). Per CLAUDE.md’s “Choose the right template block” section, meta-analysis uses the Experiment data callout, not Review scope — this is what was used above, matching the precedent set by notes/gargaroLetUsInvestigate2023.md (another meta-analysis note in this vault) and notes/montanhiniNutrientLoadEstimation2015.md (a modelling-type note using the same callout).

No CSV rows produced. Per SCHEMA.md’s explicit warning (“Never create trial rows from a meta-analysis or review… If a meta-analysis cites a primary study you do not have, add it to Citations to chase — do not enter its numbers”), out/ayipioComparisonsAquaponicConventional2019.trials.csv and .plant.csv were written with header rows only (87 and 11 columns respectively, verified with a Python csv-module row/column count) and zero data rows. The 22 primary studies this paper pools are treated exactly as SCHEMA.md instructs: as Citations to chase entries where not already in the vault, never as trial rows here.

⚠️MATERIAL — Number of studies pooled, p.3–4: §2.4 (p.3) and the abstract both state “In total, 50 effect sizes were obtained from 22 studies.” This total (k=22) is independently corroborated by the paper’s own fail-safe-number arithmetic (p.5–6): “a fail-safe number greater than 5k + 10… would make publication bias inconsequential… a fail-safe number greater than 120,” and 5×22+10 = 120 exactly. However, §3.1 (p.4) gives a year-by-year breakdown of “number of publications (K)” that reads: “2009 (K = 1), 2011 (K = 10), 2012 (K = 2), 2014 (K = 1), 2015 (K = 1), 2016 (K = 4), 2017 (K = 4), and 2018 (K = 8)” — summing to 31, not 22. By contrast, several other categorical breakdowns in the same section do sum correctly to 22 (age composition 12+9+1=22; AP coupling 20+1+1=22; HP system type 9+6+7=22; grow media 1+3+3+15=22), which suggests the year-distribution row specifically is where the error sits (plausibly the “2011 (K=10)” figure, an unusually large jump from the surrounding single-digit years, though the paper gives no way to confirm which digit is wrong). Recorded above as 22 studies (the headline figure, cross-validated by the FSN calculation); the 31-summing year breakdown is presented in this note as-printed but not treated as authoritative. This does not affect any CSV cell (meta-analysis produces no rows) but affects how the “steady growth 2009→2018” narrative in §3.1 should be read.

⚠️MATERIAL — Aquatic-organism heterogeneity contribution, p.8 vs. p.11: §3.4 Results (p.8, “1. Aquatic Organism”) states: “Aquatic organisms accounted for 43.9% of the heterogeneity in estimating the overall effect size.” §4.2 Discussion (p.11, “Subgroup Analysis”) states: “aquatic organism type accounted for 37% of the heterogeneity in the effect sizes obtained.” Both sentences describe what reads as the identical statistic (aquatic-organism moderator’s share of total heterogeneity) with no distinguishing qualifier (e.g. before/after outlier removal — the outlier removal is mentioned only for the stocking-density/feed-CP meta-regression in §3.5/Figure 7 caption, not tied explicitly to either heterogeneity percentage). Recorded both values above with their locations; neither is preferred over the other, and this does not affect any CSV cell since the paper contributes no trial rows.

⚠️MINOR — Author-name rendering, p.10–11 vs. References list: In-text citation on p.10 reads “the cutoff by Brown and Sutton [49],” but the References list (p.11, item 49) gives the authors as “Bown, M.J.; Sutton, A.J.” (“Bown,” not “Brown”). Likely a running-text typo rather than a reference-list error, since “Bown” is a real, more distinctive surname and the reference entry is internally consistent (initials, journal, year all check out for a real 2010 Eur. J. Vasc. Endovasc. Surg. paper on systematic-review quality control). Recorded as [[Bown and Sutton 2010]] in Extraction notes only — this is a methodological citation, not a primary aquaponics study, so it was not added to Citations to chase.

[not reported] fields, grouped (note-only, no CSV cells affected):

Zotero “Date added” (no zotero-export.csv present for this batch).
An explicit numbered list or table of exactly which 22 studies (by reference number) were pooled into the meta-analysis. The paper’s Introduction and Discussion cite roughly 20+ individual primary studies by reference number in the context of AP/cHP yield comparisons (refs [10]–[33], [50], [51], [53], [56] are the ones treated as likely pool members above), but no PRISMA-style appendix table in the main text names the definitive 22. The Supplementary Materials (referenced at http://www.mdpi.com/2071-1050/11/22/6511/s1) were not fetched or read as part of this extraction — the “Citations to chase” list above should be treated as probable pool members inferred from in-text discussion, not a confirmed enumeration.
Exact geographic distribution of the 22 pooled primary studies (not stated as a summary anywhere in the main text).

[unclear] fields:

Whether the printed “2011 (K = 10)” figure in the year-distribution breakdown (p.4) is the actual source of the 22-vs-31 discrepancy noted above, versus some other year’s figure, or a double-counting of one/more studies across adjacent years — the paper gives no way to determine which.
Whether the 43.9% vs. 37% aquatic-organism heterogeneity figures (see above) reflect two different calculation methods (e.g. with vs. without the Pangasius or catfish outlier explicitly called out in Figure 7’s caption) or are simply a typo/revision-inconsistency between the Results and Discussion sections written at different times.

Tags judgment call — no Meta/Fish/ or Meta/Plant/ facet applied: Per CLAUDE.md, “Only tag an organism if the paper studied it.” This meta-analysis did not raise fish or grow any crop itself; it pooled effect sizes from 22 other studies spanning at least 7 fish/aquatic taxa (tilapia, carp, rainbow trout, Pangasius, crayfish, shrimp, perch, catfish) and at least 8 crop species (lettuce, spinach, strawberry, tomato, basil, cucumber, eggplant, babyleaf), with no single organism dominant enough to characterize the paper as being “about” it in the sense the existing Meta/Fish/Tilapia or Meta/Plant/Lettuce tags are used elsewhere in the vault (e.g. in andersonGrowthTissueElemental2017.md, where lettuce was the one crop actually grown). This mirrors the precedent in notes/gargaroLetUsInvestigate2023.md (a closely comparable meta-analysis pooling multiple lettuce-adjacent CEA studies), which likewise carries no Meta/Plant/ tag. Meta/Region/Global was applied because the underlying evidence base spans many countries with no single-site fieldwork of the authors’ own.

Cross-references already pointing at this paper, now resolvable: Three existing vault notes cite this exact paper via #todo placeholders that can now be linked to this note instead: notes/gargaroLetUsInvestigate2023.md (“Ayipio Wells McQuilling Wilson 2019… not yet in vault”), notes/baniowdehBarleyHordeumVulgare2025.md (“Ayipio Wells McQuilling Wilson 2019” re: aquaponics achieving comparable yields despite lower nutrient concentrations), and notes/emerencianoMineralSupplementationJade2025.md (“Ayipio et al. 2019” re: AP yield lower than HP on average but not significantly so). Caution: notes/baniowdehBarleyHordeumVulgare2025.md also cites a different paper as “Ayipio Kim Kim Roh 2019” (a distinct paper with co-authors Kim, Kim, and Roh — not this one, which has co-authors Wells, McQuilling, and Wilson). These two “Ayipio 2019” citations must not be conflated when resolving the vault’s #todo links.

PDF quality: Clean, fully extractable text layer (MDPI-typeset); no OCR artifacts; all 15 pages (main text + references) read in full. All in-text figures (1–9) are decorative/data-visualisation only — no numeric table of raw study-level data was present in the main text to extract beyond what is summarised above.

Source: sustainability-11-06511.pdf

Quartz 5

Explorer

Comparisons between Aquaponic and Conventional Hydroponic Crop Yields: A Meta-Analysis

Comparisons between Aquaponic and Conventional Hydroponic Crop Yields: A Meta-Analysis

Metadata

Abstract

Summary

Overall AP vs. cHP yield comparison and heterogeneity

Nutrient supplementation

Hydroponic system type and grow media

Crop species

Aquatic organism / fish species and age composition

Aquaponic coupling type (recirculating vs. decoupled)

Stocking density and feed crude protein content (meta-regression)

Publication bias

Linked claims

Citations to chase

Extraction notes

Graph View

Table of Contents

Backlinks