Closing Nutrient Cycles through the Use of System-Internal Resource Streams: Implications for Circular Multitrophic Food Production Systems and Aquaponic Feed Development

Abstract

In order to further close nutrient cycles of aquaponic systems, it could be possible to integrate a third trophic level in the form of insect larvae production (i.e., black soldier fly larvae) to recycle internal waste streams into valuable nutrients. This would present opportunities to formulate sustainable circular aquafeeds that combine these internally available nutrients with complementary external raw materials. The ingredient composition of feeds for such circular multitrophic food production systems (CMFS) may affect fish performance as well as excretion of important dissolved plant nutrients such as N, P and K. Hence, fish meal from catfish processing (CM) as base ingredient was combined with variable levels of poultry by-product meal (PM) and black soldier fly larvae meal (BSFM) into three marine-ingredient-free experimental diets corresponding to hypothetical production scenarios of a CMFS that aims to integrate aquaponics with insect larvae production. These experimental diets and a commercial diet (COM) were compared using isonitrogenous and isolipidic formulations. They were fed to African catfish (Clarias gariepinus) in recirculating aquaculture systems (RAS) and evaluated concerning growth performance and nutrient excretion. All diets resulted in similar total inorganic nitrogen (TIN) excretion, whereas the increase of dietary PM inclusion from 0% (BSF diet) to 20% (MIX diet) and to 41% (PM diet) and concomitant reduction of BSFM inclusion led to increasingly higher soluble reactive phosphorus (SRP) excretion per unit of feed compared to the COM diet. While the PM diet enabled the best growth and feed conversion performance, the MIX and especially the BSF diet produced more similar performance to the COM diet, which generated the highest dissolved K excretion. The MIX and the PM diet resulted in the highest Ca and P, yet lower N content in the fish feces. Results indicate that combining CM with elevated levels of PM in the diet of African catfish could improve growth performance and reduce the need for P fertilization in aquaponics when compared to industrial diets optimized for low environmental impact. Findings are discussed regarding their implications for CMFS and aquaponic feed formulation.

1. Introduction

In order to sustainably achieve the projected growth of the global aquaculture industry, environmentally sound and resource-efficient practices are required [1]. Land-based closed recirculating aquaculture systems (RAS) are advocated as one of the technological alternatives to open system aquaculture, mitigating nutrient-induced environmental pollution while enabling intensive production with greatly reduced water consumption [2,3,4]. Aquaponic systems, particularly on-demand coupled systems [5], which use the nutrient-rich RAS water for the irrigation and nutrition of plants in hydroponics, are a sustainable extension of RAS technology. However, despite making dual use of the process water as well as productively taking advantage of otherwise wasted dissolved nutrients by combining fish and plant production, aquaponic systems still produce biogenic waste streams such as aquaculture sludge and harvest wastes from plant and fish production. In order to further close the nutrient cycle of aquaponic systems and improve their overall sustainability, such resource streams could be internally recycled into higher value nutrients by a third trophic production level in the form of insect larvae culture of, e.g., the black soldier fly (BSF) (Hermetia illucens), a species known to be an efficient converter of a wide array of biowastes [6,7,8]. These higher value nutrients could then be reintroduced into the internal nutrient cycle of the system via the fish feed, and/or removed from the system as a value-added product. Therefore, integrating fish, plant and insect larvae production into a circular multitrophic food production system (CMFS) [9,10,11] may present a potential pathway towards further optimizing the use of input water, energy and nutrients beyond aquaponic systems.

In order to achieve true circularity, ensure optimal organism health and performance and offset the nutrient loss from harvest within such a system, internally available nutrient streams have to be combined with complementary nutrients from system-external sources. Subsequently, this combination of resources has to be recycled as efficiently as possible between the three trophic production levels. On the fish production side, this means internally produced insect larvae—e.g., fed a blend of aquaculture sludge, harvest wastes from plant production and complementary external feed substrates—could be combined with the fish processing waste—the nutritionally most valuable internal waste stream [12,13,14,15]—and finally complemented with externally sourced feed ingredients in a way that creates a complete feed that covers the nutritional needs of the fish. Sensibly selecting these externally sourced feed ingredients can have positive impacts on overall system performance. Since most plant nutrients accumulate insufficiently in RAS water for optimal aquaponic plant production, additional mineral fertilization is required [16,17,18,19,20,21,22,23,24,25,26]. This opens up the opportunity to complement insect larvae and fish processing waste with external feed ingredients, which not only ensure good fish performance, but particularly also increase the excretion of important plant nutrients as suggested by recent work on improving diets for use in aquaponics [10,11,27,28,29,30,31]. In this sense, the traditional objective in aquaculture feed formulation of minimizing the excretion of eutrophication-inducing nutrients such as P [32,33,34,35] does not apply and aquaponic or CMFS diets can feature a substantially higher inclusion of, e.g., phosphorus-rich ingredients such as poultry by-product meal (PM). PM is an established alternative to fish meal [36,37,38,39], and has not only been proven to enable good fish performance even when used as the sole protein source in diets for Nile tilapia (Oreochromis niloticus) and African catfish (Clarias gariepinus), but also to increase soluble reactive phosphorus (SRP) excretion versus an entirely marine fish meal-based diet [10,11].

Therefore, this study set out to formulate exemplary experimental diets for African catfish that follow the logic of CMFS as outlined above, i.e., incorporating BSF larvae meal (BSFM) and catfish by-product meal (CM) as internally available protein sources together with poultry blood meal (PBM) and poultry by-product meal (PM) as externally sourced protein ingredients according to hypothetical CMFS production scenarios. The objective was then to compare these diets to a commercial diet in terms of (1) the growth performance achieved by African catfish reared in RAS; (2) the dissolved nutrient profiles established in the RAS water, particularly with regard to the major plant macronutrients N, P and K; (3) the partitioning between solid (feces) and dissolved nutrient excretion. In this sense, the goal of the study was to evaluate exemplary purpose-specific circular fish feeds for aquaponics and CMFS, specifically with regard to fish performance and improvement of dissolved nutrient profiles for aquaponic plant production, and to interpret results in terms of their significance with respect to the relative size and/or allocation of system-internal insect larvae production. Findings could foster progress in the developing field of aquaponic feed research and help to identify effective and efficient recycling pathways in CMFS.

2. Materials and Methods

2.1. Experimental Diets and Formulation Rationale

Under the conservative assumptions that commercial African catfish production can operate at a feed conversion ratio (FCR) of 0.9 [40,41] and achieve a filet yield of 40% [42,43,44,45,46] and that dry fish meal can be produced from fresh processing waste with a ratio of 5:1 [46,47], catfish by-product meal produced internally within a CMFS could provide 13.3% of the ingredients required per unit of feed and serve as the base ingredient for CMFS diets. This notion can be extended by establishing hypothetical operational scenarios as outlined above and graphically represented in Figure 1. This vision of a CMFS explores the extent to which internal BSF larvae production is capable of providing a supplementary feed ingredient source and, vice versa, the extent to which fish by-product meal and insect larvae need to be complemented by external resources.

Benchmark scenarios include: (1) the insect larvae cultivation based on internal waste streams and perhaps external supplementation is large and productive enough such that it covers most of the remaining protein requirement of the fish diet beyond the 13.3% catfish by-product meal, i.e., primarily additional carbohydrate and lipid ingredients have to be provided by system-external resources; (2) insect larvae cultivation based on internal waste streams and perhaps external supplementation is reasonably large and productive, and/or harvest is partly rededicated to a higher value use and removed from the system such that it covers a part, e.g., 50%, of the remaining protein requirement of the fish diet beyond the catfish by-product meal (i.e., 50% of the additionally required dietary protein as well as the carbohydrate and lipid ingredients have to be provided by system-external resources); or (3) a higher value use justifies the rededication of the entire insect larvae production and thus removal from the system, i.e., all the remaining dietary ingredient requirements beyond the fish by-product meal have to be provided by system-external sources. This logic was applied to the ingredient composition of three experimental diets which were formulated in an isonitrogenous and isolipidic manner according to the analytically determined crude protein (CP) and crude fat (CF) content of a commercially available diet for juvenile African catfish (COM) (CP: 49.2%; CF: 18.8%). Each experimental diet included a fixed amount of 13.3% CM (processing by-products of catfish species from the family Siluridae) as a proxy for African catfish by-product meal as well as variable amounts of commercially available BSFM as a proxy for CMFS internally produced BSFM and PM. The BSF diet was devoid of PM (production scenario 1), the MIX diet featured a 50:50 mix of BSFM and PM (production scenario 2) and the PM diet was devoid of BSFM (production scenario 3).

Monoammonium phosphate and a vitamin/mineral premix were added equally to all diets according to the inclusion level of the COM diet (personal communication with producer) and also pea starch and rapeseed oil inclusion was kept constant, while the inclusion levels of corn meal and poultry fat were adjusted between diets to achieve the isonitrogenous and isocaloric objective. PBM needed to be added equally to all diets in order to reach the CP content of the COM diet and ensure a sufficiently high starch level for proper extrusion. Diets were extruded at SPAROS I&D, Olhão, Portugal, and stored at −20 °C until use. Diet formulation and proximate composition are presented in

Table 1. Experimental diet formulation and proximate composition.

	Experimental Diets
	PM	MIX	BSF	COM 14
Ingredient composition (%)
Black soldier fly larvae meal 1	-	20.35	40.00	n.a. 15
Poultry by-product meal 2	41.30	20.35	-
Catfish by-product meal 3	13.30	13.30	13.30
Poultry blood meal 4	10.00	10.00	10.00
Corn meal 5	11.15	10.55	10.05
Pea starch 6	8.00	8.00	8.00
Vitamin and mineral premix 7	0.65	0.65	0.65
Monoammonium phosphate 8	0.50	0.50	0.50
Rapeseed oil 9	7.50	7.50	7.50
Poultry fat 10	7.60	8.80	10.00
Proximate composition (%-as fed) 11
Dry matter (DM)	96.00	95.70	96.05	96.10
Moisture	4.00	4.30	3.95	3.90
Crude protein (CP) (N × 6.25)	50.70	50.40	50.70	49.20
Crude fat (CF)	20.30	20.05	20.05	18.80
Crude fibre	1.25	1.40	1.55	2.05
Ash	8.00	7.30	6.70	6.70
Starch	11.50	10.90	11.65	11.55
Nitrogen-free extract 12	15.8	16.6	17.1	19.4
Gross energy (MJ/kg) 13	22.71	22.68	22.84	22.38
P/E ratio (g protein/MJ GE)	22.32	22.22	22.20	21.98

¹ Protein X (defatted Hermetia illucens meal): 58% CP, 9% CF, Protix B.V., The Netherlands. ² Poultry by-product meal 65: 65.4% CP, 11.9% CF, SAVINOR UTS, Portugal. ³ Catfish by-product meal: 57.8% CP, 7.6% CF, Bioceval GmbH & Co. KG, Germany. ⁴ Poultry blood meal: 89% CP, 0.47% CF, SONAC, The Netherlands. ⁵ Corn meal: 8.6% CP; 4.3% CF, Casa Lanchinha, Portugal. ⁶ Pea starch: 90% starch, COSUCRA, Belgium. ⁷ PREMIX Lda, Portugal: Vitamins (IU or mg/kg diet): DL-alpha tocopherol acetate, 100 mg; sodium menadione bisulphate, 25 mg; retinyl acetate, 20,000 IU; DL-cholecalciferol, 2000 IU; thiamin, 30 mg; riboflavin, 30 mg; pyridoxine, 20 mg; cyanocobalamin, 0.1 mg; nicotinic acid, 200 mg; folic acid, 15 mg; ascorbic acid, 500 mg; inositol, 500 mg; biotin, 3 mg; calcium pantothenate, 100 mg; choline chloride, 1000 mg, betaine, 500 mg. Minerals (g or mg/kg diet): copper sulfate, 9 mg; ferric sulfate, 6 mg; potassium iodide, 0.5 mg; manganese oxide, 9.6 mg; sodium selenite, 0.01 mg; zinc sulfate, 7.5 mg; sodium chloride, 400 mg. ⁸ Windmill AQUAPHOS: 26% P, ALIPHOS, The Netherlands. ⁹ J.C. Coimbra, Portugal. ¹⁰ SAVINOR UTS, Portugal. ¹¹ Analyzed in duplicate; values represent the mean. ¹² NFE = 100% − (% CP + % CF + % CFB + % ash + % moisture). ¹³ Calculated using the factors 17.15, 23.64, 39.54 MJ/kg for NFE (carbohydrates), CP and CF, respectively [48]. ¹⁴ Ingredients according to producer: Soybean meal, fish meal, wheat, rapeseed oil, hydrolyzed feather meal, sunflower protein concentrate, poultry meal, hemoglobin meal (porcine), fish oil, monoammonium phosphate (0.5%), vitamin/mineral premix (0.65%). ¹⁵ Producer does not provide the percentage inclusion of ingredients.

, while amino acid and mineral composition are shown in

Table 2. Experimental diet amino acid and mineral composition.

	Experimental Diets
	PM	MIX	BSF	COM
Essential amino acids (EAAs) (% as fed)
Arginine (Arg)	2.68	2.68	2.69	2.94
Histidine (His)	1.31	1.28	1.24	1.22
Isoleucine (Ile)	1.67	1.68	1.67	1.78
Leucine (Leu)	3.89	3.93	4.03	3.83
Lysine (Lys)	2.51	2.46	2.48	2.58
Methionine (Met)	0.76	0.76	0.76	0.76
Phenylalanine (Phe)	2.27	2.31	2.33	2.22
Threonine (Thr)	1.63	1.63	1.61	1.79
Tryptophan (Trp)	0.50	0.50	0.49	0.50
Valine (Val)	2.19	2.17	2.21	2.53
Met + Cys	1.30	1.38	1.41	1.62
Phe + Tyr	3.48	3.61	3.73	3.41
Sum EAAs	19.38	19.37	19.49	20.11
Non-essential amino acids (NEAAs) (% as fed)
Alanine (Ala)	2.43	2.39	2.38	2.54
Cysteine (Cys)	0.54	0.62	0.65	0.86
Glycine (Gly)	2.58	2.49	2.40	2.90
Proline (Pro)	3.25	3.35	3.17	2.77
Serine (Ser)	2.20	2.30	2.42	2.75
Aspartic acid (Asp) + asparagine (Asn)	3.89	3.85	3.81	3.96
Glutamic acid (Glu) + glutamine (Gln)	9.06	9.02	9.04	6.83
Tyrosine (Tyr)	1.21	1.30	1.40	1.19
Sum NEAAs	25.15	25.31	25.26	23.78
Minerals (g/kg as fed) 1
P	20.21	16.76	13.90	11.98
Ca	16.93	15.02	13.04	11.66
K	6.21	5.94	5.58	10.74
S	4.80	5.41	6.08	6.86
Na	2.56	2.74	2.88	3.30
Mg	1.42	1.43	1.43	2.24
Fe	0.57	0.61	0.67	0.34
Zn	0.29	0.28	0.27	0.11
Mn	0.17	0.19	0.15	0.04
Al	0.11	0.18	0.24	0.07
Cu	0.03	0.03	0.03	0.01

¹ Analyzed in duplicate; values represent the means.

.

2.2. Experimental System Setup

The experimental system comprised 16 RAS (Figure 2) with an individual system water volume of 392 L. Each system consisted of a circular rearing tank featuring a conical bottom including a standpipe with a bottom outlet (160 L), a settling tank for feces collection (60 L), a moving-bed bioreactor (MBBR) (172 L) equipped with 35 L of biocarriers (Hel-X HXF12KLL, Christian Stöhr, Marktrodach, Germany) including clear water chambers before and after the biomedia-containing center section, which were separated by 5 cm thick filter sponge walls (PPI 20, Schaumstoff-Meister, Straelen, Germany). Each system had a heating system for temperature control (600 W titanium tube heater, SCHEGO Schemel & Goetz, Offenbach, Germany and T controller twin, AB Aqua Medic, Bissendorf, Germany), a frequency converter-controlled pump for water circulation (DC Runner 1.3, AB Aqua Medic, Bissendorf, Germany), a float valve for fresh water supply (Rojo RJV15-1/2”, Jobe Valves, Matamata, New Zealand) and a drip applicator for base dosing into the pump sump (HydroBag 1,5 L and GraviSet Easy Bag ENFit, Fresenius Kabi, Bad Homburg, Germany). An air stone in the fish tank as well as an aerator in the MBBR oxygenated the water and enabled continuous biomedia mixing. Fish tanks were covered with amply perforated lids made from nontransparent plastic to minimize stress and prevent fish from escaping, while general lighting was provided by overhead LED lamps.

The entire setup featured the option to connect all 16 RAS in series in order to synchronize biofilters and water quality before initiating trials so that starting conditions could be equalized between the individual RAS. This was achieved by the adjustment of ball valve settings such that each pump supplied water to the fish tank of the following RAS instead of recirculating the water of its own RAS, while the last pump again supplied the fish tank of the first RAS. To counteract possible discrepancies in flow rate between pumps, all biofilters featured a closable opening leading to a shared water distribution pipe which, when open, allowed water levels between the biofilters to freely equalize. To mature and synchronize the biofilters, the entire system was stocked with Nile tilapia for 6 weeks prior to starting the trial and all 16 RAS were connected in series. One day before initiating the trial, the system was emptied, cleaned, and refilled with fresh tap water (pH 8.1–8.3) and run in series for another 24 h before finally disconnecting the individual RAS and introducing the experimental fish.

2.3. Experimental Design

Mixed-sex African catfish fry were acquired from Nutrition and Food—Bioenergie Lüchow, Altkalen, Germany, and reared on a commercial standard diet at 26 °C in a flow-through system prior to the trial. Not being fed for 24 h before transfer, 58 fish were randomly assigned to each individual experimental RAS, of which 30 fish were individually weighed and measured (initial body weight: 8.8 ± 2.3 g; initial total length: 11.0 ± 0.1 cm).

Each dietary treatment was replicated in four randomly assigned RAS and total trial duration amounted to 49 days. Biomass per RAS was determined at the start and the end of the trial as well as at the end of week 2, 4 and 6 which enabled the calculation of biomass-based FCRs for the time intervals from week 0 to 2, 2 to 4, 4 to 6 and 6 to the end of the trial for each RAS. Fish were hand-fed twice per day with a total daily ration of 3.5% of the biomass and rations were increased daily and individually per RAS according to the projected biomass development calculated from the regularly determined biomass and the FCR of the preceding time interval. During the first two weeks in which FCRs were unavailable, rations across all RAS were increased based on the initial biomass of each RAS and a conservatively assumed FCR of 1. Fish were only fed half of the total ration on the day before biomass determination (in the morning) and on the day itself (in the evening). No feed was administered the day before concluding the trial. Diligent hand feeding made sure that all feed was consumed and none exited the fish tank. Mortalities were recorded daily, and all remaining fish were measured and weighed at the end of the trial.

Feces was removed daily from each RAS by siphoning from the sedimentation tanks and the sludge water mix was filtered through 90 µm nylon mesh. The resulting wet feces was pooled per RAS over the entire period of the trial and stored at −80 °C. A daily water exchange of 10% of the RAS volume (39.2 L) was performed by keeping the circulation pumps running and opening the exchange water outlets until the volumetrically defined markings in the biofilters were reached. The removed water was afterwards replenished with fresh tap water by opening the float valves until the water levels again reached the biofilters’ maximum determined capacity, whereupon the valves were closed again to prevent system water from being replenished in an uncontrolled manner. Water recirculation in the RAS was set at 240 L/h. This was controlled weekly and in case of deviation, pump settings were readjusted. In order to stabilize system water pH, sodium hydroxide (NaOH, ≥99.5% purity, Carl Roth, Karlsruhe, Germany) was introduced into the pump sump daily after water exchange from day 30 onward by dissolving it with 1 L of deionized water and adding it into the pump sump via the drip applicator over a time frame of roughly 12–18 h. The daily dosage of NaOH was increased equally in all tanks over the course of the trial and a total of 154.2 g of NaOH was added per tank. Prior to feeding and water exchange, oxygen, pH, temperature (HQ40d, Hach Lange, Berlin, Germany), and electrical conductivity (EC) (pH/Cond 740i, WTW, Weilheim in Oberbayern, Germany) were measured in each fish tank. Starting from day 1 of the trial, the RAS water and the tap water were sampled on a weekly basis for analysis of dissolved nutrient concentrations. The trial was conducted at the Leibniz Institute of Freshwater Ecology and Inland Fisheries in Berlin, Germany, from October to December 2022.

2.4. Sample Preparation and Chemical Analysis

Experimental diets and the COM diet were analyzed for proximate composition, i.e., dry matter (DM), crude protein (CP), crude fat (CF), crude fiber (CFB), starch and ash, as well as for amino acid composition, at SGS Analytics Germany (Augsburg, Germany) in accordance with official standard methods [49,50,51,52]. For mineral composition analysis of diets and feces (Ca, P, S, Mg, Fe, Na, K, Al, Zn, Mn, Cu), samples were freeze-dried, homogenized under liquid nitrogen in a mortar and re-dried until constant weight was achieved. Before taking samples of the feces for freeze-drying and further analysis, the pooled feces material was homogenized with a blender. Microwave-digestion of 150 mg dry sample was performed in aqua regia according to DIN EN 16,174 [53] with a ratio of HCl to HNO₃ of 1:3 before subsequently determining mineral composition by inductively coupled plasma optical emission spectrometry (ICP-OES) (iCAP 7400 ICP-OES, Thermo Fisher Scientific, Waltham, MA, USA) according to DIN EN ISO 11,885 [54]. In addition to the CP determination of the diets, the Kjeldahl method [50] was also applied to determine the nitrogen (N) content of the feces. After taking the samples for N and mineral analysis, all the remaining feces material was used for DM determination by freeze-drying.

Water samples (15 mL) were filtered (0.45 µm, Sartorius, Göttingen, Germany), fixed with 150 µL 2 M HCl and subsequently analyzed for K, Mg, Ca, S, B, Fe, Na, Cu, Mn, and Zn through inductively coupled plasma optical emission spectrometry (ICP-OES) (iCAP 7400 ICP-OES, Thermo Fisher Scientific, Waltham, MA, USA) [54] while NH₄-N, NO₂-N, NO₃-N (as a sum referred to as total inorganic nitrogen—TIN) and PO₄-P (soluble reactive phosphorus—SRP) were determined via continuous flow analysis (CFA) (FSR Seal High-Resolution AA3 chemical analyzer, Seal Analytical, Norderstedt, Germany).

2.5. Calculations

The formulas for calculating nitrogen-free extract (NFE), gross energy (GE), and protein-to-energy ratio (P/E ratio) of the diets as well as for the fish performance indicators of body weight gain (BWG), total length gain (LG), condition factor (CF), biomass gain (BMG), mean daily ration (MDR), feed conversion ratio (FCR), specific growth rate (SGR), protein efficiency ratio (PER), and thermal growth coefficient (TGC) are provided in the footnotes of the respective tables. To normalize the potentially distorting effect of differential mortality between tanks, final FCR and PER were calculated based on the average total amount of feed/CP that each individual remaining at the end of the trial had received, meaning the sum of the average daily individual feed/CP rations.

Since the amount of feed administered in each RAS differed due to the applied feeding regime and in order to make results directly comparable to similar studies irrespective of the influence of tap water nutrient concentrations, dissolved nutrient concentrations needed to be normalized to make valid comparisons between dietary treatments. Normalization resulted in a measure of mg of a dissolved nutrient produced per g of feed input (as fed). This was achieved by subtracting the initial dissolved nutrient load in the RAS, entirely stemming from tap water, as well as the load originating from the daily tap water introduction over a certain period of time from the final load in the RAS at the end of that period and then adding the load which was withdrawn from the RAS over the same period. Tap water nutrient load and the nutrient load from RAS water withdrawal were calculated on a weekly basis according to mean concentrations between sampling days, and multiplied by the corresponding water volume introduced or withdrawn over the period in question. Finally, the feed input nutrient load was divided by the feed amount administered, for each tank.

The partitioning of nutrient excretion between dissolved excretion into the RAS water and solid excretion in the form of feces is expressed as a percentage of the total amount of a nutrient supplied through feed input. The reference point for dissolved excretion is the total load of a specific nutrient stemming from feed input (corrected for tap water influence) and for solid excretion it is the quantity of a specific nutrient in the total amount of recovered feces DM.

2.6. Statistical Analysis

Data are presented as mean ± standard deviation. One-way analysis of variance (ANOVA) was applied to determine differences in the means between treatments followed by a Bonferroni post-hoc test for multiple comparisons or, in case of a significant Leven’s test, a Games–Howell post-hoc test. Statistical significance was set at p < 0.05. Analyses were performed using IBM SPSS Statistics 22.0.

3. Results

3.1. Rearing Conditions

With regard to oxygen concentration, temperature, pH and EC as well as ammonium and nitrite concentrations, conditions were suitable for rearing African catfish and homogenous between tanks and dietary treatments (

Table 3. Experimental rearing conditions.

	Rearing Conditions
	PM	MIX	BSF	COM
O2 (mg/L) 1	7.13 ± 0.22	7.14 ± 0.25	7.16 ± 0.25	7.11 ± 0.23
Temperature (°C) 1	27.0 ± 0.1	27.0 ± 0.7	27.0 ± 0.6	27.0 ± 0.3
pH 1	7.43 ± 0.64	7.68 ± 0.39	7.74 ± 0.33	7.8 ± 0.27
Conductivity (EC) (µS/cm) 1	1012 ± 182	1012 ± 177	1015 ± 176	1035 ± 189
NH4+-N (mg/L) 2	0.24 ± 0.14	0.27 ± 0.14	0.29 ± 0.14	0.28 ± 0.17
NO2−-N (mg/L) 2	<0.05	<0.05	<0.05	<0.05

Values represent means ± standard deviations. ¹ Measured once daily before water exchange and feeding (n = 196). ² Sampled once weekly before water exchange and feeding (n = 32).

).

RAS water pH decreased consistently for all treatments over the course of the trial and from day 30 onward when NaOH dosing started, pH roughly stabilized for the COM, MIX and BSF treatments between 7.2 and 7.6, with the COM treatment occupying the higher end and the MIX treatment the lower end of this range (Figure S1). Midway through the trial, the pH of the PM treatment began to decouple from the other treatments, finally stabilizing roughly around a pH of 6.5. In contrast to pH, the EC of all RAS waters increased steadily until day 30 up to 950–975 µS/cm and thereafter more rapidly due to the addition of NaOH and the increasing feed input up to 1500–1550 µS/cm (Figure S2). While the PM, BSF and MIX diets produced similar EC throughout the trial, the COM diet resulted in a consistently higher EC in the RAS water, particularly up until the last week of the trial. The 10% daily water exchange resulted in a mean water exchange for the entire trial of 1184 ± 48 (PM), 1246 ± 35 (BSF), 1249 ± 16 (MIX) and 1291 ± 40 L/kg feed/day (COM) as well as a final water exchange at the end of the trial of 197 ± 8 (PM), 235 ± 11 (MIX), 237 ± 5 (BSF) and 260 ± 9 L/kg feed/day (COM).

3.2. Growth Performance

Initial body weight, initial total length, initial biomass, and final CF did not differ significantly between dietary treatments (

Table 4. Fish performance indices.

	Fish Performance
	PM	MIX	BSF	COM
Survival (%) A	86.2 ± 2.4 b	91.4 ± 2.4 ab	90.9 ± 1.7 ab	93.1 ± 3.1 a
Initial body weight (g) B	9.1 ± 2.4 a	8.5 ± 2.2 a	9.0 ± 2.2 a	8.6 ± 2.3 a
Final body weight (g) C	115.8 ± 43.4 a	90.4 ± 38.7 b	88.9 ± 33.0 bc	80.3 ± 37.4 c
BWG (%) 1A	1233 ± 29.0 a	976 ± 35.1 b	938 ± 23.7 b	865 ± 34.9 c
Initial total length (cm) B	11.1 ± 1.0 a	10.8 ± 1.0 a	11.0 ± 0.9 a	10.9 ± 1.0 a
Final total length (cm) C	23.3 ± 2.5 a	21.6 ± 2.5 b	21.6 ± 2.3 b	20.7 ± 2.5 c
LG (%) 2A	109.9 ± 4.5 a	100.1 ± 3.3 b	95.3 ± 2.0 bc	91.1 ± 3.0 c
Condition factor (CF) 3C	0.87 ± 0.12 a	0.85 ± 0.09 a	0.85 ± 0.12 a	0.85 ± 0.13 a
Initial biomass (g) A	504.4 ± 27.0 a	487.3 ± 4.2 a	497.1 ± 16.6 a	483.1 ± 18.1 a
Final biomass (g) A	5790 ± 200 a	4793 ± 235 b	4689 ± 163 bc	4355 ± 166 c
BMG (%) 4A	1049 ± 34 a	883 ± 43 b	844 ± 25 bc	802 ± 26 c
Total feed administered (g, as is) A	3329 ± 137 a	2955 ± 112 b	2906 ± 89 b	2718 ± 113 b
MDR (% biomass/d) 5D	3.06 ± 0.83 a	3.07 ± 0.84 a	3.06 ± 0.83 a	3.07 ± 0.83 a
FCR (as fed) 6A	0.60 ± 0.01 c	0.66 ± 0.01 b	0.66 ± 0.01 b	0.68 ± 0.01 a
FCR (DM basis) 6A	0.57 ± 0.01 c	0.63 ± 0.01 b	0.63 ± 0.01 b	0.66 ± 0.01 a
SGR 7A	5.18 ± 0.04 a	4.75 ± 0.07 b	4.68 ± 0.05 b	4.59 ± 0.13 b
PER 8A	3.30 ± 0.04 a	3.01 ± 0.05 b	2.99 ± 0.03 b	2.96 ± 0.05 b
TGC 9A	2.09 ± 0.03 a	1.82 ± 0.04 b	1.79 ± 0.01 b	1.70 ± 0.03 c

Values represent means ± standard deviations; means in rows with different superscript letters are significantly different (p < 0.05). ^A n = 4; ^B n = 120; ^C n = 200–217; ^D n = 196. ¹ BWG—mean body weight gain (%) = (final mean body weight (g)−initial mean body weight (g))/initial mean body weight (g)) × 100. ² LG—body length gain (%) = (final mean body length (cm)−initial mean body length (cm))/initial mean body length (cm) × 100. ³ CF—condition factor = (body weight (g)/total length (cm)3) × 100. ⁴ BMG—biomass gain (%) = (final biomass (g)—initial biomass (g))/initial biomass (g)) × 100. ⁵ MDR—mean daily ration (%/d) = sum (ri × 100)/trial duration (days); where i = day of trial (1–49) such that ri = ration (g)/biomass (g) of day i. ⁶ FCR—feed conversion ration = total feed fed per individual (g as fed or DM)/[final mean body weight (g)−initial mean body weight (g)]. ⁷ SGR—specific growth rate = [[ln (final mean body weight(g))−ln (initial mean body weight (g))]/trial duration (days)] × 100. ⁸ PER—protein efficiency ratio = (mean individual body weight gain (g)/CP fed per individual (g)). ⁹ TGC—thermal growth coefficient = 1000 × (final body weight (g) 1/3−initial body weight (g) 1/3) × (trial duration (days) × average temperature (°C)) [55].

). Fish fed the PM diet demonstrated significantly better growth and feed conversion performance across all recorded measures compared to fish fed the other diets. Fish fed the MIX and the BSF diet performed similarly to each other; even though the BSF diet resulted in slightly lower mean final body weight, BWG, LG, final biomass, BMG, SGR, PER, and TGC, differences were non-significant across all of these measures. The COM diet, however, resulted in the overall significantly lowest BWG, final total length and TGC as well as the significantly highest FCR (as fed + DM basis) among all diets. Nevertheless, final body weight, final biomass, BMG and LG did not differ between the COM and the BSF treatments, while SGR and PER were not significantly different between the COM, the BSF and the MIX treatments. Survival was mostly similar between dietary treatments, with only the PM treatment showing significantly lower survival compared to the COM treatment.

While the MDR as a percentage of the biomass per day did not differ between treatments, the fish in the PM treatment received a significantly higher total feed amount due to the faster growth and thus biomass increase compared to the other treatments, which did not differ significantly from each other in this respect. Total biomass of the fish fed with the PM diet increased consistently faster than in fish fed the other diets (Figure 3).

3.3. Dissolved Nutrient Excretion

With regard to dissolved plant macronutrients in the RAS water (Figure 4), a clear increase in concentrations and concomitant decoupling from tap water concentrations was observed over the course of the trial for TIN, SRP, K, Mg and S in all dietary treatments, while Ca concentrations did not appear to noticeably rise in any of the treatments and remained closely related to tap water levels. By the end of the trial, the highest TIN concentrations were reached in the PM treatment (122 mg/L), followed by the MIX (106 mg/L), the BSF (102 mg/L) and the COM treatment (95 mg/L). Final Ca (PM—104 mg/L; MIX—104 mg/L; BSF—105 mg/L; COM: 109 mg/L) and S concentrations (PM—63 mg/L; MIX—64 mg/L; BSF—65 mg/L; COM: 66 mg/L) were similar between treatments. This was also the case for final Mg concentrations in the PM (15.7 mg/L), MIX (15.4 mg/L) and BSF treatment (15.6 mg/L), whereas the concentration in the COM treatment was slightly higher (16.8 mg/L). Noticeable differences in final concentrations were recorded for K and SRP. While the PM (9.9 mg/L), MIX (9.5 mg/L) and BSF treatment (8.7 mg/L) produced similar final K concentrations, more than double the concentration was recorded in the COM treatment (20.5 mg/L). All treatments showed a clear partitioning with regard to final SRP concentrations. The COM treatment showed a comparably slow accumulation of SRP throughout the trial, achieving the lowest concentration of all treatments (1.1 mg/L). Even though the BSF treatment already reached more than double this concentration (2.8 mg/L), the MIX (7.3 mg/L) and especially the PM treatment (13.7 mg/L) reached substantially higher SRP levels.

Most of the dissolved plant micronutrients (Figure 5), including Na, Cu, Mn and Zn, did not show a feed-related accumulative trend above tap water concentrations and final concentrations were mostly similar between treatments. However, from week 5 onward, Mn and Zn concentrations rose in the PM treatment, finally reaching higher levels compared to the other treatments, albeit not consistently across replicates. Although the PM, MIX and BSF treatments showed similar Cu concentrations roughly consistent with tap water levels, the COM treatment resulted in lower Cu concentrations throughout. Na concentrations in all treatments did not appear to decouple from tap water concentrations until dosing of NaOH started, upon which a clear and consistent increase was recorded for all treatments. Although still low in absolute terms, Fe concentrations showed an upward trend unrelated to tap water concentrations in all treatments, with the COM and the BSF treatment reaching roughly double the level compared to the MIX and PM treatment. B concentrations for the most part seemed strongly associated with tap water levels. However, in contrast to the other treatments, the COM treatment showed a fairly consistent upward decoupling from tap water concentrations.

When normalized for the variation in feed input between tanks and excluding the influence of tap water nutrient load, the release of the major dissolved plant nutrients TIN, P and K into the RAS water per g of feed and as a percentage of dietary N, P and K input showed consistent patterns over the course of the trial (Figure 6). From the start of the trial, TIN release was similar between treatments; after week 3, it stabilized at 37–39 mg TIN/g of feed (as fed) which equaled 46–49% of dietary N released as dissolved TIN. With regard to both measures and considering the entire duration of the trial, dissolved TIN release was not significantly different between treatments (A in

Table 5. (A) Dissolved nutrients excreted and (B) solid nutrients collected over the course of the trial. Expressed as (1) mg/g of feed (excluding the influence of tap water nutrients) and (2) percentage of the total dietary input of a specific nutrient.

	Nutrient Partitioning between Water and Feces
	PM	MIX	BSF	COM	PM	MIX	BSF	COM
	Dissolved Nutrients (RAS Water) (A)				Solid Nutrients (Collected Feces) (B)
	mg/g of feed (as fed) (1)				mg/g of feed (as fed) (1)
N	38.59 ± 0.63 a	38.83 ± 0.51 a	38.58 ± 0.33 a	38.31 ± 0.62 a	1.64 ± 0.04 c	2.53 ± 0.07 b	3.27 ± 0.21 a	3.09 ± 0.20 a
P	3.99 ± 0.27 a	2.44 ± 0.10 b	0.95 ± 0.04 c	0.52 ± 0.08 d	2.58 ± 0.11 a	2.30 ± 0.10 b	1.94 ± 0.15 c	1.78 ± 0.11 c
K	1.15 ± 0.06 b	1.16 ± 0.05 b	0.93 ± 0.05 c	5.98 ± 0.10 a	0.04 ± 0.00 c	0.06 ± 0.00 b	0.08 ± 0.01 a	0.07 ± 0.00 a
Mg	1.15 ± 0.02 c	1.21 ± 0.04 bc	1.28 ± 0.04 b	1.98 ± 0.08 a	0.12 ± 0.00 b	0.14 ± 0.01 a	0.14 ± 0.01 ab	0.16 ± 0.01 a
S	4.25 ± 0.07 d	5.04 ± 0.20 c	5.60 ± 0.10 b	6.27 ± 0.26 a	0.20 ± 0.00 b	0.36 ± 0.01 a	0.45 ± 0.04 a	0.39 ± 0.03 a
Ca	−0.57 ± 0.57 b	−0.47 ± 0.29 b	0.01 ± 0.08 b	2.12 ± 0.54 a	4.64 ± 0.21 a	4.33 ± 0.19 a	3.49 ± 0.29 b	3.13 ± 0.19 b
Fe	0.0036 ± 0.0003 d	0.0051 ± 0.0004 c	0.0075 ± 0.0001 b	0.0099 ± 0.0004 a	0.19 ± 0.00 b	0.24 ± 0.02 a	0.26 ± 0.04 ab	0.12 ± 0.02 c
Mn	0.0134 ± 0.0232 a	−0.0002 ± 0.0001 b	−0.0003 ± 0.0000 b	−0.0008 ± 0.0001 ac	0.04 ± 0.00 a	0.04 ± 0.00 a	0.03 ± 0.01 a	0.00 ± 0.00 b
Zn	−0.0136 ± 0.0269 a	−0.0350 ± 0.0041 b	−0.0320 ± 0.0015 b	−0.0447 ± 0.0021 ac	0.11 ± 0.01 a	0.11 ± 0.01 a	0.11 ± 0.01 a	0.04 ± 0.00 b
Cu	0.0014 ± 0.0007 a	0.0003 ± 0.0004 a	0.0012 ± 0.0006 a	−0.0067 ± 0.0006 b	0.01 ± 0.00 a	0.01 ± 0.00 a	0.01 ± 0.00 a	0.00 ± 0.00 b
	% of nutrient fed (2)				% of nutrient fed (2)
N	47.6 ± 0.8 a	48.1 ± 0.6 a	47.6 ± 0.4 a	48.7 ± 0.8 a	2.0 ± 0.0 c	3.1 ± 0.1 b	4.0 ± 0.3 a	3.9 ± 0.3 a
P	20.5 ± 1.4 a	15.2 ± 0.6 b	7.1 ± 0.3 c	4.5 ± 0.7 d	14.2 ± 0.6 a	15.1 ± 0.6 a	15.3 ± 1.2 a	16.1 ± 1.0 a
K	19.3 ± 1.0 bc	20.4 ± 0.9 b	17.3 ± 0.9 c	57.9 ± 1.0 a	0.7 ± 0.0 c	1.1 ± 0.0 b	1.4 ± 0.1 a	0.7 ± 0.0 c
Mg	83.8 ± 1.6 b	88.3 ± 2.7 ab	92.8 ± 3.1 a	91.7 ± 3.5 a	8.7 ± 0.3 b	10.1 ± 0.4 a	10.5 ± 0.9 ab	7.3 ± 0.5 c
S	92.3 ± 1.5 a	97.3 ± 3.9 a	95.8 ± 1.6 a	95.1 ± 4.0 a	4.5 ± 0.1 c	7.0 ± 0.3 a	7.7 ± 0.7 a	5.9 ± 0.4 b
Ca	−3.5 ± 3.5 b	−3.2 ± 2.0 b	0.1 ± 0.7 b	18.9 ± 4.8 a	28.6 ± 1.3 a	30.1 ± 1.3 a	27.8 ± 2.3 a	27.9 ± 1.7 a
Fe	0.7 ± 0.1 d	0.9 ± 0.1 c	1.2 ± 0.0 b	3.0 ± 0.1 a	34.6 ± 0.8 a	40.5 ± 2.7 a	40.6 ± 6.7 a	37.8 ± 5.4 a
Mn	8.0 ± 13.8 a	−0.1 ± 0.0 b	−0.2 ± 0.0 b	−2.2 ± 0.3 ac	22.1 ± 0.7 a	21.8 ± 2.1 a	23.2 ± 5.0 ab	14.1 ± 2.3 b
Zn	−4.9 ± 9.7 a	−13.3 ± 1.6 a	−12.3 ± 0.6 a	−40.9 ± 1.9 b	37.9 ± 1.9 a	41.3 ± 2.7 a	41.9 ± 4.2 a	35.0 ± 3.8 a
Cu	4.7 ± 2.4 a	0.9 ± 1.2 a	3.9 ± 2.1 a	−49.8 ± 4.4 b	27.9 ± 1.4 b	26.5 ± 1.9 b	25.8 ± 4.0 b	37.1 ± 4.0 a

Values represent means ± standard deviations; means in rows with different superscript letters are significantly different (p < 0.05); n = 4.

). Contrary to TIN, release of SRP showed distinct patterns between the different treatments (Figure 6). From the start, SRP release was lowest in the COM treatment, remaining well below 1 mg SRP/g of feed after week 2, while the BSF treatment stabilized at around 1 mg SRP/g of feed midway through the trial. Considerably higher release of SRP was consistently recorded in the MIX treatment, stabilizing in a range of 2.1–2.4 mg SRP/g of feed in the latter half of the trial, and especially the PM treatment, stabilizing at 3.8–4.0 mg SRP/g of feed. Displaying the same consistent pattern throughout the trial, the percentage of dietary P released as SRP into the RAS water was highest in the PM treatment reaching 20.5% of dietary P, followed by the MIX (15.2%) and then BSF treatment (7.1%), while in the COM treatment, the lowest percentage of dietary P was again released (4.5%). Total SRP release over the course of the trial was significantly different between all treatments, both when measured per g of feed and as a percentage of dietary P (A in Table 5). Dissolved K release per g of feed and as percentage of dietary K was similar throughout the trial in the PM, MIX and BSF treatments, with a consistent range of 0.6–1.2 mg/g of feed and 11.5–19.3% of dietary K after week 3 (Figure 6). However, in the COM treatment, K release stabilized at substantially higher levels around 5.9–6.1 mg/g of feed and 57.2–59.0% of dietary K after week 3. Total K release over the course of the trial was significantly higher for the COM diet with respect to both measures compared to all other treatments, which, notwithstanding some significant differences, were in a very similar low range (A in Table 5).

The significantly highest release of dissolved Mg and S per g of feed was recorded in the COM treatment and the significantly lowest release of S in the PM treatment (A in Table 5). Mg release per g of feed was also lowest in the PM treatment, yet differences were only significant compared to the BSF and the COM treatment. Differences between the MIX and the BSF treatment were non-significant for Mg but significant for S. In all treatments, above 80% of the dietary Mg and S were released in dissolved form (A in Table 5).

For S, there were no significant differences between treatments in this regard, while also for Mg there were no significant differences between the MIX, BSF and COM treatments and only the PM treatment showed significantly lower release as a percentage of dietary Mg compared to the BSF and the COM treatments. Dissolved Ca release was not detectable for the PM, MIX and BSF treatments, whereas the COM treatment showed a low, but significantly higher, Ca release (A in Table 5).

In general, the release of dissolved Fe, Mn, Zn and Cu was considerably lower, especially compared to TIN, SRP, K, Mg and S, and oftentimes negative (A in Table 5). Mn and Zn release per g of feed and as percentage of dietary input was negative in all treatments, except for Mn release in the PM treatment which was positive and significantly higher than in all other treatments (A in Table 5). While Cu release was positive with regard to both measures in the PM, MIX and BSF treatments and not significantly different, the COM treatment showed negative and significantly lower values. Fe release was positive throughout and differences were significant between all treatments, with the COM treatment exhibiting the highest release per g of feed and as a percentage of dietary Fe input followed by the BSF, the MIX and the PM treatments in that order.

3.4. Solid Nutrient Excretion

Over the course of the trial, between 55.5 and 69.4 g feces DM were collected per kg of feed (as fed) which equaled 5.8 to 7.2% of feed DM recovered as feces DM (

Table 6. Recovered amount of feces matter and mineral composition of feces DM.

	Feces Mineral Composition
	PM	MIX	BSF	COM
Feces collected
g, as is	2327 ± 77	2412 ± 120	2571 ± 252	2343 ± 184
g DM	185 ± 8	188 ± 6	202 ± 16	161 ± 14
g DM/kg of feed (as fed)	55.5 ± 1.0 c	63.8 ± 2.1 ab	69.4 ± 3.8 a	59.3 ± 3.3 bc
% feed DM recovered as feces DM	5.8 ± 0.1 c	6.7 ± 0.2 ab	7.2 ± 0.4 a	6.2 ± 0.3 bc
mg/g feces DM
N	29.52 ± 0.40 d	39.68 ± 1.20 c	47.20 ± 1.36 b	52.20 ± 1.18 a
P	46.50 ± 1.15 a	36.12 ± 0.49 b	27.93 ± 0.81 d	29.97 ± 0.21 c
K	0.77 ± 0.02 c	1.02 ± 0.03 b	1.08 ± 0.05 b	1.25 ± 0.06 a
Mg	2.14 ± 0.05 b	2.17 ± 0.05 b	2.09 ± 0.08 b	2.65 ± 0.04 a
S	3.69 ± 0.04 c	5.68 ± 0.08 b	6.47 ± 0.27 a	6.55 ± 0.17 a
Ca	83.62 ± 2.44 a	67.85 ± 0.86 b	50.24 ± 1.98 c	52.78 ± 0.54 c
Fe	3.40 ± 0.02 a	3.69 ± 0.19 a	3.73 ± 0.43 a	2.08 ± 0.21 b
Mn	0.67 ± 0.02 a	0.63 ± 0.05 a	0.49 ± 0.08 a	0.08 ± 0.01 b
Zn	1.90 ± 0.09 a	1.71 ± 0.08 b	1.58 ± 0.08 b	0.65 ± 0.05 c
Cu	0.14 ± 0.01 a	0.13 ± 0.01 b	0.11 ± 0.01 b	0.08 ± 0.01 c

Values represent means ± standard deviations; means in rows with different superscript letters are significantly different (p < 0.05); n = 4.

). Both measures were lowest for the PM treatment and significantly so compared to the MIX and the BSF treatments, which did not differ significantly from each other. The second lowest values for these measures were recorded for the COM treatment, with differences being significant in comparison to the BSF treatment; the latter featured the highest feces recovery per kg of feed and as a percentage of feed DM.

Ca, N and P were by far the most prevalent mineral components of all feces types, reaching levels of 50.2–83.6, 29.5–52.2 and 27.9–46.5 mg/g DM, respectively, while S as the next most important mineral contributor only reached levels of 3.7–6.6 mg/g DM (Table 6). In the MIX and especially the BSF and COM feces, N content was higher than P content, whereas the reverse was true in the PM feces which showed a substantially higher P than N content. K content in all feces types was among the lowest of all minerals, with only Mn and Cu content being lower in the PM, MIX and BSF feces and only Mn, Cu and Zn content being lower in the COM feces.

The COM feces showed the significantly highest N, K and Mg content compared to the other feces types, while featuring the significantly lowest Fe, Mn, Zn and Cu content. The PM feces on the other hand featured the significantly lowest N, K and S content and the significantly highest P, Ca, Zn and Cu content. No significant differences were found between the PM, MIX and BSF feces with regard to Mg, Fe and Mn content. Even though the BSF feces had a significantly higher N and S content together with a significantly lower P and Ca content compared to the MIX feces, both of these types of feces shared certain similarities, with K, Mg, Fe, Mn, Zn and Cu content not differing significantly. Although similar in absolute terms to the COM feces with respect to P content, the BSF feces did feature the overall significantly lowest P content.

For all diets, substantially less N, K, Mg and S were found in the collected feces per unit of feed and as a percentage of dietary input of the respective nutrients compared to what was released as dissolved nutrients into the RAS water (A/B in Table 5). The contrary was the case for Ca, Fe, Mn, Zn and Cu, of which substantially more was found in the collected feces per unit of feed and as a percentage of dietary nutrient input than was released in dissolved form. The partitioning between dissolved and solid excretion was not as clear and consistent for P between dietary treatments as for the other nutrients mentioned above. Similar to what was found for dissolved P release, the P recovered in the collected feces per unit of feed was also significantly highest for the PM treatment followed by the MIX, the BSF and the COM treatments in this order, with only the BSF and COM treatments not differing significantly. However, in the BSF and COM treatments, more P per unit of feed was recovered in the collected feces than was released in dissolved form, whereas in the MIX treatment, dissolved release and solid P collection per unit of feed were similar. In the PM treatment, in contrast, dissolved P release was considerably higher than solid P recovery in the feces. Despite the differences between the dietary treatments in terms of dissolved and solid P excretion per unit of feed as well as the significantly different release of dissolved P as percentage of dietary P input alluded to in the previous section, no significant differences were found for the amount of P collected in the feces relative to dietary P input which ranged from 14.2% to 16.1% for all treatments (A/B in Table 5).

As suggested by feces N content (Table 6), the recovery of N per unit of feed through feces collection was significantly lowest in the PM treatment while the MIX treatment also showed significantly lower N recovery compared to the BSF and the COM treatments, which did not differ significantly from each other (B in Table 5). This was similarly reflected when expressed as a percentage of dietary N input. With only 2–4% of feed N found in the feces, N recovery as percentage of dietary nutrient input was low compared to other nutrients and recovery was only lower for K, of which only 0.7–1.4% was recovered in the collected feces.

For the other recovered nutrients in the feces, only a few major differences were recorded between the treatments (B in Table 5).

4. Discussion

4.1. Fish Performance

The results on the growth and feed conversion performance of African catfish show that, regardless of diet, fish generally performed well with a 9 to 12-fold increase in body weight within a 7-week period and that none of the experimental diets were demonstrably inferior to the COM diet. In fact, the PM diet enabled significantly better fish performance than the COM diet across all recorded measures while also the MIX diet facilitated better results across several recorded performance measures, whereas the BSF diet produced a more similar fish performance to the COM diet with various performance indicators being non-significantly different.

First of all, this indicates, regardless of the inclusion of PM and BSFM and given that fish processing waste is adequately processed, that CM could be a suitable protein ingredient at the applied inclusion level in diets for African catfish, which may at least be capable of rivaling the growth and feed conversion performance achievable with certain commercial diets. These results are in accordance with other studies incorporating fish processing waste as a suitable protein source in experimental diets for African catfish [56,57,58,59] and support the general notion of aquaculture by-products as a valuable (yet underutilized) ingredient in aquaculture feeds [12,60].

Fish performance was similar and non-significantly different between fish fed the MIX and the BSF diet regarding all recorded indicators, which means that an increase of dietary BSFM inclusion from 20.4% to 40.0% and the concomitant elimination of PM as an ingredient did not negatively affect fish performance or, conversely, that a 50% replacement of the BSFM with PM did not result in better fish performance. However, a 100% replacement of BSFM with PM in the PM diet did in fact lead to a significantly better growth and feed conversion performance compared to both the MIX and the BSF diet. On the one hand, this matches the consensus of poultry by-product meal being a nutritionally highly valuable protein ingredient in aquaculture diets [36,38,39] which can replace up to 100% of the fish meal in control diets of various freshwater fish [37] and serve as an effective and cheap protein ingredient in African catfish diets [10,61]. On the other hand, these results illustrate the apparent superiority that PM can have over BSFM at high dietary inclusion levels.

While considered to have a comparably balanced amino acid profile [62,63] resembling that of fish meal [64], meta-analyses suggest that high dietary inclusion of insect meals in aquafeeds at levels of above 25–30% [65] or 29% [66] are likely to compromise fish performance. In general, this is also reflected in studies on African catfish, which found uncompromised growth and feed conversion up to a BSFM inclusion of 7.5% (50% fish meal replacement) [67], 12.3% (20% fish meal replacement) [68], 17.18% (75% fish meal replacement) [69], and 20.2% (25% fish meal replacement) [70], whereas higher inclusion levels (15% [67]; 29.3% [68]; 33% [70]) lead to impaired performance compared to control diets. In terms of BSFM inclusion level, the MIX diet (20.4%) was below the thresholds suggested by [65,66] and roughly similar to the upper ranges successfully tested by the above cited studies involving African catfish. In contrast, the BSF diet (40%) clearly exceeded both. In case of the present study, the high BSFM inclusion of 40% in the BSF diet did not cause reduced performance compared to the COM diet and, furthermore, it appears that any relative difference between the PM and the BSFM with regard to supporting growth performance of African catfish did not surface when replacing 50% of the BSFM with PM in the MIX diet. One reason for this could have been the supplementation of the BSFM with high quality animal protein ingredients, i.e., CM and PBM, at adequate levels [56,57,59,71,72,73] in already highly energy- and protein-dense diets. This is also supported by the amino acid profile not having been negatively affected with increased inclusion of BSFM in the MIX and BSF diet. In essence, this suggests that elevated BSFM inclusion levels above what is recommended in aquafeeds by [65,66] may generally be possible without exaggerated negative effects on the growth performance of African catfish when combining BSFM with suitable complementary protein sources; this would agree with the more positive outlook on BSFM inclusion in aquafeeds given by [74] through their meta-analysis. Nevertheless, a complete replacement of the BSFM in the PM diet crystalized the relative superiority of PM over BSFM with respect to fish performance, despite the similar amino acid profiles of the respective diets. This indicates a better protein and/or energy digestibility and subsequent utilization of the PM diet versus the MIX and the BSF diets.

Since the exact formulation of the COM diet was unavailable, it is difficult to pinpoint the reasons for the MIX and particularly the PM diet outperforming the COM diet, not least due to essential amino acid profiles again being very similar between all diets. Nevertheless, the COM diet did have a slightly lower CP content, mostly due to lower levels of the non-essential amino acids Pro, Glu + Gln and Tyr, as well as a lower CF and GE content compared to the other diets; this may have contributed to the results and is, furthermore, indicated by the similar and non-significantly different PER and SGR achieved with the COM, MIX and BSF diets. It should be noted that, according to the producers’ specifications, the COM diet included soybean meal and sunflower protein concentrate as plant protein sources which, considering the limitations of plant proteins in diets especially for more carnivorous fish [12,75], could have potentially been an additional factor accounting for the reduced growth performance observed in comparison to a plant protein-devoid diet such as the PM diet.

4.2. Dissolved Nutrients

By manipulating water exchange as well as stocking density and feeding levels in the RAS unit of on-demand coupled aquaponic systems [5], it is generally possible to reach sufficiently high dissolved nitrogen levels in the process water, in the form of predominantly nitrate, even for plants with high nitrogen demand such as tomatoes [17,21,76,77]. This opens up the possibility of strongly reducing the use of energy- and carbon dioxide-intensive mineral nitrogen fertilizer in plant production [78]. However, as many studies show, the important plant macronutrients P and K can mostly not be sufficiently supplied in plant-available dissolved form via the process water [18,24,25,26,79], as is often also the case for various other nutrients [16,17,19,20,21,22,23,26]. In order to achieve optimal plant production in aquaponics, supplementation with increasingly scarce [80,81] and greenhouse-gas-intensive phosphate as well as potassium fertilizers is thus required [78]. The aquafeed industry has transitioned to diets that minimize the excretion of eutrophication-inducing nutrients, especially P, by formulating highly digestible and energy-dense diets with lower inclusion of P-rich animal proteins such as fish meal and terrestrial animal by-products in favor of plant protein sources [32,33,34,35]; this exacerbates the need for mineral fertilizer supplementation in aquaponics [18]. Therefore, a different formulation strategy may be beneficial for aquaponic-specific diets.

The excretion and accumulation of the major plant nutrients N, P and K in the RAS process water recorded in the present trial showed a clearly distinguishable pattern. On the one hand, TIN excreted per unit of feed and as a percentage of dietary N was comparable for all diets throughout the trial; with 38.3–38.8 mg/g of feed and 47.6–48.7% of dietary N, it was not significantly different between any of the diets, suggesting that amino acid catabolism and subsequent branchial nitrogen excretion as well as nitrification in the biofilters was similar between diets. On the other hand, P excretion in the form of SRP was significantly elevated with all experimental diets compared to the COM diet. Each increase in the level of poultry meal inclusion led to higher SRP concentrations in the RAS water as well as a significant increase in SRP excretion per unit of feed, with the PM diet facilitating a 4-fold higher SRP excretion (3.99 mg/g of feed) compared to the BSF diet (0.95 mg/g of feed) and an almost 8-fold higher SRP excretion compared to the COM diet (0.52 mg/g of feed). Protein ingredients from processing waste of animal origin (terrestrial or aquatic) tend to have higher phosphorus levels than, e.g., whole fish meals or plant protein ingredients, due to the elevated proportion of bones they contain [82,83,84,85]. Accordingly, it appears that firstly the inclusion of CM increased the overall phosphorus level of the experimental diets as exhibited by the higher P content of the BSF diet compared to the COM diet, which in turn significantly increased the resulting SRP excretion per unit of feed. Secondly, the increasing inclusion of PM not only enabled improved growth but also again resulted in higher dietary P content which translated into higher SRP excretion and this disproportionally so, as evidenced by the considerably higher percentage of dietary P excreted in dissolved form with the PM (20.5%) and the MIX diet (15.2%) versus the BSF (7.1%) and especially the COM diet (4.5%). Fish tend to excrete very little dissolved P when only supplied with the digestible P level required for optimal growth (as seen with the COM diet) [86]. It stands to reason that with the increasing PM inclusion in this study, digestible P levels exceeded the levels required for maximum growth and likely even the levels required for maximum bone mineralization, which would explain the disproportionately increasing excretion of P with increasing levels of dietary P [86]. Overall, the results for SRP excretion on the PM diet closely match what was previously found for African catfish fed a single protein ingredient diet based on PM with a comparable dietary P content (18.6 g/kg) to the present study (20.2 g/kg), which produced an SRP excretion of 4.2 mg/g of feed [10]. These results suggest that PM and CM represent protein ingredients capable of ensuring optimal growth in African catfish as well as substantially elevating the excretion of dissolved SRP in comparison to certain commercially available diets, making them suitable ingredients for aquaponic diets that aim to minimize the need for mineral P fertilization in the hydroponic unit of on-demand coupled aquaponic systems.

In the case of dissolved K excretion, however, the opposite was observed with the roughly twice as high K content of the COM diet (10.7 g/kg) compared to the other diets (5.6–6.2 g/kg) leading to by far the highest K concentration in the RAS water and a roughly 6-fold higher K excretion per unit of feed (5.98 mg/g) in comparison to the other diets (0.93–1.16 mg/g). While in absolute terms little difference was observed between the PM, the MIX and the BSF diet with respect to the percentage of dietary K excreted into the process water (17.3–20.4%), the substantially higher percentage recorded for the COM diet (57.9%) suggests that elevated dietary K levels can be translated well into plant available dissolved K. However, this also indicates, similar to the reasoning on P, that dietary K levels close to the requirement level of the fish tend to result in comparatively low dissolved K excretion, whereas an increase above the requirements disproportionately increases it. Interestingly, a similar study found that BSFM inclusion resulted in a considerably higher dietary K content (12.6 g/kg) and subsequently dissolved K excretion per unit of feed in African catfish (11.8 mg/g) compared to diets based on fish meal, PBM and PM, which corroborated prior findings for Nile tilapia [11]. In the present study, BSFM inclusion unexpectedly, despite using BSFM from the same supplier, neither resulted in a higher dietary potassium content nor excretion, particularly when compared to the COM diet. This appears to be evidence of the high variability of larvae mineral composition depending on the larvae production parameters including feeding substrates, the size and life stage of harvested larvae, and various other process parameters [64,87,88,89,90,91,92]. Thus, if BSFM is used as a dietary ingredient in aquaponic diets with the objective of increasing K excretion, larvae require highly digestible feed substrates rich in K [87]. Therefore, the modulation of BSF larvae mineral composition and its effect on dietary content and subsequent excretion of important plant nutrients such as K should be further investigated in the context of aquaponic feed development. Nevertheless, since K is the most abundant cation in plants [93], certain raw materials of plant origin such as soybeans can have a comparatively high K content while fish meal and other animal meals often feature a somewhat lower K content [85,94]. Accordingly, the integration of several plant ingredients into the COM diet, i.e., soybean meal, wheat and sunflower protein concentrate, may have contributed to its elevated K content and the subsequently higher excretion in comparison to the other diets. In consequence, closer investigation of plant ingredients as protein and/or carbohydrate sources as well as other non-plant ingredients naturally rich in K should be in focus when looking to increase the provisioning of dissolved K to hydroponic plant production through specialized aquaponic diets.

A consistent diet-related upward decoupling of concentrations from tap water levels was also recorded for Mg, S and Fe. Roughly proportional to its higher Mg content, the COM diet resulted in higher Mg concentrations in the RAS water as well as higher excretion of dissolved Mg per unit of feed throughout the trial (1.98 mg/g of feed) in comparison to the experimental diets, which, in line with their uniform Mg content, produced similar Mg excretion (1.15–1.28 mg/g). Moreover, the increasing dietary S content from the PM (4.8 g/kg) to the MIX (5.4 g/kg), to the BSF (6.1 g/kg) and to the COM diet (6.9 g/kg) was equally reflected in increasing S excretion per unit of feed with significant differences between all diets. Together with the load in the tap water, excreted Mg and S can accumulate to adequate levels; however, concentrations tend more often to be insufficient for optimal plant growth in aquaponic systems [16,17,21,22,79,95]. Results of the present study indicate, especially considering that for all diets a majority of dietary Mg was excreted in dissolved form, that feed formulations with higher Mg content could help to alleviate potential Mg constraints in aquaponics if tap water does not supply sufficient additional Mg. This is again in line with prior results involving African catfish which showed significantly higher Mg excretion with a BSFM-based diet which featured the highest Mg content [10]. Similar to what was discussed above for K with regard to the prior results on a BSFM-based diet, higher dietary Mg content and subsequent excretion were also not observed in the MIX or BSF diet compared to the PM diet of the present study. Thus, a comparably novel and qualitatively variable ingredient such as BSFM should not generally be equated with a specific mineral profile.

Similar to Mg, results suggest that potential S limitations in aquaponics could be addressed by an increase of dietary S content. While in previous trials a feed-related increase of S concentrations in RAS water was also observed, differences in dietary S content (3–5 g/kg) did not result in consistently significant differences in excretion [10,11]. This may have been a result of the lower dietary S contents compared to the present trial (4.8–6.9 g/kg) as well as the differing feeding regime, i.e., administering equal amounts of feed in all treatments throughout the trial; this leads fish in different treatments to receive unequal rations as a percentage of body weight as a result of differences in growth. However, since the sum of total dissolved S and S collected in the fish feces in some cases exceeded what was administered through feeding and the S content in the recovered feces was very much comparable to what was found in other studies with African catfish [10,79,96], it appears that the level of dissolved S excretion found in this study may be somewhat exaggerated. This may have been caused by a source of S that was unaccounted for (e.g., residues in the biofilter/filter mats not having been entirely removed during pre-trial cleaning of the RAS or higher S concentration in the tap water between sampling points) and/or measuring inaccuracy but, nevertheless, does not negate the general reasoning that dietary S content above a certain level affects S excretion.

The neutral pH value required for adequate nitrification in the biofilter of the RAS unit of an on-demand coupled aquaponic system is not conducive to Fe solubility and thus also makes it mostly insufficiently available for optimal plant production [28,97]. Considering the approximately neutral pH maintained in all treatments (pH 6.5–7.6), this was also the case in the present trial with only 0.7–3.0% of dietary Fe ending up in dissolved form. While the PM, MIX and BSF diets featured a higher Fe content (0.57–0.67 g/kg) than the COM diet (0.34 g/kg) and an increase in dietary Fe with increasing BSFM inclusion was equally reflected in higher excretion of dissolved Fe per unit of feed, the COM diet still facilitated the highest Fe concentrations in the RAS water and the significantly highest Fe excretion per unit of feed. Since excess Fe can be directly excreted in solid form without being absorbed in the gut in the ferrous form, as supported by the high percentage of dietary Fe recollected in the feces in the present trial, and Fe absorption by fish is influenced by its chemical forms present in the diet (Fe³⁺, Fe²⁺, inorganic or inorganic Fe complexes), it appears that Fe availability to fish was higher in the COM diet which may have facilitated transferrin-mediated branchial excretion [85,97,98]. However, this requires further investigation. Nevertheless, the results suggest that higher dietary Fe content does not generally lead to an increase in the excretion of plant-available dissolved Fe and that other dietary factors likely play a role, as also indicated by prior studies not consistently finding an increase in dissolved Fe excretion with higher Fe content in the diet [10,11].

Despite notable differences in the content of Ca, Cu, Mn and Zn between the diets, concentrations did not show a consistent feed-related increase in the RAS water above tap water levels for any of the diets and excretion per unit of feed was for the most part low to not detectable, indicating the low solubility of these minerals at neutral pH levels. Accordingly, a majority of excess dietary Ca, Cu, Mn and Zn was excreted in undissolved form as seen in the comparatively high amounts of these minerals recovered in the collected feces. At least under similar rearing conditions and RAS management (e.g., frequency of sludge removal, pH stabilization), this illustrates a low potential to modify dissolved Ca, Cu, Mn and Zn levels by diet formulation alone in comparison to other nutrients such as N, P, K, Mg or S, and points to the dependence of their concentrations in aquaponics on the levels introduced through the tap water [16,17]. However, it should be noted that prior studies found significantly higher Ca and Cu concentration exceeding tap water concentrations with diets respectively featuring the highest Ca (PM-based) and Cu content (BSF-based) versus the other experimental diets (fish meal-based, PBM-based) [10,11]. Considering that in freshwater fish much of the Ca requirement is met via branchial absorption from the surrounding water [85] and tap water Ca concentrations were comparably high in the present study, overall lower Ca levels in the rearing water may have an effect on the excretion of dietary Ca in dissolved form into the rearing water. Further research with respect to environmental conditions (e.g., pH levels, background tap water concentrations) as well as dietary composition and subsequent digestion, absorption and excretion is required concerning minerals such as Ca, Cu, Mn and Zn.

4.3. Solid Nutrients

Contrasting the excretion of dissolved nutrients with the collection of nutrients in solid form in the feces clearly shows the dichotomy in the partitioning of excretion between solid and dissolved. While for all diets considerably higher amounts of N, K, Mg and S per unit of feed and as a percentage of their dietary input ended up in dissolved form in the RAS water versus in solid form in the feces, the opposite was true for Ca, Fe, Mn, Zn and Cu which were primarily recovered in the collected feces while at maximum negligible amounts were found in dissolved form. Regarding P, results were not as bifurcated, with an increase in dietary P from the COM to the PM diet shifting P excretion from being mainly dominated by solid excretion (COM, BSF), to being balanced between the solid and dissolved excretion (MIX) and then to being more dominated by dissolved excretion (PM). In a similar vein to the discussion in the previous section, this illustrates the comparatively higher potential to manipulate the nutrient profile of the process water in aquaponic systems through diet formulation and ingredient choice under applied conditions with regard to N, K, Mg, S and P in contrast to Ca, Fe, Mn, Zn and Cu.

Taking the variation in ingredient choice in the present trial into account, the mineral profile of the collected feces was roughly comparable to what was found by other authors for African catfish [10,79,96], with Ca (50.2–83.6 g/kg DM), N (29.5–52.2 g/kg DM) and P (27.9–46.5 g/kg DM) being the predominant mineral components, which reflected their relative dietary abundance, whereas K content (0.8–1.3 g/kg DM) was among the lowest of all recorded minerals. N content of the feces tended to decrease with better growth performance of the fish, especially from the BSF (47.2 g/kg DM) to the MIX (39.7 g/kg DM) and to the PM diet (29.5 g/kg DM), which appears sensible considering the uniformity of dissolved TIN excretion as a proxy for amino acid catabolism. As previous studies demonstrated, a sizeable fraction of feces N can still be present as amino acids and further nutrients such as carbohydrates and small amounts of fat can constitute parts of the feces as well [10,11]. This may especially be true in sludge from commercial RAS operation, where a certain amount of feed loss is practically inevitable. Currently, aquaculture sludge is predominantly used for biogas production or as agricultural fertilizer [99,100]. However, various studies have begun investigating the potential of recycling aquaculture sludge into higher value nutrients through vermicomposting [101,102], polychaetes [103,104] or insect larvae [105]. Hence, extending aquaponics by such a third trophic production level in the form of Hermetia illucens larvae production, for example, could help minimize waste by transforming aquaculture sludge, together with harvest wastes from plant production, into valuable biomass which could subsequently either be allocated to external uses such as animal feeds (proteins, lipids) and the extraction of valuable biopolymers (chitin, chitosan) [106] or used to further close internal nutrient cycles of such CMFS by reintegrating the larvae into the fish feed. However, comparing the proximate composition, gross energy and amino acid content of feces from Nile tilapia fed well-digested diets to the body composition of BSF larvae as well as poultry feed as a high quality substrate for larvae rearing [11], suggests that aquaculture sludge is likely not sufficient to support optimal larvae growth when used as the sole nutrient source for BSF larvae cultivation and may also not be recommended due to the risk of potential heavy metal accumulation in the larvae [105]. In animal production, optimal organism growth and health should be of utmost importance, not least due to economic reasons. Therefore, also in insect larvae culture, the animal’s requirements for individual nutrients need to be identified and thereupon the suitability of available resources practically assessed, i.e., the combination of aquaculture sludge and plant wastes in the case of the outlined CMFS. Like in feed formulation for aquaculture as an exercise of determining combinations of complementary raw materials that meet the requirements of a certain species [107], these internally available resource streams could then, if necessary, be supplemented by suitable external biogenic waste streams to create a complete insect larvae diet. In that sense, aquaculture sludge could be viewed as a mineral-rich dietary raw material (especially regarding Ca, N and P) featuring some additional amino acids and other macronutrients. This raw material can further be variable in its composition depending on fish feed formulation and ingredient choice. This was illustrated by the considerably higher Ca and P content in the PM feces originating from the bone fraction of the PM compared to the COM diet which exhibited a higher N (and potentially amino acid) content. Considering the results on dissolved K and Mg excretion in this and prior studies [10,11] as well as the propensity of BSF larvae to accumulate certain minerals [64,89,90,105] including Mg and K [87], aquaculture sludge is likely an insufficient source of Mg and especially K to produce larvae suitable for aquaponic diets that aim to reduce the dependence on externally-sourced K fertilizer and improve Mg supply in the hydroponic unit. Therefore, aquaculture sludge should be complemented with K- and Mg-rich raw materials in the process of insect larvae diet formulation in CMFS such as the one presented in this study. Following the logic above, utilizing aquaculture sludge as outlined would on the one hand need to be evaluated in comparison to alternative uses such as biogas production, use as agricultural fertilizer or sludge remineralization as often advocated for in aquaponic research [18,24,96]. On the other hand, remineralization of frass from BSF larvae production partly based on P-rich aquaculture sludge, e.g., resulting from feeds such as the PM or MIX diet, could present a further avenue to improve P recycling in a CMFS and reduce the dependence on external mineral P fertilizers.

4.4. Implications for Fish Feeds in CMFS and Aquaponics

Hypothetical CMFS production scenarios particularly pertaining to the relative contribution of internal insect larvae production were translated into respective diet formulations which supplement potentially internally available protein sources—meal from fish processing wastes and variable levels of BSFM—with externally sourced protein ingredients in the form of PBM as well as variable levels of PM. Even at higher levels of BSFM inclusion, results suggest that the applied combination of CM and PBM with variable levels of PM and BSFM cannot only produce growth and feed conversion performance in juvenile African catfish that can rival commercially available feed options but also increase the excretion of dissolved SRP and thereby potentially decrease the need for P fertilization in aquaponics. An increase in PM inclusion and concomitant elimination of BSFM in the PM diet enabled improved fish performance and elevated dissolved SRP excretion. With respect to the allocation of internally produced BSF larvae and the consequent feed formulation in CMFS involving African catfish, this suggests that a lower BSFM and a higher PM inclusion than in the MIX diet could strike an optimal balance between maximizing fish performance and SRP excretion and optimizing the recycling of internally produced BSF larvae. In consequence and in reference to the initially outlined production scenarios, this would imply (a) that internal BSF larvae production should only be supplemented with complementary external feed substrates to a level which ensures optimal larvae growth and health and is sufficient to ensure this lower level of BSFM inclusion in the feed for African catfish, or (b) that in case of a substantially larger and thus possibly economically more sensible internal BSF larvae production, a sizeable portion of the harvest may need to be removed from the CMFS and allocated to other external use cases, or (c) that if the entire recycling of internal waste streams and thus BSF harvest into the fish feed is the objective, a potentially reduced fish performance and SRP excretion may need to be accepted in African catfish, at least in comparison to a feed such as the PM diet. However, it may also be possible that rededicating the internally produced BSF larvae entirely to other high-value external use cases, e.g., the extraction of biopolymers or the use as feed ingredient for other animals, outweighs the benefits of including them in the fish feed in the first place. Such a scenario would then require an increase of externally sourced protein ingredients in the fish feed as exemplified by the PM diet. Taking into account that K excretion can be substantially increased with elevated dietary K content, as evidenced by results produced with the COM diet and suggested by previous studies [10,11], it stands to reason that with feeds such as the PM, MIX and BSF diets it may be possible to simultaneously allow for high SRP as well as K excretion while likely ensuring similar growth performance by (a) replacing the employed carbohydrate ingredients by others richer in K, or by (b) increasing the K content of internally produced BSFM, e.g., by complementing aquaculture sludge and wastes from plant production with K-rich external feed substrates.

In any case, it has to be considered that this study utilized proxies for the internally recycled ingredients (BSFM, CM) and results have to be interpreted with these restrictions in mind, which primarily pertain to the quality and composition of the ingredients that would be available in a CMFS, e.g., the protein content and the level of the bone fraction in the fish processing waste meal [12,83] or the amino acid and particularly fatty acid and mineral composition of the BSFM [63,87,108,109]. Furthermore, since defatted BSFM and CM were used in the present study and thus primarily the protein fraction of the underlying resources was utilized, future trials would have to identify what effect a concomitant replacement of the employed poultry fat and rapeseed oil with BSF larvae oil and oil from fresh water fish processing could have on fish performance and eventually also product quality. In a similar manner, legal considerations in the context of CMFS have to be taken into account. Depending on national jurisdiction, the use of insect larvae and by-products from fish processing can be regulated differently. In the EU, for instance, BSF larvae [110] and animal by-products such as from fish processing [111,112] have been allowed in aquafeeds for some time, but feeding fish meal from one species back to the same species as simulated with the CM in the present study is prohibited. This possible limitation on the recycling of fish by-products within single-species fish production can be circumvented by producing two species in parallel such as African catfish and Nile tilapia and reallocating the resulting processing wastes accordingly. However, despite the generally recognized potential of BSF larvae for sustainable biowaste recycling [7,113], which is also one of the foundations of the CMFS concept presented in this study, stricter jurisdictions such as in the EU or the USA up to the present date do not allow some biowastes (e.g., manure) to be fed to larvae intended as an ingredient for animal feeds. This must be borne in mind throughout the process of investigating novel food production systems such as the outlined CMFS [6,110,114].

When interpreting the findings of this study detached from the CMFS point of view and purely through the lens of aquaponic diet development, the results are nonetheless encouraging. The tested ingredient combinations, particularly regarding PM and freshwater fish processing meal, may not only enable adequate performance of African catfish juveniles without the inclusion of any marine ingredients but also substantially increase SRP excretion while not increasing TIN excretion compared to a commercial diet. Based on the dissolved release rates per unit of feed, the N:P ratios achievable with, e.g., the PM (10:1) or the MIX diet (10:0.6) much more closely resemble those of hydroponic standard solutions (10:1.5–10:3) [115] than would be achievable with the COM diet (10:0.1). Hence, sufficient inclusion of PM and fish processing waste meal in specialized aquaponic feeds could reduce the need for supplementary phosphate fertilization in on-demand coupled aquaponic systems and help to accomplish the dual objective of good fish performance and improved nutrient supply for hydroponic plant production. As alluded to previously in the context of the COM diet, future research should identify feed ingredients that in tandem with P-rich raw materials such as PM and fish processing waste meals could enhance the excretion of other important and in aquaponics insufficiently available plant nutrients; this applies especially to dissolved K, as one of the most deficient yet also dietarily manipulable nutrients. Further upcoming steps would include the investigation of optimized aquaponic diets for all life stages of African catfish and other species in order to evaluate their effects on growth and nutrient excretion and ultimately judge their potential for optimal fish performance and fertilizer savings in full-fledged staggered industrial aquaponic production.

5. Conclusions

This study outlined the potential of purpose-specific diets for CMFS and aquaponics that recycle system-internal nutrient streams, optimize fish performance and improve nutrient availability in aquaponic plant production. This could be one pathway towards improving the sustainability of such food production systems. Results indicate that combining CM with PBM and, depending on the production scenario of the hypothesized CMFS, variable amounts of PM and BSFM, can enable similar or, in case of entirely substituting BSFM with PM, even better fish performance and higher dissolved SRP excretion in African catfish juveniles versus a comparable commercial feed that is optimized for low environmental impact in open system aquaculture. Depending on the objective of the CMFS as well as economic and production-related factors, differences in fish performance and dissolved SRP excretion between experimental diets could influence decisions regarding the sizing of the respective insect larvae production units and/or the allocation of the BSF larvae harvest between use as an ingredient of the fish feed or removal from the internal nutrient cycle; as appropriate, larvae could be rededicated to system-external uses such as biopolymer extraction or production of feeds for other animals. Experimental results generally emphasize the high suitability of bone mineral-rich PM and fish meal from processing wastes such as CM for aquaponic diets and, if the goal is to maximize fish performance as well as SRP excretion, lower BSFM and higher PM inclusion should be favored as illustrated by the differences between the MIX and the PM diet.

Supporting prior findings, the considerably higher K excretion caused by a higher dietary K content observed with the COM diet points to the potential of simultaneously enhancing K excretion by strategically complementing ingredients such as PM and fish meal from processing wastes with K-rich raw materials. In any case, bone mineral-rich feeds along the lines of the PM and the MIX diet will tend to result in fish feces highly abundant in Ca and P with moderate levels of N, while K content in the sludge appears comparatively less affected by dietary K content and only accumulates to low levels. This has to be taken into consideration irrespective whether the sludge is used as agricultural fertilizer, is remineralized or recycled as an ingredient of BSF larvae feed substrates.