Get the Fish Out: Black Soldier Fly Larvae and Marine Macro-algae as Feed Ingredient Replacements for Small Land-based Aquaculture Operations

Brown trout were stocked in six 120 gal systems at 1 – 1.2 kg per tank. The average starting length was 205.6mm (±16.8) and 92.6g (±27.4). At the experiment conclusion the 20% Fish and Control treatments showed a significant positive change (Bonferroni corrected one tail T-test < 0.008) in average fish weight compared to day 1 of the 90 day experiment (Fig. 1). Variability was high for the final average weight of all treatments. The Trout feed had the highest final average weight, followed by Trout+BSFW and 50% fish. Of the remaining treatments 0%, 20%, and 100% Fish were all very similar. The high variability of all treatments except 20% and Control may be an indicator of palatability issues.

Figure 1: Average fish weight by treatment at the beginning and end of the experiment. Trout+BSFL treatment was stopped early due to a large number of mortalities. Difference in avg. start and end weights were significantly different only for the Trout (Control) and 20% Fish treatments.

Mortality was highest for the Trout+BSFL treatment. By the end of the experiment nearly half of the starting biomass (47%) had been removed as mortalities (Fig. 2). The mortalities were due to a mixture of bacterial infection and intraspecies aggression. It appeared that some fish found the food mixture palatable and others did not. Those that did grew faster and became more aggressive towards those fish that did not eat as well. Much of the BSFL feed was left uneaten as well. The Trout+BSFL treatment was removed from the experiment after 32 days as it did not appear that a solution for the poor performance was readily available. Incidentally, aggression associated with the uneven growth tracks persisted even after this tank was switched to all trout feed. Half of the number of fish that started the experiment had died after 90 days.

The 0% Fish treatment had the second highest mortality by weight. Six fish, or 468 g (43%) of the starting biomass was removed from the tank by end of the experiment. Most of the moralities happened early, possibly associated with a palatability problem with the feed as the fish that died did so within the first 30 days of the experiment and tended to be small (µ= 78g±20). Bacterial infection issues were found in this tank as well. Spontaneous mortalities declined after two sequential salt baths that occurred for all tanks around day 38 of the experiment.

Fish growth was highest for the 20% Fish treatment, after the Control. The 20% Fish treatment had zero mortalities a positive weight change of 277g of mass, followed by the 100% Fish treatment at 265g. The Control treatment added 362g of positive weight change in comparison, a 23% improvement over the next best treatment diet.

Figure 2: Change in fish weight over the course of the experiment. Weight change for Trout+BSFL and 0% Fish were negative on account of mortalities during the experiment. Other treatments showed positive growth, with 20% Fish highest, followed by 100% Fish.

The second best feed conversion ratio was for the 20% Fish diet (Fig. 3). The Trout (Control) diet had a FCR = 1.02 and was the lowest of the tanks that performed well. The 0% Fish diet was less than one (FCR = 0.66) but this was in part because of the high mortality in the 0% Fish treatment. The 20% Fish diet had an FCR = 1.33; the Trout (Control) diet was 23% more efficient at producing fish mass. The 100% Fish diet had an FCR = 1.5 and the 50% Fish diet had an FCR = 1.83.

Figure 3: Feed conversion ratio for Trout+BSFL and 0% Fish were less than one owed to mortalities during the experiment. The 20% Fish diet was the next lowest FCR. The Trout (control) diet had the highest FCR.

Three sample fish from each tank were sent to Kennebec BioSciences and UNH New Hampshire Veterinary Diagnostic Laboratory for bacterial, liver and intestinal analysis. Sections of liver looked normal and were similar between groups. Bacterial kindney analysis results included one fish from the 100% Fish tank with a 5K count of Shewanella sp. Two fish from the 20% Fish tank had 12K count of Bacillus sp. Two fish from the 0% Fish tank had a 15K count of Shewanella sp. which was considered high. Shewanella is a ubiquitous genera of bacteria found in aquatic habitats. They are facultative anaerobes and are common component of the surface flora of fish. Bacillus is a large genus of endospore forming bacteria. They are resistant to most treatments (heat, antibiotics, desiccation, radiation and chemicals) in their endospore form. Both bacteria are suspected contributors to and accelerants of food spoilage.

Histopathology of the sample fish consisted of a qualitative assessment of gut health. Overall, there appeared to be an increasing degree of enteritis with decreasing amounts of fish meal in the diet (Table 3). Control fish had patchy to diffuse mucosal autolysis and sloughing in multiple sections of the intestine, which was described as moderate to marked where they occurred. Mucosal exocytosis by lymphocytes was described as moderate in the Control fish. Fish on the 100% Fish diet had the same patchy to diffuse autolysis and sloughing, described as moderate in severity. Mucosal exocytosis by lymphocytes was described as sparse to mild. Fish on the 50% Fish and 20% Fish diets had the same patchy to diffuse autolysis and sloughing, described as moderate to marked in severity. Mucosal exocytosis by lymphocytes was described as moderate and increased relative to the Control fish. Fish on the 0% Fish diet had widespread mild autolysis and sloughing, described as moderate to marked in severity. There were moderate to large numbers of lymphocytes present with moderate mucosal exocytosis.

Table 3: Diagnostic results of intestinal health inspection.

Fish food quality can have drastic effects on water quality in recirculating systems. Turbidity is a measure of water clarity, which largely affects the visual aspects of the growing area. Visitors cannot perceive pH or conductivity of the growing environment but they can readily tell if the water is stained. Feed that holds together better has a more positive visual effect on growing systems. We measured turbidity as three samples per tank at the end of the experiment (Fig. 4). The Trout+BSFL treatment had the lowest turbidity. The Control tank had the highest turbidity. In general, the less fish meal included in the diet the lower the turbidity, however, a statistical difference could not be detected between the 0%,20%, 50% and 100% treatments.

Figure 4: Turbidity comparison between treatments. There was a significant difference (Tukey Post hoc) between Trt+BSFL vs. 100% Fish, Trt+BSFL vs. Trout and Trt+BSFL vs. 0% Fish. There was a marginal difference between 0% Fish vs. 100% Fish and 0% Fish vs. Control.

pH is both a controlling and indicator variable for recirculating aquaculture systems. Bacteria that colonize the biofilter are highly reactive outside of their preferred pH range and may stop functioning. This in turn will cause the pH to decline even more. pH also affects the ability of plants in aquaponic systems to sequester nutrients. Prior experience with the aquaponics systems at USM, given the produce planted for the experiment, cucumbers and bell peppers, indicated that pH between 6.2 and 6.6 was optimal. The Trt+BSFL and 0% Fish tanks deviated outside of this range the most of the six treatments (Fig. 5). This could be a reflection of poor palatability leading to left over food, resulting in ammonia spikes that contribute to declining pH. Instances where pH>6.6 were a reaction to repeated buffer additions to the systems to bring the pH back into the optimal range.

Figure 5: Fluctuations in pH of different treatments over the course of the experiment.

Of the six treatments, the Trt+BSFL and 0% Fish treatments had the highest average pH and were the least stable around that average (Fig. 6). Their average pH=6.45 but ranged as high as 6.8 and as low as 6.1. The 20%, 100% and Control treatments were very similar in average pH, 6.3±0.08.

Figure 6: Average pH and standard deviation of treatment pH over the course of the experiment. There was no significant difference in pH between treatments. Treatments 1 and 2 showed the highest average pH as well as the highest variation, aka least stability in pH.

Conductivity is a general measure of ion concentration in water. Most freshwater fish are only affected by extremely low and extremely high conductivities, and often only when being moved between tanks or water sources where one is high and the other much lower. Plants are more directly affected by conductivity as ion concentration affects their ability to sequester nutrients from the liquid medium. High conductivity water has lots of nutrients but the plants may be saturated and unable to absorb the nutrients they actually need. Low conductivity water may not have enough nutrients overall. Cucumbers and bell peppers like conductivities between 800 and 1200 µS/cm, generally. The 20% fish treatment was near the top of this conductivity range (Fig. 7). The 50% fish treatment was near the bottom of this range. All other treatments were above (100% Fish and Control) or Below (0%Fish, Trt+BSFL) this range.

Figure 7: Average conductivity of treatments across experimental period. Conductivity is most relevant to plant growth. Cucumber and non-spicy peppers do well at conductivities between 1000 and 2000 micro siemens.

Canopy Farm, as an aquaponic grower, is as interested in vegetable produce production as it is in fish production. For aquaponic operations, produce makes up ½ - ²/₃ of the revenue from the operation. In short, it takes 12-18 months to grow a fish to market size but it takes 12-18 weeks to grow a head of lettuce or a batch of cucumbers from seed. Cucumbers and bell peppers will fruit repeatedly over the life of a plant as well, resulting in a six week interval from crop to crop. Plants, especially leafy greens, also do not have the same density restrictions as fish so more mass can be produced per unit area.

The 50% Fish treatment produced the most vegetables of any system (Fig. 8). The 50% Fish treatment exceed the production of the best experimental diet, 20% Fish, by nearly a factor of 3. The Control and 100% Fish treatments production was exceeded by more than a factor of 5.

Figure 8: Produce production was clearly highest for the 50% Fish feed recipe. The 100% Fish feed recipe was associated with the least amount of produce grown.

We performed an economic analysis of the cost of making feed vs buying and shipping feed from an online retailer. The items were all ordered from the web at the best price available from reputable retailers that could be found within 5-10 minutes of searching. Shipping costs were included in the price of the item. Savings from bulk shipping of mulitple items from one retailer was not factored in, only the cost of shipping individual items. Locally available ingredients, Distillers Waste Grain, Beer Waste Grain and Soy Okara were considered free and had the effect of reducing feed costs by reducing the amount of fish meal and BSFL meal required. Feed ingredients were convered to $/g to calculate feed cost per 1000 lbs (453.6 kg), or 22 feed bags at 44 lbs each.

We considered the cost of our three mixes with BSFL, our 100% Fish mix and a readily available commercial feed (Control). Feed cost was sensitive to the cost of BSFL meal. The BSFL we used was donated by a manufacturer. At the start of the experiment no commercial sources of BSFL were available on the internet so we had to approach a manufacturer/ wholesaler directly. At the time of writing this report BSFL meal was available at a few internet retailers but most required contacting them to get a current price. The one location that simply listed a price, a feed ingredient site for pet reptiles, provided a price that worked out to $0.05 per gram. At the internet price our 0% Fish feed cost 10x as much as the commercial (Control) feed, before adding about $300 to cover personnel time to prepare, make and bag the homemade feed. Furthermore, the menhaden fish meal we used cost $0.003 per gram, over an order of magnitude less than the BSFL meal. Overall, the Control feed was the best buy at half the cost of even the 100% Fish feed we made from scratch. To be competitive with the 100% Fish feed made from scratch, BSFL meal would need to cost $0.005/g, similar to the cost of the menhaden fish meal (Entosystems estimates selling BSFL meal at $4/kg). At that price 50% Fish feed is the best value among the three BSFL meal feeds, though it would still cost about $372 more per 1000 lbs. Even then, the Control feed would still cost 41% less than the least expensive feed made from scratch. At a cost of $0.001/g BSFL meal is 36% more expensive than commercial (Control) feed and 6.6% less expensive than the 100% Fish feed made from scratch. At $0.005/g BSFL meal, soy protein isolate becomes the most expensive ingredient. A less expensive source of this protein content enhancer might bring the cost down sufficiently to make hand made feed competitive. At an estimated cost of $0.003/g for soy protein isolate, 100% Fish feed is only 10% more expensive than the commercial (Control) feed.

Table 4: Costs estimates for 1000 lbs (453.6 kg) of feed at different black soldier fly larvae meal (BSFL) costs. Prices are expressed on per gram basis. The $0.05 line represents a real price per gram found at an internet retailer (accessed 6/11/2023). Other costs lines are estimates to explore scenarios that would make reduced fish meal feed more economical. All non-control feeds would have $272-$300 added to the production costs to account for labor of two people to prepare the feed.