System components

Each trial used replicated aquaponic systems (Fig. 1); trial 1 used six separate systems and trial 2 used four separate systems. Each aquaponic system consisted of one 110-gal fish tank, one 50-gal cone-bottom clarifier, one 30-gal cone-bottom mineralization tank, one 50-gal sump (Polytank, Litchfield, MN, USA), and two 120-gal hydroponic troughs (Red Ewald, Karnes City, TX, USA). There was one floating raft (2-inch polystyrene foam board) in each hydroponic trough for a total area of 32 ft² of raft space per system. Each raft was then subdivided into two sections that each had a total growing area of 6 ft² and held 15 plants, for a total of four growing chambers. To isolate each growing chamber, plastic paneling (Sequentia® Embossed White Wall Panel; Crane Composites, Channahon, IL, USA) partitioned the hydroponic troughs into four growing chambers. In trial 1, the partitioning resulted in a total of 24 growing chambers across the six aquaponic systems; one experimental light was placed in each chamber. Each type of experimental light was placed in a total of three growing chambers, giving each treatment three replicates. In trial 2, the partitioning resulted in a total of 16 growing chambers across the four aquaponic systems; one experimental light was placed in each growing chamber. Each type of experimental light was placed in a total of four growing chambers (one replicate per aquaponic system), giving each treatment three replicates. Between systems, 4-mil black plastic sheeting was used to isolate each system from light contamination by neighboring systems. Each system was driven by a 1022 gal/h submersible pump (Model 4000 Quiet One Lifeguard Pump; Lifeguard Aquatics, Santa Fe Springs, CA, USA) located in the sump. The fish tank was the highest point of the system and water gravity fed into filters, hydroponic troughs, and the sump. Water was then pumped from the sump back into the fish tank. The flow rate exiting the trough into the sump was standardized to 19.5 L·min⁻¹.

View Figure

• Download as Powerpoint
• Download Figure

Light technologies

The height of the lights was adjusted twice per week to maintain a photosynthetic photon flux density (PPFD) of 200 µmol·cm⁻²·s⁻¹ at the top leaves across the plant canopy (LI-180 Spectrometer; LI-COR Biosciences, Lincoln, NE, USA). Seven points of PPFD were captured from each system quadrant and averaged together to arrive at an average of 200 ± 5 µmol·cm⁻²·s⁻¹ PPFD across the raft. Lights with adjustable intensity were maintained at the same intensity over the course of the study to avoid disrupting electrical use data. Electrical use of each light was monitored using a kilowatt-hour monitor (Kill-A-Watt p4400; P3 International, New York, NY, USA) for the collection of total run time, total kilowatt-hour, and current wattage at the end of day. Photoperiod was 16 h of light, followed by 8 h of dark.

Trial 1: light technologies

Eight LED light treatments were evaluated and compared in trial 1. The control was a high-end LED grow light [NEO (NeoSol DS; Illumitex, Austin, TX, USA)], a full-spectrum LED that had been used in several previous studies at KSU (Oliver et al. 2018; Schultz et al. 2021). The seven experimental light treatments were as follows: Designers Fountain [DES (6-Light 5000 K LED High Bay Light; Designers Fountain, Rancho Dominguez, CA, USA)], Worldwide Lighting [WIDE (1-Light 4000K LED High Bay Light; Worldwide Lighting Corp, Hayward, CA, USA)], Spider Farmer [SPI (SF-2000; Spider Farmer, Alhambra, CA, USA)], Fluence [FLU (RAZRx Fluence; Osram, Austin, TX, USA)], King Plus [KING (Kinging Plus 1000 W LED Grow Light; Shenzheng King Lighting Co, Shenzhen, Guangdong, China)], Growstar [GROW (Growstar 600W LED Plant Light; Growstar Store, ON, Canada)], and Rural King [RK (4’ LED 10000 Lumen Shop Light; Rural King, Mattoon, IL, USA)].

Criteria for choosing lights in trial 2

Using data from trial 1, experimental light treatments were ranked in terms of total plant biomass (grams) produced and plant production per unit of electricity (grams per square meter per kilowatt-hour). These criteria were chosen because they address issues experienced by aquaponic producers including profitability and electrical use (Pattillo et al. 2022). Those rankings were combined into a summed rank and those with the lowest summed rank moved on to trial 2.

Trial 2: light technologies

The NEO light was the control and DES, SPI, and FLU were the three experimental light treatments. This experiment used a complete block design (Steel and Torrie 1980). All four lights were represented in each of the four replicate systems, with each light located in a different quadrant across systems (Fig. 1).

The light spectra for each of the four lights in trial 2 were recorded at 200 µmol·m⁻²·s⁻¹ PPFD. Percentage of each spectrum range, 400 to 499 nm (blue light), 500 to 599 nm (green light), and 600 to 699 nm (red light), were used to determine photon efficiency (μmol per watt), of each light.

Fish and plant materials

Before this research, the aquaponic systems were operated for at least 3 months to ensure a well-established microbial community.

Approximately 2 weeks before the start of each trial, ‘Buttercrunch’ bibb lettuce (Lactuca sativa var. capitata) seeds (Eden Brother, Arden, NC, USA) were placed in 1.5-inch synthetic cubes (Rockwool; Rockwool International, Hedenhusene, Denmark) and raised under the control LED lights inside a climate-controlled room, with an ambient temperature of 22.2 °C. On emergence of four true leaves (∼14 d), seedlings were placed into net pots and randomly assigned to treatments across systems. Lettuce was stocked at 15 seedlings per light chamber, five plants by three rows (60 plants per system or 26.7 plants/m²).

To ensure similar water quality and nutrient load across systems, water was homogenized between systems 3 d before fish stocking. Water was homogenized by moving water from one system to the adjacent system for 3 h, using the 1022 gal/h pump native to each system. Water was considered homogenized when nitrate (NO₃-N), pH, and electrical conductivity (EC) were within ± 0.1 mg·L⁻¹, pH, and mS·cm⁻¹, respectively. Nile tilapia (Oreochromis niloticus) were placed in a single 450-gal tank and held at 3 parts-per-thousand salt (NaCl) for 3 d before stocking. Stocking density was based on a target feeding rate of 60 g·m⁻² fish feed per vegetable grow space per day (Rakocy et al. 2003). Fish were fed 160 g of floating 32% protein commercial fish feed (Prod 32; Rangen Inc., Buhl, ID, USA) split into two feedings per day, morning and afternoon.

Measurements and calculations

Water-quality parameters such as dissolved oxygen (DO), pH, and water temperature (°C) were measured daily from the sump with a digital probe (ProPlus; YSI, Yellowsprings, OH, USA). EC was measured with a commercial EC meter (Truncheon; Bluelab Corporation Limited, Tauranga, New Zealand). Air temperature (°C) and humidity (percent) were monitored daily using scientific hygrometers hung 8 inches above the rafts (Dial Hygrometer; Sper Scientific, Scottsdale, AZ, USA). Twice weekly, total ammonia-nitrogen (TAN), nitrite (NO₂-N), NO₃-N, and total iron (Fe) were measured using a spectrophotometer (Hach DR/2000; Hach Company, Loveland, CO, USA). Alkalinity was measured by titration (Digital Titrator; Hach Company). When the pH measured below 6.8, calcium carbonate (CaCO₃) and potassium carbonate (K₂CO₃) were alternately added to maintain pH at 6.8. Iron was maintained at 2.0 mg·L⁻¹, based on the biweekly water-quality analysis and adjusted using 10% diethylenetriamine pentaacetic acid (DTPA) chelated Fe.

The harvest date for each trial was based on plant growth. When leaves of plants in any treatment began to overlap, the trial was ended. At that time, all plants were removed with the stalks cut just above the rockwool cubes and roots removed from below the rockwool cubes. The shoots (leaves and stems) and roots were weighed separately to the nearest 0.1 g using a digital scale (New Classic model MS12001L; Mettler Toledo, Columbus, OH, USA).

The center row of five plants had their individual wet and dry shoot and root weights measured and served as the representative samples. Data taken from representative samples included chlorophyll content (CCM-200 Plus Chlorophyll Meter; Opti-Sciences Inc., Hudson, NH, USA), number of leaves, and leaf surface area (cm²).

At the conclusion of each trial, all the fish were removed, counted, measured, and weighed. Representative fish samples in trial 1 (n = 21) and trial 2 (n = 10) were taken to determine individual weight and length differences from original stocking. Fish performance was based on average daily weight gain per fish (grams), survival (percent), feed conversion ratio (FCR = grams of dry diet feed/grams live weight gain), and specific growth rate {[SGR (percent per day)] = [ln(harvest weight) − ln(stock weight)/number of days in study] × 100}.

Trial 1: fish stocking

The fish tanks were stocked at 45.1 fish/m³. The tilapia had an average initial body weight of 278 g. Tilapia were all-male hybrids (Americulture Inc., Animas, NM, USA). Before the experiment, fish were maintained in recirculating systems at KSU.

Trial 1: environmental conditions

Over the 15-d trial, water-quality variables in the six systems averaged (± SD): temperature 25.9 ± 0.3 °C; DO 7.0 ± 0.4 mg·L⁻¹; pH 7.8 ± 0.3; EC 0.73 ± 0.06 mS·cm⁻¹; TAN 0.19 ± 0.137 mg·L⁻¹; NO₂-N 0.35 ± 0.137 mg·L⁻¹; NO₃-N 9.06 ± 5.44 mg·L⁻¹; alkalinity 58.2 ± 8.7 mg·L⁻¹; and Fe 1.83 ± 1.03 mg·L⁻¹. These values represented suitable conditions for culturing Nile tilapia and lettuce (Rakocy et al. 2003). The average flow rate for the systems in trial 1 was 5.28 gal/min.

Trial 2: stocking

The fish tanks were stocked at 21.6 fish/m³. The tilapia had an average initial body weight of 586 ± 162 g.

Trial 2: environmental conditions

Over the 17-d trial, the water-quality variables in the four systems averaged (± SD): temperature 26.0 ± 0.5 °C; DO 6.9 ± 0.2 mg·L⁻¹; pH 7.2 ± 0.4; EC 0.6 ± 0.0 mS·cm⁻¹; TAN 0.18 ± 0.14 mg·L⁻¹; NO₂-N 0.16 ± 0.14 mg·L⁻¹; NO₃-N 32.35 ± 3.69 mg·L⁻¹; alkalinity 20.14 ± 6.40 mg·L⁻¹; and Fe 2.09 ± 0.23 mg·L⁻¹. These values represented suitable conditions for culturing Nile tilapia and lettuce (Rakocy et al. 2003). The average flow rate for the systems in trial 2 was 19.5 ± 1.5 L·min⁻¹.

Trial 1: statistical analysis

The effects of the different lights on plant growth in trial 1 were analyzed in statistical software (Statistix ver. 10; Statistix, Tallahassee, FL, USA) based on a completely randomized design and compared using an analysis of variance (ANOVA) with significance set at P ≤ 0.05. If significant differences were indicated by the ANOVA, Fisher’s least significant difference test (LSD) was used to separate means (Steel and Torrie 1980).

Trial 2: statistical analysis

The effects of the different lights on plant growth in trial 2 were analyzed in Statistix 10 as a complete block design and compared using ANOVA with significance set at P ≤ 0.05. If significant differences were indicated by ANOVA, Fisher’s LSD test was used to separate means (Steel and Torrie 1980).

Trial 2: plant nutrient analysis

Nutrient concentrations in the representative samples’ leaves were also determined. Leaves from the five representative sample plants were dried using a convection drying oven (DEC5-32; Hobart, Troy, OH, USA) at 70 °C for 72 h (Arshadullah et al. 2011). The dried plant tissues were analyzed for nitrogen (N), phosphorous (P), potassium (K), calcium (Ca), magnesium (Mg), sulfur (S), Fe, molybdenum (Mo), manganese (Mn), boron (B), copper (Cu), and zinc (Zn) by a commercial laboratory (Micro Macro International, Athens, GA, USA). Water samples were taken from each system at the start and conclusion of each trial to be analyzed for NO₃-N, NH₄-N, P, K, Ca, Mg, Bo, Cu, Fe, Mn, Mo, salt (Na), sulfate (SO₂⁻⁴), fluoride (F⁻), chloride (Cl⁻), aluminum (Al), Zn, carbonates, bicarbonates, pH, EC, alkalinity, total dissolved solids, and hardness (Micro Macro International).

Trial 2: partial budget analysis

A partial budget analysis was used to evaluate the net change in profits and the cost/benefit ratio for the three experimental lights (DES, SPI, and FLU), each compared with the control treatment (NEO) (Table 1) (Rabin et al. 2007). A value of $2.64/lb was used based on market value reported for bibb lettuce at a farmers market in Fayette County, Kentucky, USA on 3 Jul 2023 (Wolff 2023), assuming a head of lettuce weighs 300 g. Average total biomass (grams) for each light treatment was extrapolated out to a full year supposing the same average production per light over 20.8 grow out cycles, or 17 d per grow out cycle as reported in trial 2. Electrical prices were calculated using the rate of $0.11/kWh for residential electricity in Kentucky, USA. This value was multiplied by the number of kilowatt-hours used per day per light and multiplied by 365 to obtain electrical usage for a full year. For a partial budget analysis, the additional purchase costs (i.e., increased fixture cost) and reduced returns (decreased production) were quantified as the costs of the change. Any additional revenue and reduced costs were quantified as the benefits of the change. In this study, the change was from NEO to a less expensive LED light. The costs were summed together and the benefits were summed together to create total costs and total benefit amounts. The total costs were subtracted from the total benefits to produce a net change in profits (NCP). The total benefits were also divided by the total costs to produce a BCR (Table 1) (Rabin et al. 2007).

Table 1. Method for calculating partial budget analysis of bibb lettuce (Lactuca sativa var. capitata) grown in aquaponic systems under four light-emitting diode (LED) lights in a complete block design. In this study, the additional purchase costs (i.e., increased fixture cost) and reduced returns (decreased production) were quantified as the costs of the change. Any additional revenue and reduced costs were quantified as the benefits of the change. In this study, the change was from a high-end LED light to a less expensive LED light. The costs were summed together, and the benefits were summed together to create total costs and total benefit amounts. The total costs were subtracted from the total benefits to produce a net change in profits. The total benefits were also divided by the total costs to produce a benefit/cost ratio.

View Fullscreen

Trial 1: fish production

The average individual harvest weight of the fish for the six systems was 325.3 ± 0.9 g with 97.3% survival. The SGR was 0.85% ± 0.14% per day and FCR was 2.15 ± 0.42 g·g⁻¹. The average daily gain per fish was 2.49 ± 0.42 g.

Trial 1: plant production

Bibb lettuce grown under NEO, FLU, SPI, GROW, and WIDE achieved significantly higher (P ≤ 0.05) total biomass weights than the KING (Table 2). Total biomass weights of bibb lettuce grown under RK and DES did not differ significantly (P > 0.05) from KING or from the NEO, FLU, SPI, and WIDE (Table 2). Bibb lettuce grown under FLU achieved significantly higher (P ≤ 0.05) representative individual lettuce weights than DES and KING. Average individual weights of bibb lettuce grown under NEO, SPI GROW, WIDE, and RK did not significantly differ (P > 0.05) from FLU or DES but were significantly higher (P ≤ 0.05) than KING. Lettuce plants grown under FLU, DES, and RK achieved significantly greater (P ≤ 0.05) production per unit energy (grams per square meter per kilowatt-hour) than NEO, SPI, GROW, KING, and WIDE. Statistical differences between lights are reported in Table 2.

Table 2. Production values from trial 1. Seven low-cost light-emitting diode (LED) lights were compared with a high-end horticulture LED grow light and evaluated in a complete randomized design for growth of bibb lettuce (Lactuca sativa var. capitata) in six indoor aquaponic systems. The plant beds from each aquaponics system were divided into four sections separated by vertical barriers, creating a total of 24 sections of plant bed. Each plant bed section was randomly assigned to one of the eight treatments; each treatment was given three replicates.

View Fullscreen

Characteristics for experimental lights

The ratio of red PPFD to blue PPFD (R:B), photon efficiency (PE) μmol per watt, and Kelvin (K) rating [measured as color correlated temperature (CCT)] were recorded for each light using the previously described spectrometer and are reported in Table 3. Kelvin measurements for KING did not register on the spectrometer so the following equation was used to calculate K using (x,y) coordinates from the Commission Internationale de l'Elcairage (CIE) 1931 color map (McCamy 1992): CCT = 437n³ + 3601n² + 6861n + 5517, where n = (x − 0.3320)/(0.1858 − y). Coordinate values for KING were x = 0.2305 and y = 0.1020.

Table 3. Rankings of lights from trial 1 used to select lights for trial 2. In trial 1, seven low-cost light-emitting diode (LED) lights were compared with a high-end horticulture LED grow light and evaluated in a complete randomized design for growth of bibb lettuce (Lactuca sativa var. capitata) in six indoor aquaponic systems. The plant beds from each aquaponics system were divided into four sections separated by vertical barriers, creating a total of 24 sections of plant bed. Each plant bed section was randomly assigned to one of the eight treatments; each treatment was given three replicates. Ranks were ordered based on the highest total plant biomass production at harvest and highest efficiency, which was calculated using total plant biomass produced per square meter of growing space per kilowatt-hour.

View Fullscreen

Ranking results: tie breaker

As in trial 1, the control for trial 2 was NEO. Two experimental lights, DES and FLU, were clear choices based on total biomass and plant production per unit of energy rankings. There was a three-way tie for the third experimental light. The third light was chosen by considering the standard deviation of the PPFD values as a tiebreaker. A lower standard deviation would indicate the most consistent spread of light across the plant rafts and a more consistent growth pattern among the plants (Table 3). Based on this, SPI was chosen as the third experimental light. The four lights used for trial 2 were NEO (control), DES, FLU, and SPI.

Trial 2: fish production

Over the 17-d trial, tilapia had a 97.2% survival. The average harvest weight of the fish for the four systems was 728.0 ± 175.7 g. The average SGR was 1.08% ± 0.08% per day, the average FCR was 1.67 ± 0.11 g·g⁻¹, and the average daily gain per fish was 0.84 ± 0.13 g.

Trial 2: plant production

In trial 2, lettuce grown under NEO achieved significantly higher (P ≤ 0.05) total biomass (939 g) than lettuce grown under DES (812 g), and FLU (913 g) and SPI (917 g) were intermediate and did not differ (P > 0.05) from NEO or DES (Table 4). Lettuce grown under DES had significantly less energy consumption per day (1.24 kWh) than SPI (1.50 kWh), FLU (1.49 kWh), and NEO (1.72 kWh), and NEO was significantly higher in energy consumption per day than SPI and FLU. Lettuce grown under DES had significantly greater production per unit of energy (36.6 g·m⁻² per kilowatt-hour) than lettuce grown under the control (NEO) (30.5 g·m⁻² per kilowatt-hour), while the lettuce grown under the FLU (34.4 g·m⁻² per kilowatt-hour and SPI (34.3 g·m⁻² per kilowatt-hour) did not significantly differ from DES or NEO (Table 4). There was no significant difference (P ≤ 0.05) between light treatments in terms of average individual plant weight, CCI, LSA, or root to shoot ratio (Table 4).

Table 4. Production values for trial 2. Bibb lettuce (Lactuca sativa var. capitata) was grown in aquaponic systems under four different light-emitting diode (LED) lights. Effects on plant performance between each light were compared in a complete block design in four aquaponics systems. The plant beds from each aquaponics system were divided into four sections separated by vertical barriers, creating a total of 16 sections of plant bed. Each plant bed section in each aquaponics system was randomly assigned to one of the four treatments; each treatment was given four replicates.

View Fullscreen

Trial 2: plant nutrient analysis

The compositional analysis of bibb leaf samples grown under the different test lights is presented in Table 5. When comparing the nutrient analysis of the leaves with published expected values (Bryson et al. 2014), all treatments had values lower than expected in K, Ca, Mg, Cu, and Al. All treatments had values higher than expected for N, Fe, Mo, Si, and Na. All treatments also had values within the expected range for Mn, B, and Zi. In terms of S content, NEO was outside the expected range, while all other light treatments were within the expected range. For Ni, NEO and SPI had lower, FLU had higher, and DES was within the range of expected values. Out of all nutrients analyzed, Si and Na levels were those most outside the anticipated range. Silicon in lettuce normally ranges between 0.2% and 6.5%, but the samples for NEO, FLU, DES, and SPI were 40.7%, 37.9%, 43.7%, and 38.7%, respectively. Sodium has an expected range of 200 to 800 ppm, but the samples for NEO, FLU, DES, and SPI were 2024, 1681, 1796, and 1607 ppm, respectively.

Table 5. Nutrient analysis for trial 2. Mean (± standard deviation) of nutrients in bibb lettuce (Lactuca sativa var. capitata) grown in aquaponic systems under four different light-emitting diode (LED) lights. Effects on plant performance between each light were compared in a complete block design in four aquaponics systems. The plant beds from each aquaponics system were divided into four sections separated by vertical barriers, creating a total of 16 sections of plant bed. Each plant bed section in each aquaponics system was randomly assigned to one of the four treatments; each treatment was given four replicates.

View Fullscreen

Trial 2: partial budget analysis

A partial budget analysis (Rabin et al. 2007) was used to evaluate the NCP and the cost/benefit ratio for the three experimental lights from trial 2 (DES, SPI, and FLU), each compared with the control [NEO (Table 6)]. When comparing the less expensive LED lights with the expensive control LED, the results for the NCP over 1 year for the less expensive test lights were all positive: $1305 per year for DES, $1107 per year for SPI, and $1042 per year for FLU. This net change in profit is the difference in total added income and total reduced income caused by one of the three experimental lights instead of the NEO. A partial budget analysis does not determine the profitability of an enterprise. It only determines what additional profit (or loss) could be realized if an alternative light is used in place of the NEO, with all other factors remaining the same.

Table 6. Partial budget analysis for trial 2. Bibb lettuce (Lactuca sativa var. capitata) was grown in aquaponic systems under four different light-emitting diode (LED) lights. Effects on plant performance between each light were compared in a complete block design in four aquaponics systems. The plant beds from each aquaponics system were divided into four sections separated by vertical barriers, creating a total of 16 sections of plant bed. Each plant bed section in each aquaponics system was randomly assigned to one of the four treatments; each treatment was given four replicates. Net change in profits, benefit/cost ratio, yearly production values, yearly electrical costs, value of 1 lb (0.45 kg) of lettuce, and electrical cost per kWh are presented to highlight potential change in profit between different light types.

View Fullscreen

A BCR of 1 indicates that the profits will equal the cost of the change. The greater a ratio is above 1, the greater yield a producer would see by making that change. When comparing the less expensive LED lights with the expensive control LED, the results for the BCR were 6.07 for DES, 3.37 for SPI, and 2.96 for FLU. BCR shows whether an expected change will yield a positive net present value.

Trial 1 discussion

The primary of trial 1 was to screen low-cost LED lights in terms of plant growth and select the most promising candidates for a more thorough investigation and comparison in trial 2. The seven lights were selected based on information available from the manufacturers and included a combination of horticulture LED grow lights and retail LED lights for residential or commercial uses. Nonhorticulture grow lights were selected for comparison because of the wide range of price points, sizes, and features available in the retail market. For lights not designed and marketed for plant growth, detailed spectra are not usually provided, so K ratings and other information were used to infer a range of R:B ratios in the lights. Research by Pennisi et al. (2019) had shown that R:B ratios of ∼3:1 performed well for basil, and Schultz et al. (2021) reported that a similar R:B ratio performed well for bibb lettuce in aquaponics. Similarly, Kong et al. (2019) reported that a higher R:B ratio resulted in a greater production efficiency for lettuce grown indoors, with a target of 4.0 to 4.5. The R:B of the experimental lights screened in trial 1 ranged from 0.51 to 2.93, and the control had an R:B ratio of 6.64 (Table 7).

Table 7. Light characterization for lights in trial 2. Bibb lettuce (Lactuca sativa var. capitata) was grown in aquaponic systems under four different light-emitting diode (LED) lights. Effects on plant performance between each light were compared in a complete block design in four aquaponics systems. The plant beds from each aquaponics system were divided into four sections separated by vertical barriers, creating a total of 16 sections of plant bed. Each plant bed section in each aquaponics system was randomly assigned to one of the four treatments; each treatment was given four replicates. Blue light percentage, green light percentage, red light percentage, ratio of red PPFD (photosynthetic photon flux density) to blue PPFD, photon efficiency, Kelvin rating, and purchase cost in USD per light fixture are presented here.

View Fullscreen

Trial 1 demonstrated that low-cost horticultural and even nonhorticultural LED lights could support lettuce production similar to the expensive control light treatment at similar or even lower levels of energy consumption. Kelvin ratings of the experimental lights varied across the different lights to evaluate the effects of K rating on plant production and to test if this metric would be a good value in deciding which light to choose if a spectrometer was not available. The K rating was not a good indicator for the success of a light, as the K rating fluctuated from 1713 K to 5300 K (Table 7), with plant production values showing no trends when looking at K alone and controlling for all other parameters.

Trial 2 discussion

In trial 2, the SPI ($300) and FLU ($364) treatments achieved similar levels of biomass production (917 and 913 g, respectively) and plant production per unit energy (34 g·m⁻² per kilowatt-hour) as the control NEO ($1400), with a total biomass production of 939 g and plant production per unit energy 31 g·m⁻² per kilowatt-hour. The DES ($100) treatment was the most efficient at converting electrical energy into plant biomass (37 g·m⁻² per kilowatt-hour), better even than the control (NEO), but significantly less biomass was produced, 812 g.

The NEO and DES had the greatest difference in R:B ratios (6.6 and 1.3, respectively) and the greatest differences in total production (939 g and 812 g, respectively). According to previous research at this laboratory, R:B ratio appeared to be a primary determinant of light performance (Oliver et al. 2018); however, when the average individual weights of the plants in the representative samples of the two treatments are compared, they are not different. This indicated that the plants directly under the lights are receiving adequate light intensity and proper wavelengths for efficient photosynthesis, despite the difference in R:B ratios. The higher total biomass production in the NEO can be attributed to better light distribution over the whole bed compared with the DES. This is supported by a lower standard deviation for PPFD readings in trial 1 for NEO (21.3 µmol·m⁻²·s⁻¹) compared with the DES (37.7 µmol·m⁻²·s⁻¹).

In terms of economics, all three of the experimental light treatments provided positive results with regard to NCP and BCR when compared with the control (NEO). Over the course of 1 year, all three experimental lights would yield/save more than $1000 compared with the NEO. In previous research, Nelson and Bugbee (2014) reported that although LED lights yielded good overall plant production and production per unit energy, LEDs were less cost effective than high-pressure sodium lights over 5 years of operation. This was primarily because of the higher initial purchase cost price of the LED fixtures. In the current research, the partial budget analysis showed that the three low-cost LED lights had relatively large economic benefits compared with the control light (NEO), much of which was because of lower initial purchase costs.

In terms of the best-performing light, a single “best” light cannot be separated out. Based on total biomass, production per unit energy, average representative weight, and partial budget analysis, each of the three experimental lights could potentially be a sufficient replacement for the control (NEO). A situation when one experimental light would be “better” than the other would come down to system design and budget of the producers. For example, light used for supplemental lighting in a greenhouse would not necessarily be the best light for a vertically stacked system. The NEO and SPI lights may be better suited for large plant beds, like a University of the Virgin Islands–style system (Rakocy et al. 2006), because of their adjustable intensity output, uniform PPFD spread, and increased wattage. Because they are adjustable, the output can be modified, and their intensity can be adjusted to match the required light with the available natural sunlight. With their adjustable light intensity, these lights can be hung at the desired height and adjusted to fit the needs in that specific environment. The DES and FLU may be better for more compact units, like vertical systems, because of their fixed light intensity output and low wattage usage. Most vertical systems receive no natural sunlight and would not need adjustable light outputs, as the lights could be set to their needed height once and would not need to be adjusted again. The length and width of the lights is such that a singular light can be used to cover the entire plant bed surface.

In terms of commercial application of these results, additional research would be needed in terms of plant bed area covered by each light and the distribution uniformity over the bed. This would require evaluation in commercial scale or near-commercial scale facilities. Differences in economic returns are sufficiently large that using multiple low-cost lights to provide adequate coverage could still be cost effective when compared with one expensive light with a broader light distribution.