LEDs and control system.

Photosynthetic light was provided using a custom-built 400-W LED array consisting of four 100-W warm-white LED modules (3000 K; EpiLEDs, Tainan, Taiwan) mounted on aluminum heat sinks with air circulation provided by two 120-mm cooling fans. Glass lenses (Satistronix Group, Shenzhen, China) were attached below the LED modules to focus the light on the plant production area below the LEDs. The lights were powered using a 30-V DC power supply (Fig. 1). The LED array lighted an area of ≈0.75 m².

View Figure

• Download as Powerpoint
• Download Figure

A custom-made control board was used to alter the duty cycle of the LEDs (Fig. 1). The LEDs were continuously turned on and off (at ≈1000 Hz in this study; i.e., each on/off cycle lasted 1 ms) and duty cycle is the proportion of time that the diodes are energized during each cycle. Thus, the intensity of the LED light was modified by changing the duty cycle. The control board altered the duty cycle of the LEDs proportionally in response to an external 0- to 2-V DC signal supplied to the board. To provide this voltage signal, an analog output module (SDM-A04; Campbell Scientific, Logan, UT) was connected to a datalogger (CR1000; Campbell Scientific).

Chlorophyll fluorescence measurements.

Chlorophyll fluorescence was measured with a pulse-amplitude modulated fluorometer (MINI-PAM; Heinz Walz, Effeltrich, Germany). The fluorometer provided an analog voltage signal to the datalogger, proportional to the measured fluorescence. Following measurements of steady-state fluorescence (F₀ and F_s in the dark and light, respectively), the datalogger sent a serial signal to the fluorometer to trigger a saturating light pulse and then measured the maximal fluorescence yield (F_m and F_m′, in the dark and light, respectively). The intensity of the saturating light pulse was not monitored, but adjusted as needed to assure that the fluorescence signal truly saturated during F_m and F_m′ measurements. Chlorophyll fluorescence parameters were calculated by the datalogger using the equations developed by Genty et al. (1989) and Bilger and Björkman (1990). The datalogger also collected and stored all data.

Pulse-amplitude modulated chlorophyll fluorometers measure the difference in chlorophyll fluorescence coming from a leaf exposed to actinic light vs. fluorescence induced by actinic light plus a modulated, low-intensity measuring light. These measurements, therefore, depend on a steady-state actinic light. The pulsing nature of the actinic LED lights resulted in high variability in the fluorescence data. To reduce this variability, F_s was estimated as the average of 1000 measurements collected at 20-ms intervals. Note that the variability in F_s can likely be reduced if the light output from the LEDs were altered by controlling the current going to the LEDs, rather than by changing the duty cycle. We choose to use duty cycle control, because of its ease of implementation.

Maximal fluorescence yield was taken as the highest fluorescence value of 50 measurements collected at 20-ms intervals during a 0.8- to 1.0-s saturating pulse. To avoid possible interference from pulsed light with the measurement of F_m′, the LEDs were turned off while F_m′ was measured. Quantum yield of PSII was calculated as Φ_PSII = (F_m′ − F_s)/F_m′ (Genty et al., 1989).

At the onset of an experiment, plants were dark adapted overnight, after which dark-adapted chlorophyll fluorescence measurements were collected. A single measurement was used to determine F_o and maximal chlorophyll fluorescence (F_m) was taken as the highest of 50 measurements collected at 20-ms intervals during a 0.8- to 1.0-s saturating pulse. The dark-adapted quantum efficiency of photochemical energy conversion by PSII (F_v/F_m) was calculated as (F_m − F_o)/F_m (Genty et al., 1989).

The combination of maximum fluorescence measurements on dark-adapted leaves and under actinic light also allows for the calculation of the apparent rate constant of NPQ [(F_m − F_m′)/F_m′; Bilger and Björkman, 1990]. This provides an estimate of the magnitude of thermal dissipation based on the assumption that NPQ is due solely to thermal energy dissipation (Adams et al., 1999; Baker and Rosenqvist, 2004; Bilger and Björkman, 1990; Demmig-Adams, 1990, 2012).

To estimate the rate of photosynthetic electron transport through PSII, a quantum sensor (LI-190; LI-COR BioSciences, Lincoln, NE) connected to the datalogger was used to measure PPFD. This sensor was mounted adjacent to the spot where chlorophyll fluorescence was measured. To estimate ETR, we assumed that photons were equally partitioned between PSI and PSII, and that 84% of incident PPFD was absorbed by the leaf (Baker, 2008). Thus, ETR is calculated as Φ_PSII × PPFD × 0.5 × 0.84. The assumption that the leaves absorb 84% of incident light is based on data from a wide range of species (Baker, 2008; Björkman and Demmig, 1987). Since plants can alter the ratio of photons absorbed by PSI and PSII through state transitions of the PSII light-harvesting complex (Allen, 2003; Rochaix, 2014), the assumption that electrons are evenly partitioned between the photosystems may not always be correct. However, state transitions mainly occur at very low light levels (Ruban, 2015) and were likely unimportant in this study. Regardless of their potential shortcomings, calculated ETR values provide a useful way of determining relative changes in ETR (Maxwell and Johnson, 2000).

Biofeedback control.

The duty cycle of the LEDs was controlled by the datalogger via the control board based on a specified target ETR. Duty cycle, and thus PPFD, was adjusted based on the deviation of the observed ETR from ETR_T as follows: new duty cycle = (ETR_T/ETR) × old duty cycle. Thus, the biofeedback system compensated for low ETR by increasing PPFD and decreased PPFD whenever ETR exceeded ETR_T. An example program of the biofeedback system is available on-line (University of Georgia, 2015).

Plant material.

Lettuce ‘Green Towers’ (for studies with 16 h constant ETR_T) and ‘Green Ice’ (for stepwise changes in ETR_T) was grown from seed in 10-cm containers and used for trials 4 to 6 weeks after germination. These cultivars were used because their leaf morphology makes it easy to use the fluorometer leaf clip. Pothos and sweetpotato ‘Marguerite’ were propagated from stem cuttings and grown in 15-cm containers. Experiments were conducted 6 to 8 weeks after rooting. Greenhouse temperature averaged 22 °C and the photoperiod was ≈11 h. The daily light integral inside the greenhouse in the 2 weeks before data collection was (mean ± sd) 8.4 ± 4.0, 9.8 ± 3.3, and 9.3 ± 3.5 mol·m⁻²·d⁻¹ for lettuce, sweetpotato, and pothos, respectively. These three species were selected because of their different light requirements. Pothos is a tropical understory species, well adapted to low light environments, lettuce grows well at intermediate light levels, whereas sweetpotato performs best under high light. All plants were grown in a peat-based soilless substrate (Fafard 2P; Sun Gro Horticulture, Agawam, MA) in a greenhouse on ebb-and-flow benches, and irrigated daily with a 15N–2.2P–12.5K water-soluble fertilizer (15-5-15 Cal-Mag, Everris, Marysville, OH) solution containing 100 mg·L⁻¹ N.

Experimental setup.

The LED array was mounted in a growth chamber (E15; Conviron, Winnipeg, MB, Canada), which was set to maintain a constant temperature of 25 °C. During each experiment, a single plant was placed beneath the LED array, the fluorometer leaf clip was attached to the uppermost fully expanded leaf, and the quantum sensor was positioned next to the leaf clip. Sensors were positioned ≈55 cm below the LED light. Two types of trials were conducted: ETR_T was maintained at a constant value (70 or 100 µmol·m⁻²·s⁻¹) for 16 h or ETR_T was increased from 0 to a maximum ETR_T in seven steps, and then decreased again, maintaining each ETR_T for 1 h. All trials were conducted under ambient CO₂ concentrations.

Trials with a constant ETR_T were conducted with lettuce ‘Green Towers’, after the plants had acclimated to growth chamber conditions (16-h photoperiod, 25 °C in the light and 20 °C in the dark, PPFD ≈240 µmol·m⁻²·s⁻¹) for over 2 weeks. During these trials, F_m′ was measured every 15 min, whereas F_s was measured every 5 min. Because changes in F_m′ were slow, except for immediately after turning on the LED lights, Φ_PSII and ETR were calculated assuming that F_m′ did not change appreciably in a 15-min period, thereby reducing the frequency of saturating light pulses. Dark-adapted chlorophyll fluorescence measurements were taken hourly for 4 h before and after the 16-h periods with constant ETR_T.

For stepwise ETR_T trials, the maximum ETR_T for each species (pothos, lettuce, and sweetpotato) was chosen based on preliminary studies. Plants were moved from the greenhouse into the growth camber the evening before the start of the measurements and dark adapted in the growth chamber overnight. For lettuce, ETR_T was increased from 0 to 70 µmol·m⁻²·s⁻¹ in 10-µmol·m⁻²·s⁻¹ steps, then decreased at the same rate. For sweetpotato, ETR_T was increased from 0 to 98 µmol·m⁻²·s⁻¹ in 14-µmol·m⁻²·s⁻¹ steps and for pothos from 0 to 49 µmol·m⁻²·s⁻¹ in 7-µmol·m⁻²·s⁻¹ steps. Each experiment was repeated at least three times with each species. Data presented are from representative runs. Lettuce trials were run using a power supply of ≈240 W, which resulted in a maximum PPFD of ≈560 µmol·m⁻²·s⁻¹. A 400-W power supply was used for the pothos and sweetpotato trials, resulting in maximum PPFD of ≈940 µmol·m⁻²·s⁻¹.

Because each ETR_T was maintained for only 60 min, F_s and F_m′ were both measured every 2 min. More frequent measurements were needed than in the trials with constant ETR_T, because of the relative large changes in duty cycle and PPFD needed to reach the various ETR_T levels, and to assure we collected enough data during each 1-h period to determine whether ETR was stable. Changes in plant physiological responses over time and relationships among the different physiological variables were analyzed using regression analysis with P < 0.05 considered to be statistically significant.

Maintaining a stable ETR.

Experiments were conducted during which the biofeedback system was programmed to maintain the ETR of lettuce at 70 or 100 µmol·m⁻²·s⁻¹ for 16 h. Afterward, the LEDs were turned off and fluorescence data were collected hourly for an additional 4 h in the dark.

During the trial in which ETR_T was 70 µmol·m⁻²·s⁻¹, F_v/F_m was ≈0.82 in the 4 h before the lights were turned on (Fig. 2), normal for healthy leaves (Björkman and Demmig, 1987). The first F_s and F_m′ data were collected 5 min after the LED lights were turned on (at 232 µmol·m⁻²·s⁻¹) and the observed Φ_PSII was relatively low (0.516). This low Φ_PSII shortly after the start of the light period was likely due to the reduction of electron receptors in the electron transport pathway. This results in a closure of PSII reaction centers since the primary PSII electron acceptors are unable to transfer absorbed electrons to the next carrier in the electron transport chain. These reaction centers re-open after light-induced activation of Calvin cycle enzymes allows the reduced nicotinamide adenine dinucleotide phosphate and adenosine triphosphate produced by the light reactions to be used for carbon assimilation (Maxwell and Johnson, 2000). Due to the low Φ_PSII, the initial ETR was 41.5 µmol·m⁻²·s⁻¹, well below 70 µmol·m⁻²·s⁻¹. Thus, PPFD was increased by the biofeedback system. After the second F_m′ measurement, Φ_PSII increased to 0.643 and ETR was 91.6 µmol·m⁻²·s⁻¹. The biofeedback system then downregulated PPFD and was able to maintain a stable ETR (70.0 ± 0.8 µmol·m⁻²·s⁻¹, mean ± sd) during the remainder of the 16-h period (Fig. 2). Because Φ_PSII changed little during this 16-h period (0.687 ± 0.006), the biofeedback system did not need to make substantial changes in PPFD (243 ± 2 µmol·m⁻²·s⁻¹) to maintain ETR close to 70 µmol·m⁻²·s⁻¹. NPQ during this period remained steady at 0.383 ± 0.007. Following the 16 h light period F_v/F_m was ≈0.83, indicating that the fluorescence measurements did not cause damage to PSII (Fig. 2). We subsequently have seen reductions in F_v/F_m of lettuce after 16 h of more frequent F_m′ measurements (every 5 min or less), indicating that these measurements can induce photoinhinbition. This appears to be the result of the frequent application of saturating light pulses needed for the measurement of F_m′.

View Figure

• Download as Powerpoint
• Download Figure

With an ETR_T of 100 µmol·m⁻²·s⁻¹, the initial Φ_PSII was low and NPQ was high (Fig. 2), similar to the response observed at an ETR_T of 70 µmol·m⁻²·s⁻¹. Once again, the low Φ_PSII resulted in a low ETR and a high initial PPFD. However, after the second F_m′ measurement, the system was able to adjust the PPFD and maintain the ETR close to 100 µmol·m⁻²·s⁻¹ (100.0 ± 1.7 µmol·m⁻²·s⁻¹) for the remainder of the 16-h period. To maintain this ETR, the biofeedback system had to gradually, and linearly, increase the PPFD, from 389 to 409 µmol·m⁻²·s⁻¹ (r = 0.68, P < 0.0001). This increase in PPFD was needed, because Φ_PSII decreased gradually over time, from 0.612 to 0.582 (r = −0.72, P < 0.0001), whereas NPQ increased (r = 0.90, P < 0.0001). The decrease in Φ_PSII was negatively correlated with the increase in NPQ (r = −0.73, P < 0.0001). This increase in NPQ was likely due to the upregulation of the xanthophyll cycle: lumen acidification triggers the de-epoxidation of violaxanthin to form antheraxanthin and zeaxanthin, which leads to the dissipation of excess absorbed light energy as heat (Demmig-Adams et al., 2012; Horton, 2012; Rochaix, 2014). In the dark period following the light cycle, F_v/F_m was ≈0.815. This high F_v/F_m indicates that there was no damage to PSII (Björkman and Demmig, 1987).

Controlling ETR in a stepwise pattern.

A series of experiments was conducted to test the ability of the biofeedback system to effectively alter and control ETR over time. In these studies, the system was programmed to change ETR_T once every hour. The ETR_T was gradually increased and then decreased in a stepwise pattern over the course of 15 h. Lettuce and sweetpotato showed similar responses in these studies. Control of ETR was good when ETR_T was low, but the maximum ETR that could be maintained was ≈60 and 84 µmol·m⁻²·s⁻¹ for lettuce and sweetpotato, respectively (Fig. 3). This lower maximum ETR for lettuce was partly due to the lower maximum PPFD levels during the lettuce trials (maximum PPFD of ≈560 µmol·m⁻²·s⁻¹) than during the sweetpotato trials (maximum PPFD of ≈940 µmol·m⁻²·s⁻¹) This difference in maximum PPFD resulted from the use of different power supplies in the lettuce and sweetpotato studies. The ETR of pothos was controlled over the entire range tested, but ETR became increasingly more variable at higher ETR_T levels. To a lesser extent, this was true for lettuce and sweetpotato as well, and this response is similar to that seen when the ETR_T of lettuce was maintained steady for 16 h; ETR was more variable with an ETR_T of 100 than with an ETR_T of 70 µmol·m⁻²·s⁻¹ (Fig. 2).

View Figure

• Download as Powerpoint
• Download Figure

As expected, PPFD had to be gradually increased to achieve higher ETR levels. As ETR_T was gradually increased, PPFD even needed to be increased during the 1-h periods in which ETR_T was unchanged to maintain that steady ETR level. For example, to maintain a lettuce ETR of 30 µmol·m⁻²·s⁻¹, PPFD needed to be increased from 135 to 155 µmol·m⁻²·s⁻¹. This response was most notable in sweetpotato and lettuce and was due to a decrease in Φ_PSII during this time (Fig. 4). Conversely, as ETR_T was gradually decreased (8–15 hours), no large changes in PPFD were needed to maintain a stable ETR. All three plant species required a higher PPFD to achieve the same ETR_T during the latter half of these studies, when ETR_T was gradually decreased, than during the first half when ETR_T was increased (Fig. 5). For example, during the period of increasing ETR_T with lettuce, the average PPFD required to maintain an ETR of 30 µmol·m⁻²·s⁻¹ was 145 µmol·m⁻²·s⁻¹, as compared with 265 µmol·m⁻²·s⁻¹ during the period that ETR_T was decreased (Figs. 3 and 5). This is a direct consequence of the higher Φ_PSII during the period of increasing ETR_T than during decreasing ETR_T.

View Figure

• Download as Powerpoint
• Download Figure

View Figure

• Download as Powerpoint
• Download Figure

There was a large decrease in Φ_PSII during the initial 8 h of these trials, as ETR_T and PPFD were increased. This was most pronounced in pothos and sweetpotato, possibly because these plants were exposed to higher light levels than lettuce. In all three species, this decrease in Φ_PSII was associated with an increase in NPQ (Fig. 4), a typical response to increasing PPFD (e.g., Demmig-Adams et al., 1996). There was also a strong relationship between decreasing Φ_PSII and increasing NPQ during the period that ETR_T was increased for all species (Fig. 6), as was recently also found for Arabidopsis thaliana (Ware et al., 2015). The relationship between NPQ and Φ_PSII of both lettuce and pothos differed during the period of increasing ETR_T, as compared with the period of decreasing ETR_T (Fig. 6). As ETR_T was decreased, NPQ of lettuce and pothos also decreased rapidly. However, with sweetpotato, the pattern of the change in NPQ differed, and NPQ increased throughout the first 11 h of the trial. This different response of sweetpotato is consistent with the notion that the regulation of short- and long-term changes of Φ_PSII and xanthophyll cycle activity is species specific (Demmig-Adams et al., 2012).

View Figure

• Download as Powerpoint
• Download Figure

The decrease in NPQ of lettuce and pothos during the latter half of these trials resulted in only small increases in Φ_PSII (Figs. 4 and 6). The same level of NPQ during the period of decreasing ETR_T resulted in lower Φ_PSII than during the initial 8 h of increasing ETR_T. This effect was less pronounced in sweetpotato, which showed a slower downregulation of NPQ as ETR_T was decreased. The relationship between NPQ and Φ_PSII of sweetpotato was similar during the periods of increasing and decreasing ETR_T.

There was a slow and gradual decline in F_v/F_m during the initial 1 h of darkness, regardless of species (Fig. 3). This indicates that measuring chlorophyll fluorescence every 2 min had a significant impact on the physiology of the measured leaves. This was also reflected in a slow, gradual increase in NPQ of lettuce and pothos during this period. When the lights were turned off again after 14 h, F_v/F_m did not recover to the initial F_v/F_m, and was well below the 0.82 that is considered typical for healthy leaves (Björkman and Demmig, 1987). This lasting reduction in F_v/F_m was not due to NPQ, which did decrease during the final dark period to levels close to those at the start of these experiments (Fig. 4).

The low F_v/F_m at the end of these trials is consistent with the poor recovery of Φ_PSII as ETR_T and PPFD were gradually decreased during the latter part of these studies. Long-lasting depressions in Φ_PSII can be due to elevated levels of xanthophylls cycle pigments; prolonged light stress can promote the accumulation and retention of zeaxanthin (Demmig-Adams and Adams, 2006; Demmig-Adams et al., 2012; Horton, 2012; Ruban, 2015). An accumulation of zeaxanthin should be evident from prolonged high NPQ levels (i.e., sweetpotato), which was not seen in lettuce or pothos. The different relationship between NPQ and Φ_PSII of lettuce and pothos during the periods of increasing and decreasing ETR_T indicates that the low Φ_PSII during the latter half of these studies was not caused by high NPQ. Instead, the low Φ_PSII of lettuce and pothos during the period of decreasing ETR_T was likely due to photoinhibition. High light levels can induce damage to the reaction centers, making them nonfunctional and decrease Φ_PSII and thus ETR (Ruban, 2015). Photosystem II is particularly sensitive to photoinhibition, which typically results in degradation of the D1 protein, an integral core PSII reaction center protein (Horton, 2012; Horton and Ruban, 2005; Rochaix, 2014; Ruban, 2015). Since the repair of D1 proteins can take several hours (Ohad et al., 1984), photoinhibition can reduce Φ_PSII and ETR for prolonged periods (Demmig-Adams et al., 2012; Horton, 2012; Ruban, 2015). A decrease in Φ_PSII as a result of upregulation of NPQ is easily distinguished from photoinhibition using chlorophyll fluorescence measurements, since upregulation of NPQ is observed as a decrease in F_m′, but a decline in Φ_PSII can be due to a decrease in F_m′ or an increase in F_s. The low Φ_PSII observed in the trials with pothos and lettuce was associated with a prolonged high level of F_s even while PPFD was decreasing (results not shown), which indicates that photoinhibition was the cause. This is consistent with the findings by Ruban (2015), who showed that photoinhibition can be determined from F_s measurements. Recently, we have determined that photoinhibition in lettuce can result from frequent F_m′ measurements: measuring F_m′ at intervals of 5 min or shorter induces an increase in F_s, whereas measurement intervals of 15 min or longer do not (unpublished results). Apparently, exposing leaves to the very high PPFD of a saturating light pulse too frequently induces photoinhibition, even if the saturating light is applied for as little as 1 s every 5 min. F_m′ generally does not change rapidly under conditions with relative stable PPFD, as is evident from the slow changes in NPQ when ETR is maintained at a stable level for prolonged periods (Fig. 2). Under such conditions, it is possible to measure F_s more frequently and to calculate Φ_PSII using less frequent F_m′ measurements.

Prospects for optimizing lighting.

The goal of the reported work was to develop a biofeedback system that can monitor Φ_PSII and calculate ETR, and then use that information to adjust the PPFD to maintain a specific ETR. Future work will focus on ways to use this system to optimize crop production. This work will need to include both a plant physiology and an economic component. In principle, the biofeedback system can be used not only to control ETR, but also to maintain a specific Φ_PSII or, perhaps, NPQ. To achieve optimal efficiency, it may be necessary to develop algorithms that take into account ETR, Φ_PSII, and NPQ. From an economic perspective, it may be necessary to take the cost of electricity into account as well, since the economically optimal light level will be higher when electricity prices are lower. In the case of variable electricity prices, growers may be able to aim for higher ETRs during times that the cost of electricity is low. To automate such an approach, real-time electricity prices would need to be incorporated into the control algorithm. This approach is not limited to vertical farms, but can be used in greenhouses as well, applying supplemental light only as needed. This may be especially beneficial under conditions with fluctuating PPFD from sun light (e.g., in greenhouses), which results in fluctuating Φ_PSII (Janka et al., 2015) and ETR. The biofeedback system can automatically adjust the supplemental light levels to assure that ETR is maintained at or above a specific minimum threshold. Alternatively, the supplemental light could be provided only when Φ_PSII is relatively high and light can be used efficiently.