Indoor grown cannabis yield increased proportionally with light intensity, but ultraviolet radiation did not affect yield or cannabinoid content

Lighting treatments

Pairs of LED bars (Toplight-Targeted Spectrum; LumiGrow, Emeryville, CA, USA) were spaced 0.4 m apart, on-center, over each plot. One plot in each enclosure had an additional pair of Toplight LED bars, evenly spaced between the first pair of LED bars, to facilitate higher light intensities in this plot. These fixtures were comprised of a combination of blue, phosphor-converted white (5,000 K), and red LEDs. The blue and red LEDs had peak wavelengths [± half-width at half maximum (HWHM)] of 445 ± 8.5 nm and 665 ± 8.0 nm, respectively. When all channels were operated at maximum intensity, the native spectrum (Figure 1A) of the Toplight fixtures had a photon flux ratio of blue (B, 400–500 nm), green (G, 500–600 nm) and red (R, 600–700 nm) of B17:G7:R76. This spectrum was maintained in all cases where dimming was used to reduce intensity. Initial fixture hang heights were 0.9 m above the bench. The UVA spectrum (Figure 1B) was provided by custom-made LED bars (size 0.05 m × 0.6 m; Yunustech, Mississauga, ON, Canada) which had a peak wavelength (± HWHM) of 385 ± 5.5 nm. In one plot in each enclosure, two UVA LED bars were centered 0.24 m apart, at the same height of the LED Toplight fixtures. The UVA + UVB spectrum (Figure 1C) was provided by a broad-band fluorescent lighting technology (SolarSystem UVB, California LightWorks, Canoga Park, CA, USA). A single SolarSystem UVB fixture was centered over one plot in each enclosure, positioned at the same height as the LED Toplight fixtures.

FIGURE 1

Spectrum distributions of: (A) Lumigrow Toplight LED source of photosynthetically active radiation; used in all ultraviolet (UV, 280–400 nm) spectrum and intensity treatments, (B) 385 nm peak UVA LEDs; used in UVA treatment (photon flux density of 50 μmol m^–2 s^–1), and (C) broad-band fluorescent UV; used in the UVA + UVB treatment (photon flux density of 3 μmol m^–2 s^–1).

The experiment was arranged as a randomized complete block design (RCBD) with three LI and two UV-spectrum treatments and three concurrent replications (blocks). In each block (i.e., the enclosures), the three intensity treatment plots targeted canopy-level PPFDs of 600 (control), 800, and 1,000 μmol m^–2 s^–1, respectively. The UV-spectrum treatments had targeted canopy-level PPFDs of 600 μmol m^–2 s^–1 plus UV photon flux densities (PFD) of 50 and 3.0 μmol m^–2 s^–1 for the UVA and UVA + UVB treatments, respectively. In each enclosure (i.e., block), the locations of the five treatment plots were randomly arranged among the six bench positions, with one plot in each block remaining empty (Figure 2). All of the Toplight and UVA fixtures had a photoperiod of 12 h (06:30 to 18:30 HR). The SolarSystem UVB fixtures in UVA + UVB treatment plots had a 5-h photoperiod (13:30 to 18:30 HR) and were only operated during the last 20 days of the trial. The spectra were evaluated using a radiometrically-calibrated spectrometer (XR-Flame-S; Ocean Optics, Dunedin, FL, USA) coupled to a CC3 cosine-corrector attached to a 1.9 m × 400 μm UV-Vis optical fiber. A MS Excel tool developed by Mah et al. (2019) was used to integrate spectral irradiance data into PPFD, UV-PFD, and compute biologically-effective UV-PFD and radiant flux density (in the UV treatment plots). Intensities of the Toplight LEDs were modified using the lighting control software (smartPAR; LumiGrow) to achieve the prescribed intensity and spectra. The UV spectra and intensity levels in the UV treatment plots were characterized with the Toplight LEDs turned off. The UV intensity in the UVA treatment was adjusted using constant current dimmers. The UV intensity in the UVA + UVB treatment was adjusted using aluminum neutral density screen affixed to underside of the SolarSystem UVB fixture. The dominant peak in the UVA + UVB spectrum spanned from ≈275 to 380 nm, with additional narrow-band peaks at 312, 365, 404, 435, 545, and 579 nm (Figure 1C). On a photon-flux basis, the spectrum had a ratio of UVB to UVA of 1.07 (i.e., almost equal photon flux levels of UVB and UVA). Flux density integrals of the UV spectrum treatments, based on both raw flux density data and on conversions according to the Biological Spectral Weighting Function (BSWF; Flint and Caldwell, 2003), are presented in Table 1.

FIGURE 2

Schematic of the layout of treatment plots within each enclosure. The arrangement of the five treatments was randomized within the six plot locations in each enclosure (i.e., block) with the sixth plot in each block remaining empty.

Following spectrum characterization, the canopy-level PPFD of each plant was measured and recorded, twice weekly, using a handheld light sensor (LI-180; LI-COR Biosciences, Lincoln, NE, USA) that was spectrum-matched to the PPFD readings of the XR-Flame-S spectrometer under the PAR spectrum in Figure 1A. Light fixture hang-heights were adjusted accordingly to ensure the targeted LI levels in each plot were maintained as the plants grew.

Plant cultivation

Uniform rooted cuttings of the clonal cannabis genotype ‘Meridian’ were transplanted into rockwool cubes (0.15 m × 0.15 m × 0.15 m; Grodan, Milton, ON, Canada) and grown using the Toplight-Targeted Spectrum (LumiGrow) fixture, as described above. Transplants were grown for 21 days under PPFD of ≈600 μmol m^–2 s^–1 and were maintained in a vegetative state using a 16-h photoperiod (06:30 to 22:30 HR). The transplants selected for the trial were trimmed to uniform height of 33.2 ± 2.62 cm (mean ± SD, n = 90). In each plot, the trimmed transplants were arranged in two rows of three, with plants spaced 45 cm apart, on center (i.e., planting density of ≈10 plants/m²). The PAR photoperiod was set to 12-h in every plot to induce flowering, and plants were grown for an additional 45 days under their respective lighting treatments and then harvested. Throughout the vegetative and flowering stages, plants were drip-irrigated twice daily at 2 L h^–1 for 540 s, such that each plant received ≈0.6 L day^–1. The nutrient solution was comprised of Dutch Nutrients Gro A and Gro B (Homegrown Hydroponics, Toronto, ON, Canada) at a rate of 5 mL L^–1 in rainwater, resulting in an EC of ≈1.75 dS m^–1 and pH of ≈5.6.

Growth and yield parameters

At harvest, the stems were cut at growing medium level and the inflorescences were hand-trimmed from each plant. Selected leaves from the upper canopy of each plant in the control, UVA, and UVA + UVB treatments were photographed individually. The fresh weight of inflorescence (FW_f) and non-floral (FW_nf) aboveground tissues (i.e., stems and leaves) were recorded using a digital scale (AX622N/E Adventure Precision Balance; OHAUS Corporation, Parsippany, NJ, USA). The plants were harvested and processed one at a time and FW of aboveground tissues were taken within 5 min of each plant being cut. The separated aboveground tissues from one randomly selected plant in each plot (i.e., three plants per treatment) were oven-dried to constant weight at 65°C and re-weighed (AX622N/E Adventure Precision Balance; OHAUS Corporation) to determine water content of the respective tissues. After determining there were no treatment effects on tissue water content, the dry weight of inflorescence (DW_f) and non-floral (DW_nf) aboveground tissues were calculated for each plant using the average water content for the respective tissues. Harvest index (HI) for each plant was calculated using the following formula: HI = DW_f/(DW_f + DW_nf). The entire inflorescence tissue from one randomly selected plant in each LI treatment plot were spread out in a single layer on perforated drying trays and air dried at (mean ± SD) 19 ± 1.8°C and 51 ± 9.2 RH% for 5 days (i.e., final moisture content of ≈10–15%). Drying room air temperature and RH were recorded every 300 s using a data logger (HOBO MX1102A, Onset) centered in the drying room. Once dry, the inflorescence material from each plant was homogenized and composite ≈5 g samples were collected and submitted to a 3rd-party lab (RPC; Fredericton, NB, Canada) for analysis of cannabinoid, terpene and water content. The following cannabinoids were tested for using methanol/chloroform solvent extraction and high performance liquid chromatography (HPLC) separation with variable wavelength UV detection (VWD): cannabigerol (CBG), cannabigerolic acid (CBGA), total equivalent cannabigerol (T-CBG), cannabidiol (CBD), cannabidiolic acid (CBDA), total equivalent cannabidiol (T-CBD), Δ⁹-tetrahydrocannabinol (THC), Δ⁹-tetrahydrocannabinolic acid (THCA), total equivalent Δ⁹-tetrahydrocannabinol (T-THC), Δ⁸-tetrahydrocannabinol (D8THC), and cannabinol (CBN). The following terpenes were tested for using hexane/ethanol solvent extraction and gas chromatography (GC) separation with mass spectrometry detection (MSD): alpha pinene, beta pinene, myrcene, limonene, terpinolene, linalool, terpineol, caryophyllene, humulene, 3-carene, cis-ocimene, eucalyptol, trans-ocimene, fenchol, borneol, valencene, cis-nerolidol, trans-nerolidol, guaiol, alpha-bisabolol, and sabinene. In addition, the foliar tissues that were trimmed from the upper inflorescences (commonly called “sugar leaves” in the cannabis production industry) from two randomly selected plants from each control, UVA and, UVA + UVB treatment plot were spread out in a single layer on perforated drying trays and air dried for 5 days. Approximately 1 g of dried sugar leaves from each sample (18 samples total) were submitted to the internal lab (HEXO; Gatineau, QC, Canada) for analysis of content of the following cannabinoids: cannabichromene (CBC), CBD, CBDA, T-CBD, CBG, CBGA, T-CBG, CBN, THC, THCA, T-THC and D8THC. Analysis was done using acetonitrile solvent extraction, ultra performance liquid chromatography (UPLC) separation and VWD were extracted in acetonitrile and analyzed as per Layton and Aubin (2019). All secondary metabolite concentrations are expressed as mg g^–1 of dry biomass, with tissue water contents determined by drying to constant weight.

Increasing light intensity proportionally increases yield

Many (interrelated) environmental parameters can be optimized in order to maximize yields, including temperature, humidity, CO₂ concentration, and fertility. However, in indoor cultivation environments, LI is one of the most prominent and expensive input parameters under the complete control of the cultivator (Mills, 2012). The optimum LI in a given production scenario will depend on many economic factors, but the responses of modern cannabis genotypes’ yield and secondary metabolite composition to LI are key input factors that can only be elucidated experimentally. Compared with the 600 μmol m^–2 s^–1 treatment, increasing the PAR exposure by 1.6 times (i.e., 1,000 μmol m^–2 s^–1) increased inflorescence dry yield by the same magnitude. This implies that, for yield responses in this trial, the cannabis plants growing under 1,000 μmol m^–2 s^–1 were still on the linear phase of the light response curve (i.e., still operating at maximum quantum efficiency). This is also supported by the linear yield responses to increasing LI up to 1,800 μmol m^–2 s^–1 reported in Rodriguez-Morrison et al. (2021a). A linear yield response to a range of LIs that exceeds normal production levels (Potter and Duncombe, 2012), confers a relatively reliable and easily interpreted basic model for how cannabis yield responds to changes in LI. For example, a simple regression of the mean DW_f at the three tested PPFD levels (i.e., for calculating slope) in the present study predicts that every additional 100 μmol m^–2 s^–1 of daily PAR will increase yield by 4.6 g/plant (i.e., 51 g m^–2 at the present study’s planting density).

Although not statistically significant in this trial, the harvest index (HI, i.e., proportion of marketable aboveground biomass) also rose with increasing LI, similar to Rodriguez-Morrison et al. (2021a). This trend has also been observed in other species, but the rate of increase in cannabis HI was approximately 4-fold higher than in indoor-grown wheat (Bugbee and Salisbury, 1988) over a similar LI range. This serves as further evidence of cannabis’ enormous phenotypic plasticity in response to LI. Higher HI could enhance harvest efficiency by reducing the non-marketable proportion of the biomass, all of which needs to be removed at harvest and disposed (EMCDDA, 2012). Further, since inflorescence tissues have substantially higher cannabinoid contents than other aboveground tissues (Richins et al., 2018), plants that produce proportionally higher inflorescence biomass under higher LI may also increase overall cannabinoid yield. While this was not evaluated in the present study, it is also likely that the increased HI was associated with larger inflorescences and increased floral density, such as reported in Rodriguez-Morrison et al. (2021a). These attributes are generally highly valued by the industry, particularly when the crop production is targeted toward the dry inflorescence market which presently accounts for over 60% of total cannabis product sales in Canada (Health Canada, 2021).

Potential costs and benefits of increasing light levels in commercial production

The slope of the yield response curve relates to a crop’s phenotypic plasticity to respond to changes in environmental inputs – light intensity in this case – which will of course vary by genotype and production environment (Backer et al., 2019; Zheng and Llewellyn, 2022). Nevertheless, using the present study’s genotype as a proxy, one can estimate the payoff associated with increases in LI. For example, increasing the PPFD by 100 μmol m^–2 s^–1 increases the total light integral over 45 days by 195 mol m^–2. The estimated increase in yield of 51 g m^–2 for 195 mol m^–2 of additional lighting corresponds to a light use efficiency of 0.26 g mol^–1. The energy cost for additional lighting relies heavily on fixture efficacy, light distribution, and local cost of electricity. Most modern horticultural LED fixtures have efficacy values exceeding 2.5 μmol J^–1 (i.e., 9 mol kWh^–1) (Design Lights Consortium, 2022). If electricity cost was 0.10 $CAD/kWh, the estimated energy cost in this scenario would be approximately 0.042 $CAD/g or an energy use efficiency of approximately 2.4 g kWh^–1 (for lighting). The increases in fixture efficacy in modern horticultural LEDs (Kusuma et al., 2020) may explain two-times higher estimated energy use efficiency in the present study vs. estimates from scientific studies in past decades (EMCDDA, 2012). At the current wholesale price for dried inflorescence of 4.00 $CAD/g (Cannabis Benchmarks, 2022), the added electricity cost of increasing yield by increasing LI comprises only ≈1% of the total price, and therefore may make economic sense. However, the costs of additional lighting infrastructure and ancillary costs such as heat management and higher crop production inputs must also be considered when assessing the potential profitability associated with any lighting strategy.

Light intensity did not substantively affect chemical composition

The lack of LI effects on inflorescence cannabinoid content was consistent with other studies (Vanhove et al., 2011; Potter and Duncombe, 2012; Rodriguez-Morrison et al., 2021a). The lack of LI effects on inflorescence terpene content was also consistent with Rodriguez-Morrison et al. (2021a). Overall, both cannabinoid and terpene yield (i.e., g/plant or g m^–2) increased concurrently with increasing inflorescence DW, which may be important for processing cannabis extracts. While many factors are involved in evaluating profitability of adopting a specific production practice, raising canopy-level LI may be an economically feasible way to increase inflorescence and secondary metabolite yield – but not concentration – in indoor cannabis production.

Secondary metabolite content was unaffected by ultraviolet radiation

Ultraviolet radiation radiation can invoke both eustress and distress responses in plants, depending on many intrinsic (e.g., genotype and ontological stage) and extrinsic factors (e.g., UV intensity, duration, and spectrum, PAR intensity). The major modes of action of UV radiation on plants are through photoreceptor-mediated responses (e.g., UVR8) and the production of reactive oxygen species (ROS), which can cause cellular damage and influence gene expression (Jansen et al., 1998; Hideg et al., 2013). One common response to UV exposure is the production of photoprotective compounds, particularly in epidermal regions, to diminish UV penetration deeper into plant tissues (Frohnmeyer and Staiger, 2003; Huché-Thélier et al., 2016). In indoor cannabis production, the primary goal of exposing flowering plants to UV radiation is to upregulate the production of cannabinoids which are predominantly synthesized in the copious glandular trichomes that are found on and around inflorescence tissues (Potter, 2014) and which have UV-photoprotective properties (Hazekamp et al., 2005). Since this trial investigated a THC-dominant cultivar, it would be expected that any UV-induced upregulation of cannabinoid synthesis would result in substantially higher THC concentrations in the inflorescence and surrounding foliar (i.e., sugar leaves) tissues.

Since there were no yield reductions, there was little evidence of spectrum-induced distress in either of the UV treatments. While we did not quantify trichomes, we noted that there appeared to be higher trichome density on the sugar leaves in the UVA + UBV vs. the UVA and control treatments, particularly in areas proximate to the petioles. This observed increase in trichome density may explain the ≈30% higher foliar THC content and trends (not statistically significant) of ≈10% higher THCA and T-THC content in the UVA + UBV treatment vs. control. However, since foliar cannabinoid content was much lower than inflorescence tissues due to lower trichome density (Small, 2017), these tissues are of relatively low value in commercial indoor cannabis production and are often discarded (Potter, 2014). Therefore, from a production perspective, the minor increases in foliar cannabinoid concentration under UV exposure were probably not commercially relevant. Since none of the inflorescence cannabinoid levels were affected by the UV treatments, the eustress levels of UV radiation in this study did not have substantial effects on the secondary metabolite composition of the cannabis genotype used in this investigation. It is possible that the relatively low cannabinoid content of genotypes used in prior studies (e.g., Pate, 1983; Lydon et al., 1987) had conferred a relatively greater potential for stress-induced cannabinoid upregulation than in modern genotypes with much higher cannabinoid content (Dujourdy and Besacier, 2017). Further, within a given genotype’s overall genetic potential for producing various cannabinoids, there may be a higher likelihood for a plant to upregulate the production of one metabolite over another if both are normally present at relatively high concentrations. Therefore, in contrast with the present study, a genotype with a characteristically more balanced ratio of THC to CBD (i.e., type II) may show higher plasticity toward modifying the cannabinoid metabolome under UV eustress conditions. This may be particularly relevant to THC and CBD since they have the same biochemical precursor. However, this was not found to be the case in the two genotypes investigated in Rodriguez-Morrison et al. (2021b), however, their UV exposure protocols had only detrimental impacts on plant health, yield, and secondary metabolite composition. The lack of UV treatment effects in the type I chemotype grown in the present study may also indicate that this cultivar was already functioning near its maximum genetic capacity for the production of THC, thus UV exposure had no promotion effects on cannabinoid composition.

Unlike the PPFD parameter in assimilation lighting, where all photons within the 400–700 nm waveband are given equal weight, plant responses to UV can be very dynamic across the UVB and UVA wavebands. According to Flint and Caldwell (2003), the efficacy of the UV spectrum for inducing plant responses decreases by more than a factor of 200 as wavelength increases from 280 to 400 nm; with almost half of this reduction occurring in the narrow 300–310 nm range. Therefore, the spectrum distribution of a UV treatment is an extremely important descriptive parameter for any UV treatment protocol. For example, the raw flux density in the UVA treatment was ≈15 times higher but the BSWF-converted daily UV dose was ≈10% lower than in the UVA + UVB treatment. In prior studies, UV exposure levels were often described according to their daily “dose”, normally expressed in units of radiant flux density (e.g., kJ m^–2), however, incomplete spectrum distribution information makes it difficult to compare UV treatment levels between studies (e.g., Fairbairn and Liebmann, 1974). We are aware of only two prior cannabis studies that provided quantitative data on UV exposure levels (summarized in Table 1). No raw flux density or spectral data was provided in Lydon et al. (1987) but their reported maximum BSWF-adjusted daily UV doses were approximately 30% higher than both treatments in the present study and are comparable to the maximum UV dose in Rodriguez-Morrison et al. (2021b). Due to the different UV exposure periods, the total BSWF-adjusted UV doses were lowest in the present study, moderate in Lydon et al. (1987) and highest in Rodriguez-Morrison et al. (2021b).

We thank HEXO Corporation for providing the research venue, experimental apparatus (i.e., enclosures), plant materials, crop inputs (e.g., water, fertilizer, electricity for lighting) and logistical support. Special thanks to Allison Slater for her technical support. California LightWorks donated the luminaires for the UVA + UVB treatment plots.