The objectives of the photosynthesis and leaf morphology investigations in this study were 2-fold: (1) to address the knowledge gap in the relationships between localized cannabis leaf photosynthesis and yield and (2) observe and report changes in physiology as the plant progresses through the flowering ontogeny.

General morphological, physiological, and yield responses of plants are well-documented across LI gradients ranging from below the compensation point to DLIs beyond 60 mol·m⁻²·d⁻¹. Recently, the LI responses of myriad plant attributes were compiled across a tremendous range species, ecotypes and growing environments, and concisely reported them in the excellent review paper by Poorter et al. (2019). The trends in their LI models align well with primary attributes measured in the present study, including morphological parameters such as plant height and internode length (data not shown), SLW (discussed below), and physiological parameters such as F_v/F_m, LNCER (i.e., photosynthesis at growth light; Phot/A^GL), and A_sat (i.e., photosynthesis at saturating light; Phot/A^SL). In general, cannabis photosynthesis, and yield responses to localized LI were linear across the APPFD range of 120–1,800 μmol·m⁻²·s⁻¹. While these results are in agreement with the contemporary literature on cannabis (Chandra et al., 2008, 2015; Potter and Duncombe, 2012; Bauerle et al., 2020; Eaves et al., 2020), we also showed substantial chronological dependencies on leaf photosynthetic indices.

By surveying the photosynthetic parameters of the upper cannabis canopy across a broad range of LPPFDs and over multiple timepoints during the generative phase, we saw evidence of both acclimation and early senescence as the crop progressed through its ontogeny. At the beginning of the trial, the plants were abruptly transitioned from a uniform PPFD (425 μmol·m⁻²·s⁻¹) and 18-h photoperiod (i.e., 27.5 mol·m⁻²·d⁻¹) and subjected to a much shorter photoperiod (12-h) and an enormous range of LI (120–1,800 μmol·m⁻²·s⁻¹), resulting in DLIs ranging from 5.2 to 78 mol·m⁻²·d⁻¹. Further, on a DLI-basis, ≈1/3 of the plants were exposed to lower LIs in the flowering vs. vegetative phase (i.e., APPFD <640 μmol·m⁻²·s⁻¹). These sudden transitions in both LI and photoperiod resulted in substantive changes in the plants' lighting environment at the start of the trial, stimulating various morphological and physiological adaptations with differing degrees of plasticity. The leaves measured in week 1 developed and expanded during the prior vegetative phase under a different lighting regimen (LI and photoperiod). The leaves measured in week 5 were developed under their respective LPPFDs during a period characterized by slowing vegetative growth and transitioning to flower development. The leaves measured in week 9 would have also developed under their respective LPPFDs, but since cannabis vegetative growth greatly diminishes after the first 5 weeks in 12-h days (Potter, 2014), these tissues were physiologically much older than the leaves measured in week 5, with concomitant reductions in photosynthetic capacity (Bielczynski et al., 2017; Bauerle et al., 2020).

These differences in leaf physiological age, plant ontogeny, and localized lighting environments during leaf expansion vs. measurement resulted in notable temporal variability in leaf-level LI responses (Figure 4). In week 1, there were no LI treatment effects on A_sat and the slopes of the LSP, LNCER, and F_v/F_m were shallower than in weeks 5 and 9. The comparatively lower LI responses in week 1 were likely due to the reduced adaptive plasticity that mature foliar tissues have vs. leaves that developed under a new lighting regime (Sims and Pearcy, 1992). Further, Y-intercepts for the A_sat, LSP, and LNCER models were higher in week 1 than weeks 5 and 9, which may be partly due to the higher LI (amplified by the longer photoperiod) that the leaves developed under, during the latter part of the vegetative phase. Moreover, the A_sat, LSP, and LNCER models in weeks 5 and 9 have comparable slopes, but there is a vertical translation in the respective models, resulting week 9 models having substantially lower Y-intercepts (i.e., approximately half) for these parameters. The interplay of physiological age of foliage and plant ontogeny (i.e., onset of senescence) on the diminished photosynthetic capacity of the leaves in week 9 is unknown, but the dynamic temporal nature of cannabis photosynthesis (during flowering) is manifest in these models.

Given these impacts of physiological age and light history, we posit that cannabis leaf photosynthesis cannot be used as a stand-alone gauge for predicting yield. Chandra et al. (2008) and Chandra et al. (2015) provided insight into the substantial capacity for drug-type strains of indoor grown cannabis leaves to respond to LI; and the results of these trials are much lauded in the industry as evidence that maximum photosynthesis and yields will be reached under canopy-level PPFDs of ≈1,500 μmol·m⁻²·s⁻¹. However, their 400–500 μmol·m⁻²·s⁻¹ increments in LPPFD does not provide sufficient granularity (particularly at low LI) to reliably model the LRCs, thus no models were provided. Further, the LRCs were made on leaves of varying and unreported physiological ages, from plants exposed to a vegetative photoperiod (18-h), and acclimated to unspecified localized LI (a canopy-level PPFD of 700 μmol·m⁻²·s⁻¹ was indicated in Chandra et al., 2015). The strong associations between a tissue's light history and its photosynthesis responses to LI, demonstrated in this trial and by others (Björkman, 1981), represent a major shortcoming of using leaf LI response models to infer crop growth and yield. To illustrate, Figure 3 shows LRCs of leaves from a single cultivar, at similar physiological ages (week 5 after transition to 12-h photoperiod) but acclimated to disparate LPPFDs: 91 and 1,238 μmol·m⁻²·s⁻¹. The relative difference in LNCER at higher LIs (≈50%) between these two curves is representative of the potential uncertainty due to just one of the uncontrolled parameters (LNCER) in these prior works. Differing physiological ages of tissues at the time of measurement may have conferred an even larger degree of uncertainty in the magnitude of leaf responses to LI (Bauerle et al., 2020) than leaf light history. Consideration must also be given to the different life stages of a photoperiodic crop (i.e., vegetative vs. generative) and the inherent impact that day length imbues on the total daily PAR exposure (i.e., DLI) which can correlate better to crop yield than PPFD. Further, for a given DLI, yields are higher under longer photoperiod (Vlahos et al., 1991; Zhang et al., 2018), ostensibly due to their relative proximity to their maximum QY (Ohyama et al., 2005). A final distinction between leaf photosynthesis and whole plant yield responses to LI is the saturating LI: the LSP for leaf photosynthesis were substantially lower than the LSP for yield, which remains undefined due to the linearity of the light response model.

Newly-expanded leaves, especially in herbaceous species, are able to vary their leaf size, thickness and chlorophyll content in response to LPPFD in order to balance myriad factors such as internal and leaf surface gas exchange (CO₂ and H₂O), internal architecture of the light-harvesting complexes, and resistance to photoinhibition (Björkman, 1981). In the present study, the effects of LI on leaf morphology was only evaluated in week 5, when the crop was still actively growing vegetative biomass. Reductions in SLW (i.e., increases in specific leaf area, SLA) in response to increasing LI are abundant in the literature (Sims and Pearcy, 1992; Fernandes et al., 2013; Gratani, 2014). In particular, Poorter et al. (2019) reported a saturating response of SLW [also known as leaf mass (per) area; LMA] to LI across 520 species (36% of which were herbaceous plants), however much of their data was at DLIs lower than the minimum DLI in the present study (5.2 mol·m⁻²·d⁻¹), which affected the shape of their SLW response model to LI. Across similar DLI ranges, the average increase in SLW across 520 species was 1.7 × in Poorter et al. (2019) vs. 1.6 × in the present study, indicating that cannabis SLW responses to LI are consistent with normal trends for this parameter.

The lack of LI treatment effects on CCI are also consistent with other studies that have shown that area-based chlorophyll content is fairly stable across a broad range of LIs (Björkman, 1981; Poorter et al., 2019), despite substantial variability in photosynthetic efficiency. However, since there were LI treatment effects on SLW, chlorophyll content on leaf volume or mass bases would likely have reduced under higher LI. The positional effects on CCI (i.e., higher in upper vs. lower canopy) were probably due to the interplay between self-shading and advancing physiological age of the lower leaves (Bauerle et al., 2020). The temporal effects on CCI, which was higher in week 1 vs. weeks 5 and 9, in both upper and lower leaves, may have been due to changes in QY over the life-cycle of the crop. Bugbee and Monje (1992) presented a similar trend; high QY during the active growth phase of a 60-d crop cycle of wheat, followed by a reduction in QY at the onset of senescence (i.e., shortly before harvest). The decline in chlorophyll content in the latter phase of the production cycle probably contributed to the reductions in the photosynthetic parameters (e.g., A_sat, LSP, LNCER) of the tissues measured in week 9 vs. week 5.

Overall, the impact that increasing LI had on cannabis morphology and yield were captured holistically in the plant sketches in Figure 6, which shows plants grown under higher LIs had shorter internodes, smaller leaves, and much larger and denser inflorescences (resulting in higher harvest index), especially at the plant apex. Like many other plant species, we have found that cannabis has immense plasticity to rapidly acclimate its morphology and physiology, both at leaf- and whole plant-levels, to changes in the growing lighting environment. Therefore, in order reliably predict cannabis growth and yield to LI, it is necessary to grow plants under a broad range of LIs through their full ontological development, as was done in this study. Without knowing the respective tissues' age and light history, instantaneous light response curves at leaf-, branch-, or even canopy-levels cannot reliably predict yield.