The goal of this study was to determine the optimal concentration of N, P, and K in the nutrient solution for the flowering stage of soilless cannabis production using RSM. The optimal concentrations of nutrient solution N and P was predicted to be approximately 194 mg L^–1 N, and 59 mg L^–1 P, respectively. Based on analysis of the response surface model, it was found N and P were the most important factors in predicting inflorescence yield. Inflorescence yield decreased markedly outside of the range of 160–230 mg L^–1 N, and 40–80 mg L^–1 P. These findings suggest that drug-type cannabis responds well to nitrogen and phosphorus during the flowering stage. Inflorescence yield did not respond to nutrient solution K concentration within the tested range, indicating the K currently supplied (300–400 mg L^–1) by some commercial cultivators are likely too high.

Inflorescence yield had a strong positive correlation with a number of vegetative growth attributes. The strong correlation between inflorescence yield and plant growth index indicates that larger plant size can result in higher inflorescence yield. Nutrient supply, especially N, can determine cannabis plant size as N is an essential component of plant chlorophyll and ribulose-1,5-bisphosphate carboxylase-oxygenase (Rubisco). Low levels of N can reduce plant photosynthetic capacity and limit plant growth (Saloner and Bernstein, 2020). For flowering drug-type cannabis in soilless culture, supply of 30 and 80 mg L^–1 N restricted whole plant and inflorescence growth, but plants performed optimally with supply of 160–320 mg L^–1 N (Saloner and Bernstein, 2021). The optimal N supply (194 mg L^–1) found in our study is within their range, despite the two studies using two different growing methods and plants with different genetic backgrounds. For drug-type cannabis during the flowering stage in an organic-based soilless production system, the optimal N supply was slightly higher (212–261 mg L^–1; Caplan et al., 2017a) than the optimal level found in the present study. A possible explanation for the higher optimal N supply in the organic fertiliser study is that N from organic-based fertilisers may not always be readily available, as the release of N from organic fertilisers depends on the speed and extent of the mineralisation process (Hartz et al., 2010; Dion et al., 2020). Though it is unclear what source of organic nitrogen was used in their study, factoring in organic N availability of around 60% would put our findings in line with those by Caplan et al. (2017a). Along with aboveground growth, root growth also contributes to overall plant size. We found that inflorescence yield had a strong positive correlation with root dry weight, supporting our conclusion that larger plants produce higher yields. The context of where plants spend their energy is important. For industrial hemp, increasing N supply increased plant growth, but this growth was partitioned more towards stem material rather than valuable inflorescence material (Campiglia et al., 2017). Further investigations of cannabis response to nitrogen should consider product quality, and the distribution of biomass to various plant organs to maximise inflorescence growth and quality.

While modelling of cannabis inflorescence yield response to N, P, and K with surface analysis accounts for interaction between nutrients, the surface response model demonstrated that K, within the tested range of 60–340 mg L^–1, had no effect on inflorescence yield. This lack of response may suggest that 60 mg L^–1 K is not low enough to cause nutrient deficiency, and 340 mg L^–1K is not high enough to cause toxicity. Moreover, cannabis responses to K may be genotype specific. Plants of one cannabis genotype Royal Medic supplied with 240 mg L^–1 K had 25% reduced fresh shoot and root biomass by compared to those fed with 175 mg L^–1, while plants of genotype Desert Queen had up to 40% increased shoot and root biomass (Saloner et al., 2019). Plant height, number of nodes on the main stem, and stem diameter of these two genotypes remained similar, so this difference in biomass was caused by one genotype becoming “bushier” than the other under high K supply. These differences in the response to K supply may be due to differences in plant tissue (e.g., main stem vs. side branch) sensitivity to K. Plant phenological stage (i.e., vegetative or flowering stage) may also be a factor in cannabis response to K supply. In a previous study of flowering aquaponic cannabis response to K, inflorescence yield increased when plants were provided with K up to 150 mg L^–1 (Yep and Zheng, 2020). Genotype and plant phenological stage should be considered in future studies looking at cannabis response to nutrients, especially K.

Many commercial cannabis cultivation operations currently use fertiliser formulations that contain very high levels of P (more than 200 mg L^–1 P in some cases). This practice is based on anecdotal evidence that P enhances inflorescence production. These concentrations are much higher than the optimal rate of 60 mg L^–1 P found in our study, and at the higher range could cause reduction of both plant growth and inflorescence yield. In addition to reducing plant growth and yield, excessive supply of nutrients is a potential source of environmental pollution. Though, cannabis does appear to have the ability to store and mobilise certain amount of P when required. When provided with P higher than 30 mg L^–1 in the vegetative stage, cannabis sequestered excess P in root tissue to prevent excess accumulation in the shoots (Shiponi and Bernstein, 2021). A greater understanding of what cannabis P requirements are, and whether there is any truth to the practice of supplying high concentrations of P, should be a priority for making cannabis production more sustainable. However, based on existing data it appears that the levels of P found in many cannabis specific commercial fertilisers are far higher than needed and could lead to negative environmental impacts.

While the cannabinoid concentrations in the floral tissues in our study did not respond to nutrient solution NPK concentrations, other studies indicate that plant mineral nutrition can affect production of secondary metabolites in cannabis (Caplan et al., 2017a; Saloner and Bernstein, 2021). There appears to be an inverse relationship between cannabis yield and potency, with cannabinoid concentrations decreasing as plant inflorescence yield increases. Inflorescence from plants supplied with 160 mg L^–1N had approximately 30 and 20% lower concentrations of THCA and CBDA than plants supplied with 30 mg L^–1N (Saloner and Bernstein, 2021). However, while nutrient stress and deficiency may enhance inflorescence cannabinoid content, this method is not ideal for optimising overall plant productivity as plants supplied with 160 mg L^–1 N yielded twice that of those supplied with 30 mg L^–1N. Cannabis grown in two organic growing media with different organic fertiliser rates (i.e., 57, 113, 170, 226, and 283 mg L^–1N) had negative linear relationships between the concentrations of inflorescence THCA and CBGA and the fertiliser application rate for some of the treatment combinations (i.e., growing media and fertiliser rate) (Caplan et al., 2017a). However, for the most of the treatment combinations, fertiliser rates from 57 to 226 mg L^–1N did not have any effects on THCA or CBGA concentrations; and the cannabinoid concentrations only dropped when the fertiliser rate increased to the highest level of 283 mg L^–1N. The context of yield is again important when analysing differences in cannabinoid content as THCA concentrations dropped by ∼20% in the highest fertiliser rate, but inflorescence yield almost doubled vs. lowest fertiliser rate. As noted by Bernstein et al. (2019), an understanding of how nutrient supply influences cannabinoid concentrations would be an important step towards controlling and standardising the cannabinoid contents of medical cannabis. Cannabinoid concentrations are also important to recreational consumers, who rank THC and CBD concentrations among the most important factors when making purchasing decisions (Zhu et al., 2020). Given that cannabinoids are the compounds that make cannabis so uniquely valuable, more work needs to be done to investigate the effect of mineral nutrition on cannabis yield, and the relationship between yield and potency. Further work should also evaluate other compounds that are known to impact product quality.

The use of central-composite design allows experimenters to account for potential interactions between the different nutrients. This is important as nutrient interactions have been shown to affect plant nutrient uptake (Fageria, 2001; Rietra et al., 2017). A recent study found that high K supply decreased concentrations of Ca and Mg in cannabis leaf tissue, indicating antagonistic relationships between these positively charged ions (Saloner et al., 2019). An understanding of how combinations of nutrients at different concentrations affect crop growth, yield, and quality is important for the development of recommendations for the commercial cannabis industry. Had the same number of nutrients and nutrient levels as were included in this study been investigated with a traditional full-factorial design, many more nutrient solution treatment groups would have been required, compared to the number of treatment groups used in this study. The difference in number of treatment groups needed can be more pronounced as more factors (i.e., Ca, Mg) are included. Considering the high cost of cannabis and growing space in controlled environments, the response surface approach allowed us to complete this study where another experimental design may have been prohibitive.

No matter the experimental design used, an inherent problem in nutrient solution experiments is that nutrients cannot be added individually but must be added as a compound containing both anions and cations. Further, the ionic balance constraint requires the sum of the charges of cations and anions in solution to be equal (De Rijck and Schrevens, 1999b). The implication for formulating experimental treatment solutions is that it is practically impossible to change the level of one nutrient while keeping concentrations of all other nutrients the same. In this study, we focused on N, P, and K concentrations while attempting to keep all other nutrients at reasonable levels using commonly available horticultural fertiliser compounds. For example, potassium nitrate and calcium nitrate usually contribute to the bulk of nitrogen, potassium, and calcium in horticultural nutrient solutions (Resh, 2012). Formulating a high N, low K nutrient solution with these fertilisers results in higher levels of Ca than other nutrient solution treatments. Likewise, a low N, high K nutrient solution necessitates an additional source of K such as KCl, which would increase solution Cl concentration. Higher concentrations of nutrients such as Ca and Cl bring the potential for nutrient interactions which may affect experimental results. The lack of response to K in the range of 60–340 mg L^–1 observed in our trial may be partially due to competition for uptake from Ca. Regarding experimental Cl levels, hydroponic cannabis has been shown to tolerate rates of 180 mg L^–1 Cl with no impact on yield or potency (Yep et al., 2020a) so it is unlikely Cl levels limited plant growth in this trial. Though less than ideal in an experimental setting, there is no perfect solution for the problem of keeping all nutrient concentrations the same when formulating treatment solutions.

While this trial determined the theoretical optimum levels of N and P for the DWC growing method, these levels may not be definitive for all production methods or genotypes. Our trial was conducted in solution culture with weekly nutrient solution changes, and the EC and pH dynamics of our DWC units are likely different than other growing methods, meaning that plant nutrient availability and overall salinity of the nutrient solution would also likely be considerably different. Many commercial cannabis operations utilise substrate-based soilless cultivation systems, such as coir in containers, that may offer more nutrient and pH buffering capacity (Zheng, 2020). Having said that, our trial does represent or closely resemble some common soilless production practices, such as growing cannabis in rockwool, in the current cannabis production industry (Zheng, 2021). The treatments were applied only during the short-day period (i.e., flowering stage), and considering that plant nutrient requirement may vary at different development stages, the same experiment may also need to be conducted for the vegetative stage. Another limitation of our study was that we only used a single cannabis cultivar. Similar experiments should be performed on different cultivars, with disparate growth habits and cannabinoid compositions to investigate how individual cultivars may respond to NPK treatment levels. Additionally, this study only looked at inflorescence yield and cannabinoid composition and did not evaluate the impact of NPK on inflorescence terpene content or organoleptic properties.

Drug-type cannabis is still a relatively new crop in the legal setting, especially for large-scale commercial production, and many aspects of its cultivation are relatively unknown. We found that response surface methodology was a suitable experimental approach for investigation of cannabis responses to NPK, and that modelling of yield response to these nutrients aided us in achieving our experimental objective. Based on the results of this study, we recommend providing plants with a nutrient solution containing N and P at approximately 194 and 59 mg L^–1, respectively, to achieve maximal inflorescence yield. Future studies should investigate the inflorescence yield and vegetative growth response of genetically diverse cultivars to macronutrients and include more quality parameters to ensure that plant yields do not compromise product quality. Improving our understanding of cannabis responses to mineral nutrients is an essential step towards the effective and sustainable cultivation of this high-value horticultural crop.