1. Introduction

The concept of indoor agriculture (IA), first introduced in the 1970s (Mitchell, 2022), offers an environmentally sustainable alternative to large-scale leafy green production and allows fresh food to be grown closer to consumers in urban areas, even in regions with less favourable climates (Despommier, 2009; Pinstrup-Andersen, 2018). Despite not being a novel concept, these systems have gained popularity within the last decade due to technological advancements that led to reduced capital investments and operating costs, giving rise to an emergent industry. These indoor farms, also referred to as vertical farms or plant factories, fall within the realm of controlled environment agriculture (CEA). However, they distinguish themselves from their counterparts by utilizing the vertical space of the building and employing stacked shelves for production, operating exclusively with sole-source lighting systems. Leveraging high-tech technology, IA farms generally utilize hydroponic systems, and comprehensive environmental controls encompassing temperature, air circulation, and humidity settings, to name a few, in a completely sealed environment.

Some of the social-economics advantages brought by these systems are enabling year-long production closer to the final consumer (Kalantari et al., 2020), requiring little to no pesticide, and improving resource use efficiency (Banerjee and Adenaeuer, 2014; Fukuda and Wada, 2019; Pennisi et al., 2019; Stein, 2021). For example, IA farm systems consume 90–95% less water than field farming operations (Stein, 2021).

Indoor farms also occupy significantly smaller building footprints while achieving productivity per unit of land 100 times higher than outdoor growing (Kozai et al., 2022) through the use of high plant density and vertical space (Armanda et al., 2019; Kozai, 2013). These enclosed structures also allow for precise control of environmental growth factors, enabling growers to enhance plant quality attributes valued by consumers such as colour, taste, and texture (Fukuda and Wada, 2019; Pennisi et al., 2019; Seong et al., 2023; Zhang et al., 2019).

Furthermore, IA can contribute to mitigating domestic supply shortages seen in the U.S. in the last decades. Lettuce availability per capita accumulates a 54.6% decline since the 2000s, showing successive annual downward movements. That suggests a dependency of domestic availability for fresh lettuce heads on imports (USDA-ERS, 2022). The capacity to grow year-round fresh food shielded from external weather fluctuations and in significantly smaller land area, allows for the installation of these indoor farms in urban areas. That has the potential to increase production using non-arable land including cities with less favourable climates. In that aspect, indoor farms could also contribute to a reduction in greenhouse gas emissions associated with ground transportation particularly in the USA, where 91% of iceberg lettuce and 97% of Romaine lettuce are grown in only three states (USDA, 2022). From a market perspective, a recent leafy green consumer survey identified urban dwellers as a potential niche market for this industry given their positive attitude towards the innovative food production system proposed by IA (Seong et al., 2023).

Despite its innate ability to produce high-quality vegetables using an environmentally sustainable system and despite having a defined potential market, the industry struggles to grow. The economic feasibility of these vertical farming systems has recently been challenged in the trade press given that not all enterprises have met investors’ high expectations of profitability (Gordon-Smith, 2022). The complexity of these systems, coupled with the absence of established production standards in this emerging industry, involves understanding the connection between capital investments in production systems and potential output. For instance, changes in harvesting time will influence space allocation and directly impact operating costs related to labour, input consumption, electricity usage in the growing area, and the overall farm size. Simultaneously, adjustments in the production schedule directly affect plant size and, consequently, annual harvest and revenues. Therefore, the first decision when designing a farm structure with the goal of enhancing economic results must involve selecting the most efficient allocation of farm growing space by assessing the economic trade-offs between capital investments and revenues associated with alternative production cycles.

By incorporating biological, technological, and economic parameters, this study approaches the complexity of the problem from a whole-system perspective. Specifically, it tackles a heated debate within the industry, underscoring the importance of identifying the intricate relationship among these system parameters. This pertains to the decision between growing leafy greens up to maturity or harvesting earlier at ‘baby’ or ‘teen’ stages (Kuack, 2017). From a production perspective, harvesting a younger plant avoids issues with tip burn and outer leaf losses. On the other hand, growing the lettuce to maturity could increase relative yield by taking advantage of the later exponential growth stage of the plant, yet requiring more space and time. While this question involves plant physiology considerations and environment control strategies to overcome tip burn (Al-Said et al., 2018; Kang et al., 2013, 2016; Lee et al., 2019), the question also involves important aspects of profitability. This paper focuses on the latter.

Using parameters estimated for efficient space allocation aiming maximum leafy green yield in an indoor setting (Shasteen et al., 2022), this paper describes the development of a foundational production function for an indoor farm that incorporates the relationship between costs associated with, and revenues resulting from, various combinations of harvesting time and efficient space utilization. The aim is to identify the lettuce production schedule that yields the highest profitability. Given the model specification, where the farm size is determined starting from a minimum unit of growing area, the farm size at the production schedule that maximizes contribution margin to profit can be considered as the minimum economically efficient farm size.

2. Methodology

A decision on harvesting time has a direct impact on plant density during the production stage. The longer the production cycle, the larger the plant grows, and the larger will be the physical growing space required after transplanting from the high-density germination trays into the low-density production space. With larger spaces comes larger investments in facilities and technology.

Specifically, the development of the proposed bioeconomic model consisted of three steps. First, a production system was designed to define the minimum dimensions of a growing area based on the space and time determinants of a production schedule. Next, the model incorporated expected plant development within the defined growing area using plant growth coefficients under various alternative production schedules. Finally, an economic module was defined. While growing area and production schedule parameters served as grounds for to the development of the production function, production costs and plant yield informed the partial budget analysis of the various production schedules. The production schedule yielding the highest results was determined by measuring the contribution margin to profitability. Subsequently, a sensitivity analysis of the farm structure producing the maximum contribution margin was conducted, considering all costs and price components of the model.

2.1 Growing area

The first step in defining this production system is establishing how time and space are considered in the setup of an indoor farm structure. A two-stage production system was defined. Initially, plants are grown in high-density trays (Den) for a given number of days (D), referred to as propagation period, which occurs from seeding to transplant (BT, before transplant). After a pre-determined number of days, these plants are then transplanted into a grow-out area, where they continue to grow until achieving the expected size at harvest. This second stage is referred to as the production period, which takes place between transplant and harvesting (AT, after transplant). The spacing between transplanted plants during this second production stage, referred to as plant density (Den_AT), depend on the size of the plant the grower intends to sell at the end of the production cycle (DAT). Consequently, the decision regarding the plant size at harvest must be made at the time of transplanting. If the decision is made to harvest mature lettuce heads, plants are grown for a longer DAT period, resulting in a greater harvest. However, the lower post-transplant density, compared to the high density before transplanting, requires a larger growing area and, as a result, additional investments in space and equipment. The number of days taken to grow leafy greens from seeding to harvest (DBT and DAT) and the equivalent plant density prescribed for an efficient space allocation (Den_BT and Den_AT) were assumed to be determinants of the size of total growing area (GA) and, consequently, farm size.

On a continuous production system, the propagation area (Spc_BT) sits apart from the production area, extending the size of the required GA. Its size (B_area) was defined as the smallest area unit used in the production, established in this example as 1 (one) m². The variable B_area was therefore a mechanism to expand set growth space in a modular manner. The size of the production area (Spc_AT) was estimated using a space ratio (Spc_R), defined as the ratio of plant density in the propagation area (Den_BT) to that of the production area (Den_AT):

View Full Size

Citation: International Food and Agribusiness Management Review 27, 2 (2024) ; 10.22434/ifamr2023.0038

• Download Figure
• Download figure as PowerPoint slide

At this point, each slot of propagation and production space would be continuously used over time and the frequency of harvest was determined by DAT alone. Furthermore, we considered that commercial growers are likely to require daily harvesting to provide a consistent supply. To obtain daily harvests, the grower requires as many growing modules as the length in days of the total growing cycle before (DBT) and after transplant (DAT), with cycles starting consecutively on a daily basis. The time during which each space will be occupied by either propagation or production stages and the time needed to allow enough space to propagate seeds and proportionally populate the production area further, defines the total growing space (TSpc) required for daily harvest. Consequently, TSpc expands as a function of a combination of the ratios between density before and after transplanting (Den_BT and Den_AT) and between the length of cycle per growing stage (DBT and DAT). The smaller the Den_AT, the larger the plants grow, requiring a longer DAT and a larger growing space.

View Full Size

Citation: International Food and Agribusiness Management Review 27, 2 (2024) ; 10.22434/ifamr2023.0038

• Download Figure
• Download figure as PowerPoint slide

In the case of a farm system using vertical growing racks, some costs such as property rent would only be applied to the floor area, including a space for corridors (cdor) between racks (Urisami, 2021). This hypothetical indoor farm is assumed to employ a hydroponic system in which plants are grown in a nutrient solution with a soilless substrate to provide physical support to the plant. The growing system is installed across four levels in the production area and eight levels in the propagation area, assuming half of canopy to light height requirement. Floor area was estimated as the ratio of the growing area before and after transplant, to the number of shelves used in the propagation (Shelf_BT) and production (Shelf_AT) areas:

View Full Size

Citation: International Food and Agribusiness Management Review 27, 2 (2024) ; 10.22434/ifamr2023.0038

• Download Figure
• Download figure as PowerPoint slide

2.2 Plant density and plant mass

As illustrated in the first equation, the ratio of plant density before transplant to the plant density after transplant is crucial in determining the total growing area required to transplant grown seedlings into the propagation area. This relationship, as explained above, depends on the time required to grow a plant to a specific size. In this context, equations (1) to (5) define the total growing area of the farm. Given the significant impact of plant density after transplant on total growing area, the computation of the radius that defines the most efficient use of growing area space becomes significant. Below, we describe how this radius, which maximizes space use efficiency, was estimated under different scenarios of plant distribution on the canopy area, using decreasing values of plant density. The relationship between plant density, time necessary to grow, and final plant mass was then estimated to allow further projection of potential harvesting weight for a given combination of plant density and DAT that maximizes space use efficiency.

Values for the most efficient space utilization plant density and resulting plant mass associated with a given DAT were estimated based on data from a space optimization experiment (Shasteen et al., 2022), which used a combination of four popular lettuce cultivars, ‘Seurat’ (area data without crowding), ‘Rouxai’, ‘Pascal’ and ‘Rex’ (mass data). In this previous analysis of density, solutions were presented for one specific growing configuration with dimensions similar to that used in the industry that included fluctuations which could shift the maximum. In that analysis, it was observed that a transformation of top projected canopy area (TPCA) measurements into “characteristic radius” showed a roughly linear expansion trend in lettuce until canopy closure prevented further horizontal expansion, which occurred around days 19 to 20. Although that original analysis proceeded to use the radius data directly to calculate maximum density and thus introduced fluctuations from the original measurements, the present analysis created a smoother, more general solution by introducing several estimating functions into the analysis. An estimating function for the characteristic radius was made by taking the line of best fit for radius data between days 0 to 19 after transplant, which excluded radius data after the time of canopy closure. This resulted in the following definition for radius and time:

View Full Size

Citation: International Food and Agribusiness Management Review 27, 2 (2024) ; 10.22434/ifamr2023.0038

• Download Figure
• Download figure as PowerPoint slide

where r(HAT) is radius in centimetres as a function of time of harvest, m is slope, b_r is y-intercept and HAT is time of harvest in hours after transplant.

To combine the effects of the relationship between yield defined by plant mass, plant density, and length of harvest cycle at optimized canopy-space-use into the bioeconomic model, an estimator function for the growth model was created by way of a Gompertz function (Shimizu et al., 2008). The parameterized form of the Gompertz function that best fit the model data was defined as

View Full Size

Citation: International Food and Agribusiness Management Review 27, 2 (2024) ; 10.22434/ifamr2023.0038

• Download Figure
• Download figure as PowerPoint slide

where g(HAT) is the Gompertz estimator for the growth function, which gives the fresh mass per head in grams, given the constant conditions of 400 ppm CO₂, 11.6 mol/m²·per day on a 16 hour photoperiod, and a 23–19°C light to dark temperature cycle, and a, b, c, d and f are parameters used to shape the function (see coefficient values in Table A1 in the Appendix). Appropriate parameters for the function were found by using the LibreOffice v7.3.6.2 Swarm Non-Linear Solver tool to minimize the sum of squared errors (SSE) of the Gompertz function against hourly output data for the fresh mass of a head of lettuce as calculated by the model. In order to generally estimate the yield rate, it is necessary to create several sets of these transition points for different arrangements of growing area. Seven sets of maxima were created by sweeping the proportions of the growing area from 1:2 to roughly 1:1, namely from 121.92 × 243.84 cm to 241.92 × 243.84 in 20 cm increments. The non-linear solver from before was then used to minimize the SSE across all seven sets of data for the domain of time from 8.33 to 28.0 DAT (HAT 200 to 672). The lower limit of 8.33 DAT was used to exclude a small ‘tail’ of outlier data points which bucked the trend. The estimator function, in this case, made use of the Gamma Distribution from statistics, which had a much lower SSE than a second order polynomial estimator. The Gamma distribution is defined as:

View Full Size

Citation: International Food and Agribusiness Management Review 27, 2 (2024) ; 10.22434/ifamr2023.0038

• Download Figure
• Download figure as PowerPoint slide

where Γ is the gamma function, x is an integration constant, γ is the gamma distribution, α and β are arbitrary shape and rate parameters, and t is time. The parameterized form of the estimator function is:

View Full Size

Citation: International Food and Agribusiness Management Review 27, 2 (2024) ; 10.22434/ifamr2023.0038

• Download Figure
• Download figure as PowerPoint slide

where h() is the yield rate in g/m².day as a function of time of harvest, HAT is the time of harvest in hours after transplant, and δ, κ, x₀ and y₀ are parameters used to shape the function. The SSE for h(HAT) was 2697.3, in contrast with the SSE for a 2nd order polynomial, which was 7690.9 and was less accurate near the maximum, justifying the more complex estimator.

The maximum for the general estimate of yield rate occurs at 474 hours after transplant (19.75 DAT). This is consistent with the conclusions of the previous analysis, which found local maxima at 15.6 and 17.6 DAT. The difference in position for these maxima is likely due largely to the specific GA selected and the combined effects of fluctuations in the radius and growth functions.

The density (Den_AT) can be derived from the earlier equations for yield rate, h(HAT), and plant mass, g(HAT):

View Full Size

Citation: International Food and Agribusiness Management Review 27, 2 (2024) ; 10.22434/ifamr2023.0038

• Download Figure
• Download figure as PowerPoint slide

In the following sections, the density experiment data, which established the relationship between plant mass (g), density (Den_AT) and the length of production cycle (DAT) were integrated to the economic model framework, providing insight into the trade-offs in required capital investment and operating expenditures associated with space and production cycles.

2.3 Harvest

Daily harvesting occurs in one module of the production area, upon the end of each consecutive cycle, with the assumption of 100% germination and zero harvest losses. As such, daily harvest (H_D) was defined as a product of the area being harvested, plant density after transplant (Den_AT), and fresh mass of individual plant (g(HAT)):

View Full Size

Citation: International Food and Agribusiness Management Review 27, 2 (2024) ; 10.22434/ifamr2023.0038

• Download Figure
• Download figure as PowerPoint slide

2.4 Revenue

Revenue was represented in the model as the product of daily harvest (H_D), converted from harvest in grams to kilograms (1000 g=1 kg), and the average retail price for single varieties adjusted by retailers’ margin (RetM) on a commercial annual basis (y).

View Full Size

Citation: International Food and Agribusiness Management Review 27, 2 (2024) ; 10.22434/ifamr2023.0038

• Download Figure
• Download figure as PowerPoint slide

2.5 Costs

Costs considered here were the most significant operational expenditures (OpEx) in an indoor farm, including variable and fixed operating costs which were associated with activities directly affected by changes in DAT and plant density, and consequently, the size of farm. Reportedly, the three most significant operational costs in an indoor farm are electricity, labour, and depreciation of the investment, accounting for approximately 65% of total operating costs in a vertical farm (Kozai et al., 2022). Water usage within this closed and controlled environment production system, such as an indoor farm, is 95% less than in other traditional growing systems, thanks to the implementation of recirculating systems and the highly efficient reabsorption of evaporated water within these enclosed systems (Kalantari et al., 2018; Kozai et al., 2022). Therefore, while water is typically a significant consideration in agricultural system analyses, it is not deemed a relevant cost in this analysis. Indoor farms also have superior control over nutrient usage compared to traditional agricultural systems. Expenditure in nutrients can be as low as 1.5% of operating costs (Lu et al., 2022), making nutrient costs irrelevant for inclusion in this analysis.

Variable operating costs (VOC) included seeds and substrate (herein called inputs), packaging costs, electricity, and labour.

Inputs: Seeds (C_Seed) and substrate costs (C_Sbs) are expected to be directly affected by changes in plant density. Substrate was assumed to be single seeded in the plant propagation stage and then transferred with the plant upon transplant into the production area. As such, along with seeds, these costs occur as a product of plant density per square meter on the first day of each consecutive cycle per module of propagation area. On an annual basis, cost of inputs was estimated as:

View Full Size

Citation: International Food and Agribusiness Management Review 27, 2 (2024) ; 10.22434/ifamr2023.0038

• Download Figure
• Download figure as PowerPoint slide

Packaging costs (C_Pck) are expected to be directly affected by yield as more packaging material would be required for increased yields. Annual cost of packaging was estimated using the annual harvest weight (H_D·y) by the size of a package (Sz):

View Full Size

Citation: International Food and Agribusiness Management Review 27, 2 (2024) ; 10.22434/ifamr2023.0038

• Download Figure
• Download figure as PowerPoint slide

Electricity costs were estimated based on photoperiod (PhPer), photosynthetic photon flux density (PPFD, μmol/m²·s), efficacy (ηPAR, μmol/J), electricity price rate, and assumed a constant loading of energy used in heating, ventilation, and air-conditioning (HVAC) systems during the photoperiod, continuous throughout the year, therefore without seasonal effect of outdoor air temperature and uniform throughout the entire growing area. The energy consumption of HVAC systems is significant in the context of overall energy usage, potentially constituting as much as 34%, depending on the efficiency levels (Gonzales-Torres et al., 2022). In this study, consultations with stakeholders in the indoor agriculture industry were conducted, leading to the inclusion of a 30% HVAC loading factor in estimating electricity usage. The Daily Light Integral (DLI), or the total number of photons of photosynthetically active radiation per day received per m² area (mol/m²· day), was estimated as a function of the photoperiod converted from hours to seconds, (1 h = 3600 s) and PPFD used in each stage (i), adjusted for unit conversion (1 mol = 1 000 000 μmol):

View Full Size

Citation: International Food and Agribusiness Management Review 27, 2 (2024) ; 10.22434/ifamr2023.0038

• Download Figure
• Download figure as PowerPoint slide

Daily energy cost (DEC) per square meter was taken as the ratio of DLI to light efficacy (ηPAR) and multiplied by the electricity rate (C_El) in US$/kWh (where 3 600 000 J = 1 kWh):

View Full Size

Citation: International Food and Agribusiness Management Review 27, 2 (2024) ; 10.22434/ifamr2023.0038

• Download Figure
• Download figure as PowerPoint slide

Annual electricity cost becomes a function of lighting intensity and duration per each relevant growing area with the additional impact of HVAC loading:

View Full Size

Citation: International Food and Agribusiness Management Review 27, 2 (2024) ; 10.22434/ifamr2023.0038

• Download Figure
• Download figure as PowerPoint slide

Labour costs were estimated as wages paid with benefits allocated to the number of hours spent per day per square meter in five general labour activities, representing the singular effect of changes in the time required for performing seeding, transplanting, and harvesting activities. Salaries paid to other personnel not directly involved in production activities were not included. This comparison should thus be taken as a marginal effect rather than actual total labour costs.

The estimation of the number of hours per activity was done in two steps. Firstly, labour hours (L) required per area (m²) were estimated from the reported fresh mass production per hour of labour (kg/person·h) and area required to produce 1 kg of fresh mass a day (m²/kg) in the context of the Japanese Plant Factory (Uraisami, 2021). Next, the ratios of one hour of labour for each of these five activities were applied to the labour hours per square meter for each day (Uraisami 2021), including seeding (L_s), transplanting (L_t), cleaning (L_c) and harvesting and packaging (L_hpk).

Total labour hours per day (L_D) for the entire growing area was estimated as the sum of the number of labour hours per square meter before (L_BT) and after transplant (L_AT). The former includes seeding activities taking place at the Spc_BT area, and the latter includes activities deemed to take place at the Spc_AT area, including transplanting, harvesting, and packaging. It is assumed that cleaning (L_c) occurs daily in the entire growing area.

View Full Size

Citation: International Food and Agribusiness Management Review 27, 2 (2024) ; 10.22434/ifamr2023.0038

• Download Figure
• Download figure as PowerPoint slide

Annual wages paid (AnnWg) was the product of daily labour in number of hours in one year by hourly wages (W_h) subject to a 20% benefit loading (Valle de Souza et al., 2022) (W_b):

View Full Size

Citation: International Food and Agribusiness Management Review 27, 2 (2024) ; 10.22434/ifamr2023.0038

• Download Figure
• Download figure as PowerPoint slide

Total annual variable operating cost (VOC) becomes:

View Full Size

Citation: International Food and Agribusiness Management Review 27, 2 (2024) ; 10.22434/ifamr2023.0038

• Download Figure
• Download figure as PowerPoint slide

Fixed operating costs (FOC) included rent and the depreciation of technology investment, as the operational recognition of the consumption of capital invested in technology. Annual rent (AnnC_r) is applied to total floor area. Depreciation of Technology costs (AnnC_t), on the other hand, was applied to the total growing area, considering that increasing the number of shelves will multiply the technology costs. Total annual fixed operating cost (FOC) becomes:

View Full Size

Citation: International Food and Agribusiness Management Review 27, 2 (2024) ; 10.22434/ifamr2023.0038

• Download Figure
• Download figure as PowerPoint slide

3. Production Model Data

In order to define the indoor farm size and lettuce production schedule that yields highest contribution margin to profitability, the above-described model used average market prices of lettuce and production costs faced by a commercial grower operating in the US to the extent of data availability. Due to the nascent nature of this industry, neither scanner data nor national economic data are presently accessible for a retrospective market analysis aimed at trend identification. Consequently, authors have relied upon a limited corpus of international publications, industry reports and personal communication, and online market price searches.

The Market prices for lettuce in this model simulation was the average market price for differentiated lettuce, ready for consumption, and sold in hard plastic packages, resulting from an online search of groceries stores’ websites (e.g., Wholefoods, Meijer, Kroger, Aldi), in the Fall 2022 (see summary of results in Table A2 in the Appendix). No significant changes in prices could be identified for different leaf sizes. Online searches are subject to IP address restrictions, potentially introducing geographical bias to the prices utilized in this simulation. Nonetheless, a subsequent sensitivity analysis, detailed in the paper’s conclusion, investigates the impact of varying price points on the results presented herein. The average package size of 5 oz (approximately 142 g) was adopted, aligning with the packaging size predominantly used by these growers and conforming to the standard practices of the few retailers on which the industry currently relies for distribution of their produce. A retailer margin was estimated using a standard industry gross margin of 40%, an approximation to USDA-published price spread from farm to consumer (USDA ERS, 2019), applied over the cost of goods sold (COGS) (Clark, 2020).

Cost of seeds was the average price for leafy lettuce cultivars, including red, green, and oakleaf varieties, bought at a medium size package from an online distributor. As for substrate, 1-inch rockwool hydroponic grow cubes starters were considered, following average market prices (see model parameters in Table 1). As for packaging costs, an average price per unit of a most observed amorphous polyethylene terephthalate (APET) clear plastic clamshell container, suitable for 5 oz (approximately 142 grams) of lettuce, sourced online at the time of this study, was used.

View Full Size

Initial growth parameters selected for this study

Citation: International Food and Agribusiness Management Review 27, 2 (2024) ; 10.22434/ifamr2023.0038

• Download Figure
• Download figure as PowerPoint slide

Average US electricity costs for the industrial sector were US$0.0845/kWh in 2022 (US-EIA, 2022), with prices as low as US$0.0694/kWh in North Dakota or US$0.0672/kWh in Idaho and as high as US$0.3671/kWh. A median price was used in the simulation (Table 1).

Wages paid in horticulture farms in the U.S. and classified under the NAICS code 111419, varied from US$6.57/h to US$31.42/h in 2022 (BLS 2022). However, these data include workers assigned to tasks not included in this analysis such as administration or marketing. An hourly wage of US$12.46 per hour was deemed to reflect production related wages (Agrilyst, 2017).

Annual rent paid for the facilities was estimated at US$921/m² based on reported average rent per square foot paid for industrial space in the U.S. in 2021 of US$7.13/ft² (JLL, 2022).

Given the lack of standardized systems for indoor agriculture in the U.S. at the time of this publication, estimates of technology investment are restricted to a few publications. For reference, a Danish study reported approximately US$535/m² of growing area (Avgoustaki and Xydis, 2020), while a technology provider quoted between US$760/m² and US$978/m² in Europe (IFarm, 2022). These included costs of lighting fixture, growth unit racks, hydroponic systems, ventilation fans, and HVAC systems, but no structural building costs. The Japanese Plant Factory Association reports initial investment in the cultivation area, including LED fixtures and other materials, to be US$1,019/m² with an annual depreciation cost of seven years (Uraisami, 2021). As the model applied labour requirement parameters in the context of Japanese Plant Factories, we assume a correlation between labour requirement and level of automation and considered more appropriate to incorporate the cost of investments in technology reported for the Japanese Plant Factories into this analysis.

4. Results

A partial budget analysis was performed for specific scenarios of DAT with the equivalent required plant density (DenAT) and resulting plant mass (g). These scenarios represented fractional increments of DAT from 7 to 28 days. The plant density required for each DAT scenario and resulting plant mass were estimated using equations (6) to (11). With the purpose of identifying the maximum contribution margin to profit considering alternative growing schedules, marginal increments to profit were examined using a normalized measure of results per square meter of growing area (GA) and per day of production cycle (DAT). Refraining from describing these results as profits, given the restricted focus of the analysis, a measure of contribution margin (CM) was used, being the contribution towards profitability rather than profit itself. Thus, the resulting CM is therefore a vector of the difference between projected annual revenues ($/m²) and the sum of annual production costs ($/m²) for each production schedule scenario (i), all normalized by of GA and length of harvesting cycle, or DAT:

View Full Size

Citation: International Food and Agribusiness Management Review 27, 2 (2024) ; 10.22434/ifamr2023.0038

• Download Figure
• Download figure as PowerPoint slide

The maximum contribution margin (V) in the vector CM is estimated as:

View Full Size

Citation: International Food and Agribusiness Management Review 27, 2 (2024) ; 10.22434/ifamr2023.0038

• Download Figure
• Download figure as PowerPoint slide

where CM(i) represents the contribution margin associated with production schedule i, and n = 504, the total number of simulations for fractional increments of DAT from 7 to 28 days. The results of the partial budget analysis, utilizing the production schedule defined as the one that yields the maximum contribution to margin, are presented in Table 2.

View Full Size

Description of the lettuce farm at maximum Contribution Margin (CM)

Citation: International Food and Agribusiness Management Review 27, 2 (2024) ; 10.22434/ifamr2023.0038

• Download Figure
• Download figure as PowerPoint slide

Total normalized revenue ($/m²·DAT) is larger at 15.3 DAT (see red line on Figure 1). However, costs are significantly higher at lower DAT, shifting the point of highest CM towards a higher DAT (see red dots on Figure 1). Considering potential revenue for plant mass and costs associated with length of harvesting cycle and total required GA, maximum CM of US$15.40/m²·DAT is achieved when harvesting cycle is 19.4 DAT.

View Full Size

Normalized revenues and costs (top) and contribution margin (bottom) resulting from varying alternative days after transplant (DAT) and use of the corresponding required plant density and resulting plant mass. Maximum revenue is represented by a red line and maximum contribution margin is indicated by the red dot.

Citation: International Food and Agribusiness Management Review 27, 2 (2024) ; 10.22434/ifamr2023.0038

• Download Figure
• Download figure as PowerPoint slide

In this scenario, total GA is 800 m² (Table 2) and a total floor area of 272.8 m². Throughout the year, propagation density is constant at 1,550 seeds/m² and this is equal to the number of plants harvested per day in the minimum size area defined in the model (i.e., propagation area = 1 m²). Daily harvest reaches 118 kg, each plant head is harvested with approximately 76 g of mass, using a production density of 38 plants/m². As previously described, harvest occurs in one growing module per day, but, for overall comparison of costs associated with total production, total annual harvest, and total annual revenue are 53 kg/m² and US$962/m², respectively. Converting into imperial system for comparison, this hypothetical farm is producing 93 722 lbs. per annum utilizing a growing area of 8,611 ft², which results in an annual yield of 11 lb/ft², slightly lower than the annual 13.03 lb/ft² projection of a high-tech automated vertical farm described in the trade press reporting on the U.S. context (Amidi-Abraham, 2021). Similarly, Amidi-Abraham reported annual OpEx of US$1.47/lb excluding costs associated with installations. Another example was reported for a vertical farm operating in Canada, with total operational variable costs at US$1.46/lb. (Eaves and Eaves, 2018). Our analysis reports a similar annual variable OpEx of US$1.74/lb. of production.

Looking at annual labour and energy costs our study reports US$84/m² and US$60/m², respectively, which are significantly higher than the values reported in a Canadian scenario of US$38.11/m² in labour costs, and US$46.16/m² in energy costs (Eaves and Eaves, 2018), but their wages and electricity costs were lower than the U.S. context depicted here, at US$10.49/h and US$0.037/kWh. Packaging costs were estimated at US$15/m², which is significantly less than the US$24.29/m² reported by Eaves and Eaves (2018), but that value is strictly related to how produce is sold, which is not specified in that paper. Packaging options are endless with an equally wide range of costs associated with packaging, but at least 60% of the industry reported packaging costs from US$0.01 to US$1.49 per kg of produce (Agritecture and WayBeyond, 2021). In this simulation, packaging costs sit at US$0.28/kg.

All variable (VOC) and fixed operating costs (FOC) were affected by plant density associated with choices of harvesting cycle length when normalized by m² and DAT. Across all scenarios of increasing DAT, the marginal increase in most costs, except for packaging material costs, diminishes at a higher rate for lower DAT than it does for DAT higher than 15 days. Packaging costs presented a projection similar to harvest weight and revenue (Figure 1). Although total harvest weight increases with the size of plants, total annual harvest per m² and DAT is larger for the density applied in growing scenarios with mid-range DAT. The expansion of GA as a function of DAT seems to occur at a faster rate than plant growth rate at later days of DAT. Input costs showed a particularly steeper slope for scenarios with lower DAT when higher plant density is used. At maximum CM, VOC were distributed as 22% in seeds and substrate, 7% in packaging materials, 41% in labour costs, and 29% in electricity costs.

In terms of scalability, the model was designed to start with a minimum size base area. Therefore, the results represented the minimum size farm that can be projected in a modular manner to proportionally expand the GA. Any increases in the base area will proportionally project an increase in both VOC and FOC that will result in the same ratios of US$203/m² and US$459/m², respectively. In a more realistic scenario, larger facilities may be able to negotiate better prices with suppliers thus achieving further cost reduction and economies of scale.

5. Sensitivity analysis

The nature of the bioeconomic modelling framework integrates the relationship between plant biology, through growth rate at conditions specified in the production module, along with economic factors arising from costs and revenue associated with production and yield. With that, any change in parameters will unequivocally affect the estimation of the contribution margin for a given production schedule. Model exogenous factors are now tested for their impact on changes in the selection of the production schedule DAT and resulting GA that generates maximum contribution margin. Using one variable at a time (OAT) sensitivity analysis methodology, we estimate results by changing one variable at a time while other parameters are kept as a baseline value (Balesdent et al., 2016). The above-described framework becomes a function which uses inputs from a vector of d values for each variable of choice X = (x₁, x₂, …, x_d) to estimate a d × k matrix of output Y = (ϒ_ij)_{1 ≤ i ≤ k, 1 ≤ j ≤ d}. Each estimated ϒ carries a set of key indicators (k), namely highest CM, and GA, annual revenue, FOC, VOC at highest CM, for a range of input values (d = 20) used for seed prices, electricity prices, wages, packaging costs and lettuce market prices. The range of Y output is presented and analysed as the lagged differences in the value of key indicators ($/m² of total GA, or in m² in the case of GA) as a rate of change (ROC) per 1% stepwise increases in the value of each variable (see values in Table A3 in the Appendix).

Increases in the price of seeds had the least impact on all key indicators (Table 3). The spread of results for all indicators informed the potential impact of changes in seed prices (Figure 2). Although a wide distribution in GA was observed, up to an increase of 1 m² in GA for a 1% change at the higher range of seed prices. Overall, CM estimates in response to changes in seeds prices varied from US$16.35/m²·DAT at lowest prices to US$15.03/m²·DAT at the end of the price range. The cumulative effect was a growing area of 768 m² at the lowest seed price expanding to 815 m² at the end of the range. The best-scenario production schedule increased from 19 to 19.5 DAT.

View Full Size

Mean rate of change in key indicators for each 1% change in prices of seeds, electricity, wages, packaging and lettuce prices

Citation: International Food and Agribusiness Management Review 27, 2 (2024) ; 10.22434/ifamr2023.0038

• Download Figure
• Download figure as PowerPoint slide

View Full Size

Distribution of changes in key indicators in response to 1% increases in the price of seeds, electricity, wages, packaging and in the price of lettuce. All key financial indicators are normalized as a measure of dollar per square meter ($/m²) of total growing area (GA).

Citation: International Food and Agribusiness Management Review 27, 2 (2024) ; 10.22434/ifamr2023.0038

• Download Figure
• Download figure as PowerPoint slide

Between all costs, electricity was second to wages in the effect on all three key indicators, affecting CM negatively by up to US$0.09/m²·DAT for 1% increase in electricity prices at the end of the range (Figure 2). Total VOC progressively increased from US$0.17/m² for each 1% increase in prices to nearly an order of magnitude larger value of US$1.65/m² at the end of the range of electricity prices. The impact on GA was positive leading to 858 m² required to achieve a CM of US$9.23/m²·DAT when electricity costs reach US$0.303/kWh. The DAT remained between 19.2 and 19.9 days in production.

On average, 1% changes in wages had the largest impact on all key indicators comparatively with changes in all other costs, increasing VOC by almost US$1/m² for each 1% change in wages (Table 3). Such a strong effect on VOC, lead to a longer DAT, which requires a larger GA causing corresponding large effect on FOC. Resulting CM falls from US$17/m²·DAT to US$11/m²·DAT at the upper end of the price range. Changes in VOC were steeper (in US$/%) as wages increased, ranging from US$0.51/m² to US$1.46/m² for each 1% change in wages at higher wage values. Production cycle increased from 19.2 to 20 DAT to compensate for higher labour costs.

The negative impact of packaging costs on CM was small, with a maximum of US$0.06/m²·DAT (Figure 2). The most significant impact of packaging costs, within this range of prices, was on VOC, which increased by $0.618/m² on average per 1% increase in packaging material prices (Table 3). At the maximum price, VOC increased to a maximum of US$1.18/m².

More significant changes in CM, annual revenue, FOC and GA were observed with changes in the market price for lettuce than any changes in costs (Table 3). On average, CM increased by US$0.62/m²·DAT for 1% change in lettuce prices. However, it took a negative value for lettuce prices below US$22/kg. Economically viable farms considering production settings and costs applied here will nevertheless observe a rapid increase in CM from US$3/m²·DAT to US$68/m²·DAT. Annual revenue increased on average by US$11/m² for each stepwise increase in lettuce market prices. At the lower range of prices, a larger GA of 1,292 m² was required to achieve higher CM (see lower tail of GA distribution plot in Figure 2). Prescribed DAT moved from 23 days at lower prices to 17.6 days at maximum prices. Market prices affected VOC by US$0.46/m² for each 1% change in prices at lower prices, reducing to US$0.23/m² at higher price range. Revenues were directly and significantly affected by changes in lettuce prices, up to US$17/m² at maximum prices.

6. Conclusions

The success of the indoor agriculture (IA) industry depends ultimately on the understanding of the synergy between technology and biological responses to inputs. This study used a bioeconomic modelling framework to define the best production schedule for minimizing initial capital investments and maximizing profitability, considering plant growth rates and efficient space utilization, and comparing results from cost minimization and revenue maximization strategies.

Using a measure of contribution margin to profit (CM), which was estimated as the difference between operating costs and total revenues, on a normalized scale per area and per day of production schedule (US$/m²·DAT), highest CM was identified to occur when plant size was 76 g, grown with a density of 38 plants/m² and harvested on the 19th day after transplanting (DAT). The model, constructed with a minimum base area of 1 m², estimated the growing area (GA) to be 800 m², and requires a floor area of 273 m², which makes it suitable for installation in urban areas. Given average market prices and U.S. average values for the most relevant operating costs, this commercial farm harvests 118 kg a day, and achieves a total revenue of US$769,804 per annum. Normalized operating costs had a more significant impact on profitability at lower DAT, while the rate of increase in revenues decelerated at mid-range DAT, as lower DAT implies higher plant density and therefore a smaller growing area. On the other hand, the speed at which increasing DAT expands growing area was faster than the actual plant growth rate, which led harvest and directly associated normalized revenue to decline for DAT larger than 20 days.

While these results are intrinsically dependent on growing methods and conditions, the bioeconomic framework described here allowed for estimating economic parameters for a hypothetical commercial lettuce indoor farm, and for a further analysis of these parameters’ sensitivity to exogenous factors, such as input costs and market lettuce prices. This sensitivity analysis gave further insights on the choice between management strategies. Lettuce market prices presented the highest impact on DAT decision, and consequent GA, and had the greatest effect on CM. On the other hand, oscillations in input costs had a small impact on the best-scenario DAT, changing GA by, on average, less than 1 m² per 1% change in costs. These results suggest that a cost reduction strategy would be less effective than a premium market targeting strategy. Within the realm of costs, labour cost, accounting for 41% of the total variable operating costs (VOC) in the hypothetical farm, had the most significant impact on all key indicators (i.e., CM, VOC, and GA) comparatively with other variable costs.

For the moment, this framework defines a foundational production function based on plant density and production schedule for lettuce in the U.S. and sets the scene for industry development from smaller sized farms, which are suitable for installation in urban areas. Given the complexity of this biological, technical, and economic system, this analysis suggested that starting at smaller scale may mitigate risks associated with oscillations in VOC, lettuce market prices, and large capital investments in technology. Further research can also explore alternative growing systems adopting extra tiers of shelving, production settings, including “crowding”, optimizing CO₂, or lighting parameters, and their specific effect on profitability, as well as analyse resource use efficiency and economies of scale. Finally, further research on returns on equity (ROE) is recommended to forecast the long-term economic sustainability in this industry, incorporating independent analyses of net profit margin, asset turnover, and financial leverage. Recognizing the substantial link between investment uncertainty and irreversibility is essential in the context of indoor agriculture.