Introduction
As worldwide participation in efforts combating climate change and rising CO2 emissions expands, there is increasing recognition of the need to encourage consumer food choices that are environmentally sound. However, within this framework of consumer choice there is a constant tug in the popular literature between focusing on a seasonal, local dietReference Pollan 1 , Reference Smith, MacKinnon and Smith 2 or more simply on a healthier diet with a greater preponderance of fruits and vegetables 3 . Developments in post-harvest management, breed development, refrigeration and other strategies have made it feasible to generate out-of-season, fresh produce from anywhere for anywhere. Consumers have largely come to expect, demand and feel entitled to this pristine produce. The complex and lengthy supply chains necessary for year round access are particularly relevant in the colder regions of the US, where winter weather seemingly prevents sustainable, local production of fresh fruits and vegetables. National and international produce sourcing has raised environmental concerns about the purely nutrition-focused diet.
The ‘eat local’ movement has gained significant strength in the past 15 years, often in the name of environmental soundness especially centered on the notion of ‘food miles’. However, this notion is itself contested, as not only the miles but the efficiency of transportation per unit of food is also a consideration in analyzing C-footprints—or overall carbon emissions. Also, the focus on ‘localization’ does not necessarily account for the food production method or the choice of food items (i.e., the overall food on the plate) in one's diet. One study reported that 83% of the greenhouse gas (GHG) emissions from US food supply are derived from the production phase, while only 11% stem from transportationReference Weber and Matthews 4 . This same study estimated that buying all local foods would only reduce GHG emissions from an average diet by 4–5%, suggesting that the best reduction strategy lies in food product choiceReference Weber and Matthews 4 .
As this debate continues, life cycle assessment (LCA) as ‘a technique to assess the environmental aspects and potential effects associated with a product, process, or service, by compiling an inventory of relevant energy and material inputs and environmental releases; evaluating the potential environmental effects associated with identified inputs and releases; and interpreting the results to help [make] a more informed decision’ has become an increasingly important tool for identifying strategies and system components for improvement 5 . As reviewed by Saunders et al., most of the early studies conducted using ‘food miles’ compare only transportation emissions between local and imported foods, while a thorough LCA involves quantification of both production and transportation effects through its ‘cradle to grave’ analytical approach, providing a more robust comparison of similar food productsReference Saunders, Barber and Taylor 6 . This study, while utilizing LCA's ‘cradle-to-grave’ framework and a leading LCA software package, takes an approach focusing on carbon footprint rather than an overall environmental impact analysis in the hope of capturing relative contributions to climate change.
Few LCAs of fruits and vegetables have been conducted worldwide. A nearly cradle-to-grave study performed in Germany comparing domestic to imported apples found imported apples to contain higher embedded energyReference Blanke and Burdick 7 . Carrots produced locally in Sweden were found to incur fewer GHGs than when importedReference Carlsson-Kanyama 8 . However, Swedish tomato production demonstrated a reverse locational effect with local tomatoes incurring higher GHGs primarily due to differences in greenhouse heating efficiencyReference Carlsson-Kanyama 8 . Another study compared salad crop production for UK markets, finding that importation from Spain demonstrated a lower global warming potential (GWP) than indoor production in the UK, but a higher GWP than outdoor UK productionReference Canals, Muñoz, Hospido, Plassmann, McLaren, Edwards-Jones and Hounsome 9 .
Lettuce production and consumption is particularly interesting to study since 98% of US lettuce consumed is sourced from domestic production, nearly all of which occurs in California and ArizonaReference Boriss and Brunke 10 , and thus large amounts of lettuce are routinely shipped across the country. This will probably increase in the absence of fundamental production-location change as per capita leaf lettuce consumption has been increasing in the past decadeReference Boriss and Brunke 10 .
One LCA of lettuce's environmental footprint has been conducted to date in a European contextReference Hospido, Milà i Canals, McLaren, Truninger, Edwards-Jones and Clift 11 . The analysis compared GHG emissions of summer UK field-grown lettuce production, winter UK heated greenhouse production and summer and winter field-grown lettuce production in Spain. All were modeled based on UK consumption. Locational effects on GHG emissions were consistent with research on other produce, where local systems proved more favorable to importation except when using a heated greenhouseReference Carlsson-Kanyama 8 , Reference Canals, Muñoz, Hospido, Plassmann, McLaren, Edwards-Jones and Hounsome 9 .
Recognizing that energy inputs in greenhouse systems contribute substantially to any analysis of local off-season production in cold climates, there has been significant work in the US to develop unheated hoop house systems for winter production of certain vegetables. This began with the work of Eliot ColemanReference Coleman and Damrosch 12 and has been significantly expanded with peer-reviewed researchReference Carey, Jett, Lamont, Nennich, Orzolek and Williams 13 . One earlier study compared energy use by tomato cropping systems in varying climates without regard to transportation, finding that highly mechanized field production in California utilized 7.5 times less energy per hectare than Israeli hoop house tomato productionReference Stanhill 14 . However, to date there has been no analysis to determine if hoop houses represent advancement from an environmental footprint standpoint over cross-country importation in production of crops that are fairly cold tolerant and do not require supplemental light. We therefore present a carbon footprint assessment of two different methods/locations of lettuce production, incorporating both the production and transportation footprints. The production methods include an organic hoop house production system in Michigan, USA and that of open-field, conventional production in a southwestern US desert climate. Hoop houses, structures made of plastic or metal pipe and covered with plastic or other sheeting, present an interesting mode of production because they do not incur the same heating costs as conventional greenhouses yet are capable of producing a variety of winter crops. We hypothesized that due to reduction in energy for heating in winter production these structures could be environmentally advantageous relative to long-distance importation.
Methods
Site and product dependency
Agricultural LCA results are highly site-dependentReference Carlsson-Kanyama 8 . Comparisons between food types are particularly difficult given differences in production methods and boundary conditions for analysis. For example, GHG emissions from fruit and vegetable production/distribution vary drastically from others, such as cereals or products associated with trophic magnification like meat and dairyReference Weber and Matthews 4 . In part because of wide variations in production inputs and shipping mechanisms for each crop or food product and in part due to the relative paucity of a robust set of LCAs across crop types, it is necessary to individually assess effects rather than extrapolating results from another crop. The same principle applies to geographical locations. The study detailed here significantly contributes to this growing dataset of agricultural LCAs by comparing the environmental effects of unheated winter lettuce production in cold and warm climates to supply a cold climate market. Thus, the sustainability of outdoor and hoop house lettuce production—as components of a supply chain complete to retail outlet—are evaluated using two International Standards Organization (ISO)—compliant environmental impact assessment methods (GHG Protocol and Eco-Indicator 99).
System boundaries
The system boundaries for this study encompass manufacturing of building materials and fertilizers through transportation of lettuce to the retailer (Fig. 1). More specifically, effects are calculated for the following:
-
• Manufacturing and transportation of the hoop house structure
-
• Manufacturing of the fertilizer (or compost in the Michigan case), including fuel and electricity use
-
• Manufacturing of irrigation materials, including fuel and electricity use
-
• Fertilizer application
-
• Use of agricultural machinery for land preparation (tractor for outdoor; rotary cultivator for indoor), including fuel use
-
• Electricity use in the hoop house (e.g., for automatic ventilation, irrigation)
-
• Irrigation, including water use and electricity
-
• Processing, transportation and storage of lettuce between the farm and the retailer
Figure 1. System boundaries of the carbon footprint analyses of leaf lettuce produced in a Michigan hoop house and California desert (thick line).
Excluded in the study are effects resulting from upstream processes such as manufacturing of agricultural machinery, seeds, lorries or passenger cars. Downstream processes not included are: storage at the retailer, transportation of lettuce from retailer to consumer, home storage and use and waste scenarios—the assumption being made that these will simultaneously be identical independent of product source and extremely variable based on location of consumer to retailer as well as transportation patterns of the consumer. For example, one study showcased this uncertainty in the emissions from the final retailer to consumer transport, demonstrating that it could range from 0 g CO2 per kg of food (if the consumer walks) to 12.1 kg CO2 per kg food (if the consumer drives 15 km for the sole purpose of shopping)Reference Van Hauwermeiren, Coene, Engelen and Mathijs 15 . Other studies have excluded final transport due to this variability in mode of transport, vehicle fuel efficiency and number of items in the consumer's hypothetical shopping cartReference Carlsson-Kanyama 8 .
Strategies and assumptions in cold-climate hoop house production scenario
We modeled this system using the hoop house most frequently recommended by the staff at Michigan State University's (MSU) Student Organic Farm (A. Montri, personal communication, October 2010; J. Biernbaum, Michigan State University, personal communication, October 2010), a 30′×96′ (9.14 m×29.26 m) steel, aluminum and wood frame with a double layer of polyethylene film. The Student Organic Farm hosts four houses of this exact type for research and production purposes, and the MSU Extension team assists in building this type of house across Michigan, making this assumption particularly relevant. One of the hoop houses produces greens for a campus dining hall through the winter. Data describing the weight of the hoop house structure's components (steel, polyethylene and aluminum) were provided by the manufacturer. These weights were then amortized over the material's projected life span: the steel (1817.09 kg) and aluminum (65.3 kg) used in the hoop house frame are assumed to have 40-year life spans, the polyethylene film (65.3 kg) used to cover the hoop house is assumed to have a 4-year life span, while the pine boards used as the structure's base posts are assumed to have a 10-year life span (A. Montri, personal communication, October 2010). The inflation fan used to separate the two layers of polyethylene film is assumed to run constantly. The specifications for the fan's electrical use were found on the manufacturer's website 16 . Fuel use during transportation of the hoop house to the farm was estimated using the location of the manufacturer and was adjusted for a 40-year structural life span and full truckload.
Hoop house input, timing of production, yields, etc. were provided by field experts based on their experiences with these systems (A. Montri, personal communication, October 2010; J. Biernbaum, Michigan State University, personal communication, October 2010). The first lettuce crop was assumed transplanted into the hoop house in East Lansing, Michigan on September 18 and harvested in early December. The second crop was planted in early February and harvested in early April, with each harvest yielding 4000 heads for a total of 8000 heads during the Michigan off-season (A. Montri, personal communication, October 2010). This yield is the result of planting 70% of the hoop house area with plants in 8″×8″ squares.
The leaf lettuce was assumed irrigated every 7–10 days between 10.00 hours and 14.00 hours using a drip irrigation system. The irrigation system included the PVC pipeline and a frost-free hydrant (J. Biernbaum, Michigan State University, personal communication, October 2010). Water use was estimated as: 1.18 m3 week−1 between September and December, 0.47 m3 week−1 between February and March, and 1.18 m3 week−1 during the first week of April (A. Montri, personal communication, October 2010).
The only land preparation in this scenario was tilling the 185.8 m2 planted area by rototiller four times over the course of the season; the estimated time was 2 h and gasoline use for this period was included (A. Montri, personal communication, October 2010). The only soil amendment used in this scenario was compost, produced at a nearby facility rather than at the farm itself. Emissions from transporting compost to the site are not calculated, nor are tractor emissions from application. This study does not calculate emissions from pest management techniques.
It should be noted that while this hoop house is presumed to only produce leaf lettuce, this is not realistic. Diversified production is much more common on farms of this sort. This assumption was made to simplify the comparison process between production methods.
Strategies and assumptions in warm climate production scenario
Data for the warm climate (California) production process was primarily based on two previously compiled reports on southwestern lettuce productionReference Smith, Cahn, Daugovish, Koike, Natwick, Smith, Subbarao, Takele and Turini 17 , Reference Takele, Aguiar and Walton 18 .
Leaf lettuce in this scenario was produced on two 121-hectare land plots in California. The first crop was grown in Imperial County, California between mid-September and early DecemberReference Smith, Cahn, Daugovish, Koike, Natwick, Smith, Subbarao, Takele and Turini 17 . The second crop was grown in the Central Valley region of California between late December and early AprilReference Smith, Cahn, Daugovish, Koike, Natwick, Smith, Subbarao, Takele and Turini 17 . These planting to harvest lengths are close to the suggested 130-day growing period for this region and time of year. The marked geographical difference complicates the impact analysis only with respect to irrigation levels, discussed below. Reported leaf lettuce harvest in the southwestern US during the winter varies between 51,900 and 66,700 heads per hectare. We assumed a 59,300 heads per hectare yield, amounting to 14.4 million heads in total over the course of the season. The farm production is assumed to use conventional practices including pre-irrigation prior to cultivation. The crop was hand-weeded after 30 days and again after 45–50 days of growthReference Smith, Cahn, Daugovish, Koike, Natwick, Smith, Subbarao, Takele and Turini 17 .
Electricity use and manufacturing costs associated with irrigation are contained within the ‘Irrigating’ process of SimaPro 7.3's US EcoInvent 2.2 database used for this analysis. Industry average water use data suggests a total of 7619.98 m3 hectare−1 for leaf lettuce production in Imperial County, California and a value of 9144.00 m3 hectare−1 for Central Valley, CaliforniaReference Smith, Cahn, Daugovish, Koike, Natwick, Smith, Subbarao, Takele and Turini 17 . Thus, the average of the two values was taken, and water use of 8381.99 m3 hectare−1 was assumed.
Average industry data suggest that 280.2 kg hectare−1 of phosphorus fertilizer and 250.1 kg hectare−1 of nitrogen fertilizer are applied to leaf lettuce crops in this region 19 . Fertilizer manufacturing is accounted for in the SimaPro database, including industry-average models of fertilizer transportation from storehouse to farm.
Diesel use during land preparation is based on the 1996 UC study on loose-leaf lettuce productionReference Takele, Aguiar and Walton 18 . Excluding the use of machinery for pesticide application, they calculated 32.79 machine-hours per hectare during the first production cycle (September to December) and 3.81 machine-hours per hectare during the second production cycle (January to April). A 200 horsepower 4-wheel drive tractor performed plowing, ripping, stubble disking, disking and listing beds prior to the first crop's seeding. After seeding, this tractor spread the fertilizer. During the second production cycle, the only use for the tractor was in disking and fertilizing.
Energy use during lettuce processing and storage was quantified based on a 2005 study comparing the energy requirements embedded in local versus imported applesReference Blanke and Burdick 7 . The authors’ recommendation for the energy requirement of initially cooling apples is 0.086 MJ kg−1. The apples’ cooling process is assumed to be comparable to the customary initial vacuum cooling stage for US leaf lettuce, and is so applied (as 0.043 MJ per head). The authors also recommend the energy use of a refrigerated truck at 0.3 MJ t−1 km−1 (0.055 MJ kg−1). This study adopts this value and considers it a contribution to transportation energy requirements. A 2008 LCA of European salad crop production cites this same total processing and storage energy contribution (0.141 MJ kg−1 or 35.5 kWh ton−1 lettuce) in the case of the study's most efficient observed farmReference Canals, Muñoz, Hospido, Plassmann, McLaren, Edwards-Jones and Hounsome 9 . Since the authors caution against citing this direct farm measurement alone, the apple study is considered a validation of the 0.141 MJ kg−1 value.
The leaf lettuce was transported 3605 km directly from the farm in southern Central Valley California to the retailer in mid-Michigan. Since this is a simplification of real-life vegetable supply chains, it probably underestimates the distance traveled. A >32-ton lorry is assumed to carry a 35-ton truckload of leaf lettuce to the retailer 19 . The route taken by the truck after the initial drop-off incurs a great deal of uncertainty, and is therefore omitted. The implications of these transportation assumptions are further discussed in the results of this study.
Results
The carbon footprint analyses were conducted using impact assessments within SimaPro 7.3. This software is designed to conduct impact assessments using four basic steps, although the last three are optional and depend on the assessment method 20 . The first, characterization, determines which substances contribute to an impact category (e.g., acidification) and at what level. The next, damage assessment, combines impact categories into damage categories (e.g., adding respiratory inorganics and carcinogens to human health damage). Normalization compares the impact contributions to some specified reference impact; this study skipped the normalization step. Lastly, weighting applies weighting factors to each category in order to obtain a total or overall impact score.
SimaPro 7.3 was chosen for its inclusion of the US-EI 2.2 database, which reports EcoInvent 2.2 industrial data using US electrical production as opposed to European. Also, the EcoInvent 2.2 and US-LCI databases have a wide selection of agricultural data. Impact assessments were performed using the GHG Protocol and Eco-Indicator 99 methods. Though the focus of this study is on the production methods’ contribution to climate change, other impact categories are used to verify the results of environmental impact rankings.
Throughout the analysis one head of leaf lettuce is assumed to weigh 0.5 kg. Results of the impact assessments performed within SimaPro 7.3 are reported in terms of this study's functional unit: 1 kg of leaf lettuce. It should be kept in mind that lettuce yields, fertilizer use and irrigation requirements are specific to the soil type and temperature characteristics of mid-Michigan and the southern deserts of California as well as precipitation levels in California. Also, the results of this study are largely only applicable to leaf lettuce and crops with similar growing requirements and yields.
GHG protocol
The GHG Protocol, while not a full life cycle impact assessment method, is the most widely used GHG accounting tool in the world, both in academic and private sectors. Since this study's focus is on overall carbon impact, GHG Protocol was the appropriate framework for analysis. The results are reported in kg CO2 per unit.
Hoop house impact assessment
Producing 1 kg of leaf lettuce using an unheated greenhouse, with the system conditions outlined earlier, results in emissions of 0.198 kg CO2. The largest contributor to this impact stemmed from the energy embedded in the hoop house structure (Fig. 2). 23.2% of total emissions came from the steel framework, and 8.31% from transport of the house from Georgia to Michigan (adjusted for a 6-house load and a 40-year life span). The hoop house framing and plastic film were of less importance, accounting for 2.58% and 5.24%, respectively. Irrigation, namely PVC pipes with 14.9% of irrigation's total 24.8% emissions contribution, and fan electricity played roughly equal parts in the impact assessment. Almost 90% of 7.45% for land preparation emissions originated with compost production rather than rototiller usage.
Figure 2. Contribution to hoop house emissions by stage.
Outdoor impact assessment
Producing 1 kg of lettuce in California according to the previously described method then shipping it to Michigan results in emission of 0.857 kg CO2. The final transport of lettuce to Michigan, which includes lorry operation, diesel use and refrigeration accounts for over three-quarters of this value. 48.2% of total emissions came from truck use, while 29.5% of the total came from refrigeration electricity. These values are most likely underestimates of the true values, as the supply chain assumed here was simplified to exclude diversions to regional hubs. Also, over three-quarters of the land preparation emissions stemmed from nitrogen fertilizer production and application. Tractor use thus represents a very small portion of the overall system emissions. A summary of the major process contributions is illustrated in Fig. 3.
Figure 3. Contribution to California emissions by stage.
Production comparison
Total emissions from California production were 4.3 times those of the hoop house production in Michigan (Table 1). In fact, emissions from refrigerated transport were greater than that of the entire Michigan hoop house system. These two systems exhibited comparable irrigation-related emissions: 0.073 kg CO2 in the California scenario and 0.05 kg CO2 in the Michigan scenario. Comparing all other production stages, the Michigan hoop house scenario was found to have a smaller carbon footprint (Table 1). In the case of added electricity, the energy needed to continuously run the hoop house fan was about one-fifth of the energy needed to vacuum cool and refrigerate the lettuce during shipment. SimaPro's ‘Lorry operation’ process represented 48.2% of the California system emissions, but only 5.15% of the hoop house system. In real terms, the per kg emissions due to lorry operation from California to Michigan was 25 times that of the lorry operation for hoop house delivery from Georgia to Michigan amortized over a 40-year span.
Table 1. Carbon emission comparison.
1 California's impact assessment displays these emissions as one ‘Transport’ category. Numbers are separated here to highlight refrigeration contribution.
Results could be altered by future additions of packaging effects. One head of local, hoop house lettuce is typically packed in a single food grade plastic bag, while shipped lettuce would be packed in a 24-heads per waxed cardboard box. While the packaging regimen is assumed to be negligibly different with respect to overall carbon footprint it would increase the nominal value of each system's emissions.
Eco-Indicator 99
The EI '99 analytical method takes a ‘damage-oriented’ approach and evaluates a ‘product or process’ sustainability based on the damages it causes in each of three categories: human health, ecosystem quality and resources. The first, human health, measures the years of life lost due to resource use or emissions related to the product (Disability Adjusted Life Years or DALYs). Damage is a function of carcinogens, respiratory inorganics and organics, climate change, radiation and ozone layer depletion 20 . Ecosystem quality measures, through ecotoxicity and acidification/eutrophication, species loss within a specified area over a certain period of time. ‘Resources’ measures the energy necessary for future extraction of fossil fuels and minerals that must compensate for the overuse of resources today 20 . Damages are weighted according to the hierarchist perspective, a long-term perspective that weights equally present and future damages. As such, ecosystem quality is weighted at 40%, human health at 30% and resources at 30%Reference Goedkoop and Spriensma 21 . The results are reported as both (1) weighted collective damage attributed to each production stage and (2) weighted damage per impact category caused by each production method.
Resource depletion, over 90% of which is fossil fuel use, was the most significant source of ecological damage in the hoop house case, whether this was from truck diesel use, electrical grid powering or rototilling. The most environmentally damaging production stage was, as the first impact assessment method predicted, construction of the hoop house. Power generation and diesel consumption emitted high levels of respiratory inorganics emissions, contributing to the majority of human health effects especially within the infrastructure stage.
The most environmentally damaging stage in the California supply chain was transportation. This was mostly due to fossil fuel consumption (Resources) and its ensuing respiratory inorganics emissions, climate change impact and carcinogenic effects (Human Health). Tractor use and power generation translated into similarly distributed categorical effects within land preparation and irrigation. However, transportation's eco-factor points in this assessment were roughly five times larger than any other stage. Cooling had a negligible impact on this assessment.
Not surprisingly, the distant, outdoor lettuce production exhibited higher eco-factor points than the local, hoop house lettuce production in every impact category, though the relative importance of each category remained consistent between production methods (Fig. 4). It is notable that the top four impact categories all relate to transportation. This is consistent with the first impact assessment, where transportation was the key production stage, and distant, outdoor production was more environmentally damaging per kg of lettuce.
Figure 4. Comparison of EI '99 by impact category for winter lettuce production.
When comparing the scenarios by damage assessment category, the distant, outdoor system incurs over four times the damage the local, indoor system does (89.11 mPt versus 21.10 mPt). Though the difference in ‘ecosystem quality’ is slight, the differences in ‘resources’ and ‘human health’ are substantial.
Discussion
This study provides a cogent analysis of two production systems—one outdoor at a distance to the consumer and one indoor in proximity to the consumer. In these analytical frameworks it is always a challenge to define the boundary conditions in such a way that the analysis is rigorous yet achievable. We purposefully did not include travel of the consumer to the point of purchase. One UK study sought to find the distance a consumer must drive to a farm such that the per unit carbon emissions were greater than that of a large-scale vegetable box delivery system that included packaging and cold storage (without regard to production method)Reference Coley, Howard and Winter 22 . The results illustrated that as long as the consumer did not drive more than 7.4 km (4.6 miles) to the farm, the system was more environmentally friendly than the box system.
In the research reported here, although transportation from retailer to consumer is not included within this study's system boundaries, we did conduct a simple hypothetical scenario using the GHG Protocol method to gauge the significance of a potential expansion in miles driven on the overall environmental impact. A process was added to each system that models a 5-mile (one way) shipment of 1 kg of lettuce in a passenger car (with no other groceries purchased or ancillary stops during the trip). Under this scenario, the discrepancy between distantly-produced and locally-produced fossil CO2 equivalents was the same as before the process was added. However, if the locally-produced lettuce travels one mile farther than the distantly-produced lettuce, the overall impact is then greater by at least 0.05 kg CO2 (4.3 kg versus 4.25 kg). The fossil fuel use (Eco-Indicator 99) is also pushed above the distantly-produced value. This implies a strong impact of consumer decisions over transportation routes, multiple use trips and sourcing decisions relative to overall system sustainability. Removing this influence enabled a clearer and more convincing comparison between upstream emissions associated with each production method.
The cold-climate model contains particularly sensitive assumptions that, when adjusted, are likely to affect the overall production system impact. For instance, the electricity mix modeled in SimaPro is an average US mix of 44% coal. Since Michigan's electricity mix utilizes closer to 66% coal, the electricity contributions (from irrigation and fan operation) are probably higher than reported. However, as the fan would probably not be running during production breaks, the reported numbers are slightly inflated. Even if the electricity contributions were doubled from 0.074 to 0.148 kg CO2 the overall local system emissions would still not compare to the distant systems. Also, since the steel framework accounted for a quarter of the overall impact, more efficient manufacturing techniques could significantly alter the local systems’ emissions. Location of production (i.e., sourcing a more local hoop house) would have a less significant effect since the hoop house's transport contributed only 8% of emissions.
Nonetheless, the results here provide a firm basis for considering greater scale off-season production in unheated hoop house systems close to points of consumption. Even considering a worst-case scenario for hoop house emissions, shipping conventionally grown leafy greens to cold-climate areas appears to incur greater environmental effects than a local system, given sufficient distance. A simple calculation reveals that the conventionally grown leafy lettuce incurs greater system-wide emissions at farm–retail transport distances above 37.9 km, assuming that the consumer drives the same distance to retail in both conventional and local cases in an ‘average’ vehicle. This indicates that conventional lettuce production, only when part of a truly localized supply chain, would incur fewer kg CO2 per hectare. This result is consistent with both the UK lettuce seasonality study, as it found local field-grown lettuce to incur fewer GHG emissions than imported lettuceReference Hospido, Milà i Canals, McLaren, Truninger, Edwards-Jones and Clift 11 , and a study finding imported apples to contain higher embedded energy than domestic applesReference Blanke and Burdick 7 .
However, this study contradicts a recent Belgian report arguing that the processing, storage and transport of cabbage lettuce (among other products) have a higher carbon footprint as part of a local food system than a mainstream food systemReference Van Hauwermeiren, Coene, Engelen and Mathijs 15 . Their results, though they do not include the final transport from retailer to consumer, would suggest that these stages have a higher carbon footprint in a non-conventional supply chain. In comparing the transport and cooling stages of the distant, outdoor system with the local, hoop house transport process in the Michigan system, the distant, outdoor system is worse by a margin of 0.650 kg CO2 per head of lettuce. Also, the Weber and Matthews study presented earlier suggested that the transportation emissions for fruits and vegetables were a very minor value compared to GHG emissions from productionReference Weber and Matthews 4 . As 77.7% of the distant, outdoor system emissions were the result of transportation, this study clearly contradicts that result in cases of leaf lettuce and crops with similar production footprints. In sum, this study found the percent of total emissions due to transport to be higher than previous studies, but the end result of a local food system having lower carbon footprint (except in the case of production in a heated greenhouse) is generally consistent with other researchReference Carlsson-Kanyama 8 , Reference Canals, Muñoz, Hospido, Plassmann, McLaren, Edwards-Jones and Hounsome 9 , Reference Hospido, Milà i Canals, McLaren, Truninger, Edwards-Jones and Clift 11 .
As for the ‘eating local’ debate, this study supports the claim that production location can be a key determinant of a supply chain's sustainability (at least with respect to environmental components and dependent on product type), but it is important to incorporate production methods within the system analyzed. It is also clear, however, that transportation issues between the retail space and the home can play a major role in tipping the balance on environmental sustainability indicators.
Conclusion
The results of this study suggest that unheated, hoop house lettuce production has less environmental impact, (and in particular carbon footprint) than outdoor, distant production, and speaks to the potential value of more localized food systems. Still, hoop house sustainability could possibly be improved by cleaner steel manufacturing or possibly by replacing steel with plastic or biobased materials. This requires more research. Although the physical and economic feasibility of large-scale hoop house production remains to be quantified, this study has demonstrated that hoop houses should be included in any discussion about reducing carbon footprints by sourcing appropriate produce locally during the winter.
